Google Git
Sign in
llvm / llvm-project / 7c706af03b8634f1f2a64599037580e5f4785082 / . / llvm / test / CodeGen / AMDGPU / GlobalISel / fshr.ll
blob: 91655dfffd7256e633577319eb57e15ee2351f95 [file] [log] [blame]
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=tahiti -o - %s | FileCheck -check-prefixes=GCN,GFX6 %s
; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=fiji -o - %s | FileCheck -check-prefixes=GCN,GFX8 %s
; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 -o - %s | FileCheck -check-prefixes=GCN,GFX9 %s
; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 -o - %s | FileCheck -check-prefixes=GCN,GFX10 %s
define amdgpu_ps i7 @s_fshr_i7(i7 inreg %lhs, i7 inreg %rhs, i7 inreg %amt) {
; GFX6-LABEL: s_fshr_i7:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_cvt_f32_ubyte0_e32 v0, 7
; GFX6-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX6-NEXT: s_sub_i32 s3, 0, 7
; GFX6-NEXT: s_lshl_b32 s0, s0, 1
; GFX6-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX6-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX6-NEXT: v_mul_lo_u32 v1, s3, v0
; GFX6-NEXT: s_movk_i32 s3, 0x7f
; GFX6-NEXT: s_and_b32 s2, s2, s3
; GFX6-NEXT: s_and_b32 s1, s1, s3
; GFX6-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX6-NEXT: v_add_i32_e32 v0, vcc, v0, v1
; GFX6-NEXT: v_mul_hi_u32 v0, s2, v0
; GFX6-NEXT: v_mul_lo_u32 v0, v0, 7
; GFX6-NEXT: v_sub_i32_e32 v0, vcc, s2, v0
; GFX6-NEXT: v_subrev_i32_e32 v1, vcc, 7, v0
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 7, v0
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX6-NEXT: v_subrev_i32_e32 v1, vcc, 7, v0
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 7, v0
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX6-NEXT: v_sub_i32_e32 v1, vcc, 6, v0
; GFX6-NEXT: v_and_b32_e32 v0, s3, v0
; GFX6-NEXT: v_and_b32_e32 v1, s3, v1
; GFX6-NEXT: v_lshl_b32_e32 v1, s0, v1
; GFX6-NEXT: v_lshr_b32_e32 v0, s1, v0
; GFX6-NEXT: v_or_b32_e32 v0, v1, v0
; GFX6-NEXT: v_readfirstlane_b32 s0, v0
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i7:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_cvt_f32_ubyte0_e32 v0, 7
; GFX8-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX8-NEXT: s_sub_i32 s3, 0, 7
; GFX8-NEXT: s_lshl_b32 s0, s0, 1
; GFX8-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX8-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX8-NEXT: v_mul_lo_u32 v1, s3, v0
; GFX8-NEXT: s_movk_i32 s3, 0x7f
; GFX8-NEXT: s_and_b32 s2, s2, s3
; GFX8-NEXT: s_and_b32 s1, s1, s3
; GFX8-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v1
; GFX8-NEXT: v_mul_hi_u32 v0, s2, v0
; GFX8-NEXT: v_mul_lo_u32 v0, v0, 7
; GFX8-NEXT: v_sub_u32_e32 v0, vcc, s2, v0
; GFX8-NEXT: v_subrev_u32_e32 v1, vcc, 7, v0
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 7, v0
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX8-NEXT: v_subrev_u32_e32 v1, vcc, 7, v0
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 7, v0
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX8-NEXT: v_sub_u16_e32 v1, 6, v0
; GFX8-NEXT: v_and_b32_e32 v0, s3, v0
; GFX8-NEXT: v_and_b32_e32 v1, s3, v1
; GFX8-NEXT: v_lshlrev_b16_e64 v1, v1, s0
; GFX8-NEXT: v_lshrrev_b16_e64 v0, v0, s1
; GFX8-NEXT: v_or_b32_e32 v0, v1, v0
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i7:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_cvt_f32_ubyte0_e32 v0, 7
; GFX9-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX9-NEXT: s_sub_i32 s3, 0, 7
; GFX9-NEXT: s_lshl_b32 s0, s0, 1
; GFX9-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX9-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX9-NEXT: v_mul_lo_u32 v1, s3, v0
; GFX9-NEXT: s_movk_i32 s3, 0x7f
; GFX9-NEXT: s_and_b32 s2, s2, s3
; GFX9-NEXT: s_and_b32 s1, s1, s3
; GFX9-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX9-NEXT: v_add_u32_e32 v0, v0, v1
; GFX9-NEXT: v_mul_hi_u32 v0, s2, v0
; GFX9-NEXT: v_mul_lo_u32 v0, v0, 7
; GFX9-NEXT: v_sub_u32_e32 v0, s2, v0
; GFX9-NEXT: v_subrev_u32_e32 v1, 7, v0
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 7, v0
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX9-NEXT: v_subrev_u32_e32 v1, 7, v0
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 7, v0
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX9-NEXT: v_sub_u16_e32 v1, 6, v0
; GFX9-NEXT: v_and_b32_e32 v0, s3, v0
; GFX9-NEXT: v_and_b32_e32 v1, s3, v1
; GFX9-NEXT: v_lshlrev_b16_e64 v1, v1, s0
; GFX9-NEXT: v_lshrrev_b16_e64 v0, v0, s1
; GFX9-NEXT: v_or_b32_e32 v0, v1, v0
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i7:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v0, 7
; GFX10-NEXT: s_sub_i32 s3, 0, 7
; GFX10-NEXT: s_lshl_b32 s0, s0, 1
; GFX10-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX10-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX10-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX10-NEXT: v_mul_lo_u32 v1, s3, v0
; GFX10-NEXT: s_movk_i32 s3, 0x7f
; GFX10-NEXT: s_and_b32 s2, s2, s3
; GFX10-NEXT: s_and_b32 s1, s1, s3
; GFX10-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX10-NEXT: v_add_nc_u32_e32 v0, v0, v1
; GFX10-NEXT: v_mul_hi_u32 v0, s2, v0
; GFX10-NEXT: v_mul_lo_u32 v0, v0, 7
; GFX10-NEXT: v_sub_nc_u32_e32 v0, s2, v0
; GFX10-NEXT: v_subrev_nc_u32_e32 v1, 7, v0
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 7, v0
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_subrev_nc_u32_e32 v1, 7, v0
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 7, v0
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_sub_nc_u16 v1, 6, v0
; GFX10-NEXT: v_and_b32_e32 v0, s3, v0
; GFX10-NEXT: v_and_b32_e32 v1, s3, v1
; GFX10-NEXT: v_lshrrev_b16 v0, v0, s1
; GFX10-NEXT: v_lshlrev_b16 v1, v1, s0
; GFX10-NEXT: v_or_b32_e32 v0, v1, v0
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call i7 @llvm.fshr.i7(i7 %lhs, i7 %rhs, i7 %amt)
ret i7 %result
}
define i7 @v_fshr_i7(i7 %lhs, i7 %rhs, i7 %amt) {
; GFX6-LABEL: v_fshr_i7:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_cvt_f32_ubyte0_e32 v3, 7
; GFX6-NEXT: v_rcp_iflag_f32_e32 v3, v3
; GFX6-NEXT: s_sub_i32 s4, 0, 7
; GFX6-NEXT: v_and_b32_e32 v2, 0x7f, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX6-NEXT: v_mul_f32_e32 v3, 0x4f7ffffe, v3
; GFX6-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX6-NEXT: v_mul_lo_u32 v4, s4, v3
; GFX6-NEXT: v_mul_hi_u32 v4, v3, v4
; GFX6-NEXT: v_add_i32_e32 v3, vcc, v3, v4
; GFX6-NEXT: v_mul_hi_u32 v3, v2, v3
; GFX6-NEXT: v_mov_b32_e32 v4, 0x7f
; GFX6-NEXT: v_and_b32_e32 v1, v1, v4
; GFX6-NEXT: v_mul_lo_u32 v3, v3, 7
; GFX6-NEXT: v_sub_i32_e32 v2, vcc, v2, v3
; GFX6-NEXT: v_subrev_i32_e32 v3, vcc, 7, v2
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 7, v2
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX6-NEXT: v_subrev_i32_e32 v3, vcc, 7, v2
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 7, v2
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 6, v2
; GFX6-NEXT: v_and_b32_e32 v2, v2, v4
; GFX6-NEXT: v_and_b32_e32 v3, v3, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v0, v3, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v1, v2, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i7:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_cvt_f32_ubyte0_e32 v3, 7
; GFX8-NEXT: v_rcp_iflag_f32_e32 v3, v3
; GFX8-NEXT: s_sub_i32 s4, 0, 7
; GFX8-NEXT: v_and_b32_e32 v2, 0x7f, v2
; GFX8-NEXT: v_lshlrev_b16_e32 v0, 1, v0
; GFX8-NEXT: v_mul_f32_e32 v3, 0x4f7ffffe, v3
; GFX8-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX8-NEXT: v_mul_lo_u32 v4, s4, v3
; GFX8-NEXT: v_mul_hi_u32 v4, v3, v4
; GFX8-NEXT: v_add_u32_e32 v3, vcc, v3, v4
; GFX8-NEXT: v_mul_hi_u32 v3, v2, v3
; GFX8-NEXT: v_mov_b32_e32 v4, 0x7f
; GFX8-NEXT: v_and_b32_e32 v1, v1, v4
; GFX8-NEXT: v_mul_lo_u32 v3, v3, 7
; GFX8-NEXT: v_sub_u32_e32 v2, vcc, v2, v3
; GFX8-NEXT: v_subrev_u32_e32 v3, vcc, 7, v2
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 7, v2
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX8-NEXT: v_subrev_u32_e32 v3, vcc, 7, v2
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 7, v2
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX8-NEXT: v_sub_u16_e32 v3, 6, v2
; GFX8-NEXT: v_and_b32_e32 v2, v2, v4
; GFX8-NEXT: v_and_b32_e32 v3, v3, v4
; GFX8-NEXT: v_lshlrev_b16_e32 v0, v3, v0
; GFX8-NEXT: v_lshrrev_b16_e32 v1, v2, v1
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i7:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_cvt_f32_ubyte0_e32 v3, 7
; GFX9-NEXT: v_rcp_iflag_f32_e32 v3, v3
; GFX9-NEXT: s_sub_i32 s4, 0, 7
; GFX9-NEXT: v_and_b32_e32 v2, 0x7f, v2
; GFX9-NEXT: v_lshlrev_b16_e32 v0, 1, v0
; GFX9-NEXT: v_mul_f32_e32 v3, 0x4f7ffffe, v3
; GFX9-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX9-NEXT: v_mul_lo_u32 v4, s4, v3
; GFX9-NEXT: v_mul_hi_u32 v4, v3, v4
; GFX9-NEXT: v_add_u32_e32 v3, v3, v4
; GFX9-NEXT: v_mul_hi_u32 v3, v2, v3
; GFX9-NEXT: v_mov_b32_e32 v4, 0x7f
; GFX9-NEXT: v_and_b32_e32 v1, v1, v4
; GFX9-NEXT: v_mul_lo_u32 v3, v3, 7
; GFX9-NEXT: v_sub_u32_e32 v2, v2, v3
; GFX9-NEXT: v_subrev_u32_e32 v3, 7, v2
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 7, v2
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX9-NEXT: v_subrev_u32_e32 v3, 7, v2
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 7, v2
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX9-NEXT: v_sub_u16_e32 v3, 6, v2
; GFX9-NEXT: v_and_b32_e32 v2, v2, v4
; GFX9-NEXT: v_and_b32_e32 v3, v3, v4
; GFX9-NEXT: v_lshlrev_b16_e32 v0, v3, v0
; GFX9-NEXT: v_lshrrev_b16_e32 v1, v2, v1
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i7:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v3, 7
; GFX10-NEXT: s_sub_i32 s4, 0, 7
; GFX10-NEXT: v_and_b32_e32 v2, 0x7f, v2
; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0
; GFX10-NEXT: v_and_b32_e32 v1, 0x7f, v1
; GFX10-NEXT: v_rcp_iflag_f32_e32 v3, v3
; GFX10-NEXT: v_mul_f32_e32 v3, 0x4f7ffffe, v3
; GFX10-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX10-NEXT: v_mul_lo_u32 v4, s4, v3
; GFX10-NEXT: v_mul_hi_u32 v4, v3, v4
; GFX10-NEXT: v_add_nc_u32_e32 v3, v3, v4
; GFX10-NEXT: v_mul_hi_u32 v3, v2, v3
; GFX10-NEXT: v_mul_lo_u32 v3, v3, 7
; GFX10-NEXT: v_sub_nc_u32_e32 v2, v2, v3
; GFX10-NEXT: v_subrev_nc_u32_e32 v3, 7, v2
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 7, v2
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
; GFX10-NEXT: v_subrev_nc_u32_e32 v3, 7, v2
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 7, v2
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
; GFX10-NEXT: v_mov_b32_e32 v3, 0x7f
; GFX10-NEXT: v_sub_nc_u16 v4, 6, v2
; GFX10-NEXT: v_and_b32_e32 v2, v2, v3
; GFX10-NEXT: v_and_b32_e32 v4, v4, v3
; GFX10-NEXT: v_lshrrev_b16 v1, v2, v1
; GFX10-NEXT: v_lshlrev_b16 v0, v4, v0
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i7 @llvm.fshr.i7(i7 %lhs, i7 %rhs, i7 %amt)
ret i7 %result
}
define amdgpu_ps i8 @s_fshr_i8(i8 inreg %lhs, i8 inreg %rhs, i8 inreg %amt) {
; GFX6-LABEL: s_fshr_i8:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b32 s3, s2, 7
; GFX6-NEXT: s_and_b32 s1, s1, 0xff
; GFX6-NEXT: s_andn2_b32 s2, 7, s2
; GFX6-NEXT: s_lshl_b32 s0, s0, 1
; GFX6-NEXT: s_lshl_b32 s0, s0, s2
; GFX6-NEXT: s_lshr_b32 s1, s1, s3
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i8:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_and_b32 s1, s1, 0xff
; GFX8-NEXT: s_and_b32 s3, s2, 7
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_andn2_b32 s2, 7, s2
; GFX8-NEXT: s_lshl_b32 s0, s0, 1
; GFX8-NEXT: s_lshl_b32 s0, s0, s2
; GFX8-NEXT: s_lshr_b32 s1, s1, s3
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i8:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_and_b32 s1, s1, 0xff
; GFX9-NEXT: s_and_b32 s3, s2, 7
; GFX9-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX9-NEXT: s_andn2_b32 s2, 7, s2
; GFX9-NEXT: s_lshl_b32 s0, s0, 1
; GFX9-NEXT: s_lshl_b32 s0, s0, s2
; GFX9-NEXT: s_lshr_b32 s1, s1, s3
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_and_b32 s1, s1, 0xff
; GFX10-NEXT: s_and_b32 s3, s2, 7
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: s_andn2_b32 s2, 7, s2
; GFX10-NEXT: s_lshl_b32 s0, s0, 1
; GFX10-NEXT: s_lshr_b32 s1, s1, s3
; GFX10-NEXT: s_lshl_b32 s0, s0, s2
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
%result = call i8 @llvm.fshr.i8(i8 %lhs, i8 %rhs, i8 %amt)
ret i8 %result
}
define i8 @v_fshr_i8(i8 %lhs, i8 %rhs, i8 %amt) {
; GFX6-LABEL: v_fshr_i8:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_and_b32_e32 v3, 7, v2
; GFX6-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX6-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX6-NEXT: v_and_b32_e32 v2, 7, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX6-NEXT: v_lshlrev_b32_e32 v0, v2, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v1, v3, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i8:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_and_b32_e32 v3, 7, v2
; GFX8-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX8-NEXT: v_and_b32_e32 v2, 7, v2
; GFX8-NEXT: v_lshlrev_b16_e32 v0, 1, v0
; GFX8-NEXT: v_lshlrev_b16_e32 v0, v2, v0
; GFX8-NEXT: v_lshrrev_b16_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i8:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_and_b32_e32 v3, 7, v2
; GFX9-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX9-NEXT: v_and_b32_e32 v2, 7, v2
; GFX9-NEXT: v_lshlrev_b16_e32 v0, 1, v0
; GFX9-NEXT: v_lshlrev_b16_e32 v0, v2, v0
; GFX9-NEXT: v_lshrrev_b16_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v3, -1, v2
; GFX10-NEXT: v_and_b32_e32 v2, 7, v2
; GFX10-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0
; GFX10-NEXT: v_and_b32_e32 v3, 7, v3
; GFX10-NEXT: v_lshrrev_b16 v1, v2, v1
; GFX10-NEXT: v_lshlrev_b16 v0, v3, v0
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i8 @llvm.fshr.i8(i8 %lhs, i8 %rhs, i8 %amt)
ret i8 %result
}
define amdgpu_ps i8 @s_fshr_i8_4(i8 inreg %lhs, i8 inreg %rhs) {
; GFX6-LABEL: s_fshr_i8_4:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b32 s1, s1, 0xff
; GFX6-NEXT: s_lshl_b32 s0, s0, 4
; GFX6-NEXT: s_lshr_b32 s1, s1, 4
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i8_4:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_and_b32 s1, s1, 0xff
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_lshl_b32 s0, s0, 4
; GFX8-NEXT: s_lshr_b32 s1, s1, 4
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i8_4:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_and_b32 s1, s1, 0xff
; GFX9-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX9-NEXT: s_lshl_b32 s0, s0, 4
; GFX9-NEXT: s_lshr_b32 s1, s1, 4
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i8_4:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_and_b32 s1, s1, 0xff
; GFX10-NEXT: s_lshl_b32 s0, s0, 4
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: s_lshr_b32 s1, s1, 4
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
%result = call i8 @llvm.fshr.i8(i8 %lhs, i8 %rhs, i8 4)
ret i8 %result
}
define i8 @v_fshr_i8_4(i8 %lhs, i8 %rhs) {
; GFX6-LABEL: v_fshr_i8_4:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 4, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v1, 4, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i8_4:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v2, 4
; GFX8-NEXT: v_lshlrev_b16_e32 v0, 4, v0
; GFX8-NEXT: v_lshrrev_b16_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i8_4:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_mov_b32 s4, 4
; GFX9-NEXT: v_lshlrev_b16_e32 v0, 4, v0
; GFX9-NEXT: v_lshrrev_b16_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i8_4:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX10-NEXT: v_lshlrev_b16 v0, 4, v0
; GFX10-NEXT: v_lshrrev_b16 v1, 4, v1
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i8 @llvm.fshr.i8(i8 %lhs, i8 %rhs, i8 4)
ret i8 %result
}
define amdgpu_ps i8 @s_fshr_i8_5(i8 inreg %lhs, i8 inreg %rhs) {
; GFX6-LABEL: s_fshr_i8_5:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b32 s1, s1, 0xff
; GFX6-NEXT: s_lshl_b32 s0, s0, 3
; GFX6-NEXT: s_lshr_b32 s1, s1, 5
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i8_5:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_and_b32 s1, s1, 0xff
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_lshl_b32 s0, s0, 3
; GFX8-NEXT: s_lshr_b32 s1, s1, 5
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i8_5:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_and_b32 s1, s1, 0xff
; GFX9-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX9-NEXT: s_lshl_b32 s0, s0, 3
; GFX9-NEXT: s_lshr_b32 s1, s1, 5
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i8_5:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_and_b32 s1, s1, 0xff
; GFX10-NEXT: s_lshl_b32 s0, s0, 3
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: s_lshr_b32 s1, s1, 5
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
%result = call i8 @llvm.fshr.i8(i8 %lhs, i8 %rhs, i8 5)
ret i8 %result
}
define i8 @v_fshr_i8_5(i8 %lhs, i8 %rhs) {
; GFX6-LABEL: v_fshr_i8_5:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v1, 5, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i8_5:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v2, 5
; GFX8-NEXT: v_lshlrev_b16_e32 v0, 3, v0
; GFX8-NEXT: v_lshrrev_b16_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i8_5:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v2, 5
; GFX9-NEXT: v_lshlrev_b16_e32 v0, 3, v0
; GFX9-NEXT: v_lshrrev_b16_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i8_5:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX10-NEXT: v_lshlrev_b16 v0, 3, v0
; GFX10-NEXT: v_lshrrev_b16 v1, 5, v1
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i8 @llvm.fshr.i8(i8 %lhs, i8 %rhs, i8 5)
ret i8 %result
}
define amdgpu_ps i16 @s_fshr_v2i8(i16 inreg %lhs.arg, i16 inreg %rhs.arg, i16 inreg %amt.arg) {
; GFX6-LABEL: s_fshr_v2i8:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_movk_i32 s7, 0xff
; GFX6-NEXT: s_lshr_b32 s3, s0, 8
; GFX6-NEXT: s_lshr_b32 s4, s1, 8
; GFX6-NEXT: s_lshr_b32 s5, s2, 8
; GFX6-NEXT: s_and_b32 s6, s2, 7
; GFX6-NEXT: s_and_b32 s1, s1, s7
; GFX6-NEXT: s_andn2_b32 s2, 7, s2
; GFX6-NEXT: s_lshl_b32 s0, s0, 1
; GFX6-NEXT: s_lshl_b32 s0, s0, s2
; GFX6-NEXT: s_lshr_b32 s1, s1, s6
; GFX6-NEXT: s_andn2_b32 s2, 7, s5
; GFX6-NEXT: s_lshl_b32 s3, s3, 1
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: s_lshl_b32 s2, s3, s2
; GFX6-NEXT: s_and_b32 s1, s5, 7
; GFX6-NEXT: s_and_b32 s3, s4, s7
; GFX6-NEXT: s_lshr_b32 s1, s3, s1
; GFX6-NEXT: s_or_b32 s1, s2, s1
; GFX6-NEXT: s_and_b32 s1, s1, s7
; GFX6-NEXT: s_and_b32 s0, s0, s7
; GFX6-NEXT: s_lshl_b32 s1, s1, 8
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_v2i8:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_lshr_b32 s3, s0, 8
; GFX8-NEXT: s_lshr_b32 s5, s2, 8
; GFX8-NEXT: s_and_b32 s6, s2, 7
; GFX8-NEXT: s_andn2_b32 s2, 7, s2
; GFX8-NEXT: s_lshl_b32 s0, s0, 1
; GFX8-NEXT: s_lshl_b32 s0, s0, s2
; GFX8-NEXT: s_movk_i32 s2, 0xff
; GFX8-NEXT: s_lshr_b32 s4, s1, 8
; GFX8-NEXT: s_and_b32 s1, s1, s2
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_and_b32 s4, s4, s2
; GFX8-NEXT: s_lshr_b32 s1, s1, s6
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: s_and_b32 s1, s5, 7
; GFX8-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX8-NEXT: s_andn2_b32 s5, 7, s5
; GFX8-NEXT: s_lshl_b32 s3, s3, 1
; GFX8-NEXT: s_lshl_b32 s3, s3, s5
; GFX8-NEXT: s_lshr_b32 s1, s4, s1
; GFX8-NEXT: s_or_b32 s1, s3, s1
; GFX8-NEXT: s_and_b32 s0, s0, s2
; GFX8-NEXT: s_and_b32 s1, s1, s2
; GFX8-NEXT: s_bfe_u32 s2, 8, 0x100000
; GFX8-NEXT: s_lshl_b32 s1, s1, s2
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_v2i8:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_lshr_b32 s3, s0, 8
; GFX9-NEXT: s_lshr_b32 s5, s2, 8
; GFX9-NEXT: s_and_b32 s6, s2, 7
; GFX9-NEXT: s_andn2_b32 s2, 7, s2
; GFX9-NEXT: s_lshl_b32 s0, s0, 1
; GFX9-NEXT: s_lshl_b32 s0, s0, s2
; GFX9-NEXT: s_movk_i32 s2, 0xff
; GFX9-NEXT: s_lshr_b32 s4, s1, 8
; GFX9-NEXT: s_and_b32 s1, s1, s2
; GFX9-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX9-NEXT: s_and_b32 s4, s4, s2
; GFX9-NEXT: s_lshr_b32 s1, s1, s6
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: s_and_b32 s1, s5, 7
; GFX9-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX9-NEXT: s_andn2_b32 s5, 7, s5
; GFX9-NEXT: s_lshl_b32 s3, s3, 1
; GFX9-NEXT: s_lshl_b32 s3, s3, s5
; GFX9-NEXT: s_lshr_b32 s1, s4, s1
; GFX9-NEXT: s_or_b32 s1, s3, s1
; GFX9-NEXT: s_and_b32 s0, s0, s2
; GFX9-NEXT: s_and_b32 s1, s1, s2
; GFX9-NEXT: s_bfe_u32 s2, 8, 0x100000
; GFX9-NEXT: s_lshl_b32 s1, s1, s2
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_v2i8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_lshr_b32 s4, s1, 8
; GFX10-NEXT: s_movk_i32 s7, 0xff
; GFX10-NEXT: s_lshr_b32 s3, s0, 8
; GFX10-NEXT: s_and_b32 s4, s4, s7
; GFX10-NEXT: s_lshr_b32 s5, s2, 8
; GFX10-NEXT: s_and_b32 s6, s2, 7
; GFX10-NEXT: s_andn2_b32 s2, 7, s2
; GFX10-NEXT: s_lshl_b32 s0, s0, 1
; GFX10-NEXT: s_and_b32 s1, s1, s7
; GFX10-NEXT: s_lshl_b32 s0, s0, s2
; GFX10-NEXT: s_and_b32 s2, s5, 7
; GFX10-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX10-NEXT: s_andn2_b32 s5, 7, s5
; GFX10-NEXT: s_lshl_b32 s3, s3, 1
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: s_lshl_b32 s3, s3, s5
; GFX10-NEXT: s_lshr_b32 s2, s4, s2
; GFX10-NEXT: s_lshr_b32 s1, s1, s6
; GFX10-NEXT: s_or_b32 s2, s3, s2
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: s_and_b32 s1, s2, s7
; GFX10-NEXT: s_bfe_u32 s2, 8, 0x100000
; GFX10-NEXT: s_and_b32 s0, s0, s7
; GFX10-NEXT: s_lshl_b32 s1, s1, s2
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
%lhs = bitcast i16 %lhs.arg to <2 x i8>
%rhs = bitcast i16 %rhs.arg to <2 x i8>
%amt = bitcast i16 %amt.arg to <2 x i8>
%result = call <2 x i8> @llvm.fshr.v2i8(<2 x i8> %lhs, <2 x i8> %rhs, <2 x i8> %amt)
%cast.result = bitcast <2 x i8> %result to i16
ret i16 %cast.result
}
define i16 @v_fshr_v2i8(i16 %lhs.arg, i16 %rhs.arg, i16 %amt.arg) {
; GFX6-LABEL: v_fshr_v2i8:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_lshrrev_b32_e32 v5, 8, v2
; GFX6-NEXT: v_and_b32_e32 v6, 7, v2
; GFX6-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX6-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX6-NEXT: v_and_b32_e32 v2, 7, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX6-NEXT: s_movk_i32 s4, 0xff
; GFX6-NEXT: v_lshrrev_b32_e32 v4, 8, v1
; GFX6-NEXT: v_and_b32_e32 v1, s4, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v0, v2, v0
; GFX6-NEXT: v_xor_b32_e32 v2, -1, v5
; GFX6-NEXT: v_lshrrev_b32_e32 v1, v6, v1
; GFX6-NEXT: v_and_b32_e32 v2, 7, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v3, 1, v3
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v2, v2, v3
; GFX6-NEXT: v_and_b32_e32 v1, 7, v5
; GFX6-NEXT: v_and_b32_e32 v3, s4, v4
; GFX6-NEXT: v_lshrrev_b32_e32 v1, v1, v3
; GFX6-NEXT: v_or_b32_e32 v1, v2, v1
; GFX6-NEXT: v_and_b32_e32 v1, s4, v1
; GFX6-NEXT: v_and_b32_e32 v0, 0xff, v0
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v2i8:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_lshrrev_b32_e32 v5, 8, v2
; GFX8-NEXT: v_and_b32_e32 v6, 7, v2
; GFX8-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX8-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX8-NEXT: v_and_b32_e32 v2, 7, v2
; GFX8-NEXT: v_lshlrev_b16_e32 v0, 1, v0
; GFX8-NEXT: v_lshrrev_b32_e32 v4, 8, v1
; GFX8-NEXT: v_lshlrev_b16_e32 v0, v2, v0
; GFX8-NEXT: v_xor_b32_e32 v2, -1, v5
; GFX8-NEXT: v_lshrrev_b16_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: v_and_b32_e32 v1, 7, v5
; GFX8-NEXT: v_and_b32_e32 v2, 7, v2
; GFX8-NEXT: v_lshlrev_b16_e32 v3, 1, v3
; GFX8-NEXT: v_lshlrev_b16_e32 v2, v2, v3
; GFX8-NEXT: v_lshrrev_b16_sdwa v1, v1, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX8-NEXT: v_or_b32_e32 v1, v2, v1
; GFX8-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX8-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX8-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v2i8:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_lshrrev_b32_e32 v5, 8, v2
; GFX9-NEXT: v_and_b32_e32 v6, 7, v2
; GFX9-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX9-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX9-NEXT: v_and_b32_e32 v2, 7, v2
; GFX9-NEXT: v_lshlrev_b16_e32 v0, 1, v0
; GFX9-NEXT: v_lshrrev_b32_e32 v4, 8, v1
; GFX9-NEXT: v_lshlrev_b16_e32 v0, v2, v0
; GFX9-NEXT: v_xor_b32_e32 v2, -1, v5
; GFX9-NEXT: v_lshrrev_b16_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: v_and_b32_e32 v1, 7, v5
; GFX9-NEXT: v_and_b32_e32 v2, 7, v2
; GFX9-NEXT: v_lshlrev_b16_e32 v3, 1, v3
; GFX9-NEXT: v_lshlrev_b16_e32 v2, v2, v3
; GFX9-NEXT: v_lshrrev_b16_sdwa v1, v1, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_or_b32_e32 v1, v2, v1
; GFX9-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v2i8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v1
; GFX10-NEXT: s_movk_i32 s4, 0xff
; GFX10-NEXT: v_and_b32_e32 v7, 7, v2
; GFX10-NEXT: v_xor_b32_e32 v6, -1, v3
; GFX10-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX10-NEXT: v_and_b32_e32 v3, 7, v3
; GFX10-NEXT: v_and_b32_e32 v5, s4, v5
; GFX10-NEXT: v_lshlrev_b16 v4, 1, v4
; GFX10-NEXT: v_and_b32_e32 v6, 7, v6
; GFX10-NEXT: v_and_b32_e32 v1, s4, v1
; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0
; GFX10-NEXT: v_and_b32_e32 v2, 7, v2
; GFX10-NEXT: v_lshrrev_b16 v3, v3, v5
; GFX10-NEXT: v_lshlrev_b16 v4, v6, v4
; GFX10-NEXT: v_lshrrev_b16 v1, v7, v1
; GFX10-NEXT: v_lshlrev_b16 v0, v2, v0
; GFX10-NEXT: v_or_b32_e32 v2, v4, v3
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: v_and_b32_sdwa v1, v2, s4 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT: s_setpc_b64 s[30:31]
%lhs = bitcast i16 %lhs.arg to <2 x i8>
%rhs = bitcast i16 %rhs.arg to <2 x i8>
%amt = bitcast i16 %amt.arg to <2 x i8>
%result = call <2 x i8> @llvm.fshr.v2i8(<2 x i8> %lhs, <2 x i8> %rhs, <2 x i8> %amt)
%cast.result = bitcast <2 x i8> %result to i16
ret i16 %cast.result
}
define amdgpu_ps i32 @s_fshr_v4i8(i32 inreg %lhs.arg, i32 inreg %rhs.arg, i32 inreg %amt.arg) {
; GFX6-LABEL: s_fshr_v4i8:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_movk_i32 s13, 0xff
; GFX6-NEXT: s_lshr_b32 s3, s0, 8
; GFX6-NEXT: s_lshr_b32 s4, s0, 16
; GFX6-NEXT: s_lshr_b32 s5, s0, 24
; GFX6-NEXT: s_lshr_b32 s6, s1, 8
; GFX6-NEXT: s_lshr_b32 s7, s1, 16
; GFX6-NEXT: s_lshr_b32 s8, s1, 24
; GFX6-NEXT: s_lshr_b32 s9, s2, 8
; GFX6-NEXT: s_lshr_b32 s10, s2, 16
; GFX6-NEXT: s_lshr_b32 s11, s2, 24
; GFX6-NEXT: s_and_b32 s12, s2, 7
; GFX6-NEXT: s_and_b32 s1, s1, s13
; GFX6-NEXT: s_andn2_b32 s2, 7, s2
; GFX6-NEXT: s_lshl_b32 s0, s0, 1
; GFX6-NEXT: s_lshl_b32 s0, s0, s2
; GFX6-NEXT: s_lshr_b32 s1, s1, s12
; GFX6-NEXT: s_andn2_b32 s2, 7, s9
; GFX6-NEXT: s_lshl_b32 s3, s3, 1
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: s_lshl_b32 s2, s3, s2
; GFX6-NEXT: s_and_b32 s1, s9, 7
; GFX6-NEXT: s_and_b32 s3, s6, s13
; GFX6-NEXT: s_lshr_b32 s1, s3, s1
; GFX6-NEXT: s_andn2_b32 s3, 7, s10
; GFX6-NEXT: s_lshl_b32 s4, s4, 1
; GFX6-NEXT: s_or_b32 s1, s2, s1
; GFX6-NEXT: s_lshl_b32 s3, s4, s3
; GFX6-NEXT: s_and_b32 s2, s10, 7
; GFX6-NEXT: s_and_b32 s4, s7, s13
; GFX6-NEXT: s_lshr_b32 s2, s4, s2
; GFX6-NEXT: s_and_b32 s1, s1, s13
; GFX6-NEXT: s_or_b32 s2, s3, s2
; GFX6-NEXT: s_and_b32 s3, s11, 7
; GFX6-NEXT: s_andn2_b32 s4, 7, s11
; GFX6-NEXT: s_lshl_b32 s5, s5, 1
; GFX6-NEXT: s_and_b32 s0, s0, s13
; GFX6-NEXT: s_lshl_b32 s1, s1, 8
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: s_and_b32 s1, s2, s13
; GFX6-NEXT: s_lshl_b32 s4, s5, s4
; GFX6-NEXT: s_lshr_b32 s3, s8, s3
; GFX6-NEXT: s_lshl_b32 s1, s1, 16
; GFX6-NEXT: s_or_b32 s3, s4, s3
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: s_and_b32 s1, s3, s13
; GFX6-NEXT: s_lshl_b32 s1, s1, 24
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_v4i8:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_movk_i32 s13, 0xff
; GFX8-NEXT: s_lshr_b32 s3, s0, 8
; GFX8-NEXT: s_lshr_b32 s4, s0, 16
; GFX8-NEXT: s_lshr_b32 s5, s0, 24
; GFX8-NEXT: s_lshr_b32 s6, s1, 8
; GFX8-NEXT: s_lshr_b32 s7, s1, 16
; GFX8-NEXT: s_lshr_b32 s8, s1, 24
; GFX8-NEXT: s_and_b32 s1, s1, s13
; GFX8-NEXT: s_lshr_b32 s9, s2, 8
; GFX8-NEXT: s_lshr_b32 s10, s2, 16
; GFX8-NEXT: s_lshr_b32 s11, s2, 24
; GFX8-NEXT: s_and_b32 s12, s2, 7
; GFX8-NEXT: s_andn2_b32 s2, 7, s2
; GFX8-NEXT: s_lshl_b32 s0, s0, 1
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_lshl_b32 s0, s0, s2
; GFX8-NEXT: s_andn2_b32 s2, 7, s9
; GFX8-NEXT: s_lshl_b32 s3, s3, 1
; GFX8-NEXT: s_lshr_b32 s1, s1, s12
; GFX8-NEXT: s_lshl_b32 s2, s3, s2
; GFX8-NEXT: s_and_b32 s3, s6, s13
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: s_and_b32 s1, s9, 7
; GFX8-NEXT: s_bfe_u32 s3, s3, 0x100000
; GFX8-NEXT: s_lshr_b32 s1, s3, s1
; GFX8-NEXT: s_andn2_b32 s3, 7, s10
; GFX8-NEXT: s_lshl_b32 s4, s4, 1
; GFX8-NEXT: s_lshl_b32 s3, s4, s3
; GFX8-NEXT: s_and_b32 s4, s7, s13
; GFX8-NEXT: s_or_b32 s1, s2, s1
; GFX8-NEXT: s_and_b32 s2, s10, 7
; GFX8-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX8-NEXT: s_lshr_b32 s2, s4, s2
; GFX8-NEXT: s_and_b32 s1, s1, s13
; GFX8-NEXT: s_or_b32 s2, s3, s2
; GFX8-NEXT: s_and_b32 s3, s11, 7
; GFX8-NEXT: s_andn2_b32 s4, 7, s11
; GFX8-NEXT: s_lshl_b32 s5, s5, 1
; GFX8-NEXT: s_and_b32 s0, s0, s13
; GFX8-NEXT: s_lshl_b32 s1, s1, 8
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: s_and_b32 s1, s2, s13
; GFX8-NEXT: s_lshl_b32 s4, s5, s4
; GFX8-NEXT: s_lshr_b32 s3, s8, s3
; GFX8-NEXT: s_lshl_b32 s1, s1, 16
; GFX8-NEXT: s_or_b32 s3, s4, s3
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: s_and_b32 s1, s3, s13
; GFX8-NEXT: s_lshl_b32 s1, s1, 24
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_v4i8:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_movk_i32 s13, 0xff
; GFX9-NEXT: s_lshr_b32 s3, s0, 8
; GFX9-NEXT: s_lshr_b32 s4, s0, 16
; GFX9-NEXT: s_lshr_b32 s5, s0, 24
; GFX9-NEXT: s_lshr_b32 s6, s1, 8
; GFX9-NEXT: s_lshr_b32 s7, s1, 16
; GFX9-NEXT: s_lshr_b32 s8, s1, 24
; GFX9-NEXT: s_and_b32 s1, s1, s13
; GFX9-NEXT: s_lshr_b32 s9, s2, 8
; GFX9-NEXT: s_lshr_b32 s10, s2, 16
; GFX9-NEXT: s_lshr_b32 s11, s2, 24
; GFX9-NEXT: s_and_b32 s12, s2, 7
; GFX9-NEXT: s_andn2_b32 s2, 7, s2
; GFX9-NEXT: s_lshl_b32 s0, s0, 1
; GFX9-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX9-NEXT: s_lshl_b32 s0, s0, s2
; GFX9-NEXT: s_andn2_b32 s2, 7, s9
; GFX9-NEXT: s_lshl_b32 s3, s3, 1
; GFX9-NEXT: s_lshr_b32 s1, s1, s12
; GFX9-NEXT: s_lshl_b32 s2, s3, s2
; GFX9-NEXT: s_and_b32 s3, s6, s13
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: s_and_b32 s1, s9, 7
; GFX9-NEXT: s_bfe_u32 s3, s3, 0x100000
; GFX9-NEXT: s_lshr_b32 s1, s3, s1
; GFX9-NEXT: s_andn2_b32 s3, 7, s10
; GFX9-NEXT: s_lshl_b32 s4, s4, 1
; GFX9-NEXT: s_lshl_b32 s3, s4, s3
; GFX9-NEXT: s_and_b32 s4, s7, s13
; GFX9-NEXT: s_or_b32 s1, s2, s1
; GFX9-NEXT: s_and_b32 s2, s10, 7
; GFX9-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX9-NEXT: s_lshr_b32 s2, s4, s2
; GFX9-NEXT: s_and_b32 s1, s1, s13
; GFX9-NEXT: s_or_b32 s2, s3, s2
; GFX9-NEXT: s_and_b32 s3, s11, 7
; GFX9-NEXT: s_andn2_b32 s4, 7, s11
; GFX9-NEXT: s_lshl_b32 s5, s5, 1
; GFX9-NEXT: s_and_b32 s0, s0, s13
; GFX9-NEXT: s_lshl_b32 s1, s1, 8
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: s_and_b32 s1, s2, s13
; GFX9-NEXT: s_lshl_b32 s4, s5, s4
; GFX9-NEXT: s_lshr_b32 s3, s8, s3
; GFX9-NEXT: s_lshl_b32 s1, s1, 16
; GFX9-NEXT: s_or_b32 s3, s4, s3
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: s_and_b32 s1, s3, s13
; GFX9-NEXT: s_lshl_b32 s1, s1, 24
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_v4i8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_lshr_b32 s6, s1, 8
; GFX10-NEXT: s_movk_i32 s13, 0xff
; GFX10-NEXT: s_lshr_b32 s3, s0, 8
; GFX10-NEXT: s_lshr_b32 s4, s0, 16
; GFX10-NEXT: s_lshr_b32 s5, s0, 24
; GFX10-NEXT: s_and_b32 s6, s6, s13
; GFX10-NEXT: s_lshr_b32 s7, s1, 16
; GFX10-NEXT: s_lshr_b32 s8, s1, 24
; GFX10-NEXT: s_and_b32 s1, s1, s13
; GFX10-NEXT: s_lshr_b32 s9, s2, 8
; GFX10-NEXT: s_lshr_b32 s10, s2, 16
; GFX10-NEXT: s_lshr_b32 s11, s2, 24
; GFX10-NEXT: s_and_b32 s12, s2, 7
; GFX10-NEXT: s_andn2_b32 s2, 7, s2
; GFX10-NEXT: s_lshl_b32 s0, s0, 1
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: s_lshl_b32 s0, s0, s2
; GFX10-NEXT: s_and_b32 s2, s9, 7
; GFX10-NEXT: s_bfe_u32 s6, s6, 0x100000
; GFX10-NEXT: s_andn2_b32 s9, 7, s9
; GFX10-NEXT: s_lshl_b32 s3, s3, 1
; GFX10-NEXT: s_lshr_b32 s1, s1, s12
; GFX10-NEXT: s_lshr_b32 s2, s6, s2
; GFX10-NEXT: s_and_b32 s6, s7, s13
; GFX10-NEXT: s_lshl_b32 s3, s3, s9
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: s_or_b32 s1, s3, s2
; GFX10-NEXT: s_and_b32 s2, s10, 7
; GFX10-NEXT: s_bfe_u32 s6, s6, 0x100000
; GFX10-NEXT: s_andn2_b32 s3, 7, s10
; GFX10-NEXT: s_lshl_b32 s4, s4, 1
; GFX10-NEXT: s_lshr_b32 s2, s6, s2
; GFX10-NEXT: s_lshl_b32 s3, s4, s3
; GFX10-NEXT: s_andn2_b32 s4, 7, s11
; GFX10-NEXT: s_lshl_b32 s5, s5, 1
; GFX10-NEXT: s_and_b32 s6, s11, 7
; GFX10-NEXT: s_or_b32 s2, s3, s2
; GFX10-NEXT: s_and_b32 s1, s1, s13
; GFX10-NEXT: s_lshl_b32 s4, s5, s4
; GFX10-NEXT: s_lshr_b32 s5, s8, s6
; GFX10-NEXT: s_and_b32 s0, s0, s13
; GFX10-NEXT: s_lshl_b32 s1, s1, 8
; GFX10-NEXT: s_or_b32 s3, s4, s5
; GFX10-NEXT: s_and_b32 s2, s2, s13
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: s_lshl_b32 s1, s2, 16
; GFX10-NEXT: s_and_b32 s2, s3, s13
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: s_lshl_b32 s1, s2, 24
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
%lhs = bitcast i32 %lhs.arg to <4 x i8>
%rhs = bitcast i32 %rhs.arg to <4 x i8>
%amt = bitcast i32 %amt.arg to <4 x i8>
%result = call <4 x i8> @llvm.fshr.v4i8(<4 x i8> %lhs, <4 x i8> %rhs, <4 x i8> %amt)
%cast.result = bitcast <4 x i8> %result to i32
ret i32 %cast.result
}
define i32 @v_fshr_v4i8(i32 %lhs.arg, i32 %rhs.arg, i32 %amt.arg) {
; GFX6-LABEL: v_fshr_v4i8:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_lshrrev_b32_e32 v9, 8, v2
; GFX6-NEXT: v_lshrrev_b32_e32 v10, 16, v2
; GFX6-NEXT: v_lshrrev_b32_e32 v11, 24, v2
; GFX6-NEXT: v_and_b32_e32 v12, 7, v2
; GFX6-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX6-NEXT: s_movk_i32 s4, 0xff
; GFX6-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v4, 16, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v5, 24, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v6, 8, v1
; GFX6-NEXT: v_lshrrev_b32_e32 v7, 16, v1
; GFX6-NEXT: v_lshrrev_b32_e32 v8, 24, v1
; GFX6-NEXT: v_and_b32_e32 v1, s4, v1
; GFX6-NEXT: v_and_b32_e32 v2, 7, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX6-NEXT: v_lshlrev_b32_e32 v0, v2, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v1, v12, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: v_and_b32_e32 v1, 7, v9
; GFX6-NEXT: v_xor_b32_e32 v9, -1, v9
; GFX6-NEXT: v_and_b32_e32 v6, s4, v6
; GFX6-NEXT: v_lshrrev_b32_e32 v1, v1, v6
; GFX6-NEXT: v_xor_b32_e32 v6, -1, v10
; GFX6-NEXT: v_and_b32_e32 v9, 7, v9
; GFX6-NEXT: v_lshlrev_b32_e32 v3, 1, v3
; GFX6-NEXT: v_mov_b32_e32 v2, 0xff
; GFX6-NEXT: v_lshlrev_b32_e32 v3, v9, v3
; GFX6-NEXT: v_and_b32_e32 v6, 7, v6
; GFX6-NEXT: v_lshlrev_b32_e32 v4, 1, v4
; GFX6-NEXT: v_or_b32_e32 v1, v3, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v4, v6, v4
; GFX6-NEXT: v_and_b32_e32 v3, 7, v10
; GFX6-NEXT: v_and_b32_e32 v6, v7, v2
; GFX6-NEXT: v_lshrrev_b32_e32 v3, v3, v6
; GFX6-NEXT: v_xor_b32_e32 v6, -1, v11
; GFX6-NEXT: v_and_b32_e32 v1, v1, v2
; GFX6-NEXT: v_or_b32_e32 v3, v4, v3
; GFX6-NEXT: v_and_b32_e32 v4, 7, v11
; GFX6-NEXT: v_and_b32_e32 v6, 7, v6
; GFX6-NEXT: v_lshlrev_b32_e32 v5, 1, v5
; GFX6-NEXT: v_and_b32_e32 v0, v0, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: v_and_b32_e32 v1, v3, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v5, v6, v5
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v4, v8
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX6-NEXT: v_or_b32_e32 v4, v5, v4
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: v_and_b32_e32 v1, v4, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 24, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v4i8:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_and_b32_e32 v8, 7, v2
; GFX8-NEXT: v_lshrrev_b32_e32 v5, 8, v2
; GFX8-NEXT: v_lshrrev_b32_e32 v6, 16, v2
; GFX8-NEXT: v_lshrrev_b32_e32 v7, 24, v2
; GFX8-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX8-NEXT: v_and_b32_e32 v2, 7, v2
; GFX8-NEXT: v_lshlrev_b16_e32 v9, 1, v0
; GFX8-NEXT: v_lshlrev_b16_e32 v2, v2, v9
; GFX8-NEXT: v_lshrrev_b16_sdwa v8, v8, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX8-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX8-NEXT: v_or_b32_e32 v2, v2, v8
; GFX8-NEXT: v_and_b32_e32 v8, 7, v5
; GFX8-NEXT: v_xor_b32_e32 v5, -1, v5
; GFX8-NEXT: v_lshrrev_b32_e32 v4, 8, v1
; GFX8-NEXT: v_and_b32_e32 v5, 7, v5
; GFX8-NEXT: v_lshlrev_b16_e32 v3, 1, v3
; GFX8-NEXT: v_lshlrev_b16_e32 v3, v5, v3
; GFX8-NEXT: v_lshrrev_b16_sdwa v4, v8, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX8-NEXT: v_xor_b32_e32 v5, -1, v6
; GFX8-NEXT: v_or_b32_e32 v3, v3, v4
; GFX8-NEXT: v_and_b32_e32 v4, 7, v6
; GFX8-NEXT: v_mov_b32_e32 v6, 1
; GFX8-NEXT: v_lshlrev_b16_sdwa v8, v6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_and_b32_e32 v5, 7, v5
; GFX8-NEXT: v_mov_b32_e32 v9, 0xff
; GFX8-NEXT: v_lshlrev_b16_e32 v5, v5, v8
; GFX8-NEXT: v_and_b32_sdwa v8, v1, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-NEXT: v_lshrrev_b16_e32 v4, v4, v8
; GFX8-NEXT: v_or_b32_e32 v4, v5, v4
; GFX8-NEXT: v_and_b32_e32 v5, 7, v7
; GFX8-NEXT: v_xor_b32_e32 v7, -1, v7
; GFX8-NEXT: v_and_b32_e32 v7, 7, v7
; GFX8-NEXT: v_lshlrev_b16_sdwa v0, v6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
; GFX8-NEXT: v_lshlrev_b16_e32 v0, v7, v0
; GFX8-NEXT: v_lshrrev_b16_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: v_mov_b32_e32 v1, 8
; GFX8-NEXT: v_lshlrev_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX8-NEXT: s_movk_i32 s4, 0xff
; GFX8-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX8-NEXT: v_and_b32_e32 v2, s4, v4
; GFX8-NEXT: v_and_b32_e32 v0, s4, v0
; GFX8-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX8-NEXT: v_or_b32_e32 v1, v1, v2
; GFX8-NEXT: v_lshlrev_b32_e32 v0, 24, v0
; GFX8-NEXT: v_or_b32_e32 v0, v1, v0
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v4i8:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_and_b32_e32 v8, 7, v2
; GFX9-NEXT: v_lshrrev_b32_e32 v5, 8, v2
; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v2
; GFX9-NEXT: v_lshrrev_b32_e32 v7, 24, v2
; GFX9-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX9-NEXT: v_and_b32_e32 v2, 7, v2
; GFX9-NEXT: v_lshlrev_b16_e32 v9, 1, v0
; GFX9-NEXT: v_lshlrev_b16_e32 v2, v2, v9
; GFX9-NEXT: v_lshrrev_b16_sdwa v8, v8, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX9-NEXT: v_or_b32_e32 v2, v2, v8
; GFX9-NEXT: v_and_b32_e32 v8, 7, v5
; GFX9-NEXT: v_xor_b32_e32 v5, -1, v5
; GFX9-NEXT: v_lshrrev_b32_e32 v4, 8, v1
; GFX9-NEXT: v_and_b32_e32 v5, 7, v5
; GFX9-NEXT: v_lshlrev_b16_e32 v3, 1, v3
; GFX9-NEXT: v_lshlrev_b16_e32 v3, v5, v3
; GFX9-NEXT: v_lshrrev_b16_sdwa v4, v8, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_xor_b32_e32 v5, -1, v6
; GFX9-NEXT: v_or_b32_e32 v3, v3, v4
; GFX9-NEXT: v_and_b32_e32 v4, 7, v6
; GFX9-NEXT: v_mov_b32_e32 v6, 1
; GFX9-NEXT: v_lshlrev_b16_sdwa v8, v6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX9-NEXT: v_and_b32_e32 v5, 7, v5
; GFX9-NEXT: v_mov_b32_e32 v9, 0xff
; GFX9-NEXT: v_lshlrev_b16_e32 v5, v5, v8
; GFX9-NEXT: v_and_b32_sdwa v8, v1, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT: v_lshrrev_b16_e32 v4, v4, v8
; GFX9-NEXT: v_or_b32_e32 v4, v5, v4
; GFX9-NEXT: v_and_b32_e32 v5, 7, v7
; GFX9-NEXT: v_xor_b32_e32 v7, -1, v7
; GFX9-NEXT: v_and_b32_e32 v7, 7, v7
; GFX9-NEXT: v_lshlrev_b16_sdwa v0, v6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
; GFX9-NEXT: v_lshlrev_b16_e32 v0, v7, v0
; GFX9-NEXT: v_lshrrev_b16_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: v_mov_b32_e32 v1, 8
; GFX9-NEXT: s_movk_i32 s4, 0xff
; GFX9-NEXT: v_lshlrev_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_and_b32_e32 v0, s4, v0
; GFX9-NEXT: v_and_or_b32 v1, v2, s4, v1
; GFX9-NEXT: v_and_b32_e32 v2, s4, v4
; GFX9-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX9-NEXT: v_lshlrev_b32_e32 v0, 24, v0
; GFX9-NEXT: v_or3_b32 v0, v1, v2, v0
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v4i8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 8, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX10-NEXT: v_xor_b32_e32 v8, -1, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v12, 24, v2
; GFX10-NEXT: v_lshrrev_b32_e32 v10, 16, v2
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v6
; GFX10-NEXT: v_lshlrev_b16 v3, 1, v3
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 16, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 24, v0
; GFX10-NEXT: v_xor_b32_e32 v14, -1, v12
; GFX10-NEXT: v_and_b32_e32 v11, 7, v11
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 8, v1
; GFX10-NEXT: v_and_b32_e32 v8, 7, v8
; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0
; GFX10-NEXT: v_mov_b32_e32 v13, 0xff
; GFX10-NEXT: v_lshlrev_b16 v3, v11, v3
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v10
; GFX10-NEXT: s_movk_i32 s4, 0xff
; GFX10-NEXT: v_lshlrev_b16 v0, v8, v0
; GFX10-NEXT: v_lshrrev_b32_e32 v9, 24, v1
; GFX10-NEXT: v_and_b32_e32 v8, s4, v1
; GFX10-NEXT: v_and_b32_sdwa v1, v1, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_and_b32_e32 v10, 7, v10
; GFX10-NEXT: v_and_b32_e32 v6, 7, v6
; GFX10-NEXT: v_and_b32_e32 v7, s4, v7
; GFX10-NEXT: v_and_b32_e32 v11, 7, v11
; GFX10-NEXT: v_lshlrev_b16 v4, 1, v4
; GFX10-NEXT: v_and_b32_e32 v13, 7, v14
; GFX10-NEXT: v_lshlrev_b16 v5, 1, v5
; GFX10-NEXT: v_and_b32_e32 v12, 7, v12
; GFX10-NEXT: v_and_b32_e32 v2, 7, v2
; GFX10-NEXT: v_lshrrev_b16 v6, v6, v7
; GFX10-NEXT: v_lshlrev_b16 v4, v11, v4
; GFX10-NEXT: v_lshrrev_b16 v1, v10, v1
; GFX10-NEXT: v_lshlrev_b16 v5, v13, v5
; GFX10-NEXT: v_lshrrev_b16 v7, v12, v9
; GFX10-NEXT: v_lshrrev_b16 v2, v2, v8
; GFX10-NEXT: v_or_b32_e32 v3, v3, v6
; GFX10-NEXT: v_or_b32_e32 v1, v4, v1
; GFX10-NEXT: v_mov_b32_e32 v6, 8
; GFX10-NEXT: v_or_b32_e32 v4, v5, v7
; GFX10-NEXT: v_or_b32_e32 v0, v0, v2
; GFX10-NEXT: v_and_b32_e32 v1, s4, v1
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, v6, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_e32 v3, s4, v4
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX10-NEXT: v_and_or_b32 v0, v0, s4, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 24, v3
; GFX10-NEXT: v_or3_b32 v0, v0, v1, v2
; GFX10-NEXT: s_setpc_b64 s[30:31]
%lhs = bitcast i32 %lhs.arg to <4 x i8>
%rhs = bitcast i32 %rhs.arg to <4 x i8>
%amt = bitcast i32 %amt.arg to <4 x i8>
%result = call <4 x i8> @llvm.fshr.v4i8(<4 x i8> %lhs, <4 x i8> %rhs, <4 x i8> %amt)
%cast.result = bitcast <4 x i8> %result to i32
ret i32 %cast.result
}
define amdgpu_ps i24 @s_fshr_i24(i24 inreg %lhs, i24 inreg %rhs, i24 inreg %amt) {
; GFX6-LABEL: s_fshr_i24:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_cvt_f32_ubyte0_e32 v0, 24
; GFX6-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX6-NEXT: s_sub_i32 s3, 0, 24
; GFX6-NEXT: s_lshl_b32 s0, s0, 1
; GFX6-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX6-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX6-NEXT: v_mul_lo_u32 v1, s3, v0
; GFX6-NEXT: s_mov_b32 s3, 0xffffff
; GFX6-NEXT: s_and_b32 s2, s2, s3
; GFX6-NEXT: s_and_b32 s1, s1, s3
; GFX6-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX6-NEXT: v_add_i32_e32 v0, vcc, v0, v1
; GFX6-NEXT: v_mul_hi_u32 v0, s2, v0
; GFX6-NEXT: v_mul_lo_u32 v0, v0, 24
; GFX6-NEXT: v_sub_i32_e32 v0, vcc, s2, v0
; GFX6-NEXT: v_subrev_i32_e32 v1, vcc, 24, v0
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX6-NEXT: v_subrev_i32_e32 v1, vcc, 24, v0
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX6-NEXT: v_sub_i32_e32 v1, vcc, 23, v0
; GFX6-NEXT: v_and_b32_e32 v0, s3, v0
; GFX6-NEXT: v_and_b32_e32 v1, s3, v1
; GFX6-NEXT: v_lshl_b32_e32 v1, s0, v1
; GFX6-NEXT: v_lshr_b32_e32 v0, s1, v0
; GFX6-NEXT: v_or_b32_e32 v0, v1, v0
; GFX6-NEXT: v_readfirstlane_b32 s0, v0
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i24:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_cvt_f32_ubyte0_e32 v0, 24
; GFX8-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX8-NEXT: s_sub_i32 s3, 0, 24
; GFX8-NEXT: s_lshl_b32 s0, s0, 1
; GFX8-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX8-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX8-NEXT: v_mul_lo_u32 v1, s3, v0
; GFX8-NEXT: s_mov_b32 s3, 0xffffff
; GFX8-NEXT: s_and_b32 s2, s2, s3
; GFX8-NEXT: s_and_b32 s1, s1, s3
; GFX8-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v1
; GFX8-NEXT: v_mul_hi_u32 v0, s2, v0
; GFX8-NEXT: v_mul_lo_u32 v0, v0, 24
; GFX8-NEXT: v_sub_u32_e32 v0, vcc, s2, v0
; GFX8-NEXT: v_subrev_u32_e32 v1, vcc, 24, v0
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX8-NEXT: v_subrev_u32_e32 v1, vcc, 24, v0
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX8-NEXT: v_sub_u32_e32 v1, vcc, 23, v0
; GFX8-NEXT: v_and_b32_e32 v0, s3, v0
; GFX8-NEXT: v_and_b32_e32 v1, s3, v1
; GFX8-NEXT: v_lshlrev_b32_e64 v1, v1, s0
; GFX8-NEXT: v_lshrrev_b32_e64 v0, v0, s1
; GFX8-NEXT: v_or_b32_e32 v0, v1, v0
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i24:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_cvt_f32_ubyte0_e32 v0, 24
; GFX9-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX9-NEXT: s_sub_i32 s3, 0, 24
; GFX9-NEXT: s_lshl_b32 s0, s0, 1
; GFX9-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX9-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX9-NEXT: v_mul_lo_u32 v1, s3, v0
; GFX9-NEXT: s_mov_b32 s3, 0xffffff
; GFX9-NEXT: s_and_b32 s2, s2, s3
; GFX9-NEXT: s_and_b32 s1, s1, s3
; GFX9-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX9-NEXT: v_add_u32_e32 v0, v0, v1
; GFX9-NEXT: v_mul_hi_u32 v0, s2, v0
; GFX9-NEXT: v_mul_lo_u32 v0, v0, 24
; GFX9-NEXT: v_sub_u32_e32 v0, s2, v0
; GFX9-NEXT: v_subrev_u32_e32 v1, 24, v0
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX9-NEXT: v_subrev_u32_e32 v1, 24, v0
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX9-NEXT: v_sub_u32_e32 v1, 23, v0
; GFX9-NEXT: v_and_b32_e32 v0, s3, v0
; GFX9-NEXT: v_and_b32_e32 v1, s3, v1
; GFX9-NEXT: v_lshrrev_b32_e64 v0, v0, s1
; GFX9-NEXT: v_lshl_or_b32 v0, s0, v1, v0
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i24:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v0, 24
; GFX10-NEXT: s_sub_i32 s3, 0, 24
; GFX10-NEXT: s_lshl_b32 s0, s0, 1
; GFX10-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX10-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX10-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX10-NEXT: v_mul_lo_u32 v1, s3, v0
; GFX10-NEXT: s_mov_b32 s3, 0xffffff
; GFX10-NEXT: s_and_b32 s2, s2, s3
; GFX10-NEXT: s_and_b32 s1, s1, s3
; GFX10-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX10-NEXT: v_add_nc_u32_e32 v0, v0, v1
; GFX10-NEXT: v_mul_hi_u32 v0, s2, v0
; GFX10-NEXT: v_mul_lo_u32 v0, v0, 24
; GFX10-NEXT: v_sub_nc_u32_e32 v0, s2, v0
; GFX10-NEXT: v_subrev_nc_u32_e32 v1, 24, v0
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v0
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_subrev_nc_u32_e32 v1, 24, v0
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v0
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10-NEXT: v_sub_nc_u32_e32 v1, 23, v0
; GFX10-NEXT: v_and_b32_e32 v0, s3, v0
; GFX10-NEXT: v_and_b32_e32 v1, s3, v1
; GFX10-NEXT: v_lshrrev_b32_e64 v0, v0, s1
; GFX10-NEXT: v_lshl_or_b32 v0, s0, v1, v0
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call i24 @llvm.fshr.i24(i24 %lhs, i24 %rhs, i24 %amt)
ret i24 %result
}
define i24 @v_fshr_i24(i24 %lhs, i24 %rhs, i24 %amt) {
; GFX6-LABEL: v_fshr_i24:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_cvt_f32_ubyte0_e32 v3, 24
; GFX6-NEXT: v_rcp_iflag_f32_e32 v3, v3
; GFX6-NEXT: s_sub_i32 s4, 0, 24
; GFX6-NEXT: v_and_b32_e32 v2, 0xffffff, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX6-NEXT: v_mul_f32_e32 v3, 0x4f7ffffe, v3
; GFX6-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX6-NEXT: v_mul_lo_u32 v4, s4, v3
; GFX6-NEXT: v_mul_hi_u32 v4, v3, v4
; GFX6-NEXT: v_add_i32_e32 v3, vcc, v3, v4
; GFX6-NEXT: v_mul_hi_u32 v3, v2, v3
; GFX6-NEXT: v_mov_b32_e32 v4, 0xffffff
; GFX6-NEXT: v_and_b32_e32 v1, v1, v4
; GFX6-NEXT: v_mul_lo_u32 v3, v3, 24
; GFX6-NEXT: v_sub_i32_e32 v2, vcc, v2, v3
; GFX6-NEXT: v_subrev_i32_e32 v3, vcc, 24, v2
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX6-NEXT: v_subrev_i32_e32 v3, vcc, 24, v2
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 23, v2
; GFX6-NEXT: v_and_b32_e32 v2, v2, v4
; GFX6-NEXT: v_and_b32_e32 v3, v3, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v0, v3, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v1, v2, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i24:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_cvt_f32_ubyte0_e32 v3, 24
; GFX8-NEXT: v_rcp_iflag_f32_e32 v3, v3
; GFX8-NEXT: s_sub_i32 s4, 0, 24
; GFX8-NEXT: v_and_b32_e32 v2, 0xffffff, v2
; GFX8-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX8-NEXT: v_mul_f32_e32 v3, 0x4f7ffffe, v3
; GFX8-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX8-NEXT: v_mul_lo_u32 v4, s4, v3
; GFX8-NEXT: v_mul_hi_u32 v4, v3, v4
; GFX8-NEXT: v_add_u32_e32 v3, vcc, v3, v4
; GFX8-NEXT: v_mul_hi_u32 v3, v2, v3
; GFX8-NEXT: v_mov_b32_e32 v4, 0xffffff
; GFX8-NEXT: v_and_b32_e32 v1, v1, v4
; GFX8-NEXT: v_mul_lo_u32 v3, v3, 24
; GFX8-NEXT: v_sub_u32_e32 v2, vcc, v2, v3
; GFX8-NEXT: v_subrev_u32_e32 v3, vcc, 24, v2
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX8-NEXT: v_subrev_u32_e32 v3, vcc, 24, v2
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX8-NEXT: v_sub_u32_e32 v3, vcc, 23, v2
; GFX8-NEXT: v_and_b32_e32 v2, v2, v4
; GFX8-NEXT: v_and_b32_e32 v3, v3, v4
; GFX8-NEXT: v_lshlrev_b32_e32 v0, v3, v0
; GFX8-NEXT: v_lshrrev_b32_e32 v1, v2, v1
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i24:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_cvt_f32_ubyte0_e32 v3, 24
; GFX9-NEXT: v_rcp_iflag_f32_e32 v3, v3
; GFX9-NEXT: s_sub_i32 s4, 0, 24
; GFX9-NEXT: v_and_b32_e32 v2, 0xffffff, v2
; GFX9-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX9-NEXT: v_mul_f32_e32 v3, 0x4f7ffffe, v3
; GFX9-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX9-NEXT: v_mul_lo_u32 v4, s4, v3
; GFX9-NEXT: v_mul_hi_u32 v4, v3, v4
; GFX9-NEXT: v_add_u32_e32 v3, v3, v4
; GFX9-NEXT: v_mul_hi_u32 v3, v2, v3
; GFX9-NEXT: v_mov_b32_e32 v4, 0xffffff
; GFX9-NEXT: v_and_b32_e32 v1, v1, v4
; GFX9-NEXT: v_mul_lo_u32 v3, v3, 24
; GFX9-NEXT: v_sub_u32_e32 v2, v2, v3
; GFX9-NEXT: v_subrev_u32_e32 v3, 24, v2
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX9-NEXT: v_subrev_u32_e32 v3, 24, v2
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc
; GFX9-NEXT: v_sub_u32_e32 v3, 23, v2
; GFX9-NEXT: v_and_b32_e32 v2, v2, v4
; GFX9-NEXT: v_and_b32_e32 v3, v3, v4
; GFX9-NEXT: v_lshrrev_b32_e32 v1, v2, v1
; GFX9-NEXT: v_lshl_or_b32 v0, v0, v3, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i24:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v3, 24
; GFX10-NEXT: s_sub_i32 s4, 0, 24
; GFX10-NEXT: v_and_b32_e32 v2, 0xffffff, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX10-NEXT: v_rcp_iflag_f32_e32 v3, v3
; GFX10-NEXT: v_mul_f32_e32 v3, 0x4f7ffffe, v3
; GFX10-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX10-NEXT: v_mul_lo_u32 v4, s4, v3
; GFX10-NEXT: v_mul_hi_u32 v4, v3, v4
; GFX10-NEXT: v_add_nc_u32_e32 v3, v3, v4
; GFX10-NEXT: v_mov_b32_e32 v4, 0xffffff
; GFX10-NEXT: v_mul_hi_u32 v3, v2, v3
; GFX10-NEXT: v_and_b32_e32 v1, v1, v4
; GFX10-NEXT: v_mul_lo_u32 v3, v3, 24
; GFX10-NEXT: v_sub_nc_u32_e32 v2, v2, v3
; GFX10-NEXT: v_subrev_nc_u32_e32 v3, 24, v2
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v2
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
; GFX10-NEXT: v_subrev_nc_u32_e32 v3, 24, v2
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v2
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
; GFX10-NEXT: v_sub_nc_u32_e32 v3, 23, v2
; GFX10-NEXT: v_and_b32_e32 v2, v2, v4
; GFX10-NEXT: v_and_b32_e32 v3, v3, v4
; GFX10-NEXT: v_lshrrev_b32_e32 v1, v2, v1
; GFX10-NEXT: v_lshl_or_b32 v0, v0, v3, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i24 @llvm.fshr.i24(i24 %lhs, i24 %rhs, i24 %amt)
ret i24 %result
}
define amdgpu_ps i48 @s_fshr_v2i24(i48 inreg %lhs.arg, i48 inreg %rhs.arg, i48 inreg %amt.arg) {
; GFX6-LABEL: s_fshr_v2i24:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_movk_i32 s10, 0xff
; GFX6-NEXT: s_lshr_b32 s9, s1, 8
; GFX6-NEXT: s_and_b32 s1, s1, s10
; GFX6-NEXT: s_lshr_b32 s6, s0, 8
; GFX6-NEXT: s_lshr_b32 s8, s0, 24
; GFX6-NEXT: s_lshl_b32 s1, s1, 8
; GFX6-NEXT: s_or_b32 s1, s8, s1
; GFX6-NEXT: s_and_b32 s6, s6, s10
; GFX6-NEXT: s_lshr_b32 s8, s2, 8
; GFX6-NEXT: s_and_b32 s8, s8, s10
; GFX6-NEXT: s_lshr_b32 s7, s0, 16
; GFX6-NEXT: s_and_b32 s0, s0, s10
; GFX6-NEXT: s_lshl_b32 s6, s6, 8
; GFX6-NEXT: s_or_b32 s0, s0, s6
; GFX6-NEXT: s_and_b32 s6, s7, s10
; GFX6-NEXT: s_and_b32 s7, s9, s10
; GFX6-NEXT: s_lshr_b32 s9, s2, 16
; GFX6-NEXT: s_lshr_b32 s11, s2, 24
; GFX6-NEXT: s_and_b32 s2, s2, s10
; GFX6-NEXT: s_lshl_b32 s8, s8, 8
; GFX6-NEXT: s_or_b32 s2, s2, s8
; GFX6-NEXT: s_and_b32 s8, s9, s10
; GFX6-NEXT: v_cvt_f32_ubyte0_e32 v0, 24
; GFX6-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX6-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX6-NEXT: s_lshr_b32 s12, s3, 8
; GFX6-NEXT: s_and_b32 s3, s3, s10
; GFX6-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX6-NEXT: s_lshl_b32 s8, s8, 16
; GFX6-NEXT: s_lshl_b32 s3, s3, 8
; GFX6-NEXT: s_or_b32 s2, s2, s8
; GFX6-NEXT: s_and_b32 s8, s12, s10
; GFX6-NEXT: s_or_b32 s3, s11, s3
; GFX6-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX6-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX6-NEXT: s_bfe_u32 s3, s3, 0x100000
; GFX6-NEXT: s_lshl_b32 s8, s8, 16
; GFX6-NEXT: s_or_b32 s3, s3, s8
; GFX6-NEXT: s_lshr_b32 s8, s4, 8
; GFX6-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX6-NEXT: s_and_b32 s8, s8, s10
; GFX6-NEXT: s_lshr_b32 s9, s4, 16
; GFX6-NEXT: s_lshr_b32 s11, s4, 24
; GFX6-NEXT: s_and_b32 s4, s4, s10
; GFX6-NEXT: s_lshl_b32 s8, s8, 8
; GFX6-NEXT: s_or_b32 s4, s4, s8
; GFX6-NEXT: s_and_b32 s8, s9, s10
; GFX6-NEXT: s_sub_i32 s9, 0, 24
; GFX6-NEXT: v_mul_lo_u32 v1, s9, v0
; GFX6-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX6-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX6-NEXT: s_lshl_b32 s8, s8, 16
; GFX6-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX6-NEXT: s_or_b32 s4, s4, s8
; GFX6-NEXT: s_lshr_b32 s12, s5, 8
; GFX6-NEXT: s_and_b32 s5, s5, s10
; GFX6-NEXT: v_add_i32_e32 v0, vcc, v0, v1
; GFX6-NEXT: v_mul_hi_u32 v0, s4, v0
; GFX6-NEXT: v_cvt_f32_ubyte0_e32 v1, 24
; GFX6-NEXT: v_rcp_iflag_f32_e32 v1, v1
; GFX6-NEXT: s_lshl_b32 s5, s5, 8
; GFX6-NEXT: v_mul_lo_u32 v0, v0, 24
; GFX6-NEXT: s_and_b32 s8, s12, s10
; GFX6-NEXT: v_mul_f32_e32 v1, 0x4f7ffffe, v1
; GFX6-NEXT: v_cvt_u32_f32_e32 v1, v1
; GFX6-NEXT: v_sub_i32_e32 v0, vcc, s4, v0
; GFX6-NEXT: v_subrev_i32_e32 v2, vcc, 24, v0
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX6-NEXT: v_subrev_i32_e32 v2, vcc, 24, v0
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX6-NEXT: v_mul_lo_u32 v2, s9, v1
; GFX6-NEXT: s_or_b32 s5, s11, s5
; GFX6-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX6-NEXT: s_bfe_u32 s5, s5, 0x100000
; GFX6-NEXT: v_mul_hi_u32 v2, v1, v2
; GFX6-NEXT: s_lshl_b32 s8, s8, 16
; GFX6-NEXT: s_or_b32 s5, s5, s8
; GFX6-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX6-NEXT: v_add_i32_e32 v1, vcc, v1, v2
; GFX6-NEXT: v_mul_hi_u32 v1, s5, v1
; GFX6-NEXT: s_bfe_u32 s6, s6, 0x100000
; GFX6-NEXT: s_mov_b32 s8, 0xffffff
; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 23, v0
; GFX6-NEXT: v_mul_lo_u32 v1, v1, 24
; GFX6-NEXT: s_lshl_b32 s4, s6, 17
; GFX6-NEXT: s_lshl_b32 s0, s0, 1
; GFX6-NEXT: v_and_b32_e32 v0, s8, v0
; GFX6-NEXT: s_or_b32 s0, s4, s0
; GFX6-NEXT: v_and_b32_e32 v2, s8, v3
; GFX6-NEXT: v_lshl_b32_e32 v2, s0, v2
; GFX6-NEXT: v_lshr_b32_e32 v0, s2, v0
; GFX6-NEXT: v_sub_i32_e32 v1, vcc, s5, v1
; GFX6-NEXT: v_or_b32_e32 v0, v2, v0
; GFX6-NEXT: v_subrev_i32_e32 v2, vcc, 24, v1
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v1
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
; GFX6-NEXT: v_subrev_i32_e32 v2, vcc, 24, v1
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v1
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
; GFX6-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX6-NEXT: s_bfe_u32 s7, s7, 0x100000
; GFX6-NEXT: v_mov_b32_e32 v4, 0xffffff
; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 23, v1
; GFX6-NEXT: s_lshl_b32 s0, s7, 17
; GFX6-NEXT: s_lshl_b32 s1, s1, 1
; GFX6-NEXT: v_and_b32_e32 v1, v1, v4
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: v_and_b32_e32 v2, v2, v4
; GFX6-NEXT: v_lshl_b32_e32 v2, s0, v2
; GFX6-NEXT: v_lshr_b32_e32 v1, s3, v1
; GFX6-NEXT: v_or_b32_e32 v1, v2, v1
; GFX6-NEXT: v_lshrrev_b32_e32 v2, 8, v0
; GFX6-NEXT: v_and_b32_e32 v2, s10, v2
; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v0
; GFX6-NEXT: v_and_b32_e32 v0, s10, v0
; GFX6-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; GFX6-NEXT: v_or_b32_e32 v0, v0, v2
; GFX6-NEXT: v_and_b32_e32 v2, s10, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX6-NEXT: v_lshrrev_b32_e32 v4, 8, v1
; GFX6-NEXT: v_lshrrev_b32_e32 v5, 16, v1
; GFX6-NEXT: v_and_b32_e32 v1, s10, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v2
; GFX6-NEXT: v_and_b32_e32 v2, s10, v5
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 24, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: v_and_b32_e32 v1, s10, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; GFX6-NEXT: v_or_b32_e32 v1, v1, v2
; GFX6-NEXT: v_readfirstlane_b32 s0, v0
; GFX6-NEXT: v_readfirstlane_b32 s1, v1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_v2i24:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_movk_i32 s10, 0xff
; GFX8-NEXT: s_lshr_b32 s9, s1, 8
; GFX8-NEXT: s_bfe_u32 s11, 8, 0x100000
; GFX8-NEXT: s_and_b32 s1, s1, s10
; GFX8-NEXT: s_lshr_b32 s6, s0, 8
; GFX8-NEXT: s_lshr_b32 s8, s0, 24
; GFX8-NEXT: s_lshl_b32 s1, s1, s11
; GFX8-NEXT: s_or_b32 s1, s8, s1
; GFX8-NEXT: s_and_b32 s6, s6, s10
; GFX8-NEXT: s_lshr_b32 s8, s2, 8
; GFX8-NEXT: s_and_b32 s8, s8, s10
; GFX8-NEXT: s_lshr_b32 s7, s0, 16
; GFX8-NEXT: s_and_b32 s0, s0, s10
; GFX8-NEXT: s_lshl_b32 s6, s6, s11
; GFX8-NEXT: s_or_b32 s0, s0, s6
; GFX8-NEXT: s_and_b32 s6, s7, s10
; GFX8-NEXT: s_and_b32 s7, s9, s10
; GFX8-NEXT: s_lshr_b32 s9, s2, 16
; GFX8-NEXT: s_lshr_b32 s12, s2, 24
; GFX8-NEXT: s_and_b32 s2, s2, s10
; GFX8-NEXT: s_lshl_b32 s8, s8, s11
; GFX8-NEXT: s_or_b32 s2, s2, s8
; GFX8-NEXT: s_and_b32 s8, s9, s10
; GFX8-NEXT: v_cvt_f32_ubyte0_e32 v0, 24
; GFX8-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX8-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX8-NEXT: s_lshr_b32 s13, s3, 8
; GFX8-NEXT: s_and_b32 s3, s3, s10
; GFX8-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX8-NEXT: s_lshl_b32 s8, s8, 16
; GFX8-NEXT: s_lshl_b32 s3, s3, s11
; GFX8-NEXT: s_or_b32 s2, s2, s8
; GFX8-NEXT: s_and_b32 s8, s13, s10
; GFX8-NEXT: s_or_b32 s3, s12, s3
; GFX8-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX8-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX8-NEXT: s_bfe_u32 s3, s3, 0x100000
; GFX8-NEXT: s_lshl_b32 s8, s8, 16
; GFX8-NEXT: s_or_b32 s3, s3, s8
; GFX8-NEXT: s_lshr_b32 s8, s4, 8
; GFX8-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX8-NEXT: s_and_b32 s8, s8, s10
; GFX8-NEXT: s_lshr_b32 s9, s4, 16
; GFX8-NEXT: s_lshr_b32 s12, s4, 24
; GFX8-NEXT: s_and_b32 s4, s4, s10
; GFX8-NEXT: s_lshl_b32 s8, s8, s11
; GFX8-NEXT: s_or_b32 s4, s4, s8
; GFX8-NEXT: s_and_b32 s8, s9, s10
; GFX8-NEXT: s_sub_i32 s9, 0, 24
; GFX8-NEXT: v_mul_lo_u32 v1, s9, v0
; GFX8-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX8-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX8-NEXT: s_lshl_b32 s8, s8, 16
; GFX8-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX8-NEXT: s_or_b32 s4, s4, s8
; GFX8-NEXT: s_lshr_b32 s13, s5, 8
; GFX8-NEXT: s_and_b32 s5, s5, s10
; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v1
; GFX8-NEXT: v_mul_hi_u32 v0, s4, v0
; GFX8-NEXT: v_cvt_f32_ubyte0_e32 v1, 24
; GFX8-NEXT: v_rcp_iflag_f32_e32 v1, v1
; GFX8-NEXT: s_lshl_b32 s5, s5, s11
; GFX8-NEXT: v_mul_lo_u32 v0, v0, 24
; GFX8-NEXT: s_and_b32 s8, s13, s10
; GFX8-NEXT: v_mul_f32_e32 v1, 0x4f7ffffe, v1
; GFX8-NEXT: v_cvt_u32_f32_e32 v1, v1
; GFX8-NEXT: v_sub_u32_e32 v0, vcc, s4, v0
; GFX8-NEXT: v_subrev_u32_e32 v2, vcc, 24, v0
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX8-NEXT: v_subrev_u32_e32 v2, vcc, 24, v0
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX8-NEXT: v_mul_lo_u32 v2, s9, v1
; GFX8-NEXT: s_or_b32 s5, s12, s5
; GFX8-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX8-NEXT: s_bfe_u32 s5, s5, 0x100000
; GFX8-NEXT: v_mul_hi_u32 v2, v1, v2
; GFX8-NEXT: s_lshl_b32 s8, s8, 16
; GFX8-NEXT: s_or_b32 s5, s5, s8
; GFX8-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX8-NEXT: v_add_u32_e32 v1, vcc, v1, v2
; GFX8-NEXT: v_mul_hi_u32 v1, s5, v1
; GFX8-NEXT: s_bfe_u32 s6, s6, 0x100000
; GFX8-NEXT: s_mov_b32 s8, 0xffffff
; GFX8-NEXT: v_sub_u32_e32 v3, vcc, 23, v0
; GFX8-NEXT: v_mul_lo_u32 v1, v1, 24
; GFX8-NEXT: s_lshl_b32 s4, s6, 17
; GFX8-NEXT: s_lshl_b32 s0, s0, 1
; GFX8-NEXT: v_and_b32_e32 v0, s8, v0
; GFX8-NEXT: s_or_b32 s0, s4, s0
; GFX8-NEXT: v_and_b32_e32 v2, s8, v3
; GFX8-NEXT: v_lshlrev_b32_e64 v2, v2, s0
; GFX8-NEXT: v_lshrrev_b32_e64 v0, v0, s2
; GFX8-NEXT: v_sub_u32_e32 v1, vcc, s5, v1
; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
; GFX8-NEXT: v_subrev_u32_e32 v2, vcc, 24, v1
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v1
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
; GFX8-NEXT: v_subrev_u32_e32 v2, vcc, 24, v1
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v1
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_bfe_u32 s7, s7, 0x100000
; GFX8-NEXT: v_mov_b32_e32 v4, 0xffffff
; GFX8-NEXT: v_sub_u32_e32 v2, vcc, 23, v1
; GFX8-NEXT: s_lshl_b32 s0, s7, 17
; GFX8-NEXT: s_lshl_b32 s1, s1, 1
; GFX8-NEXT: v_and_b32_e32 v1, v1, v4
; GFX8-NEXT: v_and_b32_e32 v2, v2, v4
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: v_lshlrev_b32_e64 v2, v2, s0
; GFX8-NEXT: v_lshrrev_b32_e64 v1, v1, s3
; GFX8-NEXT: v_or_b32_e32 v1, v2, v1
; GFX8-NEXT: v_lshrrev_b32_e32 v2, 8, v0
; GFX8-NEXT: v_mov_b32_e32 v4, 8
; GFX8-NEXT: v_lshlrev_b32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX8-NEXT: v_mov_b32_e32 v4, s10
; GFX8-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX8-NEXT: v_and_b32_sdwa v0, v0, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
; GFX8-NEXT: v_and_b32_e32 v2, s10, v1
; GFX8-NEXT: v_lshrrev_b32_e32 v3, 8, v1
; GFX8-NEXT: v_and_b32_sdwa v1, v1, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX8-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; GFX8-NEXT: v_lshlrev_b32_e32 v2, 24, v2
; GFX8-NEXT: v_or_b32_e32 v0, v0, v2
; GFX8-NEXT: v_or_b32_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
; GFX8-NEXT: v_readfirstlane_b32 s1, v1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_v2i24:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_movk_i32 s11, 0xff
; GFX9-NEXT: s_lshr_b32 s10, s1, 8
; GFX9-NEXT: s_bfe_u32 s12, 8, 0x100000
; GFX9-NEXT: s_and_b32 s1, s1, s11
; GFX9-NEXT: s_lshr_b32 s7, s0, 8
; GFX9-NEXT: s_lshr_b32 s9, s0, 24
; GFX9-NEXT: s_lshl_b32 s1, s1, s12
; GFX9-NEXT: s_or_b32 s1, s9, s1
; GFX9-NEXT: s_and_b32 s7, s7, s11
; GFX9-NEXT: s_lshr_b32 s9, s2, 8
; GFX9-NEXT: s_and_b32 s9, s9, s11
; GFX9-NEXT: s_lshr_b32 s8, s0, 16
; GFX9-NEXT: s_and_b32 s0, s0, s11
; GFX9-NEXT: s_lshl_b32 s7, s7, s12
; GFX9-NEXT: s_or_b32 s0, s0, s7
; GFX9-NEXT: s_and_b32 s7, s8, s11
; GFX9-NEXT: s_and_b32 s8, s10, s11
; GFX9-NEXT: s_lshr_b32 s10, s2, 16
; GFX9-NEXT: s_lshr_b32 s13, s2, 24
; GFX9-NEXT: s_and_b32 s2, s2, s11
; GFX9-NEXT: s_lshl_b32 s9, s9, s12
; GFX9-NEXT: s_or_b32 s2, s2, s9
; GFX9-NEXT: s_and_b32 s9, s10, s11
; GFX9-NEXT: v_cvt_f32_ubyte0_e32 v0, 24
; GFX9-NEXT: s_bfe_u32 s9, s9, 0x100000
; GFX9-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX9-NEXT: s_lshr_b32 s14, s3, 8
; GFX9-NEXT: s_and_b32 s3, s3, s11
; GFX9-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX9-NEXT: s_lshl_b32 s9, s9, 16
; GFX9-NEXT: s_lshl_b32 s3, s3, s12
; GFX9-NEXT: s_or_b32 s2, s2, s9
; GFX9-NEXT: s_and_b32 s9, s14, s11
; GFX9-NEXT: s_or_b32 s3, s13, s3
; GFX9-NEXT: s_bfe_u32 s9, s9, 0x100000
; GFX9-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX9-NEXT: s_bfe_u32 s3, s3, 0x100000
; GFX9-NEXT: s_lshl_b32 s9, s9, 16
; GFX9-NEXT: s_or_b32 s3, s3, s9
; GFX9-NEXT: s_lshr_b32 s9, s4, 8
; GFX9-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX9-NEXT: s_and_b32 s9, s9, s11
; GFX9-NEXT: s_lshr_b32 s10, s4, 16
; GFX9-NEXT: s_lshr_b32 s13, s4, 24
; GFX9-NEXT: s_and_b32 s4, s4, s11
; GFX9-NEXT: s_lshl_b32 s9, s9, s12
; GFX9-NEXT: s_or_b32 s4, s4, s9
; GFX9-NEXT: s_and_b32 s9, s10, s11
; GFX9-NEXT: s_sub_i32 s10, 0, 24
; GFX9-NEXT: v_mul_lo_u32 v1, s10, v0
; GFX9-NEXT: s_bfe_u32 s9, s9, 0x100000
; GFX9-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX9-NEXT: s_lshl_b32 s9, s9, 16
; GFX9-NEXT: v_mul_hi_u32 v1, v0, v1
; GFX9-NEXT: s_or_b32 s4, s4, s9
; GFX9-NEXT: s_lshr_b32 s14, s5, 8
; GFX9-NEXT: s_and_b32 s5, s5, s11
; GFX9-NEXT: v_add_u32_e32 v0, v0, v1
; GFX9-NEXT: v_mul_hi_u32 v0, s4, v0
; GFX9-NEXT: v_cvt_f32_ubyte0_e32 v1, 24
; GFX9-NEXT: v_rcp_iflag_f32_e32 v1, v1
; GFX9-NEXT: s_lshl_b32 s5, s5, s12
; GFX9-NEXT: v_mul_lo_u32 v0, v0, 24
; GFX9-NEXT: s_and_b32 s9, s14, s11
; GFX9-NEXT: v_mul_f32_e32 v1, 0x4f7ffffe, v1
; GFX9-NEXT: v_cvt_u32_f32_e32 v1, v1
; GFX9-NEXT: v_sub_u32_e32 v0, s4, v0
; GFX9-NEXT: v_subrev_u32_e32 v2, 24, v0
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX9-NEXT: v_mul_lo_u32 v2, s10, v1
; GFX9-NEXT: s_or_b32 s5, s13, s5
; GFX9-NEXT: s_bfe_u32 s9, s9, 0x100000
; GFX9-NEXT: s_bfe_u32 s5, s5, 0x100000
; GFX9-NEXT: v_mul_hi_u32 v2, v1, v2
; GFX9-NEXT: s_lshl_b32 s9, s9, 16
; GFX9-NEXT: s_or_b32 s5, s5, s9
; GFX9-NEXT: v_subrev_u32_e32 v3, 24, v0
; GFX9-NEXT: v_add_u32_e32 v1, v1, v2
; GFX9-NEXT: v_mul_hi_u32 v1, s5, v1
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v0
; GFX9-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GFX9-NEXT: v_mul_lo_u32 v1, v1, 24
; GFX9-NEXT: s_bfe_u32 s7, s7, 0x100000
; GFX9-NEXT: s_mov_b32 s9, 0xffffff
; GFX9-NEXT: v_sub_u32_e32 v3, 23, v0
; GFX9-NEXT: v_and_b32_e32 v0, s9, v0
; GFX9-NEXT: s_lshl_b32 s4, s7, 17
; GFX9-NEXT: s_lshl_b32 s0, s0, 1
; GFX9-NEXT: v_sub_u32_e32 v1, s5, v1
; GFX9-NEXT: s_or_b32 s0, s4, s0
; GFX9-NEXT: v_and_b32_e32 v3, s9, v3
; GFX9-NEXT: v_lshrrev_b32_e64 v0, v0, s2
; GFX9-NEXT: v_lshl_or_b32 v0, s0, v3, v0
; GFX9-NEXT: v_subrev_u32_e32 v3, 24, v1
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v1
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX9-NEXT: v_subrev_u32_e32 v3, 24, v1
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v1
; GFX9-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX9-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX9-NEXT: v_mov_b32_e32 v2, 0xffffff
; GFX9-NEXT: v_sub_u32_e32 v3, 23, v1
; GFX9-NEXT: v_and_b32_e32 v1, v1, v2
; GFX9-NEXT: s_lshl_b32 s0, s8, 17
; GFX9-NEXT: s_lshl_b32 s1, s1, 1
; GFX9-NEXT: v_and_b32_e32 v3, v3, v2
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: v_lshrrev_b32_e64 v1, v1, s3
; GFX9-NEXT: v_lshl_or_b32 v1, s0, v3, v1
; GFX9-NEXT: s_mov_b32 s6, 8
; GFX9-NEXT: v_lshrrev_b32_e32 v2, 8, v0
; GFX9-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_and_b32_e32 v4, s11, v1
; GFX9-NEXT: v_lshrrev_b32_e32 v3, 8, v1
; GFX9-NEXT: v_and_b32_sdwa v1, v1, s11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; GFX9-NEXT: v_and_or_b32 v2, v0, s11, v2
; GFX9-NEXT: v_and_b32_sdwa v0, v0, s11 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; GFX9-NEXT: v_or3_b32 v0, v2, v0, v4
; GFX9-NEXT: v_and_or_b32 v1, v3, s11, v1
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
; GFX9-NEXT: v_readfirstlane_b32 s1, v1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_v2i24:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v0, 24
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v1, 24
; GFX10-NEXT: s_sub_i32 s12, 0, 24
; GFX10-NEXT: s_movk_i32 s9, 0xff
; GFX10-NEXT: s_lshr_b32 s14, s4, 8
; GFX10-NEXT: v_rcp_iflag_f32_e32 v0, v0
; GFX10-NEXT: v_rcp_iflag_f32_e32 v1, v1
; GFX10-NEXT: s_lshr_b32 s15, s4, 16
; GFX10-NEXT: s_bfe_u32 s10, 8, 0x100000
; GFX10-NEXT: s_and_b32 s14, s14, s9
; GFX10-NEXT: s_and_b32 s16, s4, s9
; GFX10-NEXT: s_lshl_b32 s14, s14, s10
; GFX10-NEXT: s_and_b32 s15, s15, s9
; GFX10-NEXT: s_or_b32 s14, s16, s14
; GFX10-NEXT: s_lshr_b32 s4, s4, 24
; GFX10-NEXT: v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX10-NEXT: v_mul_f32_e32 v1, 0x4f7ffffe, v1
; GFX10-NEXT: s_bfe_u32 s14, s14, 0x100000
; GFX10-NEXT: s_lshr_b32 s6, s0, 8
; GFX10-NEXT: s_lshr_b32 s11, s1, 8
; GFX10-NEXT: v_cvt_u32_f32_e32 v0, v0
; GFX10-NEXT: v_cvt_u32_f32_e32 v1, v1
; GFX10-NEXT: s_and_b32 s1, s1, s9
; GFX10-NEXT: s_and_b32 s6, s6, s9
; GFX10-NEXT: s_lshr_b32 s8, s0, 24
; GFX10-NEXT: v_mul_lo_u32 v2, s12, v0
; GFX10-NEXT: v_mul_lo_u32 v3, s12, v1
; GFX10-NEXT: s_bfe_u32 s12, s15, 0x100000
; GFX10-NEXT: s_lshr_b32 s15, s5, 8
; GFX10-NEXT: s_lshl_b32 s12, s12, 16
; GFX10-NEXT: s_and_b32 s5, s5, s9
; GFX10-NEXT: s_or_b32 s12, s14, s12
; GFX10-NEXT: s_lshl_b32 s5, s5, s10
; GFX10-NEXT: v_mul_hi_u32 v2, v0, v2
; GFX10-NEXT: s_and_b32 s14, s15, s9
; GFX10-NEXT: s_or_b32 s4, s4, s5
; GFX10-NEXT: s_bfe_u32 s5, s14, 0x100000
; GFX10-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX10-NEXT: s_lshl_b32 s5, s5, 16
; GFX10-NEXT: s_lshl_b32 s1, s1, s10
; GFX10-NEXT: s_or_b32 s4, s4, s5
; GFX10-NEXT: v_add_nc_u32_e32 v0, v0, v2
; GFX10-NEXT: v_mul_hi_u32 v2, v1, v3
; GFX10-NEXT: s_or_b32 s1, s8, s1
; GFX10-NEXT: s_lshr_b32 s8, s2, 8
; GFX10-NEXT: s_lshr_b32 s7, s0, 16
; GFX10-NEXT: v_mul_hi_u32 v0, s12, v0
; GFX10-NEXT: s_and_b32 s0, s0, s9
; GFX10-NEXT: s_lshl_b32 s6, s6, s10
; GFX10-NEXT: s_and_b32 s8, s8, s9
; GFX10-NEXT: v_add_nc_u32_e32 v1, v1, v2
; GFX10-NEXT: s_or_b32 s0, s0, s6
; GFX10-NEXT: s_and_b32 s6, s7, s9
; GFX10-NEXT: s_and_b32 s7, s11, s9
; GFX10-NEXT: v_mul_lo_u32 v0, v0, 24
; GFX10-NEXT: v_mul_hi_u32 v1, s4, v1
; GFX10-NEXT: s_lshr_b32 s11, s2, 16
; GFX10-NEXT: s_and_b32 s13, s2, s9
; GFX10-NEXT: s_lshl_b32 s5, s8, s10
; GFX10-NEXT: s_and_b32 s8, s11, s9
; GFX10-NEXT: s_lshr_b32 s11, s3, 8
; GFX10-NEXT: s_and_b32 s3, s3, s9
; GFX10-NEXT: v_sub_nc_u32_e32 v0, s12, v0
; GFX10-NEXT: v_mul_lo_u32 v1, v1, 24
; GFX10-NEXT: s_or_b32 s5, s13, s5
; GFX10-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX10-NEXT: s_lshr_b32 s2, s2, 24
; GFX10-NEXT: v_subrev_nc_u32_e32 v2, 24, v0
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v0
; GFX10-NEXT: s_lshl_b32 s3, s3, s10
; GFX10-NEXT: v_mov_b32_e32 v4, 0xffffff
; GFX10-NEXT: v_sub_nc_u32_e32 v1, s4, v1
; GFX10-NEXT: s_mov_b32 s4, 0xffffff
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: s_bfe_u32 s5, s5, 0x100000
; GFX10-NEXT: s_lshl_b32 s8, s8, 16
; GFX10-NEXT: v_subrev_nc_u32_e32 v3, 24, v1
; GFX10-NEXT: s_or_b32 s2, s2, s3
; GFX10-NEXT: v_subrev_nc_u32_e32 v2, 24, v0
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v0
; GFX10-NEXT: s_and_b32 s3, s11, s9
; GFX10-NEXT: s_or_b32 s5, s5, s8
; GFX10-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX10-NEXT: s_bfe_u32 s6, s6, 0x100000
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v1
; GFX10-NEXT: s_bfe_u32 s3, s3, 0x100000
; GFX10-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX10-NEXT: s_lshl_b32 s3, s3, 16
; GFX10-NEXT: v_sub_nc_u32_e32 v2, 23, v0
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_and_b32_e32 v0, s4, v0
; GFX10-NEXT: s_lshl_b32 s6, s6, 17
; GFX10-NEXT: s_lshl_b32 s0, s0, 1
; GFX10-NEXT: v_and_b32_e32 v2, s4, v2
; GFX10-NEXT: v_subrev_nc_u32_e32 v3, 24, v1
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v1
; GFX10-NEXT: v_lshrrev_b32_e64 v0, v0, s5
; GFX10-NEXT: s_or_b32 s0, s6, s0
; GFX10-NEXT: s_or_b32 s2, s2, s3
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: s_bfe_u32 s7, s7, 0x100000
; GFX10-NEXT: v_lshl_or_b32 v0, s0, v2, v0
; GFX10-NEXT: s_lshl_b32 s0, s7, 17
; GFX10-NEXT: s_lshl_b32 s1, s1, 1
; GFX10-NEXT: v_sub_nc_u32_e32 v3, 23, v1
; GFX10-NEXT: v_and_b32_e32 v1, v1, v4
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: v_and_b32_e32 v2, v3, v4
; GFX10-NEXT: v_lshrrev_b32_e64 v1, v1, s2
; GFX10-NEXT: v_lshrrev_b32_e32 v3, 8, v0
; GFX10-NEXT: v_lshl_or_b32 v1, s0, v2, v1
; GFX10-NEXT: s_mov_b32 s0, 8
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX10-NEXT: v_and_b32_e32 v3, s9, v1
; GFX10-NEXT: v_and_b32_sdwa v4, v1, s9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v1
; GFX10-NEXT: v_and_or_b32 v2, v0, s9, v2
; GFX10-NEXT: v_and_b32_sdwa v0, v0, s9 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 8, v4
; GFX10-NEXT: v_or3_b32 v0, v2, v0, v3
; GFX10-NEXT: v_and_or_b32 v1, v1, s9, v4
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s1, v1
; GFX10-NEXT: ; return to shader part epilog
%lhs = bitcast i48 %lhs.arg to <2 x i24>
%rhs = bitcast i48 %rhs.arg to <2 x i24>
%amt = bitcast i48 %amt.arg to <2 x i24>
%result = call <2 x i24> @llvm.fshr.v2i24(<2 x i24> %lhs, <2 x i24> %rhs, <2 x i24> %amt)
%cast.result = bitcast <2 x i24> %result to i48
ret i48 %cast.result
}
define <2 x i24> @v_fshr_v2i24(<2 x i24> %lhs, <2 x i24> %rhs, <2 x i24> %amt) {
; GFX6-LABEL: v_fshr_v2i24:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_cvt_f32_ubyte0_e32 v6, 24
; GFX6-NEXT: v_rcp_iflag_f32_e32 v6, v6
; GFX6-NEXT: s_sub_i32 s4, 0, 24
; GFX6-NEXT: v_and_b32_e32 v4, 0xffffff, v4
; GFX6-NEXT: v_cvt_f32_ubyte0_e32 v8, 24
; GFX6-NEXT: v_mul_f32_e32 v6, 0x4f7ffffe, v6
; GFX6-NEXT: v_cvt_u32_f32_e32 v6, v6
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 1, v1
; GFX6-NEXT: v_mul_lo_u32 v7, s4, v6
; GFX6-NEXT: v_mul_hi_u32 v7, v6, v7
; GFX6-NEXT: v_add_i32_e32 v6, vcc, v6, v7
; GFX6-NEXT: v_mul_hi_u32 v6, v4, v6
; GFX6-NEXT: v_rcp_iflag_f32_e32 v7, v8
; GFX6-NEXT: v_mov_b32_e32 v8, 0xffffff
; GFX6-NEXT: v_and_b32_e32 v5, v5, v8
; GFX6-NEXT: v_mul_lo_u32 v6, v6, 24
; GFX6-NEXT: v_mul_f32_e32 v7, 0x4f7ffffe, v7
; GFX6-NEXT: v_cvt_u32_f32_e32 v7, v7
; GFX6-NEXT: v_and_b32_e32 v2, v2, v8
; GFX6-NEXT: v_sub_i32_e32 v4, vcc, v4, v6
; GFX6-NEXT: v_subrev_i32_e32 v6, vcc, 24, v4
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v4
; GFX6-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc
; GFX6-NEXT: v_subrev_i32_e32 v6, vcc, 24, v4
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v4
; GFX6-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc
; GFX6-NEXT: v_mul_lo_u32 v6, s4, v7
; GFX6-NEXT: v_sub_i32_e32 v9, vcc, 23, v4
; GFX6-NEXT: v_and_b32_e32 v9, v9, v8
; GFX6-NEXT: v_and_b32_e32 v4, v4, v8
; GFX6-NEXT: v_mul_hi_u32 v6, v7, v6
; GFX6-NEXT: v_lshlrev_b32_e32 v0, v9, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v2
; GFX6-NEXT: v_or_b32_e32 v0, v0, v2
; GFX6-NEXT: v_add_i32_e32 v6, vcc, v7, v6
; GFX6-NEXT: v_mul_hi_u32 v6, v5, v6
; GFX6-NEXT: v_and_b32_e32 v3, v3, v8
; GFX6-NEXT: v_mul_lo_u32 v6, v6, 24
; GFX6-NEXT: v_sub_i32_e32 v2, vcc, v5, v6
; GFX6-NEXT: v_subrev_i32_e32 v4, vcc, 24, v2
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX6-NEXT: v_subrev_i32_e32 v4, vcc, 24, v2
; GFX6-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX6-NEXT: v_sub_i32_e32 v4, vcc, 23, v2
; GFX6-NEXT: v_and_b32_e32 v2, v2, v8
; GFX6-NEXT: v_and_b32_e32 v4, v4, v8
; GFX6-NEXT: v_lshlrev_b32_e32 v1, v4, v1
; GFX6-NEXT: v_lshrrev_b32_e32 v2, v2, v3
; GFX6-NEXT: v_or_b32_e32 v1, v1, v2
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v2i24:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_cvt_f32_ubyte0_e32 v6, 24
; GFX8-NEXT: v_rcp_iflag_f32_e32 v6, v6
; GFX8-NEXT: s_sub_i32 s4, 0, 24
; GFX8-NEXT: v_and_b32_e32 v4, 0xffffff, v4
; GFX8-NEXT: v_cvt_f32_ubyte0_e32 v8, 24
; GFX8-NEXT: v_mul_f32_e32 v6, 0x4f7ffffe, v6
; GFX8-NEXT: v_cvt_u32_f32_e32 v6, v6
; GFX8-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX8-NEXT: v_lshlrev_b32_e32 v1, 1, v1
; GFX8-NEXT: v_mul_lo_u32 v7, s4, v6
; GFX8-NEXT: v_mul_hi_u32 v7, v6, v7
; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v7
; GFX8-NEXT: v_mul_hi_u32 v6, v4, v6
; GFX8-NEXT: v_rcp_iflag_f32_e32 v7, v8
; GFX8-NEXT: v_mov_b32_e32 v8, 0xffffff
; GFX8-NEXT: v_and_b32_e32 v5, v5, v8
; GFX8-NEXT: v_mul_lo_u32 v6, v6, 24
; GFX8-NEXT: v_mul_f32_e32 v7, 0x4f7ffffe, v7
; GFX8-NEXT: v_cvt_u32_f32_e32 v7, v7
; GFX8-NEXT: v_and_b32_e32 v2, v2, v8
; GFX8-NEXT: v_sub_u32_e32 v4, vcc, v4, v6
; GFX8-NEXT: v_subrev_u32_e32 v6, vcc, 24, v4
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v4
; GFX8-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc
; GFX8-NEXT: v_subrev_u32_e32 v6, vcc, 24, v4
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v4
; GFX8-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc
; GFX8-NEXT: v_mul_lo_u32 v6, s4, v7
; GFX8-NEXT: v_sub_u32_e32 v9, vcc, 23, v4
; GFX8-NEXT: v_and_b32_e32 v9, v9, v8
; GFX8-NEXT: v_and_b32_e32 v4, v4, v8
; GFX8-NEXT: v_mul_hi_u32 v6, v7, v6
; GFX8-NEXT: v_lshlrev_b32_e32 v0, v9, v0
; GFX8-NEXT: v_lshrrev_b32_e32 v2, v4, v2
; GFX8-NEXT: v_or_b32_e32 v0, v0, v2
; GFX8-NEXT: v_add_u32_e32 v6, vcc, v7, v6
; GFX8-NEXT: v_mul_hi_u32 v6, v5, v6
; GFX8-NEXT: v_and_b32_e32 v3, v3, v8
; GFX8-NEXT: v_mul_lo_u32 v6, v6, 24
; GFX8-NEXT: v_sub_u32_e32 v2, vcc, v5, v6
; GFX8-NEXT: v_subrev_u32_e32 v4, vcc, 24, v2
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX8-NEXT: v_subrev_u32_e32 v4, vcc, 24, v2
; GFX8-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX8-NEXT: v_sub_u32_e32 v4, vcc, 23, v2
; GFX8-NEXT: v_and_b32_e32 v2, v2, v8
; GFX8-NEXT: v_and_b32_e32 v4, v4, v8
; GFX8-NEXT: v_lshlrev_b32_e32 v1, v4, v1
; GFX8-NEXT: v_lshrrev_b32_e32 v2, v2, v3
; GFX8-NEXT: v_or_b32_e32 v1, v1, v2
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v2i24:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_cvt_f32_ubyte0_e32 v6, 24
; GFX9-NEXT: v_rcp_iflag_f32_e32 v6, v6
; GFX9-NEXT: s_sub_i32 s4, 0, 24
; GFX9-NEXT: v_cvt_f32_ubyte0_e32 v8, 24
; GFX9-NEXT: v_rcp_iflag_f32_e32 v8, v8
; GFX9-NEXT: v_mul_f32_e32 v6, 0x4f7ffffe, v6
; GFX9-NEXT: v_cvt_u32_f32_e32 v6, v6
; GFX9-NEXT: v_and_b32_e32 v4, 0xffffff, v4
; GFX9-NEXT: v_mul_f32_e32 v8, 0x4f7ffffe, v8
; GFX9-NEXT: v_mov_b32_e32 v9, 0xffffff
; GFX9-NEXT: v_mul_lo_u32 v7, s4, v6
; GFX9-NEXT: v_and_b32_e32 v5, v5, v9
; GFX9-NEXT: v_and_b32_e32 v2, v2, v9
; GFX9-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX9-NEXT: v_mul_hi_u32 v7, v6, v7
; GFX9-NEXT: v_and_b32_e32 v3, v3, v9
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 1, v1
; GFX9-NEXT: v_add_u32_e32 v6, v6, v7
; GFX9-NEXT: v_cvt_u32_f32_e32 v7, v8
; GFX9-NEXT: v_mul_hi_u32 v6, v4, v6
; GFX9-NEXT: v_mul_lo_u32 v8, s4, v7
; GFX9-NEXT: v_mul_lo_u32 v6, v6, 24
; GFX9-NEXT: v_mul_hi_u32 v8, v7, v8
; GFX9-NEXT: v_sub_u32_e32 v4, v4, v6
; GFX9-NEXT: v_subrev_u32_e32 v6, 24, v4
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v4
; GFX9-NEXT: v_add_u32_e32 v7, v7, v8
; GFX9-NEXT: v_mul_hi_u32 v7, v5, v7
; GFX9-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc
; GFX9-NEXT: v_subrev_u32_e32 v6, 24, v4
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v4
; GFX9-NEXT: v_mul_lo_u32 v7, v7, 24
; GFX9-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc
; GFX9-NEXT: v_sub_u32_e32 v6, 23, v4
; GFX9-NEXT: v_and_b32_e32 v4, v4, v9
; GFX9-NEXT: v_and_b32_e32 v6, v6, v9
; GFX9-NEXT: v_lshrrev_b32_e32 v2, v4, v2
; GFX9-NEXT: v_lshl_or_b32 v0, v0, v6, v2
; GFX9-NEXT: v_sub_u32_e32 v2, v5, v7
; GFX9-NEXT: v_subrev_u32_e32 v4, 24, v2
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX9-NEXT: v_subrev_u32_e32 v4, 24, v2
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 24, v2
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX9-NEXT: v_sub_u32_e32 v4, 23, v2
; GFX9-NEXT: v_and_b32_e32 v2, v2, v9
; GFX9-NEXT: v_and_b32_e32 v4, v4, v9
; GFX9-NEXT: v_lshrrev_b32_e32 v2, v2, v3
; GFX9-NEXT: v_lshl_or_b32 v1, v1, v4, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v2i24:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v6, 24
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v7, 24
; GFX10-NEXT: s_sub_i32 s4, 0, 24
; GFX10-NEXT: v_mov_b32_e32 v10, 0xffffff
; GFX10-NEXT: v_and_b32_e32 v4, 0xffffff, v4
; GFX10-NEXT: v_rcp_iflag_f32_e32 v6, v6
; GFX10-NEXT: v_rcp_iflag_f32_e32 v7, v7
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX10-NEXT: v_and_b32_e32 v5, v5, v10
; GFX10-NEXT: v_and_b32_e32 v2, v2, v10
; GFX10-NEXT: v_and_b32_e32 v3, v3, v10
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 1, v1
; GFX10-NEXT: v_mul_f32_e32 v6, 0x4f7ffffe, v6
; GFX10-NEXT: v_mul_f32_e32 v7, 0x4f7ffffe, v7
; GFX10-NEXT: v_cvt_u32_f32_e32 v6, v6
; GFX10-NEXT: v_cvt_u32_f32_e32 v7, v7
; GFX10-NEXT: v_mul_lo_u32 v8, s4, v6
; GFX10-NEXT: v_mul_lo_u32 v9, s4, v7
; GFX10-NEXT: v_mul_hi_u32 v8, v6, v8
; GFX10-NEXT: v_mul_hi_u32 v9, v7, v9
; GFX10-NEXT: v_add_nc_u32_e32 v6, v6, v8
; GFX10-NEXT: v_add_nc_u32_e32 v7, v7, v9
; GFX10-NEXT: v_mul_hi_u32 v6, v4, v6
; GFX10-NEXT: v_mul_hi_u32 v7, v5, v7
; GFX10-NEXT: v_mul_lo_u32 v6, v6, 24
; GFX10-NEXT: v_mul_lo_u32 v7, v7, 24
; GFX10-NEXT: v_sub_nc_u32_e32 v4, v4, v6
; GFX10-NEXT: v_sub_nc_u32_e32 v5, v5, v7
; GFX10-NEXT: v_subrev_nc_u32_e32 v6, 24, v4
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v4
; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 24, v5
; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v5
; GFX10-NEXT: v_subrev_nc_u32_e32 v6, 24, v4
; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v4
; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 24, v5
; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 24, v5
; GFX10-NEXT: v_sub_nc_u32_e32 v6, 23, v4
; GFX10-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc_lo
; GFX10-NEXT: v_and_b32_e32 v4, v4, v10
; GFX10-NEXT: v_and_b32_e32 v6, v6, v10
; GFX10-NEXT: v_sub_nc_u32_e32 v7, 23, v5
; GFX10-NEXT: v_and_b32_e32 v5, v5, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v2, v4, v2
; GFX10-NEXT: v_and_b32_e32 v4, v7, v10
; GFX10-NEXT: v_lshrrev_b32_e32 v3, v5, v3
; GFX10-NEXT: v_lshl_or_b32 v0, v0, v6, v2
; GFX10-NEXT: v_lshl_or_b32 v1, v1, v4, v3
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i24> @llvm.fshr.v2i24(<2 x i24> %lhs, <2 x i24> %rhs, <2 x i24> %amt)
ret <2 x i24> %result
}
define amdgpu_ps i32 @s_fshr_i32(i32 inreg %lhs, i32 inreg %rhs, i32 inreg %amt) {
; GFX6-LABEL: s_fshr_i32:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v0, s1
; GFX6-NEXT: v_mov_b32_e32 v1, s2
; GFX6-NEXT: v_alignbit_b32 v0, s0, v0, v1
; GFX6-NEXT: v_readfirstlane_b32 s0, v0
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i32:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_mov_b32_e32 v0, s1
; GFX8-NEXT: v_mov_b32_e32 v1, s2
; GFX8-NEXT: v_alignbit_b32 v0, s0, v0, v1
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i32:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_mov_b32_e32 v0, s1
; GFX9-NEXT: v_mov_b32_e32 v1, s2
; GFX9-NEXT: v_alignbit_b32 v0, s0, v0, v1
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i32:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_mov_b32_e32 v0, s2
; GFX10-NEXT: v_alignbit_b32 v0, s0, s1, v0
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 %amt)
ret i32 %result
}
define amdgpu_ps i32 @s_fshr_i32_5(i32 inreg %lhs, i32 inreg %rhs) {
; GFX6-LABEL: s_fshr_i32_5:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v0, s1
; GFX6-NEXT: v_alignbit_b32 v0, s0, v0, 5
; GFX6-NEXT: v_readfirstlane_b32 s0, v0
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i32_5:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_mov_b32_e32 v0, s1
; GFX8-NEXT: v_alignbit_b32 v0, s0, v0, 5
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i32_5:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_mov_b32_e32 v0, s1
; GFX9-NEXT: v_alignbit_b32 v0, s0, v0, 5
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i32_5:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_alignbit_b32 v0, s0, s1, 5
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 5)
ret i32 %result
}
define amdgpu_ps i32 @s_fshr_i32_8(i32 inreg %lhs, i32 inreg %rhs) {
; GFX6-LABEL: s_fshr_i32_8:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v0, s1
; GFX6-NEXT: v_alignbit_b32 v0, s0, v0, 8
; GFX6-NEXT: v_readfirstlane_b32 s0, v0
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i32_8:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_mov_b32_e32 v0, s1
; GFX8-NEXT: v_alignbit_b32 v0, s0, v0, 8
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i32_8:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_mov_b32_e32 v0, s1
; GFX9-NEXT: v_alignbit_b32 v0, s0, v0, 8
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i32_8:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_alignbit_b32 v0, s0, s1, 8
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 8)
ret i32 %result
}
define i32 @v_fshr_i32(i32 %lhs, i32 %rhs, i32 %amt) {
; GFX6-LABEL: v_fshr_i32:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_alignbit_b32 v0, v0, v1, v2
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i32:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_alignbit_b32 v0, v0, v1, v2
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i32:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_alignbit_b32 v0, v0, v1, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i32:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_alignbit_b32 v0, v0, v1, v2
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 %amt)
ret i32 %result
}
define i32 @v_fshr_i32_5(i32 %lhs, i32 %rhs) {
; GFX6-LABEL: v_fshr_i32_5:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_alignbit_b32 v0, v0, v1, 5
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i32_5:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_alignbit_b32 v0, v0, v1, 5
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i32_5:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_alignbit_b32 v0, v0, v1, 5
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i32_5:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_alignbit_b32 v0, v0, v1, 5
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 5)
ret i32 %result
}
define i32 @v_fshr_i32_8(i32 %lhs, i32 %rhs) {
; GFX6-LABEL: v_fshr_i32_8:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_alignbit_b32 v0, v0, v1, 8
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i32_8:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_alignbit_b32 v0, v0, v1, 8
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i32_8:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_alignbit_b32 v0, v0, v1, 8
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i32_8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_alignbit_b32 v0, v0, v1, 8
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 8)
ret i32 %result
}
define amdgpu_ps float @v_fshr_i32_ssv(i32 inreg %lhs, i32 inreg %rhs, i32 %amt) {
; GFX6-LABEL: v_fshr_i32_ssv:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v1, s1
; GFX6-NEXT: v_alignbit_b32 v0, s0, v1, v0
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i32_ssv:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_mov_b32_e32 v1, s1
; GFX8-NEXT: v_alignbit_b32 v0, s0, v1, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i32_ssv:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_mov_b32_e32 v1, s1
; GFX9-NEXT: v_alignbit_b32 v0, s0, v1, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i32_ssv:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_alignbit_b32 v0, s0, s1, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 %amt)
%cast.result = bitcast i32 %result to float
ret float %cast.result
}
define amdgpu_ps float @v_fshr_i32_svs(i32 inreg %lhs, i32 %rhs, i32 inreg %amt) {
; GFX6-LABEL: v_fshr_i32_svs:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v1, s1
; GFX6-NEXT: v_alignbit_b32 v0, s0, v0, v1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i32_svs:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_mov_b32_e32 v1, s1
; GFX8-NEXT: v_alignbit_b32 v0, s0, v0, v1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i32_svs:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_mov_b32_e32 v1, s1
; GFX9-NEXT: v_alignbit_b32 v0, s0, v0, v1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i32_svs:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_alignbit_b32 v0, s0, v0, s1
; GFX10-NEXT: ; return to shader part epilog
%result = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 %amt)
%cast.result = bitcast i32 %result to float
ret float %cast.result
}
define amdgpu_ps float @v_fshr_i32_vss(i32 inreg %lhs, i32 inreg %rhs, i32 inreg %amt) {
; GFX6-LABEL: v_fshr_i32_vss:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v0, s1
; GFX6-NEXT: v_mov_b32_e32 v1, s2
; GFX6-NEXT: v_alignbit_b32 v0, s0, v0, v1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i32_vss:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_mov_b32_e32 v0, s1
; GFX8-NEXT: v_mov_b32_e32 v1, s2
; GFX8-NEXT: v_alignbit_b32 v0, s0, v0, v1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i32_vss:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_mov_b32_e32 v0, s1
; GFX9-NEXT: v_mov_b32_e32 v1, s2
; GFX9-NEXT: v_alignbit_b32 v0, s0, v0, v1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i32_vss:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_mov_b32_e32 v0, s2
; GFX10-NEXT: v_alignbit_b32 v0, s0, s1, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 %amt)
%cast.result = bitcast i32 %result to float
ret float %cast.result
}
define <2 x i32> @v_fshr_v2i32(<2 x i32> %lhs, <2 x i32> %rhs, <2 x i32> %amt) {
; GFX6-LABEL: v_fshr_v2i32:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_alignbit_b32 v0, v0, v2, v4
; GFX6-NEXT: v_alignbit_b32 v1, v1, v3, v5
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v2i32:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_alignbit_b32 v0, v0, v2, v4
; GFX8-NEXT: v_alignbit_b32 v1, v1, v3, v5
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v2i32:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_alignbit_b32 v0, v0, v2, v4
; GFX9-NEXT: v_alignbit_b32 v1, v1, v3, v5
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v2i32:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_alignbit_b32 v0, v0, v2, v4
; GFX10-NEXT: v_alignbit_b32 v1, v1, v3, v5
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %lhs, <2 x i32> %rhs, <2 x i32> %amt)
ret <2 x i32> %result
}
define <3 x i32> @v_fshr_v3i32(<3 x i32> %lhs, <3 x i32> %rhs, <3 x i32> %amt) {
; GFX6-LABEL: v_fshr_v3i32:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_alignbit_b32 v0, v0, v3, v6
; GFX6-NEXT: v_alignbit_b32 v1, v1, v4, v7
; GFX6-NEXT: v_alignbit_b32 v2, v2, v5, v8
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v3i32:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_alignbit_b32 v0, v0, v3, v6
; GFX8-NEXT: v_alignbit_b32 v1, v1, v4, v7
; GFX8-NEXT: v_alignbit_b32 v2, v2, v5, v8
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v3i32:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_alignbit_b32 v0, v0, v3, v6
; GFX9-NEXT: v_alignbit_b32 v1, v1, v4, v7
; GFX9-NEXT: v_alignbit_b32 v2, v2, v5, v8
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v3i32:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_alignbit_b32 v0, v0, v3, v6
; GFX10-NEXT: v_alignbit_b32 v1, v1, v4, v7
; GFX10-NEXT: v_alignbit_b32 v2, v2, v5, v8
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <3 x i32> @llvm.fshr.v3i32(<3 x i32> %lhs, <3 x i32> %rhs, <3 x i32> %amt)
ret <3 x i32> %result
}
define <4 x i32> @v_fshr_v4i32(<4 x i32> %lhs, <4 x i32> %rhs, <4 x i32> %amt) {
; GFX6-LABEL: v_fshr_v4i32:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_alignbit_b32 v0, v0, v4, v8
; GFX6-NEXT: v_alignbit_b32 v1, v1, v5, v9
; GFX6-NEXT: v_alignbit_b32 v2, v2, v6, v10
; GFX6-NEXT: v_alignbit_b32 v3, v3, v7, v11
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v4i32:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_alignbit_b32 v0, v0, v4, v8
; GFX8-NEXT: v_alignbit_b32 v1, v1, v5, v9
; GFX8-NEXT: v_alignbit_b32 v2, v2, v6, v10
; GFX8-NEXT: v_alignbit_b32 v3, v3, v7, v11
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v4i32:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_alignbit_b32 v0, v0, v4, v8
; GFX9-NEXT: v_alignbit_b32 v1, v1, v5, v9
; GFX9-NEXT: v_alignbit_b32 v2, v2, v6, v10
; GFX9-NEXT: v_alignbit_b32 v3, v3, v7, v11
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v4i32:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_alignbit_b32 v0, v0, v4, v8
; GFX10-NEXT: v_alignbit_b32 v1, v1, v5, v9
; GFX10-NEXT: v_alignbit_b32 v2, v2, v6, v10
; GFX10-NEXT: v_alignbit_b32 v3, v3, v7, v11
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %lhs, <4 x i32> %rhs, <4 x i32> %amt)
ret <4 x i32> %result
}
define amdgpu_ps i16 @s_fshr_i16(i16 inreg %lhs, i16 inreg %rhs, i16 inreg %amt) {
; GFX6-LABEL: s_fshr_i16:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b32 s3, s2, 15
; GFX6-NEXT: s_andn2_b32 s2, 15, s2
; GFX6-NEXT: s_lshl_b32 s0, s0, 1
; GFX6-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX6-NEXT: s_lshl_b32 s0, s0, s2
; GFX6-NEXT: s_bfe_u32 s2, s3, 0x100000
; GFX6-NEXT: s_and_b32 s1, s1, 0xffff
; GFX6-NEXT: s_lshr_b32 s1, s1, s2
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_and_b32 s3, s2, 15
; GFX8-NEXT: s_andn2_b32 s2, 15, s2
; GFX8-NEXT: s_bfe_u32 s4, 1, 0x100000
; GFX8-NEXT: s_lshl_b32 s0, s0, s4
; GFX8-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX8-NEXT: s_lshl_b32 s0, s0, s2
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_bfe_u32 s2, s3, 0x100000
; GFX8-NEXT: s_lshr_b32 s1, s1, s2
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_and_b32 s3, s2, 15
; GFX9-NEXT: s_andn2_b32 s2, 15, s2
; GFX9-NEXT: s_bfe_u32 s4, 1, 0x100000
; GFX9-NEXT: s_lshl_b32 s0, s0, s4
; GFX9-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX9-NEXT: s_lshl_b32 s0, s0, s2
; GFX9-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX9-NEXT: s_bfe_u32 s2, s3, 0x100000
; GFX9-NEXT: s_lshr_b32 s1, s1, s2
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_and_b32 s3, s2, 15
; GFX10-NEXT: s_andn2_b32 s2, 15, s2
; GFX10-NEXT: s_bfe_u32 s4, 1, 0x100000
; GFX10-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX10-NEXT: s_lshl_b32 s0, s0, s4
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: s_bfe_u32 s3, s3, 0x100000
; GFX10-NEXT: s_lshl_b32 s0, s0, s2
; GFX10-NEXT: s_lshr_b32 s1, s1, s3
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
%result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 %amt)
ret i16 %result
}
define amdgpu_ps i16 @s_fshr_i16_4(i16 inreg %lhs, i16 inreg %rhs) {
; GFX6-LABEL: s_fshr_i16_4:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b32 s1, s1, 0xffff
; GFX6-NEXT: s_lshl_b32 s0, s0, 12
; GFX6-NEXT: s_lshr_b32 s1, s1, 4
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i16_4:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_bfe_u32 s2, 12, 0x100000
; GFX8-NEXT: s_lshl_b32 s0, s0, s2
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_bfe_u32 s2, 4, 0x100000
; GFX8-NEXT: s_lshr_b32 s1, s1, s2
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i16_4:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_bfe_u32 s2, 12, 0x100000
; GFX9-NEXT: s_lshl_b32 s0, s0, s2
; GFX9-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX9-NEXT: s_bfe_u32 s2, 4, 0x100000
; GFX9-NEXT: s_lshr_b32 s1, s1, s2
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i16_4:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_bfe_u32 s2, 12, 0x100000
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: s_bfe_u32 s3, 4, 0x100000
; GFX10-NEXT: s_lshl_b32 s0, s0, s2
; GFX10-NEXT: s_lshr_b32 s1, s1, s3
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
%result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 4)
ret i16 %result
}
define amdgpu_ps i16 @s_fshr_i16_5(i16 inreg %lhs, i16 inreg %rhs) {
; GFX6-LABEL: s_fshr_i16_5:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b32 s1, s1, 0xffff
; GFX6-NEXT: s_lshl_b32 s0, s0, 11
; GFX6-NEXT: s_lshr_b32 s1, s1, 5
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i16_5:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_bfe_u32 s2, 11, 0x100000
; GFX8-NEXT: s_lshl_b32 s0, s0, s2
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_bfe_u32 s2, 5, 0x100000
; GFX8-NEXT: s_lshr_b32 s1, s1, s2
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i16_5:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_bfe_u32 s2, 11, 0x100000
; GFX9-NEXT: s_lshl_b32 s0, s0, s2
; GFX9-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX9-NEXT: s_bfe_u32 s2, 5, 0x100000
; GFX9-NEXT: s_lshr_b32 s1, s1, s2
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i16_5:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_bfe_u32 s2, 11, 0x100000
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: s_bfe_u32 s3, 5, 0x100000
; GFX10-NEXT: s_lshl_b32 s0, s0, s2
; GFX10-NEXT: s_lshr_b32 s1, s1, s3
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
%result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 5)
ret i16 %result
}
define i16 @v_fshr_i16(i16 %lhs, i16 %rhs, i16 %amt) {
; GFX6-LABEL: v_fshr_i16:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_and_b32_e32 v3, 15, v2
; GFX6-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX6-NEXT: v_and_b32_e32 v2, 15, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX6-NEXT: v_bfe_u32 v2, v2, 0, 16
; GFX6-NEXT: v_lshlrev_b32_e32 v0, v2, v0
; GFX6-NEXT: v_bfe_u32 v2, v3, 0, 16
; GFX6-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX6-NEXT: v_lshrrev_b32_e32 v1, v2, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_and_b32_e32 v3, 15, v2
; GFX8-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX8-NEXT: v_and_b32_e32 v2, 15, v2
; GFX8-NEXT: v_lshlrev_b16_e32 v0, 1, v0
; GFX8-NEXT: v_lshlrev_b16_e32 v0, v2, v0
; GFX8-NEXT: v_lshrrev_b16_e32 v1, v3, v1
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_and_b32_e32 v3, 15, v2
; GFX9-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX9-NEXT: v_and_b32_e32 v2, 15, v2
; GFX9-NEXT: v_lshlrev_b16_e32 v0, 1, v0
; GFX9-NEXT: v_lshlrev_b16_e32 v0, v2, v0
; GFX9-NEXT: v_lshrrev_b16_e32 v1, v3, v1
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v3, -1, v2
; GFX10-NEXT: v_and_b32_e32 v2, 15, v2
; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0
; GFX10-NEXT: v_and_b32_e32 v3, 15, v3
; GFX10-NEXT: v_lshrrev_b16 v1, v2, v1
; GFX10-NEXT: v_lshlrev_b16 v0, v3, v0
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 %amt)
ret i16 %result
}
define i16 @v_fshr_i16_4(i16 %lhs, i16 %rhs) {
; GFX6-LABEL: v_fshr_i16_4:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 12, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v1, 4, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i16_4:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_lshlrev_b16_e32 v0, 12, v0
; GFX8-NEXT: v_lshrrev_b16_e32 v1, 4, v1
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i16_4:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_lshlrev_b16_e32 v0, 12, v0
; GFX9-NEXT: v_lshrrev_b16_e32 v1, 4, v1
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i16_4:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_lshlrev_b16 v0, 12, v0
; GFX10-NEXT: v_lshrrev_b16 v1, 4, v1
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 4)
ret i16 %result
}
define i16 @v_fshr_i16_5(i16 %lhs, i16 %rhs) {
; GFX6-LABEL: v_fshr_i16_5:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 11, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v1, 5, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i16_5:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_lshlrev_b16_e32 v0, 11, v0
; GFX8-NEXT: v_lshrrev_b16_e32 v1, 5, v1
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i16_5:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_lshlrev_b16_e32 v0, 11, v0
; GFX9-NEXT: v_lshrrev_b16_e32 v1, 5, v1
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i16_5:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_lshlrev_b16 v0, 11, v0
; GFX10-NEXT: v_lshrrev_b16 v1, 5, v1
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 5)
ret i16 %result
}
define amdgpu_ps half @v_fshr_i16_ssv(i16 inreg %lhs, i16 inreg %rhs, i16 %amt) {
; GFX6-LABEL: v_fshr_i16_ssv:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_and_b32_e32 v1, 15, v0
; GFX6-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX6-NEXT: v_and_b32_e32 v0, 15, v0
; GFX6-NEXT: s_lshl_b32 s0, s0, 1
; GFX6-NEXT: v_bfe_u32 v0, v0, 0, 16
; GFX6-NEXT: v_lshl_b32_e32 v0, s0, v0
; GFX6-NEXT: v_bfe_u32 v1, v1, 0, 16
; GFX6-NEXT: s_and_b32 s0, s1, 0xffff
; GFX6-NEXT: v_lshr_b32_e32 v1, s0, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i16_ssv:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_and_b32_e32 v1, 15, v0
; GFX8-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX8-NEXT: s_bfe_u32 s2, 1, 0x100000
; GFX8-NEXT: v_and_b32_e32 v0, 15, v0
; GFX8-NEXT: s_lshl_b32 s0, s0, s2
; GFX8-NEXT: v_lshlrev_b16_e64 v0, v0, s0
; GFX8-NEXT: v_lshrrev_b16_e64 v1, v1, s1
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i16_ssv:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_and_b32_e32 v1, 15, v0
; GFX9-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX9-NEXT: s_bfe_u32 s2, 1, 0x100000
; GFX9-NEXT: v_and_b32_e32 v0, 15, v0
; GFX9-NEXT: s_lshl_b32 s0, s0, s2
; GFX9-NEXT: v_lshlrev_b16_e64 v0, v0, s0
; GFX9-NEXT: v_lshrrev_b16_e64 v1, v1, s1
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i16_ssv:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_xor_b32_e32 v1, -1, v0
; GFX10-NEXT: v_and_b32_e32 v0, 15, v0
; GFX10-NEXT: s_bfe_u32 s2, 1, 0x100000
; GFX10-NEXT: s_lshl_b32 s0, s0, s2
; GFX10-NEXT: v_and_b32_e32 v1, 15, v1
; GFX10-NEXT: v_lshrrev_b16 v0, v0, s1
; GFX10-NEXT: v_lshlrev_b16 v1, v1, s0
; GFX10-NEXT: v_or_b32_e32 v0, v1, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 %amt)
%cast.result = bitcast i16 %result to half
ret half %cast.result
}
define amdgpu_ps half @v_fshr_i16_svs(i16 inreg %lhs, i16 %rhs, i16 inreg %amt) {
; GFX6-LABEL: v_fshr_i16_svs:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b32 s2, s1, 15
; GFX6-NEXT: s_andn2_b32 s1, 15, s1
; GFX6-NEXT: s_lshl_b32 s0, s0, 1
; GFX6-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX6-NEXT: s_lshl_b32 s0, s0, s1
; GFX6-NEXT: s_bfe_u32 s1, s2, 0x100000
; GFX6-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v0, s1, v0
; GFX6-NEXT: v_or_b32_e32 v0, s0, v0
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i16_svs:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_and_b32 s2, s1, 15
; GFX8-NEXT: s_andn2_b32 s1, 15, s1
; GFX8-NEXT: s_bfe_u32 s3, 1, 0x100000
; GFX8-NEXT: s_lshl_b32 s0, s0, s3
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_lshl_b32 s0, s0, s1
; GFX8-NEXT: v_lshrrev_b16_e32 v0, s2, v0
; GFX8-NEXT: v_or_b32_e32 v0, s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i16_svs:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_and_b32 s2, s1, 15
; GFX9-NEXT: s_andn2_b32 s1, 15, s1
; GFX9-NEXT: s_bfe_u32 s3, 1, 0x100000
; GFX9-NEXT: s_lshl_b32 s0, s0, s3
; GFX9-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX9-NEXT: s_lshl_b32 s0, s0, s1
; GFX9-NEXT: v_lshrrev_b16_e32 v0, s2, v0
; GFX9-NEXT: v_or_b32_e32 v0, s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i16_svs:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_and_b32 s2, s1, 15
; GFX10-NEXT: s_andn2_b32 s1, 15, s1
; GFX10-NEXT: s_bfe_u32 s3, 1, 0x100000
; GFX10-NEXT: v_lshrrev_b16 v0, s2, v0
; GFX10-NEXT: s_lshl_b32 s0, s0, s3
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: s_lshl_b32 s0, s0, s1
; GFX10-NEXT: v_or_b32_e32 v0, s0, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 %amt)
%cast.result = bitcast i16 %result to half
ret half %cast.result
}
define amdgpu_ps half @v_fshr_i16_vss(i16 %lhs, i16 inreg %rhs, i16 inreg %amt) {
; GFX6-LABEL: v_fshr_i16_vss:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b32 s2, s1, 15
; GFX6-NEXT: s_andn2_b32 s1, 15, s1
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX6-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX6-NEXT: v_lshlrev_b32_e32 v0, s1, v0
; GFX6-NEXT: s_bfe_u32 s1, s2, 0x100000
; GFX6-NEXT: s_and_b32 s0, s0, 0xffff
; GFX6-NEXT: s_lshr_b32 s0, s0, s1
; GFX6-NEXT: v_or_b32_e32 v0, s0, v0
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i16_vss:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_and_b32 s2, s1, 15
; GFX8-NEXT: s_andn2_b32 s1, 15, s1
; GFX8-NEXT: v_lshlrev_b16_e32 v0, 1, v0
; GFX8-NEXT: v_lshlrev_b16_e32 v0, s1, v0
; GFX8-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX8-NEXT: s_bfe_u32 s1, s2, 0x100000
; GFX8-NEXT: s_lshr_b32 s0, s0, s1
; GFX8-NEXT: v_or_b32_e32 v0, s0, v0
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i16_vss:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_and_b32 s2, s1, 15
; GFX9-NEXT: s_andn2_b32 s1, 15, s1
; GFX9-NEXT: v_lshlrev_b16_e32 v0, 1, v0
; GFX9-NEXT: v_lshlrev_b16_e32 v0, s1, v0
; GFX9-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX9-NEXT: s_bfe_u32 s1, s2, 0x100000
; GFX9-NEXT: s_lshr_b32 s0, s0, s1
; GFX9-NEXT: v_or_b32_e32 v0, s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i16_vss:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_lshlrev_b16 v0, 1, v0
; GFX10-NEXT: s_andn2_b32 s2, 15, s1
; GFX10-NEXT: s_and_b32 s1, s1, 15
; GFX10-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX10-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX10-NEXT: v_lshlrev_b16 v0, s2, v0
; GFX10-NEXT: s_lshr_b32 s0, s0, s1
; GFX10-NEXT: v_or_b32_e32 v0, s0, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 %amt)
%cast.result = bitcast i16 %result to half
ret half %cast.result
}
define amdgpu_ps i32 @s_fshr_v2i16(<2 x i16> inreg %lhs, <2 x i16> inreg %rhs, <2 x i16> inreg %amt) {
; GFX6-LABEL: s_fshr_v2i16:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s6, 0xffff
; GFX6-NEXT: s_lshl_b32 s5, s5, 16
; GFX6-NEXT: s_and_b32 s4, s4, s6
; GFX6-NEXT: s_or_b32 s4, s5, s4
; GFX6-NEXT: s_bfe_u32 s5, 1, 0x100000
; GFX6-NEXT: s_and_b32 s7, s2, s6
; GFX6-NEXT: s_lshl_b32 s0, s0, s5
; GFX6-NEXT: s_lshl_b32 s1, s1, s5
; GFX6-NEXT: s_and_b32 s5, s3, s6
; GFX6-NEXT: s_lshr_b32 s7, s7, 1
; GFX6-NEXT: s_bfe_u32 s8, 14, 0x100000
; GFX6-NEXT: s_lshr_b32 s5, s5, 1
; GFX6-NEXT: s_lshl_b32 s2, s2, 1
; GFX6-NEXT: s_lshr_b32 s7, s7, s8
; GFX6-NEXT: s_lshr_b32 s5, s5, s8
; GFX6-NEXT: s_xor_b32 s4, s4, -1
; GFX6-NEXT: s_and_b32 s2, s2, s6
; GFX6-NEXT: s_or_b32 s0, s0, s7
; GFX6-NEXT: s_and_b32 s7, s4, 15
; GFX6-NEXT: s_or_b32 s1, s1, s5
; GFX6-NEXT: s_lshr_b32 s5, s4, 16
; GFX6-NEXT: s_andn2_b32 s4, 15, s4
; GFX6-NEXT: s_bfe_u32 s7, s7, 0x100000
; GFX6-NEXT: s_lshr_b32 s2, s2, 1
; GFX6-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX6-NEXT: s_lshr_b32 s2, s2, s4
; GFX6-NEXT: s_lshl_b32 s0, s0, s7
; GFX6-NEXT: s_or_b32 s0, s0, s2
; GFX6-NEXT: s_and_b32 s2, s5, 15
; GFX6-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX6-NEXT: s_lshl_b32 s3, s3, 1
; GFX6-NEXT: s_lshl_b32 s1, s1, s2
; GFX6-NEXT: s_and_b32 s2, s3, s6
; GFX6-NEXT: s_andn2_b32 s4, 15, s5
; GFX6-NEXT: s_lshr_b32 s2, s2, 1
; GFX6-NEXT: s_bfe_u32 s3, s4, 0x100000
; GFX6-NEXT: s_lshr_b32 s2, s2, s3
; GFX6-NEXT: s_or_b32 s1, s1, s2
; GFX6-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX6-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX6-NEXT: s_lshl_b32 s1, s1, 16
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_v2i16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_bfe_u32 s5, 1, 0x100000
; GFX8-NEXT: s_bfe_u32 s6, s1, 0x100000
; GFX8-NEXT: s_lshr_b32 s6, s6, s5
; GFX8-NEXT: s_bfe_u32 s7, 14, 0x100000
; GFX8-NEXT: s_lshr_b32 s3, s0, 16
; GFX8-NEXT: s_lshr_b32 s4, s1, 16
; GFX8-NEXT: s_lshl_b32 s0, s0, s5
; GFX8-NEXT: s_lshr_b32 s6, s6, s7
; GFX8-NEXT: s_or_b32 s0, s0, s6
; GFX8-NEXT: s_lshr_b32 s6, s4, s5
; GFX8-NEXT: s_lshl_b32 s1, s1, s5
; GFX8-NEXT: s_lshl_b32 s3, s3, s5
; GFX8-NEXT: s_lshr_b32 s6, s6, s7
; GFX8-NEXT: s_xor_b32 s2, s2, -1
; GFX8-NEXT: s_and_b32 s7, s2, 15
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_or_b32 s3, s3, s6
; GFX8-NEXT: s_lshr_b32 s6, s2, 16
; GFX8-NEXT: s_andn2_b32 s2, 15, s2
; GFX8-NEXT: s_bfe_u32 s7, s7, 0x100000
; GFX8-NEXT: s_lshr_b32 s1, s1, s5
; GFX8-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX8-NEXT: s_lshr_b32 s1, s1, s2
; GFX8-NEXT: s_lshl_b32 s0, s0, s7
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: s_and_b32 s1, s6, 15
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_lshl_b32 s4, s4, s5
; GFX8-NEXT: s_andn2_b32 s2, 15, s6
; GFX8-NEXT: s_lshl_b32 s1, s3, s1
; GFX8-NEXT: s_bfe_u32 s3, s4, 0x100000
; GFX8-NEXT: s_lshr_b32 s3, s3, s5
; GFX8-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX8-NEXT: s_lshr_b32 s2, s3, s2
; GFX8-NEXT: s_or_b32 s1, s1, s2
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX8-NEXT: s_lshl_b32 s1, s1, 16
; GFX8-NEXT: s_or_b32 s0, s0, s1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_v2i16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_mov_b32 s3, 0xf000f
; GFX9-NEXT: s_and_b32 s4, s2, s3
; GFX9-NEXT: s_andn2_b32 s2, s3, s2
; GFX9-NEXT: s_lshr_b32 s3, s0, 16
; GFX9-NEXT: s_lshl_b32 s0, s0, 0x10001
; GFX9-NEXT: s_lshl_b32 s3, s3, 1
; GFX9-NEXT: s_pack_ll_b32_b16 s0, s0, s3
; GFX9-NEXT: s_lshr_b32 s3, s0, 16
; GFX9-NEXT: s_lshr_b32 s5, s2, 16
; GFX9-NEXT: s_lshl_b32 s0, s0, s2
; GFX9-NEXT: s_lshl_b32 s2, s3, s5
; GFX9-NEXT: s_mov_b32 s3, 0xffff
; GFX9-NEXT: s_pack_ll_b32_b16 s0, s0, s2
; GFX9-NEXT: s_lshr_b32 s2, s1, 16
; GFX9-NEXT: s_and_b32 s1, s1, s3
; GFX9-NEXT: s_lshr_b32 s5, s4, 16
; GFX9-NEXT: s_and_b32 s3, s4, s3
; GFX9-NEXT: s_lshr_b32 s1, s1, s3
; GFX9-NEXT: s_lshr_b32 s2, s2, s5
; GFX9-NEXT: s_pack_ll_b32_b16 s1, s1, s2
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_v2i16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_lshr_b32 s4, s0, 16
; GFX10-NEXT: s_mov_b32 s3, 0xf000f
; GFX10-NEXT: s_lshl_b32 s0, s0, 0x10001
; GFX10-NEXT: s_lshl_b32 s4, s4, 1
; GFX10-NEXT: s_and_b32 s5, s2, s3
; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s4
; GFX10-NEXT: s_andn2_b32 s2, s3, s2
; GFX10-NEXT: s_lshr_b32 s3, s0, 16
; GFX10-NEXT: s_lshr_b32 s4, s2, 16
; GFX10-NEXT: s_lshl_b32 s0, s0, s2
; GFX10-NEXT: s_lshl_b32 s2, s3, s4
; GFX10-NEXT: s_mov_b32 s3, 0xffff
; GFX10-NEXT: s_lshr_b32 s4, s1, 16
; GFX10-NEXT: s_and_b32 s1, s1, s3
; GFX10-NEXT: s_and_b32 s3, s5, s3
; GFX10-NEXT: s_lshr_b32 s5, s5, 16
; GFX10-NEXT: s_lshr_b32 s1, s1, s3
; GFX10-NEXT: s_lshr_b32 s3, s4, s5
; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s2
; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s3
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
%result = call <2 x i16> @llvm.fshr.v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt)
%cast = bitcast <2 x i16> %result to i32
ret i32 %cast
}
define <2 x i16> @v_fshr_v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt) {
; GFX6-LABEL: v_fshr_v2i16:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v6, 0xffff
; GFX6-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; GFX6-NEXT: v_and_b32_e32 v4, v4, v6
; GFX6-NEXT: s_mov_b32 s5, 0xffff
; GFX6-NEXT: v_or_b32_e32 v4, v5, v4
; GFX6-NEXT: v_and_b32_e32 v5, s5, v2
; GFX6-NEXT: s_bfe_u32 s4, 1, 0x100000
; GFX6-NEXT: v_lshrrev_b32_e32 v5, 1, v5
; GFX6-NEXT: s_bfe_u32 s6, 14, 0x100000
; GFX6-NEXT: v_lshlrev_b32_e32 v0, s4, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v5, s6, v5
; GFX6-NEXT: v_or_b32_e32 v0, v0, v5
; GFX6-NEXT: v_and_b32_e32 v5, s5, v3
; GFX6-NEXT: v_lshrrev_b32_e32 v5, 1, v5
; GFX6-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX6-NEXT: v_lshrrev_b32_e32 v5, s6, v5
; GFX6-NEXT: v_lshlrev_b32_e32 v2, 1, v2
; GFX6-NEXT: v_and_b32_e32 v7, 15, v4
; GFX6-NEXT: v_or_b32_e32 v1, v1, v5
; GFX6-NEXT: v_lshrrev_b32_e32 v5, 16, v4
; GFX6-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX6-NEXT: v_and_b32_e32 v4, 15, v4
; GFX6-NEXT: v_and_b32_e32 v2, v2, v6
; GFX6-NEXT: v_bfe_u32 v7, v7, 0, 16
; GFX6-NEXT: v_lshrrev_b32_e32 v2, 1, v2
; GFX6-NEXT: v_bfe_u32 v4, v4, 0, 16
; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v2
; GFX6-NEXT: v_lshlrev_b32_e32 v0, v7, v0
; GFX6-NEXT: v_or_b32_e32 v0, v0, v2
; GFX6-NEXT: v_and_b32_e32 v2, 15, v5
; GFX6-NEXT: v_xor_b32_e32 v4, -1, v5
; GFX6-NEXT: v_bfe_u32 v2, v2, 0, 16
; GFX6-NEXT: v_lshlrev_b32_e32 v3, 1, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v1, v2, v1
; GFX6-NEXT: v_and_b32_e32 v2, v3, v6
; GFX6-NEXT: v_and_b32_e32 v4, 15, v4
; GFX6-NEXT: v_lshrrev_b32_e32 v2, 1, v2
; GFX6-NEXT: v_bfe_u32 v3, v4, 0, 16
; GFX6-NEXT: v_lshrrev_b32_e32 v2, v3, v2
; GFX6-NEXT: v_or_b32_e32 v1, v1, v2
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v2i16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_lshrrev_b16_e32 v4, 1, v1
; GFX8-NEXT: v_lshlrev_b16_e32 v3, 1, v0
; GFX8-NEXT: v_lshrrev_b16_e32 v4, 14, v4
; GFX8-NEXT: v_or_b32_e32 v3, v3, v4
; GFX8-NEXT: v_mov_b32_e32 v4, 1
; GFX8-NEXT: v_lshrrev_b16_sdwa v5, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX8-NEXT: v_lshlrev_b16_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_lshrrev_b16_e32 v5, 14, v5
; GFX8-NEXT: v_or_b32_e32 v0, v0, v5
; GFX8-NEXT: v_lshlrev_b16_e32 v5, 1, v1
; GFX8-NEXT: v_lshlrev_b16_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_and_b32_e32 v6, 15, v2
; GFX8-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX8-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX8-NEXT: v_and_b32_e32 v2, 15, v2
; GFX8-NEXT: v_lshrrev_b16_e32 v5, 1, v5
; GFX8-NEXT: v_lshlrev_b16_e32 v3, v6, v3
; GFX8-NEXT: v_lshrrev_b16_e32 v2, v2, v5
; GFX8-NEXT: v_or_b32_e32 v2, v3, v2
; GFX8-NEXT: v_and_b32_e32 v3, 15, v4
; GFX8-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX8-NEXT: v_and_b32_e32 v4, 15, v4
; GFX8-NEXT: v_lshrrev_b16_e32 v1, 1, v1
; GFX8-NEXT: v_lshlrev_b16_e32 v0, v3, v0
; GFX8-NEXT: v_lshrrev_b16_e32 v1, v4, v1
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: v_mov_b32_e32 v1, 16
; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX8-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v2i16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_mov_b32 s4, 0xf000f
; GFX9-NEXT: v_and_b32_e32 v3, s4, v2
; GFX9-NEXT: v_xor_b32_e32 v2, -1, v2
; GFX9-NEXT: v_and_b32_e32 v2, s4, v2
; GFX9-NEXT: v_pk_lshlrev_b16 v0, 1, v0 op_sel_hi:[0,1]
; GFX9-NEXT: v_pk_lshlrev_b16 v0, v2, v0
; GFX9-NEXT: v_pk_lshrrev_b16 v1, v3, v1
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v2i16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v3, -1, v2
; GFX10-NEXT: s_mov_b32 s4, 0xf000f
; GFX10-NEXT: v_pk_lshlrev_b16 v0, 1, v0 op_sel_hi:[0,1]
; GFX10-NEXT: v_and_b32_e32 v2, s4, v2
; GFX10-NEXT: v_and_b32_e32 v3, s4, v3
; GFX10-NEXT: v_pk_lshrrev_b16 v1, v2, v1
; GFX10-NEXT: v_pk_lshlrev_b16 v0, v3, v0
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i16> @llvm.fshr.v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt)
ret <2 x i16> %result
}
define <2 x i16> @v_fshr_v2i16_4_8(<2 x i16> %lhs, <2 x i16> %rhs) {
; GFX6-LABEL: v_fshr_v2i16_4_8:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_sub_i32 s4, 0, 4
; GFX6-NEXT: s_and_b32 s6, s4, 15
; GFX6-NEXT: s_bfe_u32 s6, s6, 0x100000
; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
; GFX6-NEXT: s_mov_b32 s6, 0xffff
; GFX6-NEXT: s_xor_b32 s4, s4, -1
; GFX6-NEXT: v_and_b32_e32 v2, s6, v2
; GFX6-NEXT: s_sub_i32 s5, 0, 8
; GFX6-NEXT: v_lshrrev_b32_e32 v2, 1, v2
; GFX6-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX6-NEXT: v_lshrrev_b32_e32 v2, s4, v2
; GFX6-NEXT: s_and_b32 s4, s5, 15
; GFX6-NEXT: v_or_b32_e32 v0, v0, v2
; GFX6-NEXT: v_and_b32_e32 v2, s6, v3
; GFX6-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX6-NEXT: s_xor_b32 s5, s5, -1
; GFX6-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX6-NEXT: v_lshrrev_b32_e32 v2, 1, v2
; GFX6-NEXT: s_bfe_u32 s4, s5, 0x100000
; GFX6-NEXT: v_lshrrev_b32_e32 v2, s4, v2
; GFX6-NEXT: v_or_b32_e32 v1, v1, v2
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v2i16_4_8:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_sub_i32 s4, 0, 4
; GFX8-NEXT: s_and_b32 s6, s4, 15
; GFX8-NEXT: s_sub_i32 s5, 0, 8
; GFX8-NEXT: s_xor_b32 s4, s4, -1
; GFX8-NEXT: v_lshrrev_b16_e32 v3, 1, v1
; GFX8-NEXT: v_lshrrev_b16_e32 v3, s4, v3
; GFX8-NEXT: v_lshlrev_b16_e32 v2, s6, v0
; GFX8-NEXT: s_and_b32 s4, s5, 15
; GFX8-NEXT: v_or_b32_e32 v2, v2, v3
; GFX8-NEXT: v_mov_b32_e32 v3, s4
; GFX8-NEXT: v_lshlrev_b16_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_mov_b32_e32 v3, 1
; GFX8-NEXT: s_xor_b32 s5, s5, -1
; GFX8-NEXT: v_lshrrev_b16_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_lshrrev_b16_e32 v1, s5, v1
; GFX8-NEXT: v_or_b32_e32 v0, v0, v1
; GFX8-NEXT: v_mov_b32_e32 v1, 16
; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX8-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v2i16_4_8:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_cvt_f32_ubyte0_e32 v2, 16
; GFX9-NEXT: v_rcp_iflag_f32_e32 v2, v2
; GFX9-NEXT: v_cvt_f32_ubyte0_e32 v3, 16
; GFX9-NEXT: v_rcp_iflag_f32_e32 v3, v3
; GFX9-NEXT: s_sub_i32 s4, 0, 16
; GFX9-NEXT: v_mul_f32_e32 v2, 0x4f7ffffe, v2
; GFX9-NEXT: v_cvt_u32_f32_e32 v2, v2
; GFX9-NEXT: v_mul_f32_e32 v3, 0x4f7ffffe, v3
; GFX9-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX9-NEXT: v_mul_lo_u32 v4, s4, v2
; GFX9-NEXT: v_mul_lo_u32 v5, s4, v3
; GFX9-NEXT: v_mul_hi_u32 v4, v2, v4
; GFX9-NEXT: v_mul_hi_u32 v5, v3, v5
; GFX9-NEXT: v_add_u32_e32 v2, v2, v4
; GFX9-NEXT: v_mul_hi_u32 v2, 4, v2
; GFX9-NEXT: v_add_u32_e32 v3, v3, v5
; GFX9-NEXT: v_mul_hi_u32 v3, 8, v3
; GFX9-NEXT: v_lshlrev_b32_e32 v2, 4, v2
; GFX9-NEXT: v_sub_u32_e32 v2, 4, v2
; GFX9-NEXT: v_subrev_u32_e32 v4, 16, v2
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 16, v2
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX9-NEXT: v_lshlrev_b32_e32 v3, 4, v3
; GFX9-NEXT: v_subrev_u32_e32 v4, 16, v2
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 16, v2
; GFX9-NEXT: v_sub_u32_e32 v3, 8, v3
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX9-NEXT: v_subrev_u32_e32 v4, 16, v3
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 16, v3
; GFX9-NEXT: v_cndmask_b32_e32 v3, v3, v4, vcc
; GFX9-NEXT: v_subrev_u32_e32 v4, 16, v3
; GFX9-NEXT: v_cmp_le_u32_e32 vcc, 16, v3
; GFX9-NEXT: v_cndmask_b32_e32 v3, v3, v4, vcc
; GFX9-NEXT: v_mov_b32_e32 v4, 0xffff
; GFX9-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; GFX9-NEXT: v_and_or_b32 v2, v2, v4, v3
; GFX9-NEXT: v_pk_sub_i16 v3, 16, v2 op_sel_hi:[0,1]
; GFX9-NEXT: v_pk_lshlrev_b16 v0, v3, v0
; GFX9-NEXT: v_pk_lshrrev_b16 v1, v2, v1
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v2i16_4_8:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v2, 16
; GFX10-NEXT: v_cvt_f32_ubyte0_e32 v3, 16
; GFX10-NEXT: s_sub_i32 s4, 0, 16
; GFX10-NEXT: v_rcp_iflag_f32_e32 v2, v2
; GFX10-NEXT: v_rcp_iflag_f32_e32 v3, v3
; GFX10-NEXT: v_mul_f32_e32 v2, 0x4f7ffffe, v2
; GFX10-NEXT: v_mul_f32_e32 v3, 0x4f7ffffe, v3
; GFX10-NEXT: v_cvt_u32_f32_e32 v2, v2
; GFX10-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX10-NEXT: v_mul_lo_u32 v4, s4, v2
; GFX10-NEXT: v_mul_lo_u32 v5, s4, v3
; GFX10-NEXT: v_mul_hi_u32 v4, v2, v4
; GFX10-NEXT: v_mul_hi_u32 v5, v3, v5
; GFX10-NEXT: v_add_nc_u32_e32 v2, v2, v4
; GFX10-NEXT: v_add_nc_u32_e32 v3, v3, v5
; GFX10-NEXT: v_mul_hi_u32 v2, 8, v2
; GFX10-NEXT: v_mul_hi_u32 v3, 4, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 4, v2
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 4, v3
; GFX10-NEXT: v_sub_nc_u32_e32 v2, 8, v2
; GFX10-NEXT: v_sub_nc_u32_e32 v3, 4, v3
; GFX10-NEXT: v_subrev_nc_u32_e32 v4, 16, v2
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 16, v2
; GFX10-NEXT: v_subrev_nc_u32_e32 v5, 16, v3
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 16, v3
; GFX10-NEXT: v_subrev_nc_u32_e32 v4, 16, v2
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 16, v2
; GFX10-NEXT: v_subrev_nc_u32_e32 v5, 16, v3
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc_lo
; GFX10-NEXT: v_cmp_le_u32_e32 vcc_lo, 16, v3
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc_lo
; GFX10-NEXT: v_and_or_b32 v2, 0xffff, v3, v2
; GFX10-NEXT: v_pk_sub_i16 v3, 16, v2 op_sel_hi:[0,1]
; GFX10-NEXT: v_pk_lshrrev_b16 v1, v2, v1
; GFX10-NEXT: v_pk_lshlrev_b16 v0, v3, v0
; GFX10-NEXT: v_or_b32_e32 v0, v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i16> @llvm.fshr.v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> <i16 4, i16 8>)
ret <2 x i16> %result
}
define amdgpu_ps float @v_fshr_v2i16_ssv(<2 x i16> inreg %lhs, <2 x i16> inreg %rhs, <2 x i16> %amt) {
; GFX6-LABEL: v_fshr_v2i16_ssv:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s5, 0xffff
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX6-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX6-NEXT: s_and_b32 s6, s2, s5
; GFX6-NEXT: v_or_b32_e32 v0, v1, v0
; GFX6-NEXT: s_bfe_u32 s4, 1, 0x100000
; GFX6-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX6-NEXT: s_lshr_b32 s6, s6, 1
; GFX6-NEXT: s_bfe_u32 s7, 14, 0x100000
; GFX6-NEXT: v_and_b32_e32 v2, 15, v0
; GFX6-NEXT: s_lshl_b32 s0, s0, s4
; GFX6-NEXT: s_lshr_b32 s6, s6, s7
; GFX6-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX6-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX6-NEXT: s_or_b32 s0, s0, s6
; GFX6-NEXT: v_bfe_u32 v2, v2, 0, 16
; GFX6-NEXT: s_lshl_b32 s2, s2, 1
; GFX6-NEXT: v_and_b32_e32 v0, 15, v0
; GFX6-NEXT: v_lshl_b32_e32 v2, s0, v2
; GFX6-NEXT: s_and_b32 s0, s2, s5
; GFX6-NEXT: s_lshr_b32 s0, s0, 1
; GFX6-NEXT: v_bfe_u32 v0, v0, 0, 16
; GFX6-NEXT: v_lshr_b32_e32 v0, s0, v0
; GFX6-NEXT: s_lshl_b32 s1, s1, s4
; GFX6-NEXT: s_and_b32 s4, s3, s5
; GFX6-NEXT: v_or_b32_e32 v0, v2, v0
; GFX6-NEXT: v_and_b32_e32 v2, 15, v1
; GFX6-NEXT: s_lshr_b32 s4, s4, 1
; GFX6-NEXT: v_xor_b32_e32 v1, -1, v1
; GFX6-NEXT: s_lshl_b32 s3, s3, 1
; GFX6-NEXT: v_and_b32_e32 v1, 15, v1
; GFX6-NEXT: s_and_b32 s0, s3, s5
; GFX6-NEXT: s_lshr_b32 s4, s4, s7
; GFX6-NEXT: s_or_b32 s1, s1, s4
; GFX6-NEXT: v_bfe_u32 v2, v2, 0, 16
; GFX6-NEXT: s_lshr_b32 s0, s0, 1
; GFX6-NEXT: v_bfe_u32 v1, v1, 0, 16
; GFX6-NEXT: v_lshl_b32_e32 v2, s1, v2
; GFX6-NEXT: v_lshr_b32_e32 v1, s0, v1
; GFX6-NEXT: v_or_b32_e32 v1, v2, v1
; GFX6-NEXT: v_bfe_u32 v1, v1, 0, 16
; GFX6-NEXT: v_bfe_u32 v0, v0, 0, 16
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_v2i16_ssv:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_bfe_u32 s4, 1, 0x100000
; GFX8-NEXT: s_bfe_u32 s5, s1, 0x100000
; GFX8-NEXT: s_lshr_b32 s5, s5, s4
; GFX8-NEXT: s_bfe_u32 s6, 14, 0x100000
; GFX8-NEXT: s_lshr_b32 s2, s0, 16
; GFX8-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX8-NEXT: s_lshl_b32 s0, s0, s4
; GFX8-NEXT: s_lshr_b32 s5, s5, s6
; GFX8-NEXT: s_lshr_b32 s3, s1, 16
; GFX8-NEXT: s_or_b32 s0, s0, s5
; GFX8-NEXT: v_and_b32_e32 v2, 15, v0
; GFX8-NEXT: s_lshl_b32 s1, s1, s4
; GFX8-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX8-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX8-NEXT: v_lshlrev_b16_e64 v2, v2, s0
; GFX8-NEXT: s_bfe_u32 s0, s1, 0x100000
; GFX8-NEXT: v_and_b32_e32 v0, 15, v0
; GFX8-NEXT: s_lshr_b32 s0, s0, s4
; GFX8-NEXT: s_lshr_b32 s5, s3, s4
; GFX8-NEXT: v_lshrrev_b16_e64 v0, v0, s0
; GFX8-NEXT: s_lshl_b32 s3, s3, s4
; GFX8-NEXT: s_bfe_u32 s0, s3, 0x100000
; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
; GFX8-NEXT: v_and_b32_e32 v2, 15, v1
; GFX8-NEXT: v_xor_b32_e32 v1, -1, v1
; GFX8-NEXT: s_lshl_b32 s2, s2, s4
; GFX8-NEXT: s_lshr_b32 s5, s5, s6
; GFX8-NEXT: s_or_b32 s2, s2, s5
; GFX8-NEXT: v_and_b32_e32 v1, 15, v1
; GFX8-NEXT: s_lshr_b32 s0, s0, s4
; GFX8-NEXT: v_lshlrev_b16_e64 v2, v2, s2
; GFX8-NEXT: v_lshrrev_b16_e64 v1, v1, s0
; GFX8-NEXT: v_or_b32_e32 v1, v2, v1
; GFX8-NEXT: v_mov_b32_e32 v2, 16
; GFX8-NEXT: v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX8-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_v2i16_ssv:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_mov_b32 s2, 0xf000f
; GFX9-NEXT: v_and_b32_e32 v1, s2, v0
; GFX9-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX9-NEXT: v_and_b32_e32 v0, s2, v0
; GFX9-NEXT: s_lshr_b32 s2, s0, 16
; GFX9-NEXT: s_lshl_b32 s0, s0, 0x10001
; GFX9-NEXT: s_lshl_b32 s2, s2, 1
; GFX9-NEXT: s_pack_ll_b32_b16 s0, s0, s2
; GFX9-NEXT: v_pk_lshlrev_b16 v0, v0, s0
; GFX9-NEXT: v_pk_lshrrev_b16 v1, v1, s1
; GFX9-NEXT: v_or_b32_e32 v0, v0, v1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_v2i16_ssv:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_xor_b32_e32 v1, -1, v0
; GFX10-NEXT: s_mov_b32 s2, 0xf000f
; GFX10-NEXT: s_lshr_b32 s3, s0, 16
; GFX10-NEXT: v_and_b32_e32 v0, s2, v0
; GFX10-NEXT: s_lshl_b32 s0, s0, 0x10001
; GFX10-NEXT: v_and_b32_e32 v1, s2, v1
; GFX10-NEXT: s_lshl_b32 s2, s3, 1
; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s2
; GFX10-NEXT: v_pk_lshrrev_b16 v0, v0, s1
; GFX10-NEXT: v_pk_lshlrev_b16 v1, v1, s0
; GFX10-NEXT: v_or_b32_e32 v0, v1, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call <2 x i16> @llvm.fshr.v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt)
%cast = bitcast <2 x i16> %result to float
ret float %cast
}
define amdgpu_ps float @v_fshr_v2i16_svs(<2 x i16> inreg %lhs, <2 x i16> %rhs, <2 x i16> inreg %amt) {
; GFX6-LABEL: v_fshr_v2i16_svs:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s4, 0xffff
; GFX6-NEXT: v_and_b32_e32 v2, s4, v0
; GFX6-NEXT: s_lshl_b32 s3, s3, 16
; GFX6-NEXT: s_and_b32 s2, s2, s4
; GFX6-NEXT: v_and_b32_e32 v3, s4, v1
; GFX6-NEXT: s_or_b32 s2, s3, s2
; GFX6-NEXT: s_bfe_u32 s3, 1, 0x100000
; GFX6-NEXT: v_lshrrev_b32_e32 v2, 1, v2
; GFX6-NEXT: s_bfe_u32 s5, 14, 0x100000
; GFX6-NEXT: v_lshrrev_b32_e32 v3, 1, v3
; GFX6-NEXT: s_lshl_b32 s0, s0, s3
; GFX6-NEXT: v_lshrrev_b32_e32 v2, s5, v2
; GFX6-NEXT: v_or_b32_e32 v2, s0, v2
; GFX6-NEXT: s_lshl_b32 s0, s1, s3
; GFX6-NEXT: v_lshrrev_b32_e32 v3, s5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 1, v0
; GFX6-NEXT: v_or_b32_e32 v3, s0, v3
; GFX6-NEXT: s_xor_b32 s0, s2, -1
; GFX6-NEXT: s_and_b32 s2, s0, 15
; GFX6-NEXT: v_and_b32_e32 v0, s4, v0
; GFX6-NEXT: s_lshr_b32 s1, s0, 16
; GFX6-NEXT: s_andn2_b32 s0, 15, s0
; GFX6-NEXT: v_lshrrev_b32_e32 v0, 1, v0
; GFX6-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 1, v1
; GFX6-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX6-NEXT: v_lshrrev_b32_e32 v0, s0, v0
; GFX6-NEXT: s_and_b32 s0, s1, 15
; GFX6-NEXT: v_and_b32_e32 v1, s4, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v2, s2, v2
; GFX6-NEXT: s_andn2_b32 s1, 15, s1
; GFX6-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX6-NEXT: v_or_b32_e32 v0, v2, v0
; GFX6-NEXT: v_lshlrev_b32_e32 v2, s0, v3
; GFX6-NEXT: v_lshrrev_b32_e32 v1, 1, v1
; GFX6-NEXT: s_bfe_u32 s0, s1, 0x100000
; GFX6-NEXT: v_lshrrev_b32_e32 v1, s0, v1
; GFX6-NEXT: v_or_b32_e32 v1, v2, v1
; GFX6-NEXT: v_bfe_u32 v1, v1, 0, 16
; GFX6-NEXT: v_bfe_u32 v0, v0, 0, 16
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_v2i16_svs:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_lshrrev_b16_e32 v1, 1, v0
; GFX8-NEXT: v_mov_b32_e32 v2, 1
; GFX8-NEXT: s_bfe_u32 s3, 1, 0x100000
; GFX8-NEXT: v_lshrrev_b16_sdwa v3, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: s_lshr_b32 s2, s0, 16
; GFX8-NEXT: s_lshl_b32 s0, s0, s3
; GFX8-NEXT: v_lshrrev_b16_e32 v1, 14, v1
; GFX8-NEXT: v_or_b32_e32 v1, s0, v1
; GFX8-NEXT: s_lshl_b32 s0, s2, s3
; GFX8-NEXT: v_lshrrev_b16_e32 v3, 14, v3
; GFX8-NEXT: v_or_b32_e32 v3, s0, v3
; GFX8-NEXT: s_xor_b32 s0, s1, -1
; GFX8-NEXT: v_lshlrev_b16_e32 v4, 1, v0
; GFX8-NEXT: v_lshlrev_b16_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: s_and_b32 s2, s0, 15
; GFX8-NEXT: s_lshr_b32 s1, s0, 16
; GFX8-NEXT: s_andn2_b32 s0, 15, s0
; GFX8-NEXT: v_lshrrev_b16_e32 v2, 1, v4
; GFX8-NEXT: v_lshrrev_b16_e32 v2, s0, v2
; GFX8-NEXT: v_lshlrev_b16_e32 v1, s2, v1
; GFX8-NEXT: s_and_b32 s0, s1, 15
; GFX8-NEXT: s_andn2_b32 s1, 15, s1
; GFX8-NEXT: v_lshrrev_b16_e32 v0, 1, v0
; GFX8-NEXT: v_or_b32_e32 v1, v1, v2
; GFX8-NEXT: v_lshlrev_b16_e32 v2, s0, v3
; GFX8-NEXT: v_lshrrev_b16_e32 v0, s1, v0
; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
; GFX8-NEXT: v_mov_b32_e32 v2, 16
; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX8-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_v2i16_svs:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_mov_b32 s2, 0xf000f
; GFX9-NEXT: s_and_b32 s3, s1, s2
; GFX9-NEXT: s_andn2_b32 s1, s2, s1
; GFX9-NEXT: s_lshr_b32 s2, s0, 16
; GFX9-NEXT: s_lshl_b32 s0, s0, 0x10001
; GFX9-NEXT: s_lshl_b32 s2, s2, 1
; GFX9-NEXT: s_pack_ll_b32_b16 s0, s0, s2
; GFX9-NEXT: s_lshr_b32 s2, s0, 16
; GFX9-NEXT: s_lshr_b32 s4, s1, 16
; GFX9-NEXT: s_lshl_b32 s0, s0, s1
; GFX9-NEXT: s_lshl_b32 s1, s2, s4
; GFX9-NEXT: s_pack_ll_b32_b16 s0, s0, s1
; GFX9-NEXT: v_pk_lshrrev_b16 v0, s3, v0
; GFX9-NEXT: v_or_b32_e32 v0, s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_v2i16_svs:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_lshr_b32 s3, s0, 16
; GFX10-NEXT: s_mov_b32 s2, 0xf000f
; GFX10-NEXT: s_lshl_b32 s0, s0, 0x10001
; GFX10-NEXT: s_lshl_b32 s3, s3, 1
; GFX10-NEXT: s_and_b32 s4, s1, s2
; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s3
; GFX10-NEXT: s_andn2_b32 s1, s2, s1
; GFX10-NEXT: s_lshr_b32 s2, s0, 16
; GFX10-NEXT: s_lshr_b32 s3, s1, 16
; GFX10-NEXT: v_pk_lshrrev_b16 v0, s4, v0
; GFX10-NEXT: s_lshl_b32 s0, s0, s1
; GFX10-NEXT: s_lshl_b32 s1, s2, s3
; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s1
; GFX10-NEXT: v_or_b32_e32 v0, s0, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call <2 x i16> @llvm.fshr.v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt)
%cast = bitcast <2 x i16> %result to float
ret float %cast
}
define amdgpu_ps float @v_fshr_v2i16_vss(<2 x i16> %lhs, <2 x i16> inreg %rhs, <2 x i16> inreg %amt) {
; GFX6-LABEL: v_fshr_v2i16_vss:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s4, 0xffff
; GFX6-NEXT: s_lshl_b32 s3, s3, 16
; GFX6-NEXT: s_and_b32 s2, s2, s4
; GFX6-NEXT: s_or_b32 s2, s3, s2
; GFX6-NEXT: s_bfe_u32 s3, 1, 0x100000
; GFX6-NEXT: s_and_b32 s5, s0, s4
; GFX6-NEXT: v_lshlrev_b32_e32 v0, s3, v0
; GFX6-NEXT: v_lshlrev_b32_e32 v1, s3, v1
; GFX6-NEXT: s_and_b32 s3, s1, s4
; GFX6-NEXT: s_lshr_b32 s5, s5, 1
; GFX6-NEXT: s_bfe_u32 s6, 14, 0x100000
; GFX6-NEXT: s_lshr_b32 s3, s3, 1
; GFX6-NEXT: s_lshl_b32 s0, s0, 1
; GFX6-NEXT: s_lshr_b32 s5, s5, s6
; GFX6-NEXT: s_lshr_b32 s3, s3, s6
; GFX6-NEXT: s_xor_b32 s2, s2, -1
; GFX6-NEXT: s_and_b32 s0, s0, s4
; GFX6-NEXT: v_or_b32_e32 v0, s5, v0
; GFX6-NEXT: s_and_b32 s5, s2, 15
; GFX6-NEXT: v_or_b32_e32 v1, s3, v1
; GFX6-NEXT: s_lshr_b32 s3, s2, 16
; GFX6-NEXT: s_andn2_b32 s2, 15, s2
; GFX6-NEXT: s_bfe_u32 s5, s5, 0x100000
; GFX6-NEXT: s_lshr_b32 s0, s0, 1
; GFX6-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX6-NEXT: s_lshr_b32 s0, s0, s2
; GFX6-NEXT: v_lshlrev_b32_e32 v0, s5, v0
; GFX6-NEXT: v_or_b32_e32 v0, s0, v0
; GFX6-NEXT: s_and_b32 s0, s3, 15
; GFX6-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX6-NEXT: s_lshl_b32 s1, s1, 1
; GFX6-NEXT: v_lshlrev_b32_e32 v1, s0, v1
; GFX6-NEXT: s_and_b32 s0, s1, s4
; GFX6-NEXT: s_andn2_b32 s2, 15, s3
; GFX6-NEXT: s_lshr_b32 s0, s0, 1
; GFX6-NEXT: s_bfe_u32 s1, s2, 0x100000
; GFX6-NEXT: s_lshr_b32 s0, s0, s1
; GFX6-NEXT: v_or_b32_e32 v1, s0, v1
; GFX6-NEXT: v_bfe_u32 v1, v1, 0, 16
; GFX6-NEXT: v_bfe_u32 v0, v0, 0, 16
; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_v2i16_vss:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_bfe_u32 s3, s0, 0x100000
; GFX8-NEXT: s_bfe_u32 s4, 1, 0x100000
; GFX8-NEXT: s_lshr_b32 s3, s3, s4
; GFX8-NEXT: s_bfe_u32 s5, 14, 0x100000
; GFX8-NEXT: s_lshr_b32 s2, s0, 16
; GFX8-NEXT: v_lshlrev_b16_e32 v1, 1, v0
; GFX8-NEXT: s_lshr_b32 s3, s3, s5
; GFX8-NEXT: v_or_b32_e32 v1, s3, v1
; GFX8-NEXT: s_lshr_b32 s3, s2, s4
; GFX8-NEXT: v_mov_b32_e32 v2, 1
; GFX8-NEXT: s_lshl_b32 s0, s0, s4
; GFX8-NEXT: v_lshlrev_b16_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: s_lshr_b32 s3, s3, s5
; GFX8-NEXT: s_xor_b32 s1, s1, -1
; GFX8-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX8-NEXT: v_or_b32_e32 v0, s3, v0
; GFX8-NEXT: s_and_b32 s5, s1, 15
; GFX8-NEXT: s_lshr_b32 s3, s1, 16
; GFX8-NEXT: s_andn2_b32 s1, 15, s1
; GFX8-NEXT: s_lshr_b32 s0, s0, s4
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_lshr_b32 s0, s0, s1
; GFX8-NEXT: v_lshlrev_b16_e32 v1, s5, v1
; GFX8-NEXT: v_or_b32_e32 v1, s0, v1
; GFX8-NEXT: s_and_b32 s0, s3, 15
; GFX8-NEXT: s_lshl_b32 s2, s2, s4
; GFX8-NEXT: s_andn2_b32 s1, 15, s3
; GFX8-NEXT: v_lshlrev_b16_e32 v0, s0, v0
; GFX8-NEXT: s_bfe_u32 s0, s2, 0x100000
; GFX8-NEXT: s_lshr_b32 s0, s0, s4
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_lshr_b32 s0, s0, s1
; GFX8-NEXT: v_or_b32_e32 v0, s0, v0
; GFX8-NEXT: v_mov_b32_e32 v2, 16
; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX8-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_v2i16_vss:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_mov_b32 s2, 0xf000f
; GFX9-NEXT: s_and_b32 s3, s1, s2
; GFX9-NEXT: s_andn2_b32 s1, s2, s1
; GFX9-NEXT: v_pk_lshlrev_b16 v0, 1, v0 op_sel_hi:[0,1]
; GFX9-NEXT: s_mov_b32 s2, 0xffff
; GFX9-NEXT: v_pk_lshlrev_b16 v0, s1, v0
; GFX9-NEXT: s_lshr_b32 s1, s0, 16
; GFX9-NEXT: s_and_b32 s0, s0, s2
; GFX9-NEXT: s_lshr_b32 s4, s3, 16
; GFX9-NEXT: s_and_b32 s2, s3, s2
; GFX9-NEXT: s_lshr_b32 s0, s0, s2
; GFX9-NEXT: s_lshr_b32 s1, s1, s4
; GFX9-NEXT: s_pack_ll_b32_b16 s0, s0, s1
; GFX9-NEXT: v_or_b32_e32 v0, s0, v0
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_v2i16_vss:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_pk_lshlrev_b16 v0, 1, v0 op_sel_hi:[0,1]
; GFX10-NEXT: s_mov_b32 s2, 0xf000f
; GFX10-NEXT: s_mov_b32 s3, 0xffff
; GFX10-NEXT: s_and_b32 s4, s1, s2
; GFX10-NEXT: s_andn2_b32 s1, s2, s1
; GFX10-NEXT: s_lshr_b32 s2, s0, 16
; GFX10-NEXT: v_pk_lshlrev_b16 v0, s1, v0
; GFX10-NEXT: s_and_b32 s0, s0, s3
; GFX10-NEXT: s_and_b32 s1, s4, s3
; GFX10-NEXT: s_lshr_b32 s3, s4, 16
; GFX10-NEXT: s_lshr_b32 s0, s0, s1
; GFX10-NEXT: s_lshr_b32 s1, s2, s3
; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s1
; GFX10-NEXT: v_or_b32_e32 v0, s0, v0
; GFX10-NEXT: ; return to shader part epilog
%result = call <2 x i16> @llvm.fshr.v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt)
%cast = bitcast <2 x i16> %result to float
ret float %cast
}
; ; FIXME
; define amdgpu_ps i48 @s_fshr_v3i16(<3 x i16> inreg %lhs, <3 x i16> inreg %rhs, <3 x i16> inreg %amt) {
; %result = call <3 x i16> @llvm.fshr.v3i16(<3 x i16> %lhs, <3 x i16> %rhs, <3 x i16> %amt)
; %cast = bitcast <3 x i16> %result to i48
; ret i48 %cast
; }
; ; FIXME
; define <3 x half> @v_fshr_v3i16(<3 x i16> %lhs, <3 x i16> %rhs, <3 x i16> %amt) {
; %result = call <3 x i16> @llvm.fshr.v3i16(<3 x i16> %lhs, <3 x i16> %rhs, <3 x i16> %amt)
; %cast.result = bitcast <3 x i16> %result to <3 x half>
; ret <3 x half> %cast.result
; }
define amdgpu_ps <2 x i32> @s_fshr_v4i16(<4 x i16> inreg %lhs, <4 x i16> inreg %rhs, <4 x i16> inreg %amt) {
; GFX6-LABEL: s_fshr_v4i16:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s12, 0xffff
; GFX6-NEXT: s_lshl_b32 s9, s9, 16
; GFX6-NEXT: s_and_b32 s8, s8, s12
; GFX6-NEXT: s_or_b32 s8, s9, s8
; GFX6-NEXT: s_lshl_b32 s9, s11, 16
; GFX6-NEXT: s_and_b32 s11, s4, s12
; GFX6-NEXT: s_and_b32 s10, s10, s12
; GFX6-NEXT: s_or_b32 s9, s9, s10
; GFX6-NEXT: s_bfe_u32 s10, 1, 0x100000
; GFX6-NEXT: s_lshr_b32 s11, s11, 1
; GFX6-NEXT: s_bfe_u32 s13, 14, 0x100000
; GFX6-NEXT: s_lshl_b32 s0, s0, s10
; GFX6-NEXT: s_lshr_b32 s11, s11, s13
; GFX6-NEXT: s_or_b32 s0, s0, s11
; GFX6-NEXT: s_and_b32 s11, s5, s12
; GFX6-NEXT: s_lshr_b32 s11, s11, 1
; GFX6-NEXT: s_lshl_b32 s4, s4, 1
; GFX6-NEXT: s_xor_b32 s8, s8, -1
; GFX6-NEXT: s_lshl_b32 s1, s1, s10
; GFX6-NEXT: s_lshr_b32 s11, s11, s13
; GFX6-NEXT: s_and_b32 s14, s8, 15
; GFX6-NEXT: s_and_b32 s4, s4, s12
; GFX6-NEXT: s_or_b32 s1, s1, s11
; GFX6-NEXT: s_lshr_b32 s11, s8, 16
; GFX6-NEXT: s_andn2_b32 s8, 15, s8
; GFX6-NEXT: s_bfe_u32 s14, s14, 0x100000
; GFX6-NEXT: s_lshr_b32 s4, s4, 1
; GFX6-NEXT: s_bfe_u32 s8, s8, 0x100000
; GFX6-NEXT: s_lshr_b32 s4, s4, s8
; GFX6-NEXT: s_lshl_b32 s0, s0, s14
; GFX6-NEXT: s_or_b32 s0, s0, s4
; GFX6-NEXT: s_and_b32 s4, s11, 15
; GFX6-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX6-NEXT: s_lshl_b32 s5, s5, 1
; GFX6-NEXT: s_lshl_b32 s1, s1, s4
; GFX6-NEXT: s_and_b32 s4, s5, s12
; GFX6-NEXT: s_andn2_b32 s8, 15, s11
; GFX6-NEXT: s_lshr_b32 s4, s4, 1
; GFX6-NEXT: s_bfe_u32 s5, s8, 0x100000
; GFX6-NEXT: s_lshr_b32 s4, s4, s5
; GFX6-NEXT: s_or_b32 s1, s1, s4
; GFX6-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX6-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX6-NEXT: s_lshl_b32 s1, s1, 16
; GFX6-NEXT: s_or_b32 s0, s0, s1
; GFX6-NEXT: s_lshl_b32 s1, s2, s10
; GFX6-NEXT: s_and_b32 s2, s6, s12
; GFX6-NEXT: s_lshr_b32 s2, s2, 1
; GFX6-NEXT: s_lshr_b32 s2, s2, s13
; GFX6-NEXT: s_or_b32 s1, s1, s2
; GFX6-NEXT: s_lshl_b32 s2, s3, s10
; GFX6-NEXT: s_and_b32 s3, s7, s12
; GFX6-NEXT: s_lshr_b32 s3, s3, 1
; GFX6-NEXT: s_lshr_b32 s3, s3, s13
; GFX6-NEXT: s_or_b32 s2, s2, s3
; GFX6-NEXT: s_lshl_b32 s3, s6, 1
; GFX6-NEXT: s_xor_b32 s5, s9, -1
; GFX6-NEXT: s_and_b32 s3, s3, s12
; GFX6-NEXT: s_lshl_b32 s4, s7, 1
; GFX6-NEXT: s_and_b32 s7, s5, 15
; GFX6-NEXT: s_lshr_b32 s6, s5, 16
; GFX6-NEXT: s_andn2_b32 s5, 15, s5
; GFX6-NEXT: s_bfe_u32 s7, s7, 0x100000
; GFX6-NEXT: s_lshr_b32 s3, s3, 1
; GFX6-NEXT: s_bfe_u32 s5, s5, 0x100000
; GFX6-NEXT: s_lshr_b32 s3, s3, s5
; GFX6-NEXT: s_lshl_b32 s1, s1, s7
; GFX6-NEXT: s_or_b32 s1, s1, s3
; GFX6-NEXT: s_and_b32 s3, s6, 15
; GFX6-NEXT: s_bfe_u32 s3, s3, 0x100000
; GFX6-NEXT: s_lshl_b32 s2, s2, s3
; GFX6-NEXT: s_and_b32 s3, s4, s12
; GFX6-NEXT: s_andn2_b32 s5, 15, s6
; GFX6-NEXT: s_lshr_b32 s3, s3, 1
; GFX6-NEXT: s_bfe_u32 s4, s5, 0x100000
; GFX6-NEXT: s_lshr_b32 s3, s3, s4
; GFX6-NEXT: s_or_b32 s2, s2, s3
; GFX6-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX6-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX6-NEXT: s_lshl_b32 s2, s2, 16
; GFX6-NEXT: s_or_b32 s1, s1, s2
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_v4i16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_bfe_u32 s8, 1, 0x100000
; GFX8-NEXT: s_bfe_u32 s9, s2, 0x100000
; GFX8-NEXT: s_lshr_b32 s9, s9, s8
; GFX8-NEXT: s_bfe_u32 s10, 14, 0x100000
; GFX8-NEXT: s_lshr_b32 s6, s0, 16
; GFX8-NEXT: s_lshr_b32 s7, s2, 16
; GFX8-NEXT: s_lshl_b32 s0, s0, s8
; GFX8-NEXT: s_lshr_b32 s9, s9, s10
; GFX8-NEXT: s_or_b32 s0, s0, s9
; GFX8-NEXT: s_lshr_b32 s9, s7, s8
; GFX8-NEXT: s_lshl_b32 s2, s2, s8
; GFX8-NEXT: s_xor_b32 s4, s4, -1
; GFX8-NEXT: s_lshl_b32 s6, s6, s8
; GFX8-NEXT: s_lshr_b32 s9, s9, s10
; GFX8-NEXT: s_and_b32 s11, s4, 15
; GFX8-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX8-NEXT: s_or_b32 s6, s6, s9
; GFX8-NEXT: s_lshr_b32 s9, s4, 16
; GFX8-NEXT: s_andn2_b32 s4, 15, s4
; GFX8-NEXT: s_bfe_u32 s11, s11, 0x100000
; GFX8-NEXT: s_lshr_b32 s2, s2, s8
; GFX8-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX8-NEXT: s_lshr_b32 s2, s2, s4
; GFX8-NEXT: s_lshl_b32 s0, s0, s11
; GFX8-NEXT: s_or_b32 s0, s0, s2
; GFX8-NEXT: s_and_b32 s2, s9, 15
; GFX8-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX8-NEXT: s_lshl_b32 s7, s7, s8
; GFX8-NEXT: s_andn2_b32 s4, 15, s9
; GFX8-NEXT: s_lshl_b32 s2, s6, s2
; GFX8-NEXT: s_bfe_u32 s6, s7, 0x100000
; GFX8-NEXT: s_lshr_b32 s6, s6, s8
; GFX8-NEXT: s_bfe_u32 s4, s4, 0x100000
; GFX8-NEXT: s_lshr_b32 s4, s6, s4
; GFX8-NEXT: s_or_b32 s2, s2, s4
; GFX8-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX8-NEXT: s_bfe_u32 s6, s3, 0x100000
; GFX8-NEXT: s_lshr_b32 s6, s6, s8
; GFX8-NEXT: s_bfe_u32 s0, s0, 0x100000
; GFX8-NEXT: s_lshl_b32 s2, s2, 16
; GFX8-NEXT: s_or_b32 s0, s0, s2
; GFX8-NEXT: s_lshr_b32 s2, s1, 16
; GFX8-NEXT: s_lshr_b32 s4, s3, 16
; GFX8-NEXT: s_lshl_b32 s1, s1, s8
; GFX8-NEXT: s_lshr_b32 s6, s6, s10
; GFX8-NEXT: s_or_b32 s1, s1, s6
; GFX8-NEXT: s_lshr_b32 s6, s4, s8
; GFX8-NEXT: s_lshl_b32 s3, s3, s8
; GFX8-NEXT: s_xor_b32 s5, s5, -1
; GFX8-NEXT: s_lshl_b32 s2, s2, s8
; GFX8-NEXT: s_lshr_b32 s6, s6, s10
; GFX8-NEXT: s_and_b32 s7, s5, 15
; GFX8-NEXT: s_bfe_u32 s3, s3, 0x100000
; GFX8-NEXT: s_or_b32 s2, s2, s6
; GFX8-NEXT: s_lshr_b32 s6, s5, 16
; GFX8-NEXT: s_andn2_b32 s5, 15, s5
; GFX8-NEXT: s_bfe_u32 s7, s7, 0x100000
; GFX8-NEXT: s_lshr_b32 s3, s3, s8
; GFX8-NEXT: s_bfe_u32 s5, s5, 0x100000
; GFX8-NEXT: s_lshr_b32 s3, s3, s5
; GFX8-NEXT: s_lshl_b32 s1, s1, s7
; GFX8-NEXT: s_or_b32 s1, s1, s3
; GFX8-NEXT: s_and_b32 s3, s6, 15
; GFX8-NEXT: s_bfe_u32 s3, s3, 0x100000
; GFX8-NEXT: s_lshl_b32 s4, s4, s8
; GFX8-NEXT: s_lshl_b32 s2, s2, s3
; GFX8-NEXT: s_bfe_u32 s3, s4, 0x100000
; GFX8-NEXT: s_andn2_b32 s5, 15, s6
; GFX8-NEXT: s_lshr_b32 s3, s3, s8
; GFX8-NEXT: s_bfe_u32 s4, s5, 0x100000
; GFX8-NEXT: s_lshr_b32 s3, s3, s4
; GFX8-NEXT: s_or_b32 s2, s2, s3
; GFX8-NEXT: s_bfe_u32 s2, s2, 0x100000
; GFX8-NEXT: s_bfe_u32 s1, s1, 0x100000
; GFX8-NEXT: s_lshl_b32 s2, s2, 16
; GFX8-NEXT: s_or_b32 s1, s1, s2
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_v4i16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_lshr_b32 s9, s0, 16
; GFX9-NEXT: s_mov_b32 s8, 0x10001
; GFX9-NEXT: s_mov_b32 s6, 0xf000f
; GFX9-NEXT: s_lshl_b32 s0, s0, s8
; GFX9-NEXT: s_lshl_b32 s9, s9, 1
; GFX9-NEXT: s_pack_ll_b32_b16 s0, s0, s9
; GFX9-NEXT: s_and_b32 s7, s4, s6
; GFX9-NEXT: s_andn2_b32 s4, s6, s4
; GFX9-NEXT: s_lshr_b32 s9, s0, 16
; GFX9-NEXT: s_lshr_b32 s10, s4, 16
; GFX9-NEXT: s_lshl_b32 s0, s0, s4
; GFX9-NEXT: s_lshl_b32 s4, s9, s10
; GFX9-NEXT: s_mov_b32 s9, 0xffff
; GFX9-NEXT: s_pack_ll_b32_b16 s0, s0, s4
; GFX9-NEXT: s_lshr_b32 s4, s2, 16
; GFX9-NEXT: s_lshr_b32 s10, s7, 16
; GFX9-NEXT: s_and_b32 s2, s2, s9
; GFX9-NEXT: s_and_b32 s7, s7, s9
; GFX9-NEXT: s_lshr_b32 s2, s2, s7
; GFX9-NEXT: s_lshr_b32 s4, s4, s10
; GFX9-NEXT: s_pack_ll_b32_b16 s2, s2, s4
; GFX9-NEXT: s_or_b32 s0, s0, s2
; GFX9-NEXT: s_and_b32 s2, s5, s6
; GFX9-NEXT: s_andn2_b32 s4, s6, s5
; GFX9-NEXT: s_lshr_b32 s5, s1, 16
; GFX9-NEXT: s_lshl_b32 s1, s1, s8
; GFX9-NEXT: s_lshl_b32 s5, s5, 1
; GFX9-NEXT: s_pack_ll_b32_b16 s1, s1, s5
; GFX9-NEXT: s_lshr_b32 s5, s1, 16
; GFX9-NEXT: s_lshr_b32 s6, s4, 16
; GFX9-NEXT: s_lshl_b32 s1, s1, s4
; GFX9-NEXT: s_lshl_b32 s4, s5, s6
; GFX9-NEXT: s_pack_ll_b32_b16 s1, s1, s4
; GFX9-NEXT: s_lshr_b32 s4, s3, 16
; GFX9-NEXT: s_lshr_b32 s5, s2, 16
; GFX9-NEXT: s_and_b32 s3, s3, s9
; GFX9-NEXT: s_and_b32 s2, s2, s9
; GFX9-NEXT: s_lshr_b32 s2, s3, s2
; GFX9-NEXT: s_lshr_b32 s3, s4, s5
; GFX9-NEXT: s_pack_ll_b32_b16 s2, s2, s3
; GFX9-NEXT: s_or_b32 s1, s1, s2
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_v4i16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_lshr_b32 s8, s0, 16
; GFX10-NEXT: s_mov_b32 s7, 0x10001
; GFX10-NEXT: s_mov_b32 s6, 0xf000f
; GFX10-NEXT: s_lshl_b32 s0, s0, s7
; GFX10-NEXT: s_lshl_b32 s8, s8, 1
; GFX10-NEXT: s_and_b32 s9, s4, s6
; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s8
; GFX10-NEXT: s_andn2_b32 s4, s6, s4
; GFX10-NEXT: s_lshr_b32 s8, s0, 16
; GFX10-NEXT: s_lshr_b32 s10, s4, 16
; GFX10-NEXT: s_lshl_b32 s0, s0, s4
; GFX10-NEXT: s_lshl_b32 s4, s8, s10
; GFX10-NEXT: s_mov_b32 s8, 0xffff
; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s4
; GFX10-NEXT: s_lshr_b32 s4, s1, 16
; GFX10-NEXT: s_lshl_b32 s1, s1, s7
; GFX10-NEXT: s_lshl_b32 s4, s4, 1
; GFX10-NEXT: s_and_b32 s7, s5, s6
; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s4
; GFX10-NEXT: s_andn2_b32 s4, s6, s5
; GFX10-NEXT: s_lshr_b32 s5, s1, 16
; GFX10-NEXT: s_lshr_b32 s6, s4, 16
; GFX10-NEXT: s_lshr_b32 s10, s2, 16
; GFX10-NEXT: s_and_b32 s11, s9, s8
; GFX10-NEXT: s_and_b32 s2, s2, s8
; GFX10-NEXT: s_lshr_b32 s9, s9, 16
; GFX10-NEXT: s_lshl_b32 s1, s1, s4
; GFX10-NEXT: s_lshl_b32 s4, s5, s6
; GFX10-NEXT: s_lshr_b32 s5, s3, 16
; GFX10-NEXT: s_and_b32 s6, s7, s8
; GFX10-NEXT: s_and_b32 s3, s3, s8
; GFX10-NEXT: s_lshr_b32 s7, s7, 16
; GFX10-NEXT: s_lshr_b32 s2, s2, s11
; GFX10-NEXT: s_lshr_b32 s9, s10, s9
; GFX10-NEXT: s_lshr_b32 s3, s3, s6
; GFX10-NEXT: s_lshr_b32 s5, s5, s7
; GFX10-NEXT: s_pack_ll_b32_b16 s2, s2, s9
; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s4
; GFX10-NEXT: s_pack_ll_b32_b16 s3, s3, s5
; GFX10-NEXT: s_or_b32 s0, s0, s2
; GFX10-NEXT: s_or_b32 s1, s1, s3
; GFX10-NEXT: ; return to shader part epilog
%result = call <4 x i16> @llvm.fshr.v4i16(<4 x i16> %lhs, <4 x i16> %rhs, <4 x i16> %amt)
%cast.result = bitcast <4 x i16> %result to <2 x i32>
ret <2 x i32> %cast.result
}
define <4 x half> @v_fshr_v4i16(<4 x i16> %lhs, <4 x i16> %rhs, <4 x i16> %amt) {
; GFX6-LABEL: v_fshr_v4i16:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v12, 0xffff
; GFX6-NEXT: v_lshlrev_b32_e32 v9, 16, v9
; GFX6-NEXT: v_and_b32_e32 v8, v8, v12
; GFX6-NEXT: v_or_b32_e32 v8, v9, v8
; GFX6-NEXT: v_lshlrev_b32_e32 v9, 16, v11
; GFX6-NEXT: v_and_b32_e32 v10, v10, v12
; GFX6-NEXT: s_mov_b32 s5, 0xffff
; GFX6-NEXT: v_or_b32_e32 v9, v9, v10
; GFX6-NEXT: v_and_b32_e32 v10, s5, v4
; GFX6-NEXT: s_bfe_u32 s4, 1, 0x100000
; GFX6-NEXT: v_lshrrev_b32_e32 v10, 1, v10
; GFX6-NEXT: s_bfe_u32 s6, 14, 0x100000
; GFX6-NEXT: v_lshlrev_b32_e32 v0, s4, v0
; GFX6-NEXT: v_lshrrev_b32_e32 v10, s6, v10
; GFX6-NEXT: v_or_b32_e32 v0, v0, v10
; GFX6-NEXT: v_and_b32_e32 v10, s5, v5
; GFX6-NEXT: v_lshrrev_b32_e32 v10, 1, v10
; GFX6-NEXT: v_xor_b32_e32 v8, -1, v8
; GFX6-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX6-NEXT: v_lshrrev_b32_e32 v10, s6, v10
; GFX6-NEXT: v_lshlrev_b32_e32 v4, 1, v4
; GFX6-NEXT: v_and_b32_e32 v11, 15, v8
; GFX6-NEXT: v_or_b32_e32 v1, v1, v10
; GFX6-NEXT: v_lshrrev_b32_e32 v10, 16, v8
; GFX6-NEXT: v_xor_b32_e32 v8, -1, v8
; GFX6-NEXT: v_and_b32_e32 v8, 15, v8
; GFX6-NEXT: v_and_b32_e32 v4, v4, v12
; GFX6-NEXT: v_bfe_u32 v11, v11, 0, 16
; GFX6-NEXT: v_lshrrev_b32_e32 v4, 1, v4
; GFX6-NEXT: v_bfe_u32 v8, v8, 0, 16
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v8, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v0, v11, v0
; GFX6-NEXT: v_or_b32_e32 v0, v0, v4
; GFX6-NEXT: v_and_b32_e32 v4, 15, v10
; GFX6-NEXT: v_xor_b32_e32 v8, -1, v10
; GFX6-NEXT: v_bfe_u32 v4, v4, 0, 16
; GFX6-NEXT: v_lshlrev_b32_e32 v5, 1, v5
; GFX6-NEXT: v_lshlrev_b32_e32 v1, v4, v1
; GFX6-NEXT: v_and_b32_e32 v4, v5, v12
; GFX6-NEXT: v_and_b32_e32 v8, 15, v8
; GFX6-NEXT: v_lshrrev_b32_e32 v4, 1, v4
; GFX6-NEXT: v_bfe_u32 v5, v8, 0, 16
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v5, v4
; GFX6-NEXT: v_or_b32_e32 v1, v1, v4
; GFX6-NEXT: v_and_b32_e32 v4, v6, v12
; GFX6-NEXT: v_lshrrev_b32_e32 v4, 1, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v2, s4, v2
; GFX6-NEXT: v_lshrrev_b32_e32 v4, s6, v4
; GFX6-NEXT: v_or_b32_e32 v2, v2, v4
; GFX6-NEXT: v_and_b32_e32 v4, v7, v12
; GFX6-NEXT: v_lshrrev_b32_e32 v4, 1, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v3, s4, v3
; GFX6-NEXT: v_lshrrev_b32_e32 v4, s6, v4
; GFX6-NEXT: v_or_b32_e32 v3, v3, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v4, 1, v6
; GFX6-NEXT: v_xor_b32_e32 v6, -1, v9
; GFX6-NEXT: v_and_b32_e32 v8, 15, v6
; GFX6-NEXT: v_lshlrev_b32_e32 v5, 1, v7
; GFX6-NEXT: v_lshrrev_b32_e32 v7, 16, v6
; GFX6-NEXT: v_xor_b32_e32 v6, -1, v6
; GFX6-NEXT: v_and_b32_e32 v6, 15, v6
; GFX6-NEXT: v_and_b32_e32 v4, v4, v12
; GFX6-NEXT: v_bfe_u32 v8, v8, 0, 16
; GFX6-NEXT: v_lshrrev_b32_e32 v4, 1, v4
; GFX6-NEXT: v_bfe_u32 v6, v6, 0, 16
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v6, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v2, v8, v2
; GFX6-NEXT: v_or_b32_e32 v2, v2, v4
; GFX6-NEXT: v_and_b32_e32 v4, 15, v7
; GFX6-NEXT: v_xor_b32_e32 v6, -1, v7
; GFX6-NEXT: v_bfe_u32 v4, v4, 0, 16
; GFX6-NEXT: v_lshlrev_b32_e32 v3, v4, v3
; GFX6-NEXT: v_and_b32_e32 v4, v5, v12
; GFX6-NEXT: v_and_b32_e32 v6, 15, v6
; GFX6-NEXT: v_lshrrev_b32_e32 v4, 1, v4
; GFX6-NEXT: v_bfe_u32 v5, v6, 0, 16
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v5, v4
; GFX6-NEXT: v_or_b32_e32 v3, v3, v4
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v4i16:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_lshrrev_b16_e32 v7, 1, v2
; GFX8-NEXT: v_lshlrev_b16_e32 v6, 1, v0
; GFX8-NEXT: v_lshrrev_b16_e32 v7, 14, v7
; GFX8-NEXT: v_or_b32_e32 v6, v6, v7
; GFX8-NEXT: v_mov_b32_e32 v7, 1
; GFX8-NEXT: v_lshrrev_b16_sdwa v8, v7, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX8-NEXT: v_lshlrev_b16_sdwa v0, v7, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_lshrrev_b16_e32 v8, 14, v8
; GFX8-NEXT: v_or_b32_e32 v0, v0, v8
; GFX8-NEXT: v_lshlrev_b16_e32 v8, 1, v2
; GFX8-NEXT: v_lshlrev_b16_sdwa v2, v7, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_and_b32_e32 v9, 15, v4
; GFX8-NEXT: v_lshrrev_b32_e32 v7, 16, v4
; GFX8-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX8-NEXT: v_and_b32_e32 v4, 15, v4
; GFX8-NEXT: v_lshrrev_b16_e32 v8, 1, v8
; GFX8-NEXT: v_lshlrev_b16_e32 v6, v9, v6
; GFX8-NEXT: v_lshrrev_b16_e32 v4, v4, v8
; GFX8-NEXT: v_or_b32_e32 v4, v6, v4
; GFX8-NEXT: v_and_b32_e32 v6, 15, v7
; GFX8-NEXT: v_xor_b32_e32 v7, -1, v7
; GFX8-NEXT: v_and_b32_e32 v7, 15, v7
; GFX8-NEXT: v_lshrrev_b16_e32 v2, 1, v2
; GFX8-NEXT: v_lshlrev_b16_e32 v0, v6, v0
; GFX8-NEXT: v_lshrrev_b16_e32 v2, v7, v2
; GFX8-NEXT: v_or_b32_e32 v0, v0, v2
; GFX8-NEXT: v_mov_b32_e32 v2, 16
; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX8-NEXT: v_lshrrev_b16_e32 v6, 1, v3
; GFX8-NEXT: v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX8-NEXT: v_lshlrev_b16_e32 v4, 1, v1
; GFX8-NEXT: v_lshrrev_b16_e32 v6, 14, v6
; GFX8-NEXT: v_or_b32_e32 v4, v4, v6
; GFX8-NEXT: v_mov_b32_e32 v6, 1
; GFX8-NEXT: v_lshrrev_b16_sdwa v7, v6, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_xor_b32_e32 v5, -1, v5
; GFX8-NEXT: v_lshlrev_b16_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_lshrrev_b16_e32 v7, 14, v7
; GFX8-NEXT: v_or_b32_e32 v1, v1, v7
; GFX8-NEXT: v_lshlrev_b16_e32 v7, 1, v3
; GFX8-NEXT: v_lshlrev_b16_sdwa v3, v6, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: v_and_b32_e32 v8, 15, v5
; GFX8-NEXT: v_lshrrev_b32_e32 v6, 16, v5
; GFX8-NEXT: v_xor_b32_e32 v5, -1, v5
; GFX8-NEXT: v_and_b32_e32 v5, 15, v5
; GFX8-NEXT: v_lshrrev_b16_e32 v7, 1, v7
; GFX8-NEXT: v_lshlrev_b16_e32 v4, v8, v4
; GFX8-NEXT: v_lshrrev_b16_e32 v5, v5, v7
; GFX8-NEXT: v_or_b32_e32 v4, v4, v5
; GFX8-NEXT: v_and_b32_e32 v5, 15, v6
; GFX8-NEXT: v_xor_b32_e32 v6, -1, v6
; GFX8-NEXT: v_and_b32_e32 v6, 15, v6
; GFX8-NEXT: v_lshrrev_b16_e32 v3, 1, v3
; GFX8-NEXT: v_lshlrev_b16_e32 v1, v5, v1
; GFX8-NEXT: v_lshrrev_b16_e32 v3, v6, v3
; GFX8-NEXT: v_or_b32_e32 v1, v1, v3
; GFX8-NEXT: v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX8-NEXT: v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v4i16:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_mov_b32 s4, 0xf000f
; GFX9-NEXT: v_and_b32_e32 v6, s4, v4
; GFX9-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX9-NEXT: v_and_b32_e32 v4, s4, v4
; GFX9-NEXT: v_pk_lshlrev_b16 v0, 1, v0 op_sel_hi:[0,1]
; GFX9-NEXT: v_pk_lshlrev_b16 v0, v4, v0
; GFX9-NEXT: v_xor_b32_e32 v4, -1, v5
; GFX9-NEXT: v_pk_lshrrev_b16 v2, v6, v2
; GFX9-NEXT: v_or_b32_e32 v0, v0, v2
; GFX9-NEXT: v_and_b32_e32 v2, s4, v5
; GFX9-NEXT: v_and_b32_e32 v4, s4, v4
; GFX9-NEXT: v_pk_lshlrev_b16 v1, 1, v1 op_sel_hi:[0,1]
; GFX9-NEXT: v_pk_lshlrev_b16 v1, v4, v1
; GFX9-NEXT: v_pk_lshrrev_b16 v2, v2, v3
; GFX9-NEXT: v_or_b32_e32 v1, v1, v2
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v4i16:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v6, -1, v4
; GFX10-NEXT: v_xor_b32_e32 v7, -1, v5
; GFX10-NEXT: s_mov_b32 s4, 0xf000f
; GFX10-NEXT: v_pk_lshlrev_b16 v0, 1, v0 op_sel_hi:[0,1]
; GFX10-NEXT: v_and_b32_e32 v4, s4, v4
; GFX10-NEXT: v_and_b32_e32 v6, s4, v6
; GFX10-NEXT: v_and_b32_e32 v5, s4, v5
; GFX10-NEXT: v_and_b32_e32 v7, s4, v7
; GFX10-NEXT: v_pk_lshlrev_b16 v1, 1, v1 op_sel_hi:[0,1]
; GFX10-NEXT: v_pk_lshrrev_b16 v2, v4, v2
; GFX10-NEXT: v_pk_lshlrev_b16 v0, v6, v0
; GFX10-NEXT: v_pk_lshrrev_b16 v3, v5, v3
; GFX10-NEXT: v_pk_lshlrev_b16 v1, v7, v1
; GFX10-NEXT: v_or_b32_e32 v0, v0, v2
; GFX10-NEXT: v_or_b32_e32 v1, v1, v3
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <4 x i16> @llvm.fshr.v4i16(<4 x i16> %lhs, <4 x i16> %rhs, <4 x i16> %amt)
%cast.result = bitcast <4 x i16> %result to <4 x half>
ret <4 x half> %cast.result
}
define amdgpu_ps i64 @s_fshr_i64(i64 inreg %lhs, i64 inreg %rhs, i64 inreg %amt) {
; GFX6-LABEL: s_fshr_i64:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b64 s[6:7], s[4:5], 63
; GFX6-NEXT: s_andn2_b64 s[4:5], 63, s[4:5]
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], s4
; GFX6-NEXT: s_lshr_b64 s[2:3], s[2:3], s6
; GFX6-NEXT: s_or_b64 s[0:1], s[0:1], s[2:3]
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i64:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_and_b64 s[6:7], s[4:5], 63
; GFX8-NEXT: s_andn2_b64 s[4:5], 63, s[4:5]
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], s4
; GFX8-NEXT: s_lshr_b64 s[2:3], s[2:3], s6
; GFX8-NEXT: s_or_b64 s[0:1], s[0:1], s[2:3]
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i64:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_and_b64 s[6:7], s[4:5], 63
; GFX9-NEXT: s_andn2_b64 s[4:5], 63, s[4:5]
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], s4
; GFX9-NEXT: s_lshr_b64 s[2:3], s[2:3], s6
; GFX9-NEXT: s_or_b64 s[0:1], s[0:1], s[2:3]
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i64:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_andn2_b64 s[6:7], 63, s[4:5]
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX10-NEXT: s_and_b64 s[4:5], s[4:5], 63
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s6
; GFX10-NEXT: s_lshr_b64 s[2:3], s[2:3], s4
; GFX10-NEXT: s_or_b64 s[0:1], s[0:1], s[2:3]
; GFX10-NEXT: ; return to shader part epilog
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 %amt)
ret i64 %result
}
define amdgpu_ps i64 @s_fshr_i64_5(i64 inreg %lhs, i64 inreg %rhs) {
; GCN-LABEL: s_fshr_i64_5:
; GCN: ; %bb.0:
; GCN-NEXT: s_lshl_b32 s1, s0, 27
; GCN-NEXT: s_mov_b32 s0, 0
; GCN-NEXT: s_lshr_b64 s[2:3], s[2:3], 5
; GCN-NEXT: s_or_b64 s[0:1], s[0:1], s[2:3]
; GCN-NEXT: ; return to shader part epilog
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 5)
ret i64 %result
}
define amdgpu_ps i64 @s_fshr_i64_32(i64 inreg %lhs, i64 inreg %rhs) {
; GCN-LABEL: s_fshr_i64_32:
; GCN: ; %bb.0:
; GCN-NEXT: s_mov_b32 s1, s0
; GCN-NEXT: s_mov_b32 s0, 0
; GCN-NEXT: s_mov_b32 s2, s3
; GCN-NEXT: s_mov_b32 s3, s0
; GCN-NEXT: s_or_b64 s[0:1], s[0:1], s[2:3]
; GCN-NEXT: ; return to shader part epilog
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 32)
ret i64 %result
}
define amdgpu_ps i64 @s_fshr_i64_48(i64 inreg %lhs, i64 inreg %rhs) {
; GCN-LABEL: s_fshr_i64_48:
; GCN: ; %bb.0:
; GCN-NEXT: s_lshr_b32 s2, s3, 16
; GCN-NEXT: s_lshl_b64 s[0:1], s[0:1], 16
; GCN-NEXT: s_mov_b32 s3, 0
; GCN-NEXT: s_or_b64 s[0:1], s[0:1], s[2:3]
; GCN-NEXT: ; return to shader part epilog
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 48)
ret i64 %result
}
define i64 @v_fshr_i64(i64 %lhs, i64 %rhs, i64 %amt) {
; GFX6-LABEL: v_fshr_i64:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_and_b32_e32 v5, 63, v4
; GFX6-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 1
; GFX6-NEXT: v_and_b32_e32 v4, 63, v4
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], v4
; GFX6-NEXT: v_lshr_b64 v[2:3], v[2:3], v5
; GFX6-NEXT: v_or_b32_e32 v0, v0, v2
; GFX6-NEXT: v_or_b32_e32 v1, v1, v3
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i64:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_and_b32_e32 v5, 63, v4
; GFX8-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX8-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX8-NEXT: v_and_b32_e32 v4, 63, v4
; GFX8-NEXT: v_lshlrev_b64 v[0:1], v4, v[0:1]
; GFX8-NEXT: v_lshrrev_b64 v[2:3], v5, v[2:3]
; GFX8-NEXT: v_or_b32_e32 v0, v0, v2
; GFX8-NEXT: v_or_b32_e32 v1, v1, v3
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i64:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_and_b32_e32 v5, 63, v4
; GFX9-NEXT: v_xor_b32_e32 v4, -1, v4
; GFX9-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX9-NEXT: v_and_b32_e32 v4, 63, v4
; GFX9-NEXT: v_lshlrev_b64 v[0:1], v4, v[0:1]
; GFX9-NEXT: v_lshrrev_b64 v[2:3], v5, v[2:3]
; GFX9-NEXT: v_or_b32_e32 v0, v0, v2
; GFX9-NEXT: v_or_b32_e32 v1, v1, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i64:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v5, -1, v4
; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX10-NEXT: v_and_b32_e32 v4, 63, v4
; GFX10-NEXT: v_and_b32_e32 v5, 63, v5
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v4, v[2:3]
; GFX10-NEXT: v_lshlrev_b64 v[0:1], v5, v[0:1]
; GFX10-NEXT: v_or_b32_e32 v0, v0, v2
; GFX10-NEXT: v_or_b32_e32 v1, v1, v3
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 %amt)
ret i64 %result
}
define i64 @v_fshr_i64_5(i64 %lhs, i64 %rhs) {
; GFX6-LABEL: v_fshr_i64_5:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v4, v0
; GFX6-NEXT: v_lshr_b64 v[0:1], v[2:3], 5
; GFX6-NEXT: v_lshlrev_b32_e32 v2, 27, v4
; GFX6-NEXT: v_or_b32_e32 v1, v2, v1
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i64_5:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v4, v0
; GFX8-NEXT: v_lshrrev_b64 v[0:1], 5, v[2:3]
; GFX8-NEXT: v_lshlrev_b32_e32 v2, 27, v4
; GFX8-NEXT: v_or_b32_e32 v1, v2, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i64_5:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v4, v0
; GFX9-NEXT: v_lshrrev_b64 v[0:1], 5, v[2:3]
; GFX9-NEXT: v_lshlrev_b32_e32 v2, 27, v4
; GFX9-NEXT: v_or_b32_e32 v1, v2, v1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i64_5:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v4, v0
; GFX10-NEXT: v_lshrrev_b64 v[0:1], 5, v[2:3]
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 27, v4
; GFX10-NEXT: v_or_b32_e32 v1, v2, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 5)
ret i64 %result
}
define i64 @v_fshr_i64_32(i64 %lhs, i64 %rhs) {
; GFX6-LABEL: v_fshr_i64_32:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v1, v0
; GFX6-NEXT: v_mov_b32_e32 v0, v3
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i64_32:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v1, v0
; GFX8-NEXT: v_mov_b32_e32 v0, v3
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i64_32:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v1, v0
; GFX9-NEXT: v_mov_b32_e32 v0, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i64_32:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_mov_b32_e32 v1, v0
; GFX10-NEXT: v_mov_b32_e32 v0, v3
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 32)
ret i64 %result
}
define i64 @v_fshr_i64_48(i64 %lhs, i64 %rhs) {
; GFX6-LABEL: v_fshr_i64_48:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 16
; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v3
; GFX6-NEXT: v_or_b32_e32 v0, v0, v2
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i64_48:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_lshlrev_b64 v[0:1], 16, v[0:1]
; GFX8-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i64_48:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_lshlrev_b64 v[0:1], 16, v[0:1]
; GFX9-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i64_48:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_lshlrev_b64 v[0:1], 16, v[0:1]
; GFX10-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 48)
ret i64 %result
}
define amdgpu_ps <2 x float> @v_fshr_i64_ssv(i64 inreg %lhs, i64 inreg %rhs, i64 %amt) {
; GFX6-LABEL: v_fshr_i64_ssv:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_and_b32_e32 v2, 63, v0
; GFX6-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX6-NEXT: v_and_b32_e32 v0, 63, v0
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX6-NEXT: v_lshl_b64 v[0:1], s[0:1], v0
; GFX6-NEXT: v_lshr_b64 v[2:3], s[2:3], v2
; GFX6-NEXT: v_or_b32_e32 v0, v0, v2
; GFX6-NEXT: v_or_b32_e32 v1, v1, v3
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i64_ssv:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_and_b32_e32 v2, 63, v0
; GFX8-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX8-NEXT: v_and_b32_e32 v0, 63, v0
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX8-NEXT: v_lshlrev_b64 v[0:1], v0, s[0:1]
; GFX8-NEXT: v_lshrrev_b64 v[2:3], v2, s[2:3]
; GFX8-NEXT: v_or_b32_e32 v0, v0, v2
; GFX8-NEXT: v_or_b32_e32 v1, v1, v3
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i64_ssv:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_and_b32_e32 v2, 63, v0
; GFX9-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX9-NEXT: v_and_b32_e32 v0, 63, v0
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX9-NEXT: v_lshlrev_b64 v[0:1], v0, s[0:1]
; GFX9-NEXT: v_lshrrev_b64 v[2:3], v2, s[2:3]
; GFX9-NEXT: v_or_b32_e32 v0, v0, v2
; GFX9-NEXT: v_or_b32_e32 v1, v1, v3
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i64_ssv:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_xor_b32_e32 v1, -1, v0
; GFX10-NEXT: v_and_b32_e32 v0, 63, v0
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX10-NEXT: v_and_b32_e32 v2, 63, v1
; GFX10-NEXT: v_lshrrev_b64 v[0:1], v0, s[2:3]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], v2, s[0:1]
; GFX10-NEXT: v_or_b32_e32 v0, v2, v0
; GFX10-NEXT: v_or_b32_e32 v1, v3, v1
; GFX10-NEXT: ; return to shader part epilog
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 %amt)
%cast = bitcast i64 %result to <2 x float>
ret <2 x float> %cast
}
define amdgpu_ps <2 x float> @v_fshr_i64_svs(i64 inreg %lhs, i64 %rhs, i64 inreg %amt) {
; GFX6-LABEL: v_fshr_i64_svs:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b64 s[4:5], s[2:3], 63
; GFX6-NEXT: s_andn2_b64 s[2:3], 63, s[2:3]
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX6-NEXT: v_lshr_b64 v[0:1], v[0:1], s4
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], s2
; GFX6-NEXT: v_or_b32_e32 v0, s0, v0
; GFX6-NEXT: v_or_b32_e32 v1, s1, v1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i64_svs:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_and_b64 s[4:5], s[2:3], 63
; GFX8-NEXT: s_andn2_b64 s[2:3], 63, s[2:3]
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX8-NEXT: v_lshrrev_b64 v[0:1], s4, v[0:1]
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], s2
; GFX8-NEXT: v_or_b32_e32 v0, s0, v0
; GFX8-NEXT: v_or_b32_e32 v1, s1, v1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i64_svs:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_and_b64 s[4:5], s[2:3], 63
; GFX9-NEXT: s_andn2_b64 s[2:3], 63, s[2:3]
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX9-NEXT: v_lshrrev_b64 v[0:1], s4, v[0:1]
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], s2
; GFX9-NEXT: v_or_b32_e32 v0, s0, v0
; GFX9-NEXT: v_or_b32_e32 v1, s1, v1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i64_svs:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_and_b64 s[4:5], s[2:3], 63
; GFX10-NEXT: s_andn2_b64 s[2:3], 63, s[2:3]
; GFX10-NEXT: v_lshrrev_b64 v[0:1], s4, v[0:1]
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s2
; GFX10-NEXT: v_or_b32_e32 v0, s0, v0
; GFX10-NEXT: v_or_b32_e32 v1, s1, v1
; GFX10-NEXT: ; return to shader part epilog
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 %amt)
%cast = bitcast i64 %result to <2 x float>
ret <2 x float> %cast
}
define amdgpu_ps <2 x float> @v_fshr_i64_vss(i64 %lhs, i64 inreg %rhs, i64 inreg %amt) {
; GFX6-LABEL: v_fshr_i64_vss:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 1
; GFX6-NEXT: s_and_b64 s[4:5], s[2:3], 63
; GFX6-NEXT: s_andn2_b64 s[2:3], 63, s[2:3]
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], s2
; GFX6-NEXT: s_lshr_b64 s[0:1], s[0:1], s4
; GFX6-NEXT: v_or_b32_e32 v0, s0, v0
; GFX6-NEXT: v_or_b32_e32 v1, s1, v1
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i64_vss:
; GFX8: ; %bb.0:
; GFX8-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX8-NEXT: s_and_b64 s[4:5], s[2:3], 63
; GFX8-NEXT: s_andn2_b64 s[2:3], 63, s[2:3]
; GFX8-NEXT: v_lshlrev_b64 v[0:1], s2, v[0:1]
; GFX8-NEXT: s_lshr_b64 s[0:1], s[0:1], s4
; GFX8-NEXT: v_or_b32_e32 v0, s0, v0
; GFX8-NEXT: v_or_b32_e32 v1, s1, v1
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i64_vss:
; GFX9: ; %bb.0:
; GFX9-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX9-NEXT: s_and_b64 s[4:5], s[2:3], 63
; GFX9-NEXT: s_andn2_b64 s[2:3], 63, s[2:3]
; GFX9-NEXT: v_lshlrev_b64 v[0:1], s2, v[0:1]
; GFX9-NEXT: s_lshr_b64 s[0:1], s[0:1], s4
; GFX9-NEXT: v_or_b32_e32 v0, s0, v0
; GFX9-NEXT: v_or_b32_e32 v1, s1, v1
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i64_vss:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX10-NEXT: s_andn2_b64 s[4:5], 63, s[2:3]
; GFX10-NEXT: s_and_b64 s[2:3], s[2:3], 63
; GFX10-NEXT: s_lshr_b64 s[0:1], s[0:1], s2
; GFX10-NEXT: v_lshlrev_b64 v[0:1], s4, v[0:1]
; GFX10-NEXT: v_or_b32_e32 v0, s0, v0
; GFX10-NEXT: v_or_b32_e32 v1, s1, v1
; GFX10-NEXT: ; return to shader part epilog
%result = call i64 @llvm.fshr.i64(i64 %lhs, i64 %rhs, i64 %amt)
%cast = bitcast i64 %result to <2 x float>
ret <2 x float> %cast
}
define amdgpu_ps <2 x i64> @s_fshr_v2i64(<2 x i64> inreg %lhs, <2 x i64> inreg %rhs, <2 x i64> inreg %amt) {
; GFX6-LABEL: s_fshr_v2i64:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_and_b64 s[12:13], s[8:9], 63
; GFX6-NEXT: s_andn2_b64 s[8:9], 63, s[8:9]
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], s8
; GFX6-NEXT: s_lshr_b64 s[4:5], s[4:5], s12
; GFX6-NEXT: s_or_b64 s[0:1], s[0:1], s[4:5]
; GFX6-NEXT: s_and_b64 s[4:5], s[10:11], 63
; GFX6-NEXT: s_andn2_b64 s[8:9], 63, s[10:11]
; GFX6-NEXT: s_lshl_b64 s[2:3], s[2:3], 1
; GFX6-NEXT: s_lshl_b64 s[2:3], s[2:3], s8
; GFX6-NEXT: s_lshr_b64 s[4:5], s[6:7], s4
; GFX6-NEXT: s_or_b64 s[2:3], s[2:3], s[4:5]
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_v2i64:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_and_b64 s[12:13], s[8:9], 63
; GFX8-NEXT: s_andn2_b64 s[8:9], 63, s[8:9]
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], s8
; GFX8-NEXT: s_lshr_b64 s[4:5], s[4:5], s12
; GFX8-NEXT: s_or_b64 s[0:1], s[0:1], s[4:5]
; GFX8-NEXT: s_and_b64 s[4:5], s[10:11], 63
; GFX8-NEXT: s_andn2_b64 s[8:9], 63, s[10:11]
; GFX8-NEXT: s_lshl_b64 s[2:3], s[2:3], 1
; GFX8-NEXT: s_lshl_b64 s[2:3], s[2:3], s8
; GFX8-NEXT: s_lshr_b64 s[4:5], s[6:7], s4
; GFX8-NEXT: s_or_b64 s[2:3], s[2:3], s[4:5]
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_v2i64:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_and_b64 s[12:13], s[8:9], 63
; GFX9-NEXT: s_andn2_b64 s[8:9], 63, s[8:9]
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], s8
; GFX9-NEXT: s_lshr_b64 s[4:5], s[4:5], s12
; GFX9-NEXT: s_or_b64 s[0:1], s[0:1], s[4:5]
; GFX9-NEXT: s_and_b64 s[4:5], s[10:11], 63
; GFX9-NEXT: s_andn2_b64 s[8:9], 63, s[10:11]
; GFX9-NEXT: s_lshl_b64 s[2:3], s[2:3], 1
; GFX9-NEXT: s_lshl_b64 s[2:3], s[2:3], s8
; GFX9-NEXT: s_lshr_b64 s[4:5], s[6:7], s4
; GFX9-NEXT: s_or_b64 s[2:3], s[2:3], s[4:5]
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_v2i64:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_andn2_b64 s[12:13], 63, s[8:9]
; GFX10-NEXT: s_and_b64 s[8:9], s[8:9], 63
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], 1
; GFX10-NEXT: s_lshr_b64 s[4:5], s[4:5], s8
; GFX10-NEXT: s_andn2_b64 s[8:9], 63, s[10:11]
; GFX10-NEXT: s_lshl_b64 s[2:3], s[2:3], 1
; GFX10-NEXT: s_and_b64 s[10:11], s[10:11], 63
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s12
; GFX10-NEXT: s_lshl_b64 s[2:3], s[2:3], s8
; GFX10-NEXT: s_lshr_b64 s[6:7], s[6:7], s10
; GFX10-NEXT: s_or_b64 s[0:1], s[0:1], s[4:5]
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[6:7]
; GFX10-NEXT: ; return to shader part epilog
%result = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %lhs, <2 x i64> %rhs, <2 x i64> %amt)
ret <2 x i64> %result
}
define <2 x i64> @v_fshr_v2i64(<2 x i64> %lhs, <2 x i64> %rhs, <2 x i64> %amt) {
; GFX6-LABEL: v_fshr_v2i64:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_and_b32_e32 v9, 63, v8
; GFX6-NEXT: v_xor_b32_e32 v8, -1, v8
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 1
; GFX6-NEXT: v_and_b32_e32 v8, 63, v8
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], v8
; GFX6-NEXT: v_lshr_b64 v[4:5], v[4:5], v9
; GFX6-NEXT: v_xor_b32_e32 v8, -1, v10
; GFX6-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
; GFX6-NEXT: v_or_b32_e32 v0, v0, v4
; GFX6-NEXT: v_and_b32_e32 v4, 63, v10
; GFX6-NEXT: v_and_b32_e32 v8, 63, v8
; GFX6-NEXT: v_lshl_b64 v[2:3], v[2:3], v8
; GFX6-NEXT: v_lshr_b64 v[6:7], v[6:7], v4
; GFX6-NEXT: v_or_b32_e32 v1, v1, v5
; GFX6-NEXT: v_or_b32_e32 v2, v2, v6
; GFX6-NEXT: v_or_b32_e32 v3, v3, v7
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v2i64:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_and_b32_e32 v9, 63, v8
; GFX8-NEXT: v_xor_b32_e32 v8, -1, v8
; GFX8-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX8-NEXT: v_and_b32_e32 v8, 63, v8
; GFX8-NEXT: v_lshlrev_b64 v[0:1], v8, v[0:1]
; GFX8-NEXT: v_lshrrev_b64 v[4:5], v9, v[4:5]
; GFX8-NEXT: v_xor_b32_e32 v8, -1, v10
; GFX8-NEXT: v_lshlrev_b64 v[2:3], 1, v[2:3]
; GFX8-NEXT: v_or_b32_e32 v0, v0, v4
; GFX8-NEXT: v_and_b32_e32 v4, 63, v10
; GFX8-NEXT: v_and_b32_e32 v8, 63, v8
; GFX8-NEXT: v_lshlrev_b64 v[2:3], v8, v[2:3]
; GFX8-NEXT: v_lshrrev_b64 v[6:7], v4, v[6:7]
; GFX8-NEXT: v_or_b32_e32 v1, v1, v5
; GFX8-NEXT: v_or_b32_e32 v2, v2, v6
; GFX8-NEXT: v_or_b32_e32 v3, v3, v7
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v2i64:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_and_b32_e32 v9, 63, v8
; GFX9-NEXT: v_xor_b32_e32 v8, -1, v8
; GFX9-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX9-NEXT: v_and_b32_e32 v8, 63, v8
; GFX9-NEXT: v_lshlrev_b64 v[0:1], v8, v[0:1]
; GFX9-NEXT: v_lshrrev_b64 v[4:5], v9, v[4:5]
; GFX9-NEXT: v_xor_b32_e32 v8, -1, v10
; GFX9-NEXT: v_lshlrev_b64 v[2:3], 1, v[2:3]
; GFX9-NEXT: v_or_b32_e32 v0, v0, v4
; GFX9-NEXT: v_and_b32_e32 v4, 63, v10
; GFX9-NEXT: v_and_b32_e32 v8, 63, v8
; GFX9-NEXT: v_lshlrev_b64 v[2:3], v8, v[2:3]
; GFX9-NEXT: v_lshrrev_b64 v[6:7], v4, v[6:7]
; GFX9-NEXT: v_or_b32_e32 v1, v1, v5
; GFX9-NEXT: v_or_b32_e32 v2, v2, v6
; GFX9-NEXT: v_or_b32_e32 v3, v3, v7
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v2i64:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_xor_b32_e32 v9, -1, v8
; GFX10-NEXT: v_xor_b32_e32 v11, -1, v10
; GFX10-NEXT: v_lshlrev_b64 v[0:1], 1, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], 1, v[2:3]
; GFX10-NEXT: v_and_b32_e32 v8, 63, v8
; GFX10-NEXT: v_and_b32_e32 v9, 63, v9
; GFX10-NEXT: v_and_b32_e32 v11, 63, v11
; GFX10-NEXT: v_and_b32_e32 v10, 63, v10
; GFX10-NEXT: v_lshrrev_b64 v[4:5], v8, v[4:5]
; GFX10-NEXT: v_lshlrev_b64 v[0:1], v9, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], v11, v[2:3]
; GFX10-NEXT: v_lshrrev_b64 v[6:7], v10, v[6:7]
; GFX10-NEXT: v_or_b32_e32 v0, v0, v4
; GFX10-NEXT: v_or_b32_e32 v1, v1, v5
; GFX10-NEXT: v_or_b32_e32 v2, v2, v6
; GFX10-NEXT: v_or_b32_e32 v3, v3, v7
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %lhs, <2 x i64> %rhs, <2 x i64> %amt)
ret <2 x i64> %result
}
define amdgpu_ps i128 @s_fshr_i128(i128 inreg %lhs, i128 inreg %rhs, i128 inreg %amt) {
; GFX6-LABEL: s_fshr_i128:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_movk_i32 s10, 0x7f
; GFX6-NEXT: s_mov_b32 s11, 0
; GFX6-NEXT: s_and_b64 s[12:13], s[8:9], s[10:11]
; GFX6-NEXT: s_andn2_b64 s[8:9], s[10:11], s[8:9]
; GFX6-NEXT: s_sub_i32 s9, 1, 64
; GFX6-NEXT: s_sub_i32 s13, 64, 1
; GFX6-NEXT: s_cmp_lt_u32 1, 64
; GFX6-NEXT: s_cselect_b32 s18, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 1, 0
; GFX6-NEXT: s_cselect_b32 s19, 1, 0
; GFX6-NEXT: s_lshr_b64 s[14:15], s[0:1], s13
; GFX6-NEXT: s_lshl_b64 s[16:17], s[2:3], 1
; GFX6-NEXT: s_lshl_b64 s[10:11], s[0:1], 1
; GFX6-NEXT: s_or_b64 s[14:15], s[14:15], s[16:17]
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], s9
; GFX6-NEXT: s_cmp_lg_u32 s18, 0
; GFX6-NEXT: s_cselect_b64 s[10:11], s[10:11], 0
; GFX6-NEXT: s_cselect_b64 s[0:1], s[14:15], s[0:1]
; GFX6-NEXT: s_cmp_lg_u32 s19, 0
; GFX6-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX6-NEXT: s_sub_i32 s13, s8, 64
; GFX6-NEXT: s_sub_i32 s9, 64, s8
; GFX6-NEXT: s_cmp_lt_u32 s8, 64
; GFX6-NEXT: s_cselect_b32 s16, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s8, 0
; GFX6-NEXT: s_cselect_b32 s17, 1, 0
; GFX6-NEXT: s_lshl_b64 s[2:3], s[10:11], s8
; GFX6-NEXT: s_lshr_b64 s[14:15], s[10:11], s9
; GFX6-NEXT: s_lshl_b64 s[8:9], s[0:1], s8
; GFX6-NEXT: s_or_b64 s[8:9], s[14:15], s[8:9]
; GFX6-NEXT: s_lshl_b64 s[10:11], s[10:11], s13
; GFX6-NEXT: s_cmp_lg_u32 s16, 0
; GFX6-NEXT: s_cselect_b64 s[2:3], s[2:3], 0
; GFX6-NEXT: s_cselect_b64 s[8:9], s[8:9], s[10:11]
; GFX6-NEXT: s_cmp_lg_u32 s17, 0
; GFX6-NEXT: s_cselect_b64 s[8:9], s[0:1], s[8:9]
; GFX6-NEXT: s_sub_i32 s14, s12, 64
; GFX6-NEXT: s_sub_i32 s13, 64, s12
; GFX6-NEXT: s_cmp_lt_u32 s12, 64
; GFX6-NEXT: s_cselect_b32 s15, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s12, 0
; GFX6-NEXT: s_cselect_b32 s16, 1, 0
; GFX6-NEXT: s_lshr_b64 s[0:1], s[6:7], s12
; GFX6-NEXT: s_lshr_b64 s[10:11], s[4:5], s12
; GFX6-NEXT: s_lshl_b64 s[12:13], s[6:7], s13
; GFX6-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX6-NEXT: s_lshr_b64 s[6:7], s[6:7], s14
; GFX6-NEXT: s_cmp_lg_u32 s15, 0
; GFX6-NEXT: s_cselect_b64 s[6:7], s[10:11], s[6:7]
; GFX6-NEXT: s_cmp_lg_u32 s16, 0
; GFX6-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX6-NEXT: s_cmp_lg_u32 s15, 0
; GFX6-NEXT: s_cselect_b64 s[6:7], s[0:1], 0
; GFX6-NEXT: s_or_b64 s[0:1], s[2:3], s[4:5]
; GFX6-NEXT: s_or_b64 s[2:3], s[8:9], s[6:7]
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i128:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_movk_i32 s10, 0x7f
; GFX8-NEXT: s_mov_b32 s11, 0
; GFX8-NEXT: s_and_b64 s[12:13], s[8:9], s[10:11]
; GFX8-NEXT: s_andn2_b64 s[8:9], s[10:11], s[8:9]
; GFX8-NEXT: s_sub_i32 s9, 1, 64
; GFX8-NEXT: s_sub_i32 s13, 64, 1
; GFX8-NEXT: s_cmp_lt_u32 1, 64
; GFX8-NEXT: s_cselect_b32 s18, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 1, 0
; GFX8-NEXT: s_cselect_b32 s19, 1, 0
; GFX8-NEXT: s_lshr_b64 s[14:15], s[0:1], s13
; GFX8-NEXT: s_lshl_b64 s[16:17], s[2:3], 1
; GFX8-NEXT: s_lshl_b64 s[10:11], s[0:1], 1
; GFX8-NEXT: s_or_b64 s[14:15], s[14:15], s[16:17]
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], s9
; GFX8-NEXT: s_cmp_lg_u32 s18, 0
; GFX8-NEXT: s_cselect_b64 s[10:11], s[10:11], 0
; GFX8-NEXT: s_cselect_b64 s[0:1], s[14:15], s[0:1]
; GFX8-NEXT: s_cmp_lg_u32 s19, 0
; GFX8-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX8-NEXT: s_sub_i32 s13, s8, 64
; GFX8-NEXT: s_sub_i32 s9, 64, s8
; GFX8-NEXT: s_cmp_lt_u32 s8, 64
; GFX8-NEXT: s_cselect_b32 s16, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s8, 0
; GFX8-NEXT: s_cselect_b32 s17, 1, 0
; GFX8-NEXT: s_lshl_b64 s[2:3], s[10:11], s8
; GFX8-NEXT: s_lshr_b64 s[14:15], s[10:11], s9
; GFX8-NEXT: s_lshl_b64 s[8:9], s[0:1], s8
; GFX8-NEXT: s_or_b64 s[8:9], s[14:15], s[8:9]
; GFX8-NEXT: s_lshl_b64 s[10:11], s[10:11], s13
; GFX8-NEXT: s_cmp_lg_u32 s16, 0
; GFX8-NEXT: s_cselect_b64 s[2:3], s[2:3], 0
; GFX8-NEXT: s_cselect_b64 s[8:9], s[8:9], s[10:11]
; GFX8-NEXT: s_cmp_lg_u32 s17, 0
; GFX8-NEXT: s_cselect_b64 s[8:9], s[0:1], s[8:9]
; GFX8-NEXT: s_sub_i32 s14, s12, 64
; GFX8-NEXT: s_sub_i32 s13, 64, s12
; GFX8-NEXT: s_cmp_lt_u32 s12, 64
; GFX8-NEXT: s_cselect_b32 s15, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s12, 0
; GFX8-NEXT: s_cselect_b32 s16, 1, 0
; GFX8-NEXT: s_lshr_b64 s[0:1], s[6:7], s12
; GFX8-NEXT: s_lshr_b64 s[10:11], s[4:5], s12
; GFX8-NEXT: s_lshl_b64 s[12:13], s[6:7], s13
; GFX8-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX8-NEXT: s_lshr_b64 s[6:7], s[6:7], s14
; GFX8-NEXT: s_cmp_lg_u32 s15, 0
; GFX8-NEXT: s_cselect_b64 s[6:7], s[10:11], s[6:7]
; GFX8-NEXT: s_cmp_lg_u32 s16, 0
; GFX8-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX8-NEXT: s_cmp_lg_u32 s15, 0
; GFX8-NEXT: s_cselect_b64 s[6:7], s[0:1], 0
; GFX8-NEXT: s_or_b64 s[0:1], s[2:3], s[4:5]
; GFX8-NEXT: s_or_b64 s[2:3], s[8:9], s[6:7]
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i128:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_movk_i32 s10, 0x7f
; GFX9-NEXT: s_mov_b32 s11, 0
; GFX9-NEXT: s_and_b64 s[12:13], s[8:9], s[10:11]
; GFX9-NEXT: s_andn2_b64 s[8:9], s[10:11], s[8:9]
; GFX9-NEXT: s_sub_i32 s9, 1, 64
; GFX9-NEXT: s_sub_i32 s13, 64, 1
; GFX9-NEXT: s_cmp_lt_u32 1, 64
; GFX9-NEXT: s_cselect_b32 s18, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 1, 0
; GFX9-NEXT: s_cselect_b32 s19, 1, 0
; GFX9-NEXT: s_lshr_b64 s[14:15], s[0:1], s13
; GFX9-NEXT: s_lshl_b64 s[16:17], s[2:3], 1
; GFX9-NEXT: s_lshl_b64 s[10:11], s[0:1], 1
; GFX9-NEXT: s_or_b64 s[14:15], s[14:15], s[16:17]
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], s9
; GFX9-NEXT: s_cmp_lg_u32 s18, 0
; GFX9-NEXT: s_cselect_b64 s[10:11], s[10:11], 0
; GFX9-NEXT: s_cselect_b64 s[0:1], s[14:15], s[0:1]
; GFX9-NEXT: s_cmp_lg_u32 s19, 0
; GFX9-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX9-NEXT: s_sub_i32 s13, s8, 64
; GFX9-NEXT: s_sub_i32 s9, 64, s8
; GFX9-NEXT: s_cmp_lt_u32 s8, 64
; GFX9-NEXT: s_cselect_b32 s16, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s8, 0
; GFX9-NEXT: s_cselect_b32 s17, 1, 0
; GFX9-NEXT: s_lshl_b64 s[2:3], s[10:11], s8
; GFX9-NEXT: s_lshr_b64 s[14:15], s[10:11], s9
; GFX9-NEXT: s_lshl_b64 s[8:9], s[0:1], s8
; GFX9-NEXT: s_or_b64 s[8:9], s[14:15], s[8:9]
; GFX9-NEXT: s_lshl_b64 s[10:11], s[10:11], s13
; GFX9-NEXT: s_cmp_lg_u32 s16, 0
; GFX9-NEXT: s_cselect_b64 s[2:3], s[2:3], 0
; GFX9-NEXT: s_cselect_b64 s[8:9], s[8:9], s[10:11]
; GFX9-NEXT: s_cmp_lg_u32 s17, 0
; GFX9-NEXT: s_cselect_b64 s[8:9], s[0:1], s[8:9]
; GFX9-NEXT: s_sub_i32 s14, s12, 64
; GFX9-NEXT: s_sub_i32 s13, 64, s12
; GFX9-NEXT: s_cmp_lt_u32 s12, 64
; GFX9-NEXT: s_cselect_b32 s15, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s12, 0
; GFX9-NEXT: s_cselect_b32 s16, 1, 0
; GFX9-NEXT: s_lshr_b64 s[0:1], s[6:7], s12
; GFX9-NEXT: s_lshr_b64 s[10:11], s[4:5], s12
; GFX9-NEXT: s_lshl_b64 s[12:13], s[6:7], s13
; GFX9-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX9-NEXT: s_lshr_b64 s[6:7], s[6:7], s14
; GFX9-NEXT: s_cmp_lg_u32 s15, 0
; GFX9-NEXT: s_cselect_b64 s[6:7], s[10:11], s[6:7]
; GFX9-NEXT: s_cmp_lg_u32 s16, 0
; GFX9-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX9-NEXT: s_cmp_lg_u32 s15, 0
; GFX9-NEXT: s_cselect_b64 s[6:7], s[0:1], 0
; GFX9-NEXT: s_or_b64 s[0:1], s[2:3], s[4:5]
; GFX9-NEXT: s_or_b64 s[2:3], s[8:9], s[6:7]
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i128:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_movk_i32 s10, 0x7f
; GFX10-NEXT: s_mov_b32 s11, 0
; GFX10-NEXT: s_and_b64 s[12:13], s[8:9], s[10:11]
; GFX10-NEXT: s_andn2_b64 s[8:9], s[10:11], s[8:9]
; GFX10-NEXT: s_sub_i32 s9, 1, 64
; GFX10-NEXT: s_sub_i32 s10, 64, 1
; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: s_cselect_b32 s13, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: s_cselect_b32 s18, 1, 0
; GFX10-NEXT: s_lshr_b64 s[10:11], s[0:1], s10
; GFX10-NEXT: s_lshl_b64 s[14:15], s[2:3], 1
; GFX10-NEXT: s_lshl_b64 s[16:17], s[0:1], 1
; GFX10-NEXT: s_or_b64 s[10:11], s[10:11], s[14:15]
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s9
; GFX10-NEXT: s_cmp_lg_u32 s13, 0
; GFX10-NEXT: s_cselect_b64 s[14:15], s[16:17], 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s18, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX10-NEXT: s_sub_i32 s13, s8, 64
; GFX10-NEXT: s_sub_i32 s2, 64, s8
; GFX10-NEXT: s_cmp_lt_u32 s8, 64
; GFX10-NEXT: s_cselect_b32 s16, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s8, 0
; GFX10-NEXT: s_cselect_b32 s17, 1, 0
; GFX10-NEXT: s_lshl_b64 s[10:11], s[0:1], s8
; GFX10-NEXT: s_lshr_b64 s[2:3], s[14:15], s2
; GFX10-NEXT: s_lshl_b64 s[8:9], s[14:15], s8
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[10:11]
; GFX10-NEXT: s_lshl_b64 s[10:11], s[14:15], s13
; GFX10-NEXT: s_cmp_lg_u32 s16, 0
; GFX10-NEXT: s_cselect_b64 s[8:9], s[8:9], 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[10:11]
; GFX10-NEXT: s_cmp_lg_u32 s17, 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[0:1], s[2:3]
; GFX10-NEXT: s_sub_i32 s14, s12, 64
; GFX10-NEXT: s_sub_i32 s10, 64, s12
; GFX10-NEXT: s_cmp_lt_u32 s12, 64
; GFX10-NEXT: s_cselect_b32 s15, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s12, 0
; GFX10-NEXT: s_cselect_b32 s16, 1, 0
; GFX10-NEXT: s_lshr_b64 s[0:1], s[4:5], s12
; GFX10-NEXT: s_lshl_b64 s[10:11], s[6:7], s10
; GFX10-NEXT: s_lshr_b64 s[12:13], s[6:7], s12
; GFX10-NEXT: s_or_b64 s[0:1], s[0:1], s[10:11]
; GFX10-NEXT: s_lshr_b64 s[6:7], s[6:7], s14
; GFX10-NEXT: s_cmp_lg_u32 s15, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[6:7]
; GFX10-NEXT: s_cmp_lg_u32 s16, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[4:5], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s15, 0
; GFX10-NEXT: s_cselect_b64 s[4:5], s[12:13], 0
; GFX10-NEXT: s_or_b64 s[0:1], s[8:9], s[0:1]
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[4:5]
; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt)
ret i128 %result
}
define i128 @v_fshr_i128(i128 %lhs, i128 %rhs, i128 %amt) {
; GFX6-LABEL: v_fshr_i128:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_movk_i32 s4, 0x7f
; GFX6-NEXT: v_and_b32_e32 v14, s4, v8
; GFX6-NEXT: v_xor_b32_e32 v8, -1, v8
; GFX6-NEXT: v_and_b32_e32 v15, s4, v8
; GFX6-NEXT: s_sub_i32 s5, 64, 1
; GFX6-NEXT: s_sub_i32 s4, 1, 64
; GFX6-NEXT: s_cmp_lt_u32 1, 64
; GFX6-NEXT: s_cselect_b32 s6, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 1, 0
; GFX6-NEXT: v_lshr_b64 v[8:9], v[0:1], s5
; GFX6-NEXT: v_lshl_b64 v[10:11], v[2:3], 1
; GFX6-NEXT: v_lshl_b64 v[12:13], v[0:1], 1
; GFX6-NEXT: s_cselect_b32 s7, 1, 0
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], s4
; GFX6-NEXT: s_and_b32 s4, 1, s6
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: v_or_b32_e32 v10, v8, v10
; GFX6-NEXT: v_or_b32_e32 v11, v9, v11
; GFX6-NEXT: s_and_b32 s4, 1, s7
; GFX6-NEXT: v_cndmask_b32_e32 v8, 0, v12, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v9, 0, v13, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 64, v15
; GFX6-NEXT: v_lshr_b64 v[2:3], v[8:9], v2
; GFX6-NEXT: v_lshl_b64 v[10:11], v[0:1], v15
; GFX6-NEXT: v_subrev_i32_e32 v16, vcc, 64, v15
; GFX6-NEXT: v_lshl_b64 v[12:13], v[8:9], v15
; GFX6-NEXT: v_or_b32_e32 v10, v2, v10
; GFX6-NEXT: v_or_b32_e32 v11, v3, v11
; GFX6-NEXT: v_lshl_b64 v[2:3], v[8:9], v16
; GFX6-NEXT: v_cmp_gt_u32_e32 vcc, 64, v15
; GFX6-NEXT: v_cndmask_b32_e32 v12, 0, v12, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v13, 0, v13, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v3, v3, v11, vcc
; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, 0, v15
; GFX6-NEXT: v_cndmask_b32_e32 v10, v2, v0, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v11, v3, v1, vcc
; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 64, v14
; GFX6-NEXT: v_lshr_b64 v[0:1], v[4:5], v14
; GFX6-NEXT: v_lshl_b64 v[2:3], v[6:7], v2
; GFX6-NEXT: v_subrev_i32_e32 v15, vcc, 64, v14
; GFX6-NEXT: v_or_b32_e32 v2, v0, v2
; GFX6-NEXT: v_or_b32_e32 v3, v1, v3
; GFX6-NEXT: v_lshr_b64 v[0:1], v[6:7], v15
; GFX6-NEXT: v_cmp_gt_u32_e32 vcc, 64, v14
; GFX6-NEXT: v_lshr_b64 v[8:9], v[6:7], v14
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX6-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v14
; GFX6-NEXT: v_cndmask_b32_e64 v0, v0, v4, s[4:5]
; GFX6-NEXT: v_cndmask_b32_e64 v1, v1, v5, s[4:5]
; GFX6-NEXT: v_cndmask_b32_e32 v2, 0, v8, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v3, 0, v9, vcc
; GFX6-NEXT: v_or_b32_e32 v0, v12, v0
; GFX6-NEXT: v_or_b32_e32 v1, v13, v1
; GFX6-NEXT: v_or_b32_e32 v2, v10, v2
; GFX6-NEXT: v_or_b32_e32 v3, v11, v3
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i128:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_movk_i32 s4, 0x7f
; GFX8-NEXT: v_and_b32_e32 v14, s4, v8
; GFX8-NEXT: v_xor_b32_e32 v8, -1, v8
; GFX8-NEXT: v_and_b32_e32 v15, s4, v8
; GFX8-NEXT: s_sub_i32 s5, 64, 1
; GFX8-NEXT: s_sub_i32 s4, 1, 64
; GFX8-NEXT: s_cmp_lt_u32 1, 64
; GFX8-NEXT: s_cselect_b32 s6, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 1, 0
; GFX8-NEXT: v_lshrrev_b64 v[8:9], s5, v[0:1]
; GFX8-NEXT: v_lshlrev_b64 v[10:11], 1, v[2:3]
; GFX8-NEXT: v_lshlrev_b64 v[12:13], 1, v[0:1]
; GFX8-NEXT: s_cselect_b32 s7, 1, 0
; GFX8-NEXT: v_lshlrev_b64 v[0:1], s4, v[0:1]
; GFX8-NEXT: s_and_b32 s4, 1, s6
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: v_or_b32_e32 v10, v8, v10
; GFX8-NEXT: v_or_b32_e32 v11, v9, v11
; GFX8-NEXT: s_and_b32 s4, 1, s7
; GFX8-NEXT: v_cndmask_b32_e32 v8, 0, v12, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v9, 0, v13, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX8-NEXT: v_sub_u32_e32 v2, vcc, 64, v15
; GFX8-NEXT: v_lshrrev_b64 v[2:3], v2, v[8:9]
; GFX8-NEXT: v_lshlrev_b64 v[10:11], v15, v[0:1]
; GFX8-NEXT: v_subrev_u32_e32 v16, vcc, 64, v15
; GFX8-NEXT: v_lshlrev_b64 v[12:13], v15, v[8:9]
; GFX8-NEXT: v_or_b32_e32 v10, v2, v10
; GFX8-NEXT: v_or_b32_e32 v11, v3, v11
; GFX8-NEXT: v_lshlrev_b64 v[2:3], v16, v[8:9]
; GFX8-NEXT: v_cmp_gt_u32_e32 vcc, 64, v15
; GFX8-NEXT: v_cndmask_b32_e32 v12, 0, v12, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v13, 0, v13, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v3, v3, v11, vcc
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v15
; GFX8-NEXT: v_cndmask_b32_e32 v10, v2, v0, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v11, v3, v1, vcc
; GFX8-NEXT: v_sub_u32_e32 v2, vcc, 64, v14
; GFX8-NEXT: v_lshrrev_b64 v[0:1], v14, v[4:5]
; GFX8-NEXT: v_lshlrev_b64 v[2:3], v2, v[6:7]
; GFX8-NEXT: v_subrev_u32_e32 v15, vcc, 64, v14
; GFX8-NEXT: v_or_b32_e32 v2, v0, v2
; GFX8-NEXT: v_or_b32_e32 v3, v1, v3
; GFX8-NEXT: v_lshrrev_b64 v[0:1], v15, v[6:7]
; GFX8-NEXT: v_cmp_gt_u32_e32 vcc, 64, v14
; GFX8-NEXT: v_lshrrev_b64 v[8:9], v14, v[6:7]
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX8-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v14
; GFX8-NEXT: v_cndmask_b32_e64 v0, v0, v4, s[4:5]
; GFX8-NEXT: v_cndmask_b32_e64 v1, v1, v5, s[4:5]
; GFX8-NEXT: v_cndmask_b32_e32 v2, 0, v8, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v3, 0, v9, vcc
; GFX8-NEXT: v_or_b32_e32 v0, v12, v0
; GFX8-NEXT: v_or_b32_e32 v1, v13, v1
; GFX8-NEXT: v_or_b32_e32 v2, v10, v2
; GFX8-NEXT: v_or_b32_e32 v3, v11, v3
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i128:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_movk_i32 s4, 0x7f
; GFX9-NEXT: v_and_b32_e32 v14, s4, v8
; GFX9-NEXT: v_xor_b32_e32 v8, -1, v8
; GFX9-NEXT: v_and_b32_e32 v15, s4, v8
; GFX9-NEXT: s_sub_i32 s5, 64, 1
; GFX9-NEXT: s_sub_i32 s4, 1, 64
; GFX9-NEXT: s_cmp_lt_u32 1, 64
; GFX9-NEXT: s_cselect_b32 s6, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 1, 0
; GFX9-NEXT: v_lshrrev_b64 v[8:9], s5, v[0:1]
; GFX9-NEXT: v_lshlrev_b64 v[10:11], 1, v[2:3]
; GFX9-NEXT: v_lshlrev_b64 v[12:13], 1, v[0:1]
; GFX9-NEXT: s_cselect_b32 s7, 1, 0
; GFX9-NEXT: v_lshlrev_b64 v[0:1], s4, v[0:1]
; GFX9-NEXT: s_and_b32 s4, 1, s6
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: v_or_b32_e32 v10, v8, v10
; GFX9-NEXT: v_or_b32_e32 v11, v9, v11
; GFX9-NEXT: s_and_b32 s4, 1, s7
; GFX9-NEXT: v_cndmask_b32_e32 v8, 0, v12, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v9, 0, v13, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX9-NEXT: v_sub_u32_e32 v2, 64, v15
; GFX9-NEXT: v_lshrrev_b64 v[2:3], v2, v[8:9]
; GFX9-NEXT: v_lshlrev_b64 v[10:11], v15, v[0:1]
; GFX9-NEXT: v_subrev_u32_e32 v16, 64, v15
; GFX9-NEXT: v_lshlrev_b64 v[12:13], v15, v[8:9]
; GFX9-NEXT: v_or_b32_e32 v10, v2, v10
; GFX9-NEXT: v_or_b32_e32 v11, v3, v11
; GFX9-NEXT: v_lshlrev_b64 v[2:3], v16, v[8:9]
; GFX9-NEXT: v_cmp_gt_u32_e32 vcc, 64, v15
; GFX9-NEXT: v_cndmask_b32_e32 v12, 0, v12, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v13, 0, v13, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, v3, v11, vcc
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v15
; GFX9-NEXT: v_cndmask_b32_e32 v10, v2, v0, vcc
; GFX9-NEXT: v_sub_u32_e32 v2, 64, v14
; GFX9-NEXT: v_cndmask_b32_e32 v11, v3, v1, vcc
; GFX9-NEXT: v_lshrrev_b64 v[0:1], v14, v[4:5]
; GFX9-NEXT: v_lshlrev_b64 v[2:3], v2, v[6:7]
; GFX9-NEXT: v_subrev_u32_e32 v15, 64, v14
; GFX9-NEXT: v_or_b32_e32 v2, v0, v2
; GFX9-NEXT: v_or_b32_e32 v3, v1, v3
; GFX9-NEXT: v_lshrrev_b64 v[0:1], v15, v[6:7]
; GFX9-NEXT: v_cmp_gt_u32_e32 vcc, 64, v14
; GFX9-NEXT: v_lshrrev_b64 v[8:9], v14, v[6:7]
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX9-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v14
; GFX9-NEXT: v_cndmask_b32_e64 v0, v0, v4, s[4:5]
; GFX9-NEXT: v_cndmask_b32_e64 v1, v1, v5, s[4:5]
; GFX9-NEXT: v_cndmask_b32_e32 v2, 0, v8, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, 0, v9, vcc
; GFX9-NEXT: v_or_b32_e32 v0, v12, v0
; GFX9-NEXT: v_or_b32_e32 v1, v13, v1
; GFX9-NEXT: v_or_b32_e32 v2, v10, v2
; GFX9-NEXT: v_or_b32_e32 v3, v11, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i128:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: s_sub_i32 s4, 64, 1
; GFX10-NEXT: s_sub_i32 s6, 1, 64
; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_lshrrev_b64 v[9:10], s4, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[11:12], 1, v[2:3]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[13:14], 1, v[0:1]
; GFX10-NEXT: s_cselect_b32 s7, 1, 0
; GFX10-NEXT: s_and_b32 s4, 1, vcc_lo
; GFX10-NEXT: v_lshlrev_b64 v[0:1], s6, v[0:1]
; GFX10-NEXT: v_or_b32_e32 v11, v9, v11
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4
; GFX10-NEXT: v_or_b32_e32 v10, v10, v12
; GFX10-NEXT: v_xor_b32_e32 v15, -1, v8
; GFX10-NEXT: s_movk_i32 s5, 0x7f
; GFX10-NEXT: s_and_b32 s6, 1, s7
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v11, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v10, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s6
; GFX10-NEXT: v_and_b32_e32 v19, s5, v15
; GFX10-NEXT: v_cndmask_b32_e32 v9, 0, v13, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v10, 0, v14, vcc_lo
; GFX10-NEXT: v_and_b32_e32 v20, s5, v8
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v3, s4
; GFX10-NEXT: v_sub_nc_u32_e32 v2, 64, v19
; GFX10-NEXT: v_subrev_nc_u32_e32 v8, 64, v19
; GFX10-NEXT: v_sub_nc_u32_e32 v17, 64, v20
; GFX10-NEXT: v_lshlrev_b64 v[13:14], v19, v[9:10]
; GFX10-NEXT: v_lshlrev_b64 v[11:12], v19, v[0:1]
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v2, v[9:10]
; GFX10-NEXT: v_lshlrev_b64 v[8:9], v8, v[9:10]
; GFX10-NEXT: v_lshrrev_b64 v[15:16], v20, v[4:5]
; GFX10-NEXT: v_lshlrev_b64 v[17:18], v17, v[6:7]
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v19
; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v20
; GFX10-NEXT: v_or_b32_e32 v11, v2, v11
; GFX10-NEXT: v_subrev_nc_u32_e32 v2, 64, v20
; GFX10-NEXT: v_or_b32_e32 v10, v3, v12
; GFX10-NEXT: v_cndmask_b32_e32 v12, 0, v13, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v13, v15, v17
; GFX10-NEXT: v_cndmask_b32_e32 v8, v8, v11, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v2, v[6:7]
; GFX10-NEXT: v_cndmask_b32_e32 v9, v9, v10, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v10, v16, v18
; GFX10-NEXT: v_lshrrev_b64 v[6:7], v20, v[6:7]
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v19
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v20
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v13, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v10, s4
; GFX10-NEXT: v_cndmask_b32_e32 v10, 0, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v8, v8, v0, s6
; GFX10-NEXT: v_cndmask_b32_e64 v9, v9, v1, s6
; GFX10-NEXT: v_cndmask_b32_e64 v0, v2, v4, s5
; GFX10-NEXT: v_cndmask_b32_e64 v1, v3, v5, s5
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v6, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v7, s4
; GFX10-NEXT: v_or_b32_e32 v0, v12, v0
; GFX10-NEXT: v_or_b32_e32 v1, v10, v1
; GFX10-NEXT: v_or_b32_e32 v2, v8, v2
; GFX10-NEXT: v_or_b32_e32 v3, v9, v3
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt)
ret i128 %result
}
define amdgpu_ps <4 x float> @v_fshr_i128_ssv(i128 inreg %lhs, i128 inreg %rhs, i128 %amt) {
; GFX6-LABEL: v_fshr_i128_ssv:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_sub_i32 s14, 1, 64
; GFX6-NEXT: s_sub_i32 s10, 64, 1
; GFX6-NEXT: s_cmp_lt_u32 1, 64
; GFX6-NEXT: s_movk_i32 s8, 0x7f
; GFX6-NEXT: s_cselect_b32 s15, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 1, 0
; GFX6-NEXT: s_cselect_b32 s16, 1, 0
; GFX6-NEXT: v_and_b32_e32 v6, s8, v0
; GFX6-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX6-NEXT: s_lshr_b64 s[10:11], s[0:1], s10
; GFX6-NEXT: s_lshl_b64 s[12:13], s[2:3], 1
; GFX6-NEXT: v_and_b32_e32 v7, s8, v0
; GFX6-NEXT: s_lshl_b64 s[8:9], s[0:1], 1
; GFX6-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], s14
; GFX6-NEXT: s_cmp_lg_u32 s15, 0
; GFX6-NEXT: s_cselect_b64 s[8:9], s[8:9], 0
; GFX6-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX6-NEXT: s_cmp_lg_u32 s16, 0
; GFX6-NEXT: v_sub_i32_e32 v0, vcc, 64, v7
; GFX6-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX6-NEXT: v_lshr_b64 v[0:1], s[8:9], v0
; GFX6-NEXT: v_lshl_b64 v[2:3], s[0:1], v7
; GFX6-NEXT: v_subrev_i32_e32 v8, vcc, 64, v7
; GFX6-NEXT: v_lshl_b64 v[4:5], s[8:9], v7
; GFX6-NEXT: v_or_b32_e32 v2, v0, v2
; GFX6-NEXT: v_or_b32_e32 v3, v1, v3
; GFX6-NEXT: v_lshl_b64 v[0:1], s[8:9], v8
; GFX6-NEXT: v_cmp_gt_u32_e32 vcc, 64, v7
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v8, 0, v4, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v9, 0, v5, vcc
; GFX6-NEXT: v_mov_b32_e32 v2, s0
; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, 0, v7
; GFX6-NEXT: v_mov_b32_e32 v3, s1
; GFX6-NEXT: v_cndmask_b32_e32 v7, v0, v2, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v10, v1, v3, vcc
; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 64, v6
; GFX6-NEXT: v_lshr_b64 v[0:1], s[4:5], v6
; GFX6-NEXT: v_lshl_b64 v[2:3], s[6:7], v2
; GFX6-NEXT: v_subrev_i32_e32 v11, vcc, 64, v6
; GFX6-NEXT: v_or_b32_e32 v2, v0, v2
; GFX6-NEXT: v_or_b32_e32 v3, v1, v3
; GFX6-NEXT: v_lshr_b64 v[0:1], s[6:7], v11
; GFX6-NEXT: v_cmp_gt_u32_e32 vcc, 64, v6
; GFX6-NEXT: v_lshr_b64 v[4:5], s[6:7], v6
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX6-NEXT: v_mov_b32_e32 v2, s4
; GFX6-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v6
; GFX6-NEXT: v_mov_b32_e32 v3, s5
; GFX6-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[0:1]
; GFX6-NEXT: v_cndmask_b32_e64 v1, v1, v3, s[0:1]
; GFX6-NEXT: v_cndmask_b32_e32 v2, 0, v4, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v3, 0, v5, vcc
; GFX6-NEXT: v_or_b32_e32 v0, v8, v0
; GFX6-NEXT: v_or_b32_e32 v1, v9, v1
; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_or_b32_e32 v3, v10, v3
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i128_ssv:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_sub_i32 s14, 1, 64
; GFX8-NEXT: s_sub_i32 s10, 64, 1
; GFX8-NEXT: s_cmp_lt_u32 1, 64
; GFX8-NEXT: s_movk_i32 s8, 0x7f
; GFX8-NEXT: s_cselect_b32 s15, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 1, 0
; GFX8-NEXT: s_cselect_b32 s16, 1, 0
; GFX8-NEXT: v_and_b32_e32 v6, s8, v0
; GFX8-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX8-NEXT: s_lshr_b64 s[10:11], s[0:1], s10
; GFX8-NEXT: s_lshl_b64 s[12:13], s[2:3], 1
; GFX8-NEXT: v_and_b32_e32 v7, s8, v0
; GFX8-NEXT: s_lshl_b64 s[8:9], s[0:1], 1
; GFX8-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], s14
; GFX8-NEXT: s_cmp_lg_u32 s15, 0
; GFX8-NEXT: s_cselect_b64 s[8:9], s[8:9], 0
; GFX8-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX8-NEXT: s_cmp_lg_u32 s16, 0
; GFX8-NEXT: v_sub_u32_e32 v0, vcc, 64, v7
; GFX8-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX8-NEXT: v_lshrrev_b64 v[0:1], v0, s[8:9]
; GFX8-NEXT: v_lshlrev_b64 v[2:3], v7, s[0:1]
; GFX8-NEXT: v_subrev_u32_e32 v8, vcc, 64, v7
; GFX8-NEXT: v_lshlrev_b64 v[4:5], v7, s[8:9]
; GFX8-NEXT: v_or_b32_e32 v2, v0, v2
; GFX8-NEXT: v_or_b32_e32 v3, v1, v3
; GFX8-NEXT: v_lshlrev_b64 v[0:1], v8, s[8:9]
; GFX8-NEXT: v_cmp_gt_u32_e32 vcc, 64, v7
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v8, 0, v4, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v9, 0, v5, vcc
; GFX8-NEXT: v_mov_b32_e32 v2, s0
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v7
; GFX8-NEXT: v_mov_b32_e32 v3, s1
; GFX8-NEXT: v_cndmask_b32_e32 v7, v0, v2, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v10, v1, v3, vcc
; GFX8-NEXT: v_sub_u32_e32 v2, vcc, 64, v6
; GFX8-NEXT: v_lshrrev_b64 v[0:1], v6, s[4:5]
; GFX8-NEXT: v_lshlrev_b64 v[2:3], v2, s[6:7]
; GFX8-NEXT: v_subrev_u32_e32 v11, vcc, 64, v6
; GFX8-NEXT: v_or_b32_e32 v2, v0, v2
; GFX8-NEXT: v_or_b32_e32 v3, v1, v3
; GFX8-NEXT: v_lshrrev_b64 v[0:1], v11, s[6:7]
; GFX8-NEXT: v_cmp_gt_u32_e32 vcc, 64, v6
; GFX8-NEXT: v_lshrrev_b64 v[4:5], v6, s[6:7]
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX8-NEXT: v_mov_b32_e32 v2, s4
; GFX8-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v6
; GFX8-NEXT: v_mov_b32_e32 v3, s5
; GFX8-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[0:1]
; GFX8-NEXT: v_cndmask_b32_e64 v1, v1, v3, s[0:1]
; GFX8-NEXT: v_cndmask_b32_e32 v2, 0, v4, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v3, 0, v5, vcc
; GFX8-NEXT: v_or_b32_e32 v0, v8, v0
; GFX8-NEXT: v_or_b32_e32 v1, v9, v1
; GFX8-NEXT: v_or_b32_e32 v2, v7, v2
; GFX8-NEXT: v_or_b32_e32 v3, v10, v3
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i128_ssv:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_sub_i32 s14, 1, 64
; GFX9-NEXT: s_sub_i32 s10, 64, 1
; GFX9-NEXT: s_cmp_lt_u32 1, 64
; GFX9-NEXT: s_movk_i32 s8, 0x7f
; GFX9-NEXT: s_cselect_b32 s15, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 1, 0
; GFX9-NEXT: v_and_b32_e32 v6, s8, v0
; GFX9-NEXT: s_cselect_b32 s16, 1, 0
; GFX9-NEXT: v_xor_b32_e32 v0, -1, v0
; GFX9-NEXT: s_lshr_b64 s[10:11], s[0:1], s10
; GFX9-NEXT: s_lshl_b64 s[12:13], s[2:3], 1
; GFX9-NEXT: v_and_b32_e32 v7, s8, v0
; GFX9-NEXT: s_lshl_b64 s[8:9], s[0:1], 1
; GFX9-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], s14
; GFX9-NEXT: s_cmp_lg_u32 s15, 0
; GFX9-NEXT: s_cselect_b64 s[8:9], s[8:9], 0
; GFX9-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX9-NEXT: s_cmp_lg_u32 s16, 0
; GFX9-NEXT: v_sub_u32_e32 v0, 64, v7
; GFX9-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX9-NEXT: v_lshrrev_b64 v[0:1], v0, s[8:9]
; GFX9-NEXT: v_lshlrev_b64 v[2:3], v7, s[0:1]
; GFX9-NEXT: v_subrev_u32_e32 v8, 64, v7
; GFX9-NEXT: v_lshlrev_b64 v[4:5], v7, s[8:9]
; GFX9-NEXT: v_or_b32_e32 v2, v0, v2
; GFX9-NEXT: v_or_b32_e32 v3, v1, v3
; GFX9-NEXT: v_lshlrev_b64 v[0:1], v8, s[8:9]
; GFX9-NEXT: v_cmp_gt_u32_e32 vcc, 64, v7
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v8, 0, v4, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v9, 0, v5, vcc
; GFX9-NEXT: v_mov_b32_e32 v2, s0
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v7
; GFX9-NEXT: v_mov_b32_e32 v3, s1
; GFX9-NEXT: v_cndmask_b32_e32 v7, v0, v2, vcc
; GFX9-NEXT: v_sub_u32_e32 v2, 64, v6
; GFX9-NEXT: v_cndmask_b32_e32 v10, v1, v3, vcc
; GFX9-NEXT: v_lshrrev_b64 v[0:1], v6, s[4:5]
; GFX9-NEXT: v_lshlrev_b64 v[2:3], v2, s[6:7]
; GFX9-NEXT: v_subrev_u32_e32 v11, 64, v6
; GFX9-NEXT: v_or_b32_e32 v2, v0, v2
; GFX9-NEXT: v_or_b32_e32 v3, v1, v3
; GFX9-NEXT: v_lshrrev_b64 v[0:1], v11, s[6:7]
; GFX9-NEXT: v_cmp_gt_u32_e32 vcc, 64, v6
; GFX9-NEXT: v_lshrrev_b64 v[4:5], v6, s[6:7]
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX9-NEXT: v_mov_b32_e32 v2, s4
; GFX9-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v6
; GFX9-NEXT: v_mov_b32_e32 v3, s5
; GFX9-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[0:1]
; GFX9-NEXT: v_cndmask_b32_e64 v1, v1, v3, s[0:1]
; GFX9-NEXT: v_cndmask_b32_e32 v2, 0, v4, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, 0, v5, vcc
; GFX9-NEXT: v_or_b32_e32 v0, v8, v0
; GFX9-NEXT: v_or_b32_e32 v1, v9, v1
; GFX9-NEXT: v_or_b32_e32 v2, v7, v2
; GFX9-NEXT: v_or_b32_e32 v3, v10, v3
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i128_ssv:
; GFX10: ; %bb.0:
; GFX10-NEXT: v_xor_b32_e32 v1, -1, v0
; GFX10-NEXT: s_sub_i32 s14, 1, 64
; GFX10-NEXT: s_sub_i32 s9, 64, 1
; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: s_movk_i32 s8, 0x7f
; GFX10-NEXT: s_cselect_b32 s15, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_and_b32_e32 v13, s8, v1
; GFX10-NEXT: v_and_b32_e32 v12, s8, v0
; GFX10-NEXT: s_cselect_b32 s16, 1, 0
; GFX10-NEXT: s_lshr_b64 s[8:9], s[0:1], s9
; GFX10-NEXT: s_lshl_b64 s[10:11], s[2:3], 1
; GFX10-NEXT: s_lshl_b64 s[12:13], s[0:1], 1
; GFX10-NEXT: s_or_b64 s[8:9], s[8:9], s[10:11]
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s14
; GFX10-NEXT: s_cmp_lg_u32 s15, 0
; GFX10-NEXT: v_sub_nc_u32_e32 v0, 64, v13
; GFX10-NEXT: s_cselect_b64 s[10:11], s[12:13], 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[8:9], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s16, 0
; GFX10-NEXT: v_sub_nc_u32_e32 v8, 64, v12
; GFX10-NEXT: s_cselect_b64 s[8:9], s[2:3], s[0:1]
; GFX10-NEXT: v_lshrrev_b64 v[0:1], v0, s[10:11]
; GFX10-NEXT: v_lshlrev_b64 v[2:3], v13, s[8:9]
; GFX10-NEXT: v_subrev_nc_u32_e32 v10, 64, v13
; GFX10-NEXT: v_lshrrev_b64 v[6:7], v12, s[4:5]
; GFX10-NEXT: v_lshlrev_b64 v[8:9], v8, s[6:7]
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v13
; GFX10-NEXT: v_cmp_gt_u32_e64 s0, 64, v12
; GFX10-NEXT: v_or_b32_e32 v2, v0, v2
; GFX10-NEXT: v_subrev_nc_u32_e32 v0, 64, v12
; GFX10-NEXT: v_lshlrev_b64 v[10:11], v10, s[10:11]
; GFX10-NEXT: v_or_b32_e32 v3, v1, v3
; GFX10-NEXT: v_or_b32_e32 v6, v6, v8
; GFX10-NEXT: v_or_b32_e32 v7, v7, v9
; GFX10-NEXT: v_lshrrev_b64 v[0:1], v0, s[6:7]
; GFX10-NEXT: v_lshlrev_b64 v[4:5], v13, s[10:11]
; GFX10-NEXT: v_cndmask_b32_e32 v8, v10, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v10, v11, v3, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v12, s[6:7]
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 0, v12
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v7, s0
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v13
; GFX10-NEXT: v_cndmask_b32_e32 v4, 0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v5, 0, v5, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, s4, s1
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s5, s1
; GFX10-NEXT: v_cndmask_b32_e64 v6, v8, s8, s2
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v7, v10, s9, s2
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0
; GFX10-NEXT: v_or_b32_e32 v0, v4, v0
; GFX10-NEXT: v_or_b32_e32 v1, v5, v1
; GFX10-NEXT: v_or_b32_e32 v2, v6, v2
; GFX10-NEXT: v_or_b32_e32 v3, v7, v3
; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt)
%cast.result = bitcast i128 %result to <4 x float>
ret <4 x float> %cast.result
}
define amdgpu_ps <4 x float> @v_fshr_i128_svs(i128 inreg %lhs, i128 %rhs, i128 inreg %amt) {
; GFX6-LABEL: v_fshr_i128_svs:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_movk_i32 s6, 0x7f
; GFX6-NEXT: s_mov_b32 s7, 0
; GFX6-NEXT: s_and_b64 s[8:9], s[4:5], s[6:7]
; GFX6-NEXT: s_andn2_b64 s[4:5], s[6:7], s[4:5]
; GFX6-NEXT: s_sub_i32 s5, 1, 64
; GFX6-NEXT: s_sub_i32 s9, 64, 1
; GFX6-NEXT: s_cmp_lt_u32 1, 64
; GFX6-NEXT: s_cselect_b32 s14, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 1, 0
; GFX6-NEXT: s_cselect_b32 s15, 1, 0
; GFX6-NEXT: s_lshr_b64 s[10:11], s[0:1], s9
; GFX6-NEXT: s_lshl_b64 s[12:13], s[2:3], 1
; GFX6-NEXT: s_lshl_b64 s[6:7], s[0:1], 1
; GFX6-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], s5
; GFX6-NEXT: s_cmp_lg_u32 s14, 0
; GFX6-NEXT: s_cselect_b64 s[6:7], s[6:7], 0
; GFX6-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX6-NEXT: s_cmp_lg_u32 s15, 0
; GFX6-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX6-NEXT: s_sub_i32 s9, s4, 64
; GFX6-NEXT: s_sub_i32 s5, 64, s4
; GFX6-NEXT: s_cmp_lt_u32 s4, 64
; GFX6-NEXT: s_cselect_b32 s12, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s4, 0
; GFX6-NEXT: s_cselect_b32 s13, 1, 0
; GFX6-NEXT: s_lshl_b64 s[2:3], s[6:7], s4
; GFX6-NEXT: s_lshr_b64 s[10:11], s[6:7], s5
; GFX6-NEXT: s_lshl_b64 s[4:5], s[0:1], s4
; GFX6-NEXT: s_or_b64 s[4:5], s[10:11], s[4:5]
; GFX6-NEXT: s_lshl_b64 s[6:7], s[6:7], s9
; GFX6-NEXT: s_cmp_lg_u32 s12, 0
; GFX6-NEXT: s_cselect_b64 s[2:3], s[2:3], 0
; GFX6-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX6-NEXT: s_cmp_lg_u32 s13, 0
; GFX6-NEXT: s_cselect_b64 s[0:1], s[0:1], s[4:5]
; GFX6-NEXT: s_sub_i32 s5, 64, s8
; GFX6-NEXT: s_sub_i32 s4, s8, 64
; GFX6-NEXT: s_cmp_lt_u32 s8, 64
; GFX6-NEXT: s_cselect_b32 s6, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s8, 0
; GFX6-NEXT: v_lshr_b64 v[4:5], v[0:1], s8
; GFX6-NEXT: v_lshl_b64 v[6:7], v[2:3], s5
; GFX6-NEXT: v_lshr_b64 v[8:9], v[2:3], s8
; GFX6-NEXT: s_cselect_b32 s7, 1, 0
; GFX6-NEXT: v_lshr_b64 v[2:3], v[2:3], s4
; GFX6-NEXT: s_and_b32 s4, 1, s6
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: v_or_b32_e32 v4, v4, v6
; GFX6-NEXT: v_or_b32_e32 v5, v5, v7
; GFX6-NEXT: s_and_b32 s4, 1, s7
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: s_and_b32 s4, 1, s6
; GFX6-NEXT: v_cndmask_b32_e32 v0, v2, v0, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v1, v3, v1, vcc
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: v_cndmask_b32_e32 v2, 0, v8, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v3, 0, v9, vcc
; GFX6-NEXT: v_or_b32_e32 v0, s2, v0
; GFX6-NEXT: v_or_b32_e32 v1, s3, v1
; GFX6-NEXT: v_or_b32_e32 v2, s0, v2
; GFX6-NEXT: v_or_b32_e32 v3, s1, v3
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i128_svs:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_movk_i32 s6, 0x7f
; GFX8-NEXT: s_mov_b32 s7, 0
; GFX8-NEXT: s_and_b64 s[8:9], s[4:5], s[6:7]
; GFX8-NEXT: s_andn2_b64 s[4:5], s[6:7], s[4:5]
; GFX8-NEXT: s_sub_i32 s5, 1, 64
; GFX8-NEXT: s_sub_i32 s9, 64, 1
; GFX8-NEXT: s_cmp_lt_u32 1, 64
; GFX8-NEXT: s_cselect_b32 s14, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 1, 0
; GFX8-NEXT: s_cselect_b32 s15, 1, 0
; GFX8-NEXT: s_lshr_b64 s[10:11], s[0:1], s9
; GFX8-NEXT: s_lshl_b64 s[12:13], s[2:3], 1
; GFX8-NEXT: s_lshl_b64 s[6:7], s[0:1], 1
; GFX8-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], s5
; GFX8-NEXT: s_cmp_lg_u32 s14, 0
; GFX8-NEXT: s_cselect_b64 s[6:7], s[6:7], 0
; GFX8-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX8-NEXT: s_cmp_lg_u32 s15, 0
; GFX8-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX8-NEXT: s_sub_i32 s9, s4, 64
; GFX8-NEXT: s_sub_i32 s5, 64, s4
; GFX8-NEXT: s_cmp_lt_u32 s4, 64
; GFX8-NEXT: s_cselect_b32 s12, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s4, 0
; GFX8-NEXT: s_cselect_b32 s13, 1, 0
; GFX8-NEXT: s_lshl_b64 s[2:3], s[6:7], s4
; GFX8-NEXT: s_lshr_b64 s[10:11], s[6:7], s5
; GFX8-NEXT: s_lshl_b64 s[4:5], s[0:1], s4
; GFX8-NEXT: s_or_b64 s[4:5], s[10:11], s[4:5]
; GFX8-NEXT: s_lshl_b64 s[6:7], s[6:7], s9
; GFX8-NEXT: s_cmp_lg_u32 s12, 0
; GFX8-NEXT: s_cselect_b64 s[2:3], s[2:3], 0
; GFX8-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX8-NEXT: s_cmp_lg_u32 s13, 0
; GFX8-NEXT: s_cselect_b64 s[0:1], s[0:1], s[4:5]
; GFX8-NEXT: s_sub_i32 s5, 64, s8
; GFX8-NEXT: s_sub_i32 s4, s8, 64
; GFX8-NEXT: s_cmp_lt_u32 s8, 64
; GFX8-NEXT: s_cselect_b32 s6, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s8, 0
; GFX8-NEXT: v_lshrrev_b64 v[4:5], s8, v[0:1]
; GFX8-NEXT: v_lshlrev_b64 v[6:7], s5, v[2:3]
; GFX8-NEXT: v_lshrrev_b64 v[8:9], s8, v[2:3]
; GFX8-NEXT: s_cselect_b32 s7, 1, 0
; GFX8-NEXT: v_lshrrev_b64 v[2:3], s4, v[2:3]
; GFX8-NEXT: s_and_b32 s4, 1, s6
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: v_or_b32_e32 v4, v4, v6
; GFX8-NEXT: v_or_b32_e32 v5, v5, v7
; GFX8-NEXT: s_and_b32 s4, 1, s7
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: s_and_b32 s4, 1, s6
; GFX8-NEXT: v_cndmask_b32_e32 v0, v2, v0, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v1, v3, v1, vcc
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: v_cndmask_b32_e32 v2, 0, v8, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v3, 0, v9, vcc
; GFX8-NEXT: v_or_b32_e32 v0, s2, v0
; GFX8-NEXT: v_or_b32_e32 v1, s3, v1
; GFX8-NEXT: v_or_b32_e32 v2, s0, v2
; GFX8-NEXT: v_or_b32_e32 v3, s1, v3
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i128_svs:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_movk_i32 s6, 0x7f
; GFX9-NEXT: s_mov_b32 s7, 0
; GFX9-NEXT: s_and_b64 s[8:9], s[4:5], s[6:7]
; GFX9-NEXT: s_andn2_b64 s[4:5], s[6:7], s[4:5]
; GFX9-NEXT: s_sub_i32 s5, 1, 64
; GFX9-NEXT: s_sub_i32 s9, 64, 1
; GFX9-NEXT: s_cmp_lt_u32 1, 64
; GFX9-NEXT: s_cselect_b32 s14, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 1, 0
; GFX9-NEXT: s_cselect_b32 s15, 1, 0
; GFX9-NEXT: s_lshr_b64 s[10:11], s[0:1], s9
; GFX9-NEXT: s_lshl_b64 s[12:13], s[2:3], 1
; GFX9-NEXT: s_lshl_b64 s[6:7], s[0:1], 1
; GFX9-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], s5
; GFX9-NEXT: s_cmp_lg_u32 s14, 0
; GFX9-NEXT: s_cselect_b64 s[6:7], s[6:7], 0
; GFX9-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX9-NEXT: s_cmp_lg_u32 s15, 0
; GFX9-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX9-NEXT: s_sub_i32 s9, s4, 64
; GFX9-NEXT: s_sub_i32 s5, 64, s4
; GFX9-NEXT: s_cmp_lt_u32 s4, 64
; GFX9-NEXT: s_cselect_b32 s12, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s4, 0
; GFX9-NEXT: s_cselect_b32 s13, 1, 0
; GFX9-NEXT: s_lshl_b64 s[2:3], s[6:7], s4
; GFX9-NEXT: s_lshr_b64 s[10:11], s[6:7], s5
; GFX9-NEXT: s_lshl_b64 s[4:5], s[0:1], s4
; GFX9-NEXT: s_or_b64 s[4:5], s[10:11], s[4:5]
; GFX9-NEXT: s_lshl_b64 s[6:7], s[6:7], s9
; GFX9-NEXT: s_cmp_lg_u32 s12, 0
; GFX9-NEXT: s_cselect_b64 s[2:3], s[2:3], 0
; GFX9-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX9-NEXT: s_cmp_lg_u32 s13, 0
; GFX9-NEXT: s_cselect_b64 s[0:1], s[0:1], s[4:5]
; GFX9-NEXT: s_sub_i32 s5, 64, s8
; GFX9-NEXT: s_sub_i32 s4, s8, 64
; GFX9-NEXT: s_cmp_lt_u32 s8, 64
; GFX9-NEXT: s_cselect_b32 s6, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s8, 0
; GFX9-NEXT: v_lshrrev_b64 v[4:5], s8, v[0:1]
; GFX9-NEXT: v_lshlrev_b64 v[6:7], s5, v[2:3]
; GFX9-NEXT: v_lshrrev_b64 v[8:9], s8, v[2:3]
; GFX9-NEXT: s_cselect_b32 s7, 1, 0
; GFX9-NEXT: v_lshrrev_b64 v[2:3], s4, v[2:3]
; GFX9-NEXT: s_and_b32 s4, 1, s6
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: v_or_b32_e32 v4, v4, v6
; GFX9-NEXT: v_or_b32_e32 v5, v5, v7
; GFX9-NEXT: s_and_b32 s4, 1, s7
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, v3, v5, vcc
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: s_and_b32 s4, 1, s6
; GFX9-NEXT: v_cndmask_b32_e32 v0, v2, v0, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v3, v1, vcc
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: v_cndmask_b32_e32 v2, 0, v8, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, 0, v9, vcc
; GFX9-NEXT: v_or_b32_e32 v0, s2, v0
; GFX9-NEXT: v_or_b32_e32 v1, s3, v1
; GFX9-NEXT: v_or_b32_e32 v2, s0, v2
; GFX9-NEXT: v_or_b32_e32 v3, s1, v3
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i128_svs:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_movk_i32 s6, 0x7f
; GFX10-NEXT: s_mov_b32 s7, 0
; GFX10-NEXT: s_and_b64 s[8:9], s[4:5], s[6:7]
; GFX10-NEXT: s_andn2_b64 s[4:5], s[6:7], s[4:5]
; GFX10-NEXT: s_sub_i32 s5, 1, 64
; GFX10-NEXT: s_sub_i32 s6, 64, 1
; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_lshrrev_b64 v[4:5], s8, v[0:1]
; GFX10-NEXT: s_cselect_b32 s9, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: s_cselect_b32 s14, 1, 0
; GFX10-NEXT: s_lshr_b64 s[6:7], s[0:1], s6
; GFX10-NEXT: s_lshl_b64 s[10:11], s[2:3], 1
; GFX10-NEXT: s_lshl_b64 s[12:13], s[0:1], 1
; GFX10-NEXT: s_or_b64 s[6:7], s[6:7], s[10:11]
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s5
; GFX10-NEXT: s_cmp_lg_u32 s9, 0
; GFX10-NEXT: s_cselect_b64 s[10:11], s[12:13], 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[6:7], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s14, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX10-NEXT: s_sub_i32 s9, s4, 64
; GFX10-NEXT: s_sub_i32 s2, 64, s4
; GFX10-NEXT: s_cmp_lt_u32 s4, 64
; GFX10-NEXT: s_cselect_b32 s12, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s4, 0
; GFX10-NEXT: s_cselect_b32 s13, 1, 0
; GFX10-NEXT: s_lshl_b64 s[6:7], s[0:1], s4
; GFX10-NEXT: s_lshr_b64 s[2:3], s[10:11], s2
; GFX10-NEXT: s_lshl_b64 s[4:5], s[10:11], s4
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[6:7]
; GFX10-NEXT: s_lshl_b64 s[6:7], s[10:11], s9
; GFX10-NEXT: s_cmp_lg_u32 s12, 0
; GFX10-NEXT: s_cselect_b64 s[4:5], s[4:5], 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[6:7]
; GFX10-NEXT: s_cmp_lg_u32 s13, 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[0:1], s[2:3]
; GFX10-NEXT: s_sub_i32 s0, 64, s8
; GFX10-NEXT: v_lshlrev_b64 v[6:7], s0, v[2:3]
; GFX10-NEXT: s_sub_i32 s0, s8, 64
; GFX10-NEXT: s_cmp_lt_u32 s8, 64
; GFX10-NEXT: v_lshrrev_b64 v[8:9], s0, v[2:3]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s8, 0
; GFX10-NEXT: v_or_b32_e32 v4, v4, v6
; GFX10-NEXT: v_or_b32_e32 v5, v5, v7
; GFX10-NEXT: s_cselect_b32 s0, 1, 0
; GFX10-NEXT: s_and_b32 s1, 1, vcc_lo
; GFX10-NEXT: s_and_b32 s0, 1, s0
; GFX10-NEXT: v_lshrrev_b64 v[2:3], s8, v[2:3]
; GFX10-NEXT: v_cndmask_b32_e32 v4, v8, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v5, v9, v5, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s1
; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, v2, s0
; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, v3, s0
; GFX10-NEXT: v_or_b32_e32 v0, s4, v0
; GFX10-NEXT: v_or_b32_e32 v1, s5, v1
; GFX10-NEXT: v_or_b32_e32 v2, s2, v2
; GFX10-NEXT: v_or_b32_e32 v3, s3, v3
; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt)
%cast.result = bitcast i128 %result to <4 x float>
ret <4 x float> %cast.result
}
define amdgpu_ps <4 x float> @v_fshr_i128_vss(i128 %lhs, i128 inreg %rhs, i128 inreg %amt) {
; GFX6-LABEL: v_fshr_i128_vss:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_movk_i32 s6, 0x7f
; GFX6-NEXT: s_mov_b32 s7, 0
; GFX6-NEXT: s_and_b64 s[8:9], s[4:5], s[6:7]
; GFX6-NEXT: s_andn2_b64 s[4:5], s[6:7], s[4:5]
; GFX6-NEXT: s_sub_i32 s6, 64, 1
; GFX6-NEXT: s_sub_i32 s5, 1, 64
; GFX6-NEXT: s_cmp_lt_u32 1, 64
; GFX6-NEXT: s_cselect_b32 s7, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 1, 0
; GFX6-NEXT: v_lshr_b64 v[4:5], v[0:1], s6
; GFX6-NEXT: v_lshl_b64 v[6:7], v[2:3], 1
; GFX6-NEXT: s_cselect_b32 s9, 1, 0
; GFX6-NEXT: v_lshl_b64 v[8:9], v[0:1], 1
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], s5
; GFX6-NEXT: s_and_b32 s5, 1, s7
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s5
; GFX6-NEXT: v_or_b32_e32 v6, v4, v6
; GFX6-NEXT: v_or_b32_e32 v7, v5, v7
; GFX6-NEXT: s_and_b32 s5, 1, s9
; GFX6-NEXT: v_cndmask_b32_e32 v4, 0, v8, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v5, 0, v9, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s5
; GFX6-NEXT: s_sub_i32 s5, s4, 64
; GFX6-NEXT: s_sub_i32 s6, 64, s4
; GFX6-NEXT: s_cmp_lt_u32 s4, 64
; GFX6-NEXT: s_cselect_b32 s7, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s4, 0
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX6-NEXT: s_cselect_b32 s9, 1, 0
; GFX6-NEXT: v_lshr_b64 v[2:3], v[4:5], s6
; GFX6-NEXT: v_lshl_b64 v[6:7], v[0:1], s4
; GFX6-NEXT: v_lshl_b64 v[8:9], v[4:5], s4
; GFX6-NEXT: s_and_b32 s4, 1, s7
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: s_and_b32 s4, 1, s9
; GFX6-NEXT: s_sub_i32 s10, s8, 64
; GFX6-NEXT: s_sub_i32 s9, 64, s8
; GFX6-NEXT: s_cmp_lt_u32 s8, 64
; GFX6-NEXT: v_or_b32_e32 v6, v2, v6
; GFX6-NEXT: v_or_b32_e32 v7, v3, v7
; GFX6-NEXT: v_lshl_b64 v[2:3], v[4:5], s5
; GFX6-NEXT: s_cselect_b32 s11, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s8, 0
; GFX6-NEXT: s_cselect_b32 s12, 1, 0
; GFX6-NEXT: v_cndmask_b32_e32 v4, 0, v8, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v5, 0, v9, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v6, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v3, v3, v7, vcc
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: s_lshr_b64 s[4:5], s[2:3], s8
; GFX6-NEXT: s_lshr_b64 s[6:7], s[0:1], s8
; GFX6-NEXT: s_lshl_b64 s[8:9], s[2:3], s9
; GFX6-NEXT: s_or_b64 s[6:7], s[6:7], s[8:9]
; GFX6-NEXT: s_lshr_b64 s[2:3], s[2:3], s10
; GFX6-NEXT: s_cmp_lg_u32 s11, 0
; GFX6-NEXT: s_cselect_b64 s[2:3], s[6:7], s[2:3]
; GFX6-NEXT: s_cmp_lg_u32 s12, 0
; GFX6-NEXT: s_cselect_b64 s[0:1], s[0:1], s[2:3]
; GFX6-NEXT: s_cmp_lg_u32 s11, 0
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v3, v3, v1, vcc
; GFX6-NEXT: s_cselect_b64 s[2:3], s[4:5], 0
; GFX6-NEXT: v_or_b32_e32 v0, s0, v4
; GFX6-NEXT: v_or_b32_e32 v1, s1, v5
; GFX6-NEXT: v_or_b32_e32 v2, s2, v2
; GFX6-NEXT: v_or_b32_e32 v3, s3, v3
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: v_fshr_i128_vss:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_movk_i32 s6, 0x7f
; GFX8-NEXT: s_mov_b32 s7, 0
; GFX8-NEXT: s_and_b64 s[8:9], s[4:5], s[6:7]
; GFX8-NEXT: s_andn2_b64 s[4:5], s[6:7], s[4:5]
; GFX8-NEXT: s_sub_i32 s6, 64, 1
; GFX8-NEXT: s_sub_i32 s5, 1, 64
; GFX8-NEXT: s_cmp_lt_u32 1, 64
; GFX8-NEXT: s_cselect_b32 s7, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 1, 0
; GFX8-NEXT: v_lshrrev_b64 v[4:5], s6, v[0:1]
; GFX8-NEXT: v_lshlrev_b64 v[6:7], 1, v[2:3]
; GFX8-NEXT: s_cselect_b32 s9, 1, 0
; GFX8-NEXT: v_lshlrev_b64 v[8:9], 1, v[0:1]
; GFX8-NEXT: v_lshlrev_b64 v[0:1], s5, v[0:1]
; GFX8-NEXT: s_and_b32 s5, 1, s7
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s5
; GFX8-NEXT: v_or_b32_e32 v6, v4, v6
; GFX8-NEXT: v_or_b32_e32 v7, v5, v7
; GFX8-NEXT: s_and_b32 s5, 1, s9
; GFX8-NEXT: v_cndmask_b32_e32 v4, 0, v8, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v5, 0, v9, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s5
; GFX8-NEXT: s_sub_i32 s5, s4, 64
; GFX8-NEXT: s_sub_i32 s6, 64, s4
; GFX8-NEXT: s_cmp_lt_u32 s4, 64
; GFX8-NEXT: s_cselect_b32 s7, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s4, 0
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX8-NEXT: s_cselect_b32 s9, 1, 0
; GFX8-NEXT: v_lshrrev_b64 v[2:3], s6, v[4:5]
; GFX8-NEXT: v_lshlrev_b64 v[6:7], s4, v[0:1]
; GFX8-NEXT: v_lshlrev_b64 v[8:9], s4, v[4:5]
; GFX8-NEXT: s_and_b32 s4, 1, s7
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: s_and_b32 s4, 1, s9
; GFX8-NEXT: s_sub_i32 s10, s8, 64
; GFX8-NEXT: s_sub_i32 s9, 64, s8
; GFX8-NEXT: s_cmp_lt_u32 s8, 64
; GFX8-NEXT: v_or_b32_e32 v6, v2, v6
; GFX8-NEXT: v_or_b32_e32 v7, v3, v7
; GFX8-NEXT: v_lshlrev_b64 v[2:3], s5, v[4:5]
; GFX8-NEXT: s_cselect_b32 s11, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s8, 0
; GFX8-NEXT: s_cselect_b32 s12, 1, 0
; GFX8-NEXT: v_cndmask_b32_e32 v4, 0, v8, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v5, 0, v9, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v6, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v3, v3, v7, vcc
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: s_lshr_b64 s[4:5], s[2:3], s8
; GFX8-NEXT: s_lshr_b64 s[6:7], s[0:1], s8
; GFX8-NEXT: s_lshl_b64 s[8:9], s[2:3], s9
; GFX8-NEXT: s_or_b64 s[6:7], s[6:7], s[8:9]
; GFX8-NEXT: s_lshr_b64 s[2:3], s[2:3], s10
; GFX8-NEXT: s_cmp_lg_u32 s11, 0
; GFX8-NEXT: s_cselect_b64 s[2:3], s[6:7], s[2:3]
; GFX8-NEXT: s_cmp_lg_u32 s12, 0
; GFX8-NEXT: s_cselect_b64 s[0:1], s[0:1], s[2:3]
; GFX8-NEXT: s_cmp_lg_u32 s11, 0
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v3, v3, v1, vcc
; GFX8-NEXT: s_cselect_b64 s[2:3], s[4:5], 0
; GFX8-NEXT: v_or_b32_e32 v0, s0, v4
; GFX8-NEXT: v_or_b32_e32 v1, s1, v5
; GFX8-NEXT: v_or_b32_e32 v2, s2, v2
; GFX8-NEXT: v_or_b32_e32 v3, s3, v3
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: v_fshr_i128_vss:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_movk_i32 s6, 0x7f
; GFX9-NEXT: s_mov_b32 s7, 0
; GFX9-NEXT: s_and_b64 s[8:9], s[4:5], s[6:7]
; GFX9-NEXT: s_andn2_b64 s[4:5], s[6:7], s[4:5]
; GFX9-NEXT: s_sub_i32 s6, 64, 1
; GFX9-NEXT: s_sub_i32 s5, 1, 64
; GFX9-NEXT: s_cmp_lt_u32 1, 64
; GFX9-NEXT: s_cselect_b32 s7, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 1, 0
; GFX9-NEXT: v_lshrrev_b64 v[4:5], s6, v[0:1]
; GFX9-NEXT: v_lshlrev_b64 v[6:7], 1, v[2:3]
; GFX9-NEXT: s_cselect_b32 s9, 1, 0
; GFX9-NEXT: v_lshlrev_b64 v[8:9], 1, v[0:1]
; GFX9-NEXT: v_lshlrev_b64 v[0:1], s5, v[0:1]
; GFX9-NEXT: s_and_b32 s5, 1, s7
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s5
; GFX9-NEXT: v_or_b32_e32 v6, v4, v6
; GFX9-NEXT: v_or_b32_e32 v7, v5, v7
; GFX9-NEXT: s_and_b32 s5, 1, s9
; GFX9-NEXT: v_cndmask_b32_e32 v4, 0, v8, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v5, 0, v9, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s5
; GFX9-NEXT: s_sub_i32 s5, s4, 64
; GFX9-NEXT: s_sub_i32 s6, 64, s4
; GFX9-NEXT: s_cmp_lt_u32 s4, 64
; GFX9-NEXT: s_cselect_b32 s7, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s4, 0
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX9-NEXT: s_cselect_b32 s9, 1, 0
; GFX9-NEXT: v_lshrrev_b64 v[2:3], s6, v[4:5]
; GFX9-NEXT: v_lshlrev_b64 v[6:7], s4, v[0:1]
; GFX9-NEXT: v_lshlrev_b64 v[8:9], s4, v[4:5]
; GFX9-NEXT: s_and_b32 s4, 1, s7
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: s_and_b32 s4, 1, s9
; GFX9-NEXT: s_sub_i32 s10, s8, 64
; GFX9-NEXT: s_sub_i32 s9, 64, s8
; GFX9-NEXT: s_cmp_lt_u32 s8, 64
; GFX9-NEXT: v_or_b32_e32 v6, v2, v6
; GFX9-NEXT: v_or_b32_e32 v7, v3, v7
; GFX9-NEXT: v_lshlrev_b64 v[2:3], s5, v[4:5]
; GFX9-NEXT: s_cselect_b32 s11, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s8, 0
; GFX9-NEXT: s_cselect_b32 s12, 1, 0
; GFX9-NEXT: v_cndmask_b32_e32 v4, 0, v8, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v5, 0, v9, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v6, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, v3, v7, vcc
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: s_lshr_b64 s[4:5], s[2:3], s8
; GFX9-NEXT: s_lshr_b64 s[6:7], s[0:1], s8
; GFX9-NEXT: s_lshl_b64 s[8:9], s[2:3], s9
; GFX9-NEXT: s_or_b64 s[6:7], s[6:7], s[8:9]
; GFX9-NEXT: s_lshr_b64 s[2:3], s[2:3], s10
; GFX9-NEXT: s_cmp_lg_u32 s11, 0
; GFX9-NEXT: s_cselect_b64 s[2:3], s[6:7], s[2:3]
; GFX9-NEXT: s_cmp_lg_u32 s12, 0
; GFX9-NEXT: s_cselect_b64 s[0:1], s[0:1], s[2:3]
; GFX9-NEXT: s_cmp_lg_u32 s11, 0
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, v3, v1, vcc
; GFX9-NEXT: s_cselect_b64 s[2:3], s[4:5], 0
; GFX9-NEXT: v_or_b32_e32 v0, s0, v4
; GFX9-NEXT: v_or_b32_e32 v1, s1, v5
; GFX9-NEXT: v_or_b32_e32 v2, s2, v2
; GFX9-NEXT: v_or_b32_e32 v3, s3, v3
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: v_fshr_i128_vss:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_movk_i32 s6, 0x7f
; GFX10-NEXT: s_mov_b32 s7, 0
; GFX10-NEXT: v_lshlrev_b64 v[6:7], 1, v[2:3]
; GFX10-NEXT: s_and_b64 s[8:9], s[4:5], s[6:7]
; GFX10-NEXT: s_andn2_b64 s[6:7], s[6:7], s[4:5]
; GFX10-NEXT: s_sub_i32 s4, 64, 1
; GFX10-NEXT: s_sub_i32 s5, 1, 64
; GFX10-NEXT: v_lshrrev_b64 v[4:5], s4, v[0:1]
; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_lshlrev_b64 v[8:9], 1, v[0:1]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[0:1], s5, v[0:1]
; GFX10-NEXT: s_cselect_b32 s7, 1, 0
; GFX10-NEXT: s_and_b32 s4, 1, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v6, v4, v6
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4
; GFX10-NEXT: v_or_b32_e32 v5, v5, v7
; GFX10-NEXT: s_and_b32 s5, 1, s7
; GFX10-NEXT: v_cndmask_b32_e32 v4, 0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v6, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v5, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s5
; GFX10-NEXT: v_cndmask_b32_e32 v5, 0, v9, vcc_lo
; GFX10-NEXT: s_sub_i32 s5, s6, 64
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v3, s4
; GFX10-NEXT: s_sub_i32 s4, 64, s6
; GFX10-NEXT: s_cmp_lt_u32 s6, 64
; GFX10-NEXT: v_lshrrev_b64 v[2:3], s4, v[4:5]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[6:7], s6, v[0:1]
; GFX10-NEXT: s_cmp_eq_u32 s6, 0
; GFX10-NEXT: v_lshlrev_b64 v[8:9], s6, v[4:5]
; GFX10-NEXT: s_cselect_b32 s6, 1, 0
; GFX10-NEXT: s_and_b32 s4, 1, vcc_lo
; GFX10-NEXT: v_lshlrev_b64 v[4:5], s5, v[4:5]
; GFX10-NEXT: v_or_b32_e32 v2, v2, v6
; GFX10-NEXT: v_or_b32_e32 v3, v3, v7
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4
; GFX10-NEXT: s_sub_i32 s10, s8, 64
; GFX10-NEXT: v_cndmask_b32_e32 v6, 0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v7, 0, v9, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, v2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v5, v3, s4
; GFX10-NEXT: s_and_b32 s4, 1, s6
; GFX10-NEXT: s_sub_i32 s6, 64, s8
; GFX10-NEXT: s_cmp_lt_u32 s8, 64
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s4
; GFX10-NEXT: s_cselect_b32 s11, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s8, 0
; GFX10-NEXT: s_cselect_b32 s12, 1, 0
; GFX10-NEXT: s_lshr_b64 s[4:5], s[0:1], s8
; GFX10-NEXT: s_lshl_b64 s[6:7], s[2:3], s6
; GFX10-NEXT: s_lshr_b64 s[8:9], s[2:3], s8
; GFX10-NEXT: s_or_b64 s[4:5], s[4:5], s[6:7]
; GFX10-NEXT: s_lshr_b64 s[2:3], s[2:3], s10
; GFX10-NEXT: s_cmp_lg_u32 s11, 0
; GFX10-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc_lo
; GFX10-NEXT: s_cselect_b64 s[2:3], s[4:5], s[2:3]
; GFX10-NEXT: s_cmp_lg_u32 s12, 0
; GFX10-NEXT: v_cndmask_b32_e32 v3, v3, v1, vcc_lo
; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[2:3]
; GFX10-NEXT: s_cmp_lg_u32 s11, 0
; GFX10-NEXT: v_or_b32_e32 v0, s0, v6
; GFX10-NEXT: s_cselect_b64 s[2:3], s[8:9], 0
; GFX10-NEXT: v_or_b32_e32 v1, s1, v7
; GFX10-NEXT: v_or_b32_e32 v2, s2, v2
; GFX10-NEXT: v_or_b32_e32 v3, s3, v3
; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 %amt)
%cast.result = bitcast i128 %result to <4 x float>
ret <4 x float> %cast.result
}
define amdgpu_ps i128 @s_fshr_i128_65(i128 inreg %lhs, i128 inreg %rhs) {
; GFX6-LABEL: s_fshr_i128_65:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_sub_i32 s14, 63, 64
; GFX6-NEXT: s_sub_i32 s10, 64, 63
; GFX6-NEXT: s_cmp_lt_u32 63, 64
; GFX6-NEXT: s_mov_b32 s8, 0
; GFX6-NEXT: s_cselect_b32 s15, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 63, 0
; GFX6-NEXT: s_cselect_b32 s16, 1, 0
; GFX6-NEXT: s_lshr_b64 s[10:11], s[0:1], s10
; GFX6-NEXT: s_lshl_b32 s13, s2, 31
; GFX6-NEXT: s_mov_b32 s12, s8
; GFX6-NEXT: s_lshl_b32 s9, s0, 31
; GFX6-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], s14
; GFX6-NEXT: s_cmp_lg_u32 s15, 0
; GFX6-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX6-NEXT: s_cselect_b64 s[8:9], s[8:9], 0
; GFX6-NEXT: s_cmp_lg_u32 s16, 0
; GFX6-NEXT: s_movk_i32 s10, 0x41
; GFX6-NEXT: s_cselect_b64 s[2:3], s[2:3], s[0:1]
; GFX6-NEXT: s_sub_i32 s14, s10, 64
; GFX6-NEXT: s_sub_i32 s12, 64, s10
; GFX6-NEXT: s_cmp_lt_u32 s10, 64
; GFX6-NEXT: s_cselect_b32 s15, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s10, 0
; GFX6-NEXT: s_cselect_b32 s16, 1, 0
; GFX6-NEXT: s_lshr_b64 s[0:1], s[6:7], s10
; GFX6-NEXT: s_lshl_b64 s[12:13], s[6:7], s12
; GFX6-NEXT: s_lshr_b64 s[10:11], s[4:5], s10
; GFX6-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX6-NEXT: s_lshr_b64 s[6:7], s[6:7], s14
; GFX6-NEXT: s_cmp_lg_u32 s15, 0
; GFX6-NEXT: s_cselect_b64 s[6:7], s[10:11], s[6:7]
; GFX6-NEXT: s_cmp_lg_u32 s16, 0
; GFX6-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX6-NEXT: s_cmp_lg_u32 s15, 0
; GFX6-NEXT: s_cselect_b64 s[6:7], s[0:1], 0
; GFX6-NEXT: s_or_b64 s[0:1], s[8:9], s[4:5]
; GFX6-NEXT: s_or_b64 s[2:3], s[2:3], s[6:7]
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_i128_65:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_sub_i32 s14, 63, 64
; GFX8-NEXT: s_sub_i32 s10, 64, 63
; GFX8-NEXT: s_cmp_lt_u32 63, 64
; GFX8-NEXT: s_mov_b32 s8, 0
; GFX8-NEXT: s_cselect_b32 s15, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 63, 0
; GFX8-NEXT: s_cselect_b32 s16, 1, 0
; GFX8-NEXT: s_lshr_b64 s[10:11], s[0:1], s10
; GFX8-NEXT: s_lshl_b32 s13, s2, 31
; GFX8-NEXT: s_mov_b32 s12, s8
; GFX8-NEXT: s_lshl_b32 s9, s0, 31
; GFX8-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], s14
; GFX8-NEXT: s_cmp_lg_u32 s15, 0
; GFX8-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX8-NEXT: s_cselect_b64 s[8:9], s[8:9], 0
; GFX8-NEXT: s_cmp_lg_u32 s16, 0
; GFX8-NEXT: s_movk_i32 s10, 0x41
; GFX8-NEXT: s_cselect_b64 s[2:3], s[2:3], s[0:1]
; GFX8-NEXT: s_sub_i32 s14, s10, 64
; GFX8-NEXT: s_sub_i32 s12, 64, s10
; GFX8-NEXT: s_cmp_lt_u32 s10, 64
; GFX8-NEXT: s_cselect_b32 s15, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s10, 0
; GFX8-NEXT: s_cselect_b32 s16, 1, 0
; GFX8-NEXT: s_lshr_b64 s[0:1], s[6:7], s10
; GFX8-NEXT: s_lshl_b64 s[12:13], s[6:7], s12
; GFX8-NEXT: s_lshr_b64 s[10:11], s[4:5], s10
; GFX8-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX8-NEXT: s_lshr_b64 s[6:7], s[6:7], s14
; GFX8-NEXT: s_cmp_lg_u32 s15, 0
; GFX8-NEXT: s_cselect_b64 s[6:7], s[10:11], s[6:7]
; GFX8-NEXT: s_cmp_lg_u32 s16, 0
; GFX8-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX8-NEXT: s_cmp_lg_u32 s15, 0
; GFX8-NEXT: s_cselect_b64 s[6:7], s[0:1], 0
; GFX8-NEXT: s_or_b64 s[0:1], s[8:9], s[4:5]
; GFX8-NEXT: s_or_b64 s[2:3], s[2:3], s[6:7]
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_i128_65:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_sub_i32 s14, 63, 64
; GFX9-NEXT: s_sub_i32 s10, 64, 63
; GFX9-NEXT: s_cmp_lt_u32 63, 64
; GFX9-NEXT: s_mov_b32 s8, 0
; GFX9-NEXT: s_cselect_b32 s15, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 63, 0
; GFX9-NEXT: s_cselect_b32 s16, 1, 0
; GFX9-NEXT: s_lshr_b64 s[10:11], s[0:1], s10
; GFX9-NEXT: s_lshl_b32 s13, s2, 31
; GFX9-NEXT: s_mov_b32 s12, s8
; GFX9-NEXT: s_lshl_b32 s9, s0, 31
; GFX9-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], s14
; GFX9-NEXT: s_cmp_lg_u32 s15, 0
; GFX9-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX9-NEXT: s_cselect_b64 s[8:9], s[8:9], 0
; GFX9-NEXT: s_cmp_lg_u32 s16, 0
; GFX9-NEXT: s_movk_i32 s10, 0x41
; GFX9-NEXT: s_cselect_b64 s[2:3], s[2:3], s[0:1]
; GFX9-NEXT: s_sub_i32 s14, s10, 64
; GFX9-NEXT: s_sub_i32 s12, 64, s10
; GFX9-NEXT: s_cmp_lt_u32 s10, 64
; GFX9-NEXT: s_cselect_b32 s15, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s10, 0
; GFX9-NEXT: s_cselect_b32 s16, 1, 0
; GFX9-NEXT: s_lshr_b64 s[0:1], s[6:7], s10
; GFX9-NEXT: s_lshl_b64 s[12:13], s[6:7], s12
; GFX9-NEXT: s_lshr_b64 s[10:11], s[4:5], s10
; GFX9-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX9-NEXT: s_lshr_b64 s[6:7], s[6:7], s14
; GFX9-NEXT: s_cmp_lg_u32 s15, 0
; GFX9-NEXT: s_cselect_b64 s[6:7], s[10:11], s[6:7]
; GFX9-NEXT: s_cmp_lg_u32 s16, 0
; GFX9-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX9-NEXT: s_cmp_lg_u32 s15, 0
; GFX9-NEXT: s_cselect_b64 s[6:7], s[0:1], 0
; GFX9-NEXT: s_or_b64 s[0:1], s[8:9], s[4:5]
; GFX9-NEXT: s_or_b64 s[2:3], s[2:3], s[6:7]
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_i128_65:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_sub_i32 s14, 63, 64
; GFX10-NEXT: s_sub_i32 s9, 64, 63
; GFX10-NEXT: s_cmp_lt_u32 63, 64
; GFX10-NEXT: s_mov_b32 s8, 0
; GFX10-NEXT: s_cselect_b32 s15, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 63, 0
; GFX10-NEXT: s_mov_b32 s12, s8
; GFX10-NEXT: s_cselect_b32 s16, 1, 0
; GFX10-NEXT: s_lshr_b64 s[10:11], s[0:1], s9
; GFX10-NEXT: s_lshl_b32 s13, s2, 31
; GFX10-NEXT: s_lshl_b32 s9, s0, 31
; GFX10-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13]
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s14
; GFX10-NEXT: s_cmp_lg_u32 s15, 0
; GFX10-NEXT: s_movk_i32 s12, 0x41
; GFX10-NEXT: s_cselect_b64 s[8:9], s[8:9], 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[10:11], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s16, 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[0:1]
; GFX10-NEXT: s_sub_i32 s14, s12, 64
; GFX10-NEXT: s_sub_i32 s10, 64, s12
; GFX10-NEXT: s_cmp_lt_u32 s12, 64
; GFX10-NEXT: s_cselect_b32 s15, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s12, 0
; GFX10-NEXT: s_cselect_b32 s16, 1, 0
; GFX10-NEXT: s_lshr_b64 s[0:1], s[4:5], s12
; GFX10-NEXT: s_lshl_b64 s[10:11], s[6:7], s10
; GFX10-NEXT: s_lshr_b64 s[12:13], s[6:7], s12
; GFX10-NEXT: s_or_b64 s[0:1], s[0:1], s[10:11]
; GFX10-NEXT: s_lshr_b64 s[6:7], s[6:7], s14
; GFX10-NEXT: s_cmp_lg_u32 s15, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[6:7]
; GFX10-NEXT: s_cmp_lg_u32 s16, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[4:5], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s15, 0
; GFX10-NEXT: s_cselect_b64 s[4:5], s[12:13], 0
; GFX10-NEXT: s_or_b64 s[0:1], s[8:9], s[0:1]
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[4:5]
; GFX10-NEXT: ; return to shader part epilog
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 65)
ret i128 %result
}
define i128 @v_fshr_i128_65(i128 %lhs, i128 %rhs) {
; GFX6-LABEL: v_fshr_i128_65:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_sub_i32 s5, 64, 63
; GFX6-NEXT: s_sub_i32 s4, 63, 64
; GFX6-NEXT: s_cmp_lt_u32 63, 64
; GFX6-NEXT: s_cselect_b32 s6, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 63, 0
; GFX6-NEXT: v_lshr_b64 v[8:9], v[0:1], s5
; GFX6-NEXT: v_lshlrev_b32_e32 v10, 31, v0
; GFX6-NEXT: v_lshlrev_b32_e32 v11, 31, v2
; GFX6-NEXT: s_cselect_b32 s7, 1, 0
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], s4
; GFX6-NEXT: s_and_b32 s4, 1, s6
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: v_or_b32_e32 v9, v9, v11
; GFX6-NEXT: s_and_b32 s4, 1, s7
; GFX6-NEXT: v_cndmask_b32_e32 v10, 0, v10, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: s_movk_i32 s4, 0x41
; GFX6-NEXT: s_sub_i32 s5, s4, 64
; GFX6-NEXT: s_sub_i32 s6, 64, s4
; GFX6-NEXT: s_cmp_lt_u32 s4, 64
; GFX6-NEXT: v_cndmask_b32_e32 v11, v0, v2, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v12, v1, v3, vcc
; GFX6-NEXT: v_lshr_b64 v[0:1], v[4:5], s4
; GFX6-NEXT: v_lshl_b64 v[2:3], v[6:7], s6
; GFX6-NEXT: s_cselect_b32 s7, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s4, 0
; GFX6-NEXT: s_cselect_b32 s8, 1, 0
; GFX6-NEXT: v_lshr_b64 v[8:9], v[6:7], s4
; GFX6-NEXT: s_and_b32 s4, 1, s7
; GFX6-NEXT: v_or_b32_e32 v2, v0, v2
; GFX6-NEXT: v_or_b32_e32 v3, v1, v3
; GFX6-NEXT: v_lshr_b64 v[0:1], v[6:7], s5
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: s_and_b32 s4, 1, s8
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: s_and_b32 s4, 1, s7
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: v_cndmask_b32_e32 v2, 0, v8, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v3, 0, v9, vcc
; GFX6-NEXT: v_or_b32_e32 v1, v10, v1
; GFX6-NEXT: v_or_b32_e32 v2, v11, v2
; GFX6-NEXT: v_or_b32_e32 v3, v12, v3
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_i128_65:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_sub_i32 s5, 64, 63
; GFX8-NEXT: s_sub_i32 s4, 63, 64
; GFX8-NEXT: s_cmp_lt_u32 63, 64
; GFX8-NEXT: s_cselect_b32 s6, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 63, 0
; GFX8-NEXT: v_lshrrev_b64 v[8:9], s5, v[0:1]
; GFX8-NEXT: v_lshlrev_b32_e32 v10, 31, v0
; GFX8-NEXT: v_lshlrev_b32_e32 v11, 31, v2
; GFX8-NEXT: s_cselect_b32 s7, 1, 0
; GFX8-NEXT: v_lshlrev_b64 v[0:1], s4, v[0:1]
; GFX8-NEXT: s_and_b32 s4, 1, s6
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: v_or_b32_e32 v9, v9, v11
; GFX8-NEXT: s_and_b32 s4, 1, s7
; GFX8-NEXT: v_cndmask_b32_e32 v10, 0, v10, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: s_movk_i32 s4, 0x41
; GFX8-NEXT: s_sub_i32 s5, s4, 64
; GFX8-NEXT: s_sub_i32 s6, 64, s4
; GFX8-NEXT: s_cmp_lt_u32 s4, 64
; GFX8-NEXT: v_cndmask_b32_e32 v11, v0, v2, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v12, v1, v3, vcc
; GFX8-NEXT: v_lshrrev_b64 v[0:1], s4, v[4:5]
; GFX8-NEXT: v_lshlrev_b64 v[2:3], s6, v[6:7]
; GFX8-NEXT: s_cselect_b32 s7, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s4, 0
; GFX8-NEXT: s_cselect_b32 s8, 1, 0
; GFX8-NEXT: v_lshrrev_b64 v[8:9], s4, v[6:7]
; GFX8-NEXT: s_and_b32 s4, 1, s7
; GFX8-NEXT: v_or_b32_e32 v2, v0, v2
; GFX8-NEXT: v_or_b32_e32 v3, v1, v3
; GFX8-NEXT: v_lshrrev_b64 v[0:1], s5, v[6:7]
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: s_and_b32 s4, 1, s8
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: s_and_b32 s4, 1, s7
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: v_cndmask_b32_e32 v2, 0, v8, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v3, 0, v9, vcc
; GFX8-NEXT: v_or_b32_e32 v1, v10, v1
; GFX8-NEXT: v_or_b32_e32 v2, v11, v2
; GFX8-NEXT: v_or_b32_e32 v3, v12, v3
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_i128_65:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_sub_i32 s5, 64, 63
; GFX9-NEXT: s_sub_i32 s4, 63, 64
; GFX9-NEXT: s_cmp_lt_u32 63, 64
; GFX9-NEXT: s_cselect_b32 s6, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 63, 0
; GFX9-NEXT: v_lshrrev_b64 v[8:9], s5, v[0:1]
; GFX9-NEXT: v_lshlrev_b32_e32 v10, 31, v0
; GFX9-NEXT: v_lshlrev_b32_e32 v11, 31, v2
; GFX9-NEXT: s_cselect_b32 s7, 1, 0
; GFX9-NEXT: v_lshlrev_b64 v[0:1], s4, v[0:1]
; GFX9-NEXT: s_and_b32 s4, 1, s6
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: v_or_b32_e32 v9, v9, v11
; GFX9-NEXT: s_and_b32 s4, 1, s7
; GFX9-NEXT: v_cndmask_b32_e32 v10, 0, v10, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: s_movk_i32 s4, 0x41
; GFX9-NEXT: s_sub_i32 s5, s4, 64
; GFX9-NEXT: s_sub_i32 s6, 64, s4
; GFX9-NEXT: s_cmp_lt_u32 s4, 64
; GFX9-NEXT: v_cndmask_b32_e32 v11, v0, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v12, v1, v3, vcc
; GFX9-NEXT: v_lshrrev_b64 v[0:1], s4, v[4:5]
; GFX9-NEXT: v_lshlrev_b64 v[2:3], s6, v[6:7]
; GFX9-NEXT: s_cselect_b32 s7, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s4, 0
; GFX9-NEXT: s_cselect_b32 s8, 1, 0
; GFX9-NEXT: v_lshrrev_b64 v[8:9], s4, v[6:7]
; GFX9-NEXT: s_and_b32 s4, 1, s7
; GFX9-NEXT: v_or_b32_e32 v2, v0, v2
; GFX9-NEXT: v_or_b32_e32 v3, v1, v3
; GFX9-NEXT: v_lshrrev_b64 v[0:1], s5, v[6:7]
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: s_and_b32 s4, 1, s8
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: s_and_b32 s4, 1, s7
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: v_cndmask_b32_e32 v2, 0, v8, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, 0, v9, vcc
; GFX9-NEXT: v_or_b32_e32 v1, v10, v1
; GFX9-NEXT: v_or_b32_e32 v2, v11, v2
; GFX9-NEXT: v_or_b32_e32 v3, v12, v3
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_i128_65:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: s_sub_i32 s4, 64, 63
; GFX10-NEXT: s_sub_i32 s5, 63, 64
; GFX10-NEXT: s_cmp_lt_u32 63, 64
; GFX10-NEXT: v_lshlrev_b32_e32 v11, 31, v0
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 63, 0
; GFX10-NEXT: v_lshrrev_b64 v[8:9], s4, v[0:1]
; GFX10-NEXT: s_cselect_b32 s4, 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[0:1], s5, v[0:1]
; GFX10-NEXT: s_and_b32 s5, 1, vcc_lo
; GFX10-NEXT: v_lshlrev_b32_e32 v10, 31, v2
; GFX10-NEXT: v_cndmask_b32_e32 v13, 0, v11, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s5
; GFX10-NEXT: s_movk_i32 s6, 0x41
; GFX10-NEXT: s_and_b32 s4, 1, s4
; GFX10-NEXT: s_sub_i32 s5, 64, s6
; GFX10-NEXT: v_or_b32_e32 v12, v9, v10
; GFX10-NEXT: v_cndmask_b32_e32 v14, v0, v8, vcc_lo
; GFX10-NEXT: v_lshlrev_b64 v[10:11], s5, v[6:7]
; GFX10-NEXT: v_lshrrev_b64 v[8:9], s6, v[4:5]
; GFX10-NEXT: s_sub_i32 s5, s6, 64
; GFX10-NEXT: s_cmp_lt_u32 s6, 64
; GFX10-NEXT: v_cndmask_b32_e32 v12, v1, v12, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[0:1], s5, v[6:7]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s6, 0
; GFX10-NEXT: v_or_b32_e32 v8, v8, v10
; GFX10-NEXT: v_or_b32_e32 v9, v9, v11
; GFX10-NEXT: s_cselect_b32 s5, 1, 0
; GFX10-NEXT: v_lshrrev_b64 v[6:7], s6, v[6:7]
; GFX10-NEXT: s_and_b32 s5, 1, s5
; GFX10-NEXT: s_and_b32 s6, 1, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_ne_u32_e64 s5, 0, s5
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s6
; GFX10-NEXT: v_cndmask_b32_e64 v2, v14, v2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v3, v12, v3, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v5, s5
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v4, s5
; GFX10-NEXT: v_cndmask_b32_e32 v4, 0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v5, 0, v7, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v1, v13, v1
; GFX10-NEXT: v_or_b32_e32 v2, v2, v4
; GFX10-NEXT: v_or_b32_e32 v3, v3, v5
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call i128 @llvm.fshr.i128(i128 %lhs, i128 %rhs, i128 65)
ret i128 %result
}
define amdgpu_ps <2 x i128> @s_fshr_v2i128(<2 x i128> inreg %lhs, <2 x i128> inreg %rhs, <2 x i128> inreg %amt) {
; GFX6-LABEL: s_fshr_v2i128:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_movk_i32 s18, 0x7f
; GFX6-NEXT: s_mov_b32 s19, 0
; GFX6-NEXT: s_and_b64 s[22:23], s[16:17], s[18:19]
; GFX6-NEXT: s_andn2_b64 s[16:17], s[18:19], s[16:17]
; GFX6-NEXT: s_sub_i32 s30, 1, 64
; GFX6-NEXT: s_sub_i32 s31, 64, 1
; GFX6-NEXT: s_cmp_lt_u32 1, 64
; GFX6-NEXT: s_cselect_b32 s17, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 1, 0
; GFX6-NEXT: s_cselect_b32 s23, 1, 0
; GFX6-NEXT: s_lshr_b64 s[26:27], s[0:1], s31
; GFX6-NEXT: s_lshl_b64 s[28:29], s[2:3], 1
; GFX6-NEXT: s_lshl_b64 s[24:25], s[0:1], 1
; GFX6-NEXT: s_or_b64 s[26:27], s[26:27], s[28:29]
; GFX6-NEXT: s_lshl_b64 s[0:1], s[0:1], s30
; GFX6-NEXT: s_cmp_lg_u32 s17, 0
; GFX6-NEXT: s_cselect_b64 s[24:25], s[24:25], 0
; GFX6-NEXT: s_cselect_b64 s[0:1], s[26:27], s[0:1]
; GFX6-NEXT: s_cmp_lg_u32 s23, 0
; GFX6-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX6-NEXT: s_sub_i32 s23, s16, 64
; GFX6-NEXT: s_sub_i32 s17, 64, s16
; GFX6-NEXT: s_cmp_lt_u32 s16, 64
; GFX6-NEXT: s_cselect_b32 s28, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s16, 0
; GFX6-NEXT: s_cselect_b32 s29, 1, 0
; GFX6-NEXT: s_lshl_b64 s[2:3], s[24:25], s16
; GFX6-NEXT: s_lshr_b64 s[26:27], s[24:25], s17
; GFX6-NEXT: s_lshl_b64 s[16:17], s[0:1], s16
; GFX6-NEXT: s_or_b64 s[16:17], s[26:27], s[16:17]
; GFX6-NEXT: s_lshl_b64 s[24:25], s[24:25], s23
; GFX6-NEXT: s_cmp_lg_u32 s28, 0
; GFX6-NEXT: s_cselect_b64 s[2:3], s[2:3], 0
; GFX6-NEXT: s_cselect_b64 s[16:17], s[16:17], s[24:25]
; GFX6-NEXT: s_cmp_lg_u32 s29, 0
; GFX6-NEXT: s_cselect_b64 s[16:17], s[0:1], s[16:17]
; GFX6-NEXT: s_sub_i32 s26, s22, 64
; GFX6-NEXT: s_sub_i32 s24, 64, s22
; GFX6-NEXT: s_cmp_lt_u32 s22, 64
; GFX6-NEXT: s_cselect_b32 s27, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s22, 0
; GFX6-NEXT: s_cselect_b32 s28, 1, 0
; GFX6-NEXT: s_lshr_b64 s[0:1], s[10:11], s22
; GFX6-NEXT: s_lshl_b64 s[24:25], s[10:11], s24
; GFX6-NEXT: s_lshr_b64 s[22:23], s[8:9], s22
; GFX6-NEXT: s_or_b64 s[22:23], s[22:23], s[24:25]
; GFX6-NEXT: s_lshr_b64 s[10:11], s[10:11], s26
; GFX6-NEXT: s_cmp_lg_u32 s27, 0
; GFX6-NEXT: s_cselect_b64 s[10:11], s[22:23], s[10:11]
; GFX6-NEXT: s_cmp_lg_u32 s28, 0
; GFX6-NEXT: s_cselect_b64 s[8:9], s[8:9], s[10:11]
; GFX6-NEXT: s_cmp_lg_u32 s27, 0
; GFX6-NEXT: s_cselect_b64 s[10:11], s[0:1], 0
; GFX6-NEXT: s_or_b64 s[0:1], s[2:3], s[8:9]
; GFX6-NEXT: s_or_b64 s[2:3], s[16:17], s[10:11]
; GFX6-NEXT: s_and_b64 s[8:9], s[20:21], s[18:19]
; GFX6-NEXT: s_andn2_b64 s[10:11], s[18:19], s[20:21]
; GFX6-NEXT: s_cmp_lt_u32 1, 64
; GFX6-NEXT: s_cselect_b32 s9, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 1, 0
; GFX6-NEXT: s_cselect_b32 s11, 1, 0
; GFX6-NEXT: s_lshr_b64 s[18:19], s[4:5], s31
; GFX6-NEXT: s_lshl_b64 s[20:21], s[6:7], 1
; GFX6-NEXT: s_lshl_b64 s[16:17], s[4:5], 1
; GFX6-NEXT: s_or_b64 s[18:19], s[18:19], s[20:21]
; GFX6-NEXT: s_lshl_b64 s[4:5], s[4:5], s30
; GFX6-NEXT: s_cmp_lg_u32 s9, 0
; GFX6-NEXT: s_cselect_b64 s[16:17], s[16:17], 0
; GFX6-NEXT: s_cselect_b64 s[4:5], s[18:19], s[4:5]
; GFX6-NEXT: s_cmp_lg_u32 s11, 0
; GFX6-NEXT: s_cselect_b64 s[4:5], s[6:7], s[4:5]
; GFX6-NEXT: s_sub_i32 s9, s10, 64
; GFX6-NEXT: s_sub_i32 s11, 64, s10
; GFX6-NEXT: s_cmp_lt_u32 s10, 64
; GFX6-NEXT: s_cselect_b32 s20, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s10, 0
; GFX6-NEXT: s_cselect_b32 s21, 1, 0
; GFX6-NEXT: s_lshl_b64 s[6:7], s[16:17], s10
; GFX6-NEXT: s_lshr_b64 s[18:19], s[16:17], s11
; GFX6-NEXT: s_lshl_b64 s[10:11], s[4:5], s10
; GFX6-NEXT: s_or_b64 s[10:11], s[18:19], s[10:11]
; GFX6-NEXT: s_lshl_b64 s[16:17], s[16:17], s9
; GFX6-NEXT: s_cmp_lg_u32 s20, 0
; GFX6-NEXT: s_cselect_b64 s[6:7], s[6:7], 0
; GFX6-NEXT: s_cselect_b64 s[10:11], s[10:11], s[16:17]
; GFX6-NEXT: s_cmp_lg_u32 s21, 0
; GFX6-NEXT: s_cselect_b64 s[10:11], s[4:5], s[10:11]
; GFX6-NEXT: s_sub_i32 s18, s8, 64
; GFX6-NEXT: s_sub_i32 s16, 64, s8
; GFX6-NEXT: s_cmp_lt_u32 s8, 64
; GFX6-NEXT: s_cselect_b32 s19, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 s8, 0
; GFX6-NEXT: s_cselect_b32 s20, 1, 0
; GFX6-NEXT: s_lshr_b64 s[4:5], s[14:15], s8
; GFX6-NEXT: s_lshl_b64 s[16:17], s[14:15], s16
; GFX6-NEXT: s_lshr_b64 s[8:9], s[12:13], s8
; GFX6-NEXT: s_or_b64 s[8:9], s[8:9], s[16:17]
; GFX6-NEXT: s_lshr_b64 s[14:15], s[14:15], s18
; GFX6-NEXT: s_cmp_lg_u32 s19, 0
; GFX6-NEXT: s_cselect_b64 s[8:9], s[8:9], s[14:15]
; GFX6-NEXT: s_cmp_lg_u32 s20, 0
; GFX6-NEXT: s_cselect_b64 s[8:9], s[12:13], s[8:9]
; GFX6-NEXT: s_cmp_lg_u32 s19, 0
; GFX6-NEXT: s_cselect_b64 s[12:13], s[4:5], 0
; GFX6-NEXT: s_or_b64 s[4:5], s[6:7], s[8:9]
; GFX6-NEXT: s_or_b64 s[6:7], s[10:11], s[12:13]
; GFX6-NEXT: ; return to shader part epilog
;
; GFX8-LABEL: s_fshr_v2i128:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_movk_i32 s18, 0x7f
; GFX8-NEXT: s_mov_b32 s19, 0
; GFX8-NEXT: s_and_b64 s[22:23], s[16:17], s[18:19]
; GFX8-NEXT: s_andn2_b64 s[16:17], s[18:19], s[16:17]
; GFX8-NEXT: s_sub_i32 s30, 1, 64
; GFX8-NEXT: s_sub_i32 s31, 64, 1
; GFX8-NEXT: s_cmp_lt_u32 1, 64
; GFX8-NEXT: s_cselect_b32 s17, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 1, 0
; GFX8-NEXT: s_cselect_b32 s23, 1, 0
; GFX8-NEXT: s_lshr_b64 s[26:27], s[0:1], s31
; GFX8-NEXT: s_lshl_b64 s[28:29], s[2:3], 1
; GFX8-NEXT: s_lshl_b64 s[24:25], s[0:1], 1
; GFX8-NEXT: s_or_b64 s[26:27], s[26:27], s[28:29]
; GFX8-NEXT: s_lshl_b64 s[0:1], s[0:1], s30
; GFX8-NEXT: s_cmp_lg_u32 s17, 0
; GFX8-NEXT: s_cselect_b64 s[24:25], s[24:25], 0
; GFX8-NEXT: s_cselect_b64 s[0:1], s[26:27], s[0:1]
; GFX8-NEXT: s_cmp_lg_u32 s23, 0
; GFX8-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX8-NEXT: s_sub_i32 s23, s16, 64
; GFX8-NEXT: s_sub_i32 s17, 64, s16
; GFX8-NEXT: s_cmp_lt_u32 s16, 64
; GFX8-NEXT: s_cselect_b32 s28, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s16, 0
; GFX8-NEXT: s_cselect_b32 s29, 1, 0
; GFX8-NEXT: s_lshl_b64 s[2:3], s[24:25], s16
; GFX8-NEXT: s_lshr_b64 s[26:27], s[24:25], s17
; GFX8-NEXT: s_lshl_b64 s[16:17], s[0:1], s16
; GFX8-NEXT: s_or_b64 s[16:17], s[26:27], s[16:17]
; GFX8-NEXT: s_lshl_b64 s[24:25], s[24:25], s23
; GFX8-NEXT: s_cmp_lg_u32 s28, 0
; GFX8-NEXT: s_cselect_b64 s[2:3], s[2:3], 0
; GFX8-NEXT: s_cselect_b64 s[16:17], s[16:17], s[24:25]
; GFX8-NEXT: s_cmp_lg_u32 s29, 0
; GFX8-NEXT: s_cselect_b64 s[16:17], s[0:1], s[16:17]
; GFX8-NEXT: s_sub_i32 s26, s22, 64
; GFX8-NEXT: s_sub_i32 s24, 64, s22
; GFX8-NEXT: s_cmp_lt_u32 s22, 64
; GFX8-NEXT: s_cselect_b32 s27, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s22, 0
; GFX8-NEXT: s_cselect_b32 s28, 1, 0
; GFX8-NEXT: s_lshr_b64 s[0:1], s[10:11], s22
; GFX8-NEXT: s_lshl_b64 s[24:25], s[10:11], s24
; GFX8-NEXT: s_lshr_b64 s[22:23], s[8:9], s22
; GFX8-NEXT: s_or_b64 s[22:23], s[22:23], s[24:25]
; GFX8-NEXT: s_lshr_b64 s[10:11], s[10:11], s26
; GFX8-NEXT: s_cmp_lg_u32 s27, 0
; GFX8-NEXT: s_cselect_b64 s[10:11], s[22:23], s[10:11]
; GFX8-NEXT: s_cmp_lg_u32 s28, 0
; GFX8-NEXT: s_cselect_b64 s[8:9], s[8:9], s[10:11]
; GFX8-NEXT: s_cmp_lg_u32 s27, 0
; GFX8-NEXT: s_cselect_b64 s[10:11], s[0:1], 0
; GFX8-NEXT: s_or_b64 s[0:1], s[2:3], s[8:9]
; GFX8-NEXT: s_or_b64 s[2:3], s[16:17], s[10:11]
; GFX8-NEXT: s_and_b64 s[8:9], s[20:21], s[18:19]
; GFX8-NEXT: s_andn2_b64 s[10:11], s[18:19], s[20:21]
; GFX8-NEXT: s_cmp_lt_u32 1, 64
; GFX8-NEXT: s_cselect_b32 s9, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 1, 0
; GFX8-NEXT: s_cselect_b32 s11, 1, 0
; GFX8-NEXT: s_lshr_b64 s[18:19], s[4:5], s31
; GFX8-NEXT: s_lshl_b64 s[20:21], s[6:7], 1
; GFX8-NEXT: s_lshl_b64 s[16:17], s[4:5], 1
; GFX8-NEXT: s_or_b64 s[18:19], s[18:19], s[20:21]
; GFX8-NEXT: s_lshl_b64 s[4:5], s[4:5], s30
; GFX8-NEXT: s_cmp_lg_u32 s9, 0
; GFX8-NEXT: s_cselect_b64 s[16:17], s[16:17], 0
; GFX8-NEXT: s_cselect_b64 s[4:5], s[18:19], s[4:5]
; GFX8-NEXT: s_cmp_lg_u32 s11, 0
; GFX8-NEXT: s_cselect_b64 s[4:5], s[6:7], s[4:5]
; GFX8-NEXT: s_sub_i32 s9, s10, 64
; GFX8-NEXT: s_sub_i32 s11, 64, s10
; GFX8-NEXT: s_cmp_lt_u32 s10, 64
; GFX8-NEXT: s_cselect_b32 s20, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s10, 0
; GFX8-NEXT: s_cselect_b32 s21, 1, 0
; GFX8-NEXT: s_lshl_b64 s[6:7], s[16:17], s10
; GFX8-NEXT: s_lshr_b64 s[18:19], s[16:17], s11
; GFX8-NEXT: s_lshl_b64 s[10:11], s[4:5], s10
; GFX8-NEXT: s_or_b64 s[10:11], s[18:19], s[10:11]
; GFX8-NEXT: s_lshl_b64 s[16:17], s[16:17], s9
; GFX8-NEXT: s_cmp_lg_u32 s20, 0
; GFX8-NEXT: s_cselect_b64 s[6:7], s[6:7], 0
; GFX8-NEXT: s_cselect_b64 s[10:11], s[10:11], s[16:17]
; GFX8-NEXT: s_cmp_lg_u32 s21, 0
; GFX8-NEXT: s_cselect_b64 s[10:11], s[4:5], s[10:11]
; GFX8-NEXT: s_sub_i32 s18, s8, 64
; GFX8-NEXT: s_sub_i32 s16, 64, s8
; GFX8-NEXT: s_cmp_lt_u32 s8, 64
; GFX8-NEXT: s_cselect_b32 s19, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 s8, 0
; GFX8-NEXT: s_cselect_b32 s20, 1, 0
; GFX8-NEXT: s_lshr_b64 s[4:5], s[14:15], s8
; GFX8-NEXT: s_lshl_b64 s[16:17], s[14:15], s16
; GFX8-NEXT: s_lshr_b64 s[8:9], s[12:13], s8
; GFX8-NEXT: s_or_b64 s[8:9], s[8:9], s[16:17]
; GFX8-NEXT: s_lshr_b64 s[14:15], s[14:15], s18
; GFX8-NEXT: s_cmp_lg_u32 s19, 0
; GFX8-NEXT: s_cselect_b64 s[8:9], s[8:9], s[14:15]
; GFX8-NEXT: s_cmp_lg_u32 s20, 0
; GFX8-NEXT: s_cselect_b64 s[8:9], s[12:13], s[8:9]
; GFX8-NEXT: s_cmp_lg_u32 s19, 0
; GFX8-NEXT: s_cselect_b64 s[12:13], s[4:5], 0
; GFX8-NEXT: s_or_b64 s[4:5], s[6:7], s[8:9]
; GFX8-NEXT: s_or_b64 s[6:7], s[10:11], s[12:13]
; GFX8-NEXT: ; return to shader part epilog
;
; GFX9-LABEL: s_fshr_v2i128:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_movk_i32 s18, 0x7f
; GFX9-NEXT: s_mov_b32 s19, 0
; GFX9-NEXT: s_and_b64 s[22:23], s[16:17], s[18:19]
; GFX9-NEXT: s_andn2_b64 s[16:17], s[18:19], s[16:17]
; GFX9-NEXT: s_sub_i32 s30, 1, 64
; GFX9-NEXT: s_sub_i32 s31, 64, 1
; GFX9-NEXT: s_cmp_lt_u32 1, 64
; GFX9-NEXT: s_cselect_b32 s17, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 1, 0
; GFX9-NEXT: s_cselect_b32 s23, 1, 0
; GFX9-NEXT: s_lshr_b64 s[26:27], s[0:1], s31
; GFX9-NEXT: s_lshl_b64 s[28:29], s[2:3], 1
; GFX9-NEXT: s_lshl_b64 s[24:25], s[0:1], 1
; GFX9-NEXT: s_or_b64 s[26:27], s[26:27], s[28:29]
; GFX9-NEXT: s_lshl_b64 s[0:1], s[0:1], s30
; GFX9-NEXT: s_cmp_lg_u32 s17, 0
; GFX9-NEXT: s_cselect_b64 s[24:25], s[24:25], 0
; GFX9-NEXT: s_cselect_b64 s[0:1], s[26:27], s[0:1]
; GFX9-NEXT: s_cmp_lg_u32 s23, 0
; GFX9-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX9-NEXT: s_sub_i32 s23, s16, 64
; GFX9-NEXT: s_sub_i32 s17, 64, s16
; GFX9-NEXT: s_cmp_lt_u32 s16, 64
; GFX9-NEXT: s_cselect_b32 s28, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s16, 0
; GFX9-NEXT: s_cselect_b32 s29, 1, 0
; GFX9-NEXT: s_lshl_b64 s[2:3], s[24:25], s16
; GFX9-NEXT: s_lshr_b64 s[26:27], s[24:25], s17
; GFX9-NEXT: s_lshl_b64 s[16:17], s[0:1], s16
; GFX9-NEXT: s_or_b64 s[16:17], s[26:27], s[16:17]
; GFX9-NEXT: s_lshl_b64 s[24:25], s[24:25], s23
; GFX9-NEXT: s_cmp_lg_u32 s28, 0
; GFX9-NEXT: s_cselect_b64 s[2:3], s[2:3], 0
; GFX9-NEXT: s_cselect_b64 s[16:17], s[16:17], s[24:25]
; GFX9-NEXT: s_cmp_lg_u32 s29, 0
; GFX9-NEXT: s_cselect_b64 s[16:17], s[0:1], s[16:17]
; GFX9-NEXT: s_sub_i32 s26, s22, 64
; GFX9-NEXT: s_sub_i32 s24, 64, s22
; GFX9-NEXT: s_cmp_lt_u32 s22, 64
; GFX9-NEXT: s_cselect_b32 s27, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s22, 0
; GFX9-NEXT: s_cselect_b32 s28, 1, 0
; GFX9-NEXT: s_lshr_b64 s[0:1], s[10:11], s22
; GFX9-NEXT: s_lshl_b64 s[24:25], s[10:11], s24
; GFX9-NEXT: s_lshr_b64 s[22:23], s[8:9], s22
; GFX9-NEXT: s_or_b64 s[22:23], s[22:23], s[24:25]
; GFX9-NEXT: s_lshr_b64 s[10:11], s[10:11], s26
; GFX9-NEXT: s_cmp_lg_u32 s27, 0
; GFX9-NEXT: s_cselect_b64 s[10:11], s[22:23], s[10:11]
; GFX9-NEXT: s_cmp_lg_u32 s28, 0
; GFX9-NEXT: s_cselect_b64 s[8:9], s[8:9], s[10:11]
; GFX9-NEXT: s_cmp_lg_u32 s27, 0
; GFX9-NEXT: s_cselect_b64 s[10:11], s[0:1], 0
; GFX9-NEXT: s_or_b64 s[0:1], s[2:3], s[8:9]
; GFX9-NEXT: s_or_b64 s[2:3], s[16:17], s[10:11]
; GFX9-NEXT: s_and_b64 s[8:9], s[20:21], s[18:19]
; GFX9-NEXT: s_andn2_b64 s[10:11], s[18:19], s[20:21]
; GFX9-NEXT: s_cmp_lt_u32 1, 64
; GFX9-NEXT: s_cselect_b32 s9, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 1, 0
; GFX9-NEXT: s_cselect_b32 s11, 1, 0
; GFX9-NEXT: s_lshr_b64 s[18:19], s[4:5], s31
; GFX9-NEXT: s_lshl_b64 s[20:21], s[6:7], 1
; GFX9-NEXT: s_lshl_b64 s[16:17], s[4:5], 1
; GFX9-NEXT: s_or_b64 s[18:19], s[18:19], s[20:21]
; GFX9-NEXT: s_lshl_b64 s[4:5], s[4:5], s30
; GFX9-NEXT: s_cmp_lg_u32 s9, 0
; GFX9-NEXT: s_cselect_b64 s[16:17], s[16:17], 0
; GFX9-NEXT: s_cselect_b64 s[4:5], s[18:19], s[4:5]
; GFX9-NEXT: s_cmp_lg_u32 s11, 0
; GFX9-NEXT: s_cselect_b64 s[4:5], s[6:7], s[4:5]
; GFX9-NEXT: s_sub_i32 s9, s10, 64
; GFX9-NEXT: s_sub_i32 s11, 64, s10
; GFX9-NEXT: s_cmp_lt_u32 s10, 64
; GFX9-NEXT: s_cselect_b32 s20, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s10, 0
; GFX9-NEXT: s_cselect_b32 s21, 1, 0
; GFX9-NEXT: s_lshl_b64 s[6:7], s[16:17], s10
; GFX9-NEXT: s_lshr_b64 s[18:19], s[16:17], s11
; GFX9-NEXT: s_lshl_b64 s[10:11], s[4:5], s10
; GFX9-NEXT: s_or_b64 s[10:11], s[18:19], s[10:11]
; GFX9-NEXT: s_lshl_b64 s[16:17], s[16:17], s9
; GFX9-NEXT: s_cmp_lg_u32 s20, 0
; GFX9-NEXT: s_cselect_b64 s[6:7], s[6:7], 0
; GFX9-NEXT: s_cselect_b64 s[10:11], s[10:11], s[16:17]
; GFX9-NEXT: s_cmp_lg_u32 s21, 0
; GFX9-NEXT: s_cselect_b64 s[10:11], s[4:5], s[10:11]
; GFX9-NEXT: s_sub_i32 s18, s8, 64
; GFX9-NEXT: s_sub_i32 s16, 64, s8
; GFX9-NEXT: s_cmp_lt_u32 s8, 64
; GFX9-NEXT: s_cselect_b32 s19, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 s8, 0
; GFX9-NEXT: s_cselect_b32 s20, 1, 0
; GFX9-NEXT: s_lshr_b64 s[4:5], s[14:15], s8
; GFX9-NEXT: s_lshl_b64 s[16:17], s[14:15], s16
; GFX9-NEXT: s_lshr_b64 s[8:9], s[12:13], s8
; GFX9-NEXT: s_or_b64 s[8:9], s[8:9], s[16:17]
; GFX9-NEXT: s_lshr_b64 s[14:15], s[14:15], s18
; GFX9-NEXT: s_cmp_lg_u32 s19, 0
; GFX9-NEXT: s_cselect_b64 s[8:9], s[8:9], s[14:15]
; GFX9-NEXT: s_cmp_lg_u32 s20, 0
; GFX9-NEXT: s_cselect_b64 s[8:9], s[12:13], s[8:9]
; GFX9-NEXT: s_cmp_lg_u32 s19, 0
; GFX9-NEXT: s_cselect_b64 s[12:13], s[4:5], 0
; GFX9-NEXT: s_or_b64 s[4:5], s[6:7], s[8:9]
; GFX9-NEXT: s_or_b64 s[6:7], s[10:11], s[12:13]
; GFX9-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: s_fshr_v2i128:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_movk_i32 s18, 0x7f
; GFX10-NEXT: s_mov_b32 s19, 0
; GFX10-NEXT: s_sub_i32 s30, 1, 64
; GFX10-NEXT: s_and_b64 s[22:23], s[16:17], s[18:19]
; GFX10-NEXT: s_andn2_b64 s[16:17], s[18:19], s[16:17]
; GFX10-NEXT: s_sub_i32 s31, 64, 1
; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: s_cselect_b32 s17, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: s_cselect_b32 s23, 1, 0
; GFX10-NEXT: s_lshr_b64 s[24:25], s[0:1], s31
; GFX10-NEXT: s_lshl_b64 s[26:27], s[2:3], 1
; GFX10-NEXT: s_lshl_b64 s[28:29], s[0:1], 1
; GFX10-NEXT: s_or_b64 s[24:25], s[24:25], s[26:27]
; GFX10-NEXT: s_lshl_b64 s[0:1], s[0:1], s30
; GFX10-NEXT: s_cmp_lg_u32 s17, 0
; GFX10-NEXT: s_cselect_b64 s[26:27], s[28:29], 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[24:25], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s23, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[2:3], s[0:1]
; GFX10-NEXT: s_sub_i32 s23, s16, 64
; GFX10-NEXT: s_sub_i32 s2, 64, s16
; GFX10-NEXT: s_cmp_lt_u32 s16, 64
; GFX10-NEXT: s_cselect_b32 s28, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s16, 0
; GFX10-NEXT: s_cselect_b32 s29, 1, 0
; GFX10-NEXT: s_lshl_b64 s[24:25], s[0:1], s16
; GFX10-NEXT: s_lshr_b64 s[2:3], s[26:27], s2
; GFX10-NEXT: s_lshl_b64 s[16:17], s[26:27], s16
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[24:25]
; GFX10-NEXT: s_lshl_b64 s[24:25], s[26:27], s23
; GFX10-NEXT: s_cmp_lg_u32 s28, 0
; GFX10-NEXT: s_cselect_b64 s[16:17], s[16:17], 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[24:25]
; GFX10-NEXT: s_cmp_lg_u32 s29, 0
; GFX10-NEXT: s_cselect_b64 s[2:3], s[0:1], s[2:3]
; GFX10-NEXT: s_sub_i32 s26, s22, 64
; GFX10-NEXT: s_sub_i32 s23, 64, s22
; GFX10-NEXT: s_cmp_lt_u32 s22, 64
; GFX10-NEXT: s_cselect_b32 s27, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s22, 0
; GFX10-NEXT: s_cselect_b32 s28, 1, 0
; GFX10-NEXT: s_lshr_b64 s[0:1], s[8:9], s22
; GFX10-NEXT: s_lshl_b64 s[24:25], s[10:11], s23
; GFX10-NEXT: s_lshr_b64 s[22:23], s[10:11], s22
; GFX10-NEXT: s_or_b64 s[0:1], s[0:1], s[24:25]
; GFX10-NEXT: s_lshr_b64 s[10:11], s[10:11], s26
; GFX10-NEXT: s_cmp_lg_u32 s27, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[0:1], s[10:11]
; GFX10-NEXT: s_cmp_lg_u32 s28, 0
; GFX10-NEXT: s_cselect_b64 s[0:1], s[8:9], s[0:1]
; GFX10-NEXT: s_cmp_lg_u32 s27, 0
; GFX10-NEXT: s_cselect_b64 s[8:9], s[22:23], 0
; GFX10-NEXT: s_or_b64 s[0:1], s[16:17], s[0:1]
; GFX10-NEXT: s_or_b64 s[2:3], s[2:3], s[8:9]
; GFX10-NEXT: s_and_b64 s[8:9], s[20:21], s[18:19]
; GFX10-NEXT: s_andn2_b64 s[10:11], s[18:19], s[20:21]
; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: s_cselect_b32 s9, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: s_cselect_b32 s11, 1, 0
; GFX10-NEXT: s_lshr_b64 s[16:17], s[4:5], s31
; GFX10-NEXT: s_lshl_b64 s[18:19], s[6:7], 1
; GFX10-NEXT: s_lshl_b64 s[20:21], s[4:5], 1
; GFX10-NEXT: s_or_b64 s[16:17], s[16:17], s[18:19]
; GFX10-NEXT: s_lshl_b64 s[4:5], s[4:5], s30
; GFX10-NEXT: s_cmp_lg_u32 s9, 0
; GFX10-NEXT: s_cselect_b64 s[18:19], s[20:21], 0
; GFX10-NEXT: s_cselect_b64 s[4:5], s[16:17], s[4:5]
; GFX10-NEXT: s_cmp_lg_u32 s11, 0
; GFX10-NEXT: s_cselect_b64 s[4:5], s[6:7], s[4:5]
; GFX10-NEXT: s_sub_i32 s9, s10, 64
; GFX10-NEXT: s_sub_i32 s6, 64, s10
; GFX10-NEXT: s_cmp_lt_u32 s10, 64
; GFX10-NEXT: s_cselect_b32 s20, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s10, 0
; GFX10-NEXT: s_cselect_b32 s21, 1, 0
; GFX10-NEXT: s_lshl_b64 s[16:17], s[4:5], s10
; GFX10-NEXT: s_lshr_b64 s[6:7], s[18:19], s6
; GFX10-NEXT: s_lshl_b64 s[10:11], s[18:19], s10
; GFX10-NEXT: s_or_b64 s[6:7], s[6:7], s[16:17]
; GFX10-NEXT: s_lshl_b64 s[16:17], s[18:19], s9
; GFX10-NEXT: s_cmp_lg_u32 s20, 0
; GFX10-NEXT: s_cselect_b64 s[10:11], s[10:11], 0
; GFX10-NEXT: s_cselect_b64 s[6:7], s[6:7], s[16:17]
; GFX10-NEXT: s_cmp_lg_u32 s21, 0
; GFX10-NEXT: s_cselect_b64 s[6:7], s[4:5], s[6:7]
; GFX10-NEXT: s_sub_i32 s18, s8, 64
; GFX10-NEXT: s_sub_i32 s9, 64, s8
; GFX10-NEXT: s_cmp_lt_u32 s8, 64
; GFX10-NEXT: s_cselect_b32 s19, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 s8, 0
; GFX10-NEXT: s_cselect_b32 s20, 1, 0
; GFX10-NEXT: s_lshr_b64 s[4:5], s[12:13], s8
; GFX10-NEXT: s_lshl_b64 s[16:17], s[14:15], s9
; GFX10-NEXT: s_lshr_b64 s[8:9], s[14:15], s8
; GFX10-NEXT: s_or_b64 s[4:5], s[4:5], s[16:17]
; GFX10-NEXT: s_lshr_b64 s[14:15], s[14:15], s18
; GFX10-NEXT: s_cmp_lg_u32 s19, 0
; GFX10-NEXT: s_cselect_b64 s[4:5], s[4:5], s[14:15]
; GFX10-NEXT: s_cmp_lg_u32 s20, 0
; GFX10-NEXT: s_cselect_b64 s[4:5], s[12:13], s[4:5]
; GFX10-NEXT: s_cmp_lg_u32 s19, 0
; GFX10-NEXT: s_cselect_b64 s[8:9], s[8:9], 0
; GFX10-NEXT: s_or_b64 s[4:5], s[10:11], s[4:5]
; GFX10-NEXT: s_or_b64 s[6:7], s[6:7], s[8:9]
; GFX10-NEXT: ; return to shader part epilog
%result = call <2 x i128> @llvm.fshr.v2i128(<2 x i128> %lhs, <2 x i128> %rhs, <2 x i128> %amt)
ret <2 x i128> %result
}
define <2 x i128> @v_fshr_v2i128(<2 x i128> %lhs, <2 x i128> %rhs, <2 x i128> %amt) {
; GFX6-LABEL: v_fshr_v2i128:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_sub_i32 s6, 64, 1
; GFX6-NEXT: s_sub_i32 s7, 1, 64
; GFX6-NEXT: s_cmp_lt_u32 1, 64
; GFX6-NEXT: v_lshr_b64 v[17:18], v[0:1], s6
; GFX6-NEXT: v_lshl_b64 v[21:22], v[2:3], 1
; GFX6-NEXT: s_cselect_b32 s4, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 1, 0
; GFX6-NEXT: v_or_b32_e32 v19, v17, v21
; GFX6-NEXT: v_or_b32_e32 v21, v18, v22
; GFX6-NEXT: s_cselect_b32 s5, 1, 0
; GFX6-NEXT: v_lshl_b64 v[17:18], v[0:1], 1
; GFX6-NEXT: s_and_b32 s4, 1, s4
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], s7
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: s_and_b32 s4, 1, s5
; GFX6-NEXT: v_cndmask_b32_e32 v0, v0, v19, vcc
; GFX6-NEXT: v_cmp_ne_u32_e64 s[4:5], 0, s4
; GFX6-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[4:5]
; GFX6-NEXT: s_movk_i32 s8, 0x7f
; GFX6-NEXT: v_xor_b32_e32 v2, -1, v16
; GFX6-NEXT: v_cndmask_b32_e32 v1, v1, v21, vcc
; GFX6-NEXT: v_and_b32_e32 v19, s8, v2
; GFX6-NEXT: v_cndmask_b32_e64 v1, v1, v3, s[4:5]
; GFX6-NEXT: v_cndmask_b32_e32 v17, 0, v17, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v18, 0, v18, vcc
; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 64, v19
; GFX6-NEXT: v_lshr_b64 v[2:3], v[17:18], v2
; GFX6-NEXT: v_lshl_b64 v[21:22], v[0:1], v19
; GFX6-NEXT: v_and_b32_e32 v25, s8, v16
; GFX6-NEXT: v_or_b32_e32 v23, v2, v21
; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 64, v25
; GFX6-NEXT: v_or_b32_e32 v24, v3, v22
; GFX6-NEXT: v_lshl_b64 v[2:3], v[10:11], v2
; GFX6-NEXT: v_lshr_b64 v[21:22], v[8:9], v25
; GFX6-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v19
; GFX6-NEXT: v_or_b32_e32 v21, v21, v2
; GFX6-NEXT: v_subrev_i32_e32 v2, vcc, 64, v19
; GFX6-NEXT: v_or_b32_e32 v22, v22, v3
; GFX6-NEXT: v_lshl_b64 v[2:3], v[17:18], v2
; GFX6-NEXT: v_cmp_gt_u32_e32 vcc, 64, v19
; GFX6-NEXT: v_cndmask_b32_e32 v2, v2, v23, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v3, v3, v24, vcc
; GFX6-NEXT: v_cndmask_b32_e64 v2, v2, v0, s[4:5]
; GFX6-NEXT: v_cndmask_b32_e64 v3, v3, v1, s[4:5]
; GFX6-NEXT: v_subrev_i32_e64 v0, s[4:5], 64, v25
; GFX6-NEXT: v_lshl_b64 v[16:17], v[17:18], v19
; GFX6-NEXT: v_lshr_b64 v[0:1], v[10:11], v0
; GFX6-NEXT: v_cmp_gt_u32_e64 s[4:5], 64, v25
; GFX6-NEXT: v_cndmask_b32_e32 v16, 0, v16, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v17, 0, v17, vcc
; GFX6-NEXT: v_cndmask_b32_e64 v18, v0, v21, s[4:5]
; GFX6-NEXT: v_cndmask_b32_e64 v19, v1, v22, s[4:5]
; GFX6-NEXT: v_lshr_b64 v[0:1], v[10:11], v25
; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, 0, v25
; GFX6-NEXT: v_cndmask_b32_e32 v8, v18, v8, vcc
; GFX6-NEXT: v_cndmask_b32_e64 v10, 0, v0, s[4:5]
; GFX6-NEXT: v_or_b32_e32 v0, v16, v8
; GFX6-NEXT: v_cndmask_b32_e64 v11, 0, v1, s[4:5]
; GFX6-NEXT: v_cndmask_b32_e32 v9, v19, v9, vcc
; GFX6-NEXT: v_xor_b32_e32 v8, -1, v20
; GFX6-NEXT: v_or_b32_e32 v1, v17, v9
; GFX6-NEXT: v_and_b32_e32 v17, s8, v8
; GFX6-NEXT: s_cmp_lt_u32 1, 64
; GFX6-NEXT: v_or_b32_e32 v2, v2, v10
; GFX6-NEXT: v_or_b32_e32 v3, v3, v11
; GFX6-NEXT: v_lshr_b64 v[8:9], v[4:5], s6
; GFX6-NEXT: v_lshl_b64 v[10:11], v[6:7], 1
; GFX6-NEXT: s_cselect_b32 s4, 1, 0
; GFX6-NEXT: s_cmp_eq_u32 1, 0
; GFX6-NEXT: s_cselect_b32 s5, 1, 0
; GFX6-NEXT: s_and_b32 s4, 1, s4
; GFX6-NEXT: v_or_b32_e32 v10, v8, v10
; GFX6-NEXT: v_or_b32_e32 v11, v9, v11
; GFX6-NEXT: v_lshl_b64 v[8:9], v[4:5], 1
; GFX6-NEXT: v_lshl_b64 v[4:5], v[4:5], s7
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: s_and_b32 s4, 1, s5
; GFX6-NEXT: v_cndmask_b32_e32 v8, 0, v8, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v9, 0, v9, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v4, v4, v10, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v5, v5, v11, vcc
; GFX6-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX6-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc
; GFX6-NEXT: v_sub_i32_e32 v6, vcc, 64, v17
; GFX6-NEXT: v_lshr_b64 v[6:7], v[8:9], v6
; GFX6-NEXT: v_lshl_b64 v[10:11], v[4:5], v17
; GFX6-NEXT: v_subrev_i32_e32 v18, vcc, 64, v17
; GFX6-NEXT: v_or_b32_e32 v10, v6, v10
; GFX6-NEXT: v_or_b32_e32 v11, v7, v11
; GFX6-NEXT: v_lshl_b64 v[6:7], v[8:9], v17
; GFX6-NEXT: v_lshl_b64 v[8:9], v[8:9], v18
; GFX6-NEXT: v_cmp_gt_u32_e32 vcc, 64, v17
; GFX6-NEXT: v_and_b32_e32 v16, s8, v20
; GFX6-NEXT: v_cndmask_b32_e32 v18, 0, v6, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v19, 0, v7, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v6, v8, v10, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v7, v9, v11, vcc
; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, 0, v17
; GFX6-NEXT: v_cndmask_b32_e32 v8, v6, v4, vcc
; GFX6-NEXT: v_cndmask_b32_e32 v9, v7, v5, vcc
; GFX6-NEXT: v_sub_i32_e32 v6, vcc, 64, v16
; GFX6-NEXT: v_lshr_b64 v[4:5], v[12:13], v16
; GFX6-NEXT: v_lshl_b64 v[6:7], v[14:15], v6
; GFX6-NEXT: v_subrev_i32_e32 v10, vcc, 64, v16
; GFX6-NEXT: v_or_b32_e32 v11, v4, v6
; GFX6-NEXT: v_or_b32_e32 v17, v5, v7
; GFX6-NEXT: v_lshr_b64 v[6:7], v[14:15], v10
; GFX6-NEXT: v_cmp_gt_u32_e32 vcc, 64, v16
; GFX6-NEXT: v_lshr_b64 v[4:5], v[14:15], v16
; GFX6-NEXT: v_cndmask_b32_e32 v6, v6, v11, vcc
; GFX6-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v16
; GFX6-NEXT: v_cndmask_b32_e32 v7, v7, v17, vcc
; GFX6-NEXT: v_cndmask_b32_e64 v6, v6, v12, s[4:5]
; GFX6-NEXT: v_cndmask_b32_e32 v10, 0, v4, vcc
; GFX6-NEXT: v_cndmask_b32_e64 v7, v7, v13, s[4:5]
; GFX6-NEXT: v_cndmask_b32_e32 v11, 0, v5, vcc
; GFX6-NEXT: v_or_b32_e32 v4, v18, v6
; GFX6-NEXT: v_or_b32_e32 v5, v19, v7
; GFX6-NEXT: v_or_b32_e32 v6, v8, v10
; GFX6-NEXT: v_or_b32_e32 v7, v9, v11
; GFX6-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: v_fshr_v2i128:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_sub_i32 s6, 64, 1
; GFX8-NEXT: s_sub_i32 s7, 1, 64
; GFX8-NEXT: s_cmp_lt_u32 1, 64
; GFX8-NEXT: v_lshrrev_b64 v[17:18], s6, v[0:1]
; GFX8-NEXT: v_lshlrev_b64 v[21:22], 1, v[2:3]
; GFX8-NEXT: s_cselect_b32 s4, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 1, 0
; GFX8-NEXT: v_or_b32_e32 v19, v17, v21
; GFX8-NEXT: v_or_b32_e32 v21, v18, v22
; GFX8-NEXT: s_cselect_b32 s5, 1, 0
; GFX8-NEXT: v_lshlrev_b64 v[17:18], 1, v[0:1]
; GFX8-NEXT: s_and_b32 s4, 1, s4
; GFX8-NEXT: v_lshlrev_b64 v[0:1], s7, v[0:1]
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: s_and_b32 s4, 1, s5
; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v19, vcc
; GFX8-NEXT: v_cmp_ne_u32_e64 s[4:5], 0, s4
; GFX8-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[4:5]
; GFX8-NEXT: s_movk_i32 s8, 0x7f
; GFX8-NEXT: v_xor_b32_e32 v2, -1, v16
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v21, vcc
; GFX8-NEXT: v_and_b32_e32 v19, s8, v2
; GFX8-NEXT: v_cndmask_b32_e64 v1, v1, v3, s[4:5]
; GFX8-NEXT: v_cndmask_b32_e32 v17, 0, v17, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v18, 0, v18, vcc
; GFX8-NEXT: v_sub_u32_e32 v2, vcc, 64, v19
; GFX8-NEXT: v_lshrrev_b64 v[2:3], v2, v[17:18]
; GFX8-NEXT: v_lshlrev_b64 v[21:22], v19, v[0:1]
; GFX8-NEXT: v_and_b32_e32 v25, s8, v16
; GFX8-NEXT: v_or_b32_e32 v23, v2, v21
; GFX8-NEXT: v_sub_u32_e32 v2, vcc, 64, v25
; GFX8-NEXT: v_or_b32_e32 v24, v3, v22
; GFX8-NEXT: v_lshlrev_b64 v[2:3], v2, v[10:11]
; GFX8-NEXT: v_lshrrev_b64 v[21:22], v25, v[8:9]
; GFX8-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v19
; GFX8-NEXT: v_or_b32_e32 v21, v21, v2
; GFX8-NEXT: v_subrev_u32_e32 v2, vcc, 64, v19
; GFX8-NEXT: v_or_b32_e32 v22, v22, v3
; GFX8-NEXT: v_lshlrev_b64 v[2:3], v2, v[17:18]
; GFX8-NEXT: v_cmp_gt_u32_e32 vcc, 64, v19
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v23, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v3, v3, v24, vcc
; GFX8-NEXT: v_cndmask_b32_e64 v2, v2, v0, s[4:5]
; GFX8-NEXT: v_cndmask_b32_e64 v3, v3, v1, s[4:5]
; GFX8-NEXT: v_subrev_u32_e64 v0, s[4:5], 64, v25
; GFX8-NEXT: v_lshlrev_b64 v[16:17], v19, v[17:18]
; GFX8-NEXT: v_lshrrev_b64 v[0:1], v0, v[10:11]
; GFX8-NEXT: v_cmp_gt_u32_e64 s[4:5], 64, v25
; GFX8-NEXT: v_cndmask_b32_e32 v16, 0, v16, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v17, 0, v17, vcc
; GFX8-NEXT: v_cndmask_b32_e64 v18, v0, v21, s[4:5]
; GFX8-NEXT: v_cndmask_b32_e64 v19, v1, v22, s[4:5]
; GFX8-NEXT: v_lshrrev_b64 v[0:1], v25, v[10:11]
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v25
; GFX8-NEXT: v_cndmask_b32_e32 v8, v18, v8, vcc
; GFX8-NEXT: v_cndmask_b32_e64 v10, 0, v0, s[4:5]
; GFX8-NEXT: v_or_b32_e32 v0, v16, v8
; GFX8-NEXT: v_cndmask_b32_e64 v11, 0, v1, s[4:5]
; GFX8-NEXT: v_cndmask_b32_e32 v9, v19, v9, vcc
; GFX8-NEXT: v_xor_b32_e32 v8, -1, v20
; GFX8-NEXT: v_or_b32_e32 v1, v17, v9
; GFX8-NEXT: v_and_b32_e32 v17, s8, v8
; GFX8-NEXT: s_cmp_lt_u32 1, 64
; GFX8-NEXT: v_or_b32_e32 v2, v2, v10
; GFX8-NEXT: v_or_b32_e32 v3, v3, v11
; GFX8-NEXT: v_lshrrev_b64 v[8:9], s6, v[4:5]
; GFX8-NEXT: v_lshlrev_b64 v[10:11], 1, v[6:7]
; GFX8-NEXT: s_cselect_b32 s4, 1, 0
; GFX8-NEXT: s_cmp_eq_u32 1, 0
; GFX8-NEXT: s_cselect_b32 s5, 1, 0
; GFX8-NEXT: s_and_b32 s4, 1, s4
; GFX8-NEXT: v_or_b32_e32 v10, v8, v10
; GFX8-NEXT: v_or_b32_e32 v11, v9, v11
; GFX8-NEXT: v_lshlrev_b64 v[8:9], 1, v[4:5]
; GFX8-NEXT: v_lshlrev_b64 v[4:5], s7, v[4:5]
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: s_and_b32 s4, 1, s5
; GFX8-NEXT: v_cndmask_b32_e32 v8, 0, v8, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v9, 0, v9, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v4, v4, v10, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v5, v5, v11, vcc
; GFX8-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX8-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc
; GFX8-NEXT: v_sub_u32_e32 v6, vcc, 64, v17
; GFX8-NEXT: v_lshrrev_b64 v[6:7], v6, v[8:9]
; GFX8-NEXT: v_lshlrev_b64 v[10:11], v17, v[4:5]
; GFX8-NEXT: v_subrev_u32_e32 v18, vcc, 64, v17
; GFX8-NEXT: v_or_b32_e32 v10, v6, v10
; GFX8-NEXT: v_or_b32_e32 v11, v7, v11
; GFX8-NEXT: v_lshlrev_b64 v[6:7], v17, v[8:9]
; GFX8-NEXT: v_lshlrev_b64 v[8:9], v18, v[8:9]
; GFX8-NEXT: v_cmp_gt_u32_e32 vcc, 64, v17
; GFX8-NEXT: v_and_b32_e32 v16, s8, v20
; GFX8-NEXT: v_cndmask_b32_e32 v18, 0, v6, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v19, 0, v7, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v6, v8, v10, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v7, v9, v11, vcc
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v17
; GFX8-NEXT: v_cndmask_b32_e32 v8, v6, v4, vcc
; GFX8-NEXT: v_cndmask_b32_e32 v9, v7, v5, vcc
; GFX8-NEXT: v_sub_u32_e32 v6, vcc, 64, v16
; GFX8-NEXT: v_lshrrev_b64 v[4:5], v16, v[12:13]
; GFX8-NEXT: v_lshlrev_b64 v[6:7], v6, v[14:15]
; GFX8-NEXT: v_subrev_u32_e32 v10, vcc, 64, v16
; GFX8-NEXT: v_or_b32_e32 v11, v4, v6
; GFX8-NEXT: v_or_b32_e32 v17, v5, v7
; GFX8-NEXT: v_lshrrev_b64 v[6:7], v10, v[14:15]
; GFX8-NEXT: v_cmp_gt_u32_e32 vcc, 64, v16
; GFX8-NEXT: v_lshrrev_b64 v[4:5], v16, v[14:15]
; GFX8-NEXT: v_cndmask_b32_e32 v6, v6, v11, vcc
; GFX8-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v16
; GFX8-NEXT: v_cndmask_b32_e32 v7, v7, v17, vcc
; GFX8-NEXT: v_cndmask_b32_e64 v6, v6, v12, s[4:5]
; GFX8-NEXT: v_cndmask_b32_e32 v10, 0, v4, vcc
; GFX8-NEXT: v_cndmask_b32_e64 v7, v7, v13, s[4:5]
; GFX8-NEXT: v_cndmask_b32_e32 v11, 0, v5, vcc
; GFX8-NEXT: v_or_b32_e32 v4, v18, v6
; GFX8-NEXT: v_or_b32_e32 v5, v19, v7
; GFX8-NEXT: v_or_b32_e32 v6, v8, v10
; GFX8-NEXT: v_or_b32_e32 v7, v9, v11
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: v_fshr_v2i128:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_sub_i32 s6, 64, 1
; GFX9-NEXT: s_sub_i32 s7, 1, 64
; GFX9-NEXT: s_cmp_lt_u32 1, 64
; GFX9-NEXT: v_lshrrev_b64 v[17:18], s6, v[0:1]
; GFX9-NEXT: v_lshlrev_b64 v[21:22], 1, v[2:3]
; GFX9-NEXT: s_cselect_b32 s4, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 1, 0
; GFX9-NEXT: v_or_b32_e32 v19, v17, v21
; GFX9-NEXT: v_or_b32_e32 v21, v18, v22
; GFX9-NEXT: s_cselect_b32 s5, 1, 0
; GFX9-NEXT: v_lshlrev_b64 v[17:18], 1, v[0:1]
; GFX9-NEXT: s_and_b32 s4, 1, s4
; GFX9-NEXT: v_lshlrev_b64 v[0:1], s7, v[0:1]
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: s_and_b32 s4, 1, s5
; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v19, vcc
; GFX9-NEXT: v_cmp_ne_u32_e64 s[4:5], 0, s4
; GFX9-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[4:5]
; GFX9-NEXT: s_movk_i32 s8, 0x7f
; GFX9-NEXT: v_xor_b32_e32 v2, -1, v16
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v21, vcc
; GFX9-NEXT: v_and_b32_e32 v19, s8, v2
; GFX9-NEXT: v_cndmask_b32_e64 v1, v1, v3, s[4:5]
; GFX9-NEXT: v_cndmask_b32_e32 v17, 0, v17, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v18, 0, v18, vcc
; GFX9-NEXT: v_sub_u32_e32 v2, 64, v19
; GFX9-NEXT: v_lshrrev_b64 v[2:3], v2, v[17:18]
; GFX9-NEXT: v_lshlrev_b64 v[21:22], v19, v[0:1]
; GFX9-NEXT: v_and_b32_e32 v25, s8, v16
; GFX9-NEXT: v_or_b32_e32 v23, v2, v21
; GFX9-NEXT: v_sub_u32_e32 v2, 64, v25
; GFX9-NEXT: v_or_b32_e32 v24, v3, v22
; GFX9-NEXT: v_lshlrev_b64 v[2:3], v2, v[10:11]
; GFX9-NEXT: v_lshrrev_b64 v[21:22], v25, v[8:9]
; GFX9-NEXT: v_cmp_gt_u32_e32 vcc, 64, v19
; GFX9-NEXT: v_or_b32_e32 v21, v21, v2
; GFX9-NEXT: v_subrev_u32_e32 v2, 64, v19
; GFX9-NEXT: v_or_b32_e32 v22, v22, v3
; GFX9-NEXT: v_lshlrev_b64 v[2:3], v2, v[17:18]
; GFX9-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v19
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v23, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v3, v3, v24, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v2, v2, v0, s[4:5]
; GFX9-NEXT: v_subrev_u32_e32 v0, 64, v25
; GFX9-NEXT: v_lshlrev_b64 v[16:17], v19, v[17:18]
; GFX9-NEXT: v_cndmask_b32_e64 v3, v3, v1, s[4:5]
; GFX9-NEXT: v_lshrrev_b64 v[0:1], v0, v[10:11]
; GFX9-NEXT: v_cmp_gt_u32_e64 s[4:5], 64, v25
; GFX9-NEXT: v_cndmask_b32_e32 v16, 0, v16, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v17, 0, v17, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v18, v0, v21, s[4:5]
; GFX9-NEXT: v_cndmask_b32_e64 v19, v1, v22, s[4:5]
; GFX9-NEXT: v_lshrrev_b64 v[0:1], v25, v[10:11]
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v25
; GFX9-NEXT: v_cndmask_b32_e32 v8, v18, v8, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v10, 0, v0, s[4:5]
; GFX9-NEXT: v_or_b32_e32 v0, v16, v8
; GFX9-NEXT: v_cndmask_b32_e64 v11, 0, v1, s[4:5]
; GFX9-NEXT: v_cndmask_b32_e32 v9, v19, v9, vcc
; GFX9-NEXT: v_xor_b32_e32 v8, -1, v20
; GFX9-NEXT: v_or_b32_e32 v1, v17, v9
; GFX9-NEXT: v_and_b32_e32 v17, s8, v8
; GFX9-NEXT: s_cmp_lt_u32 1, 64
; GFX9-NEXT: v_or_b32_e32 v2, v2, v10
; GFX9-NEXT: v_or_b32_e32 v3, v3, v11
; GFX9-NEXT: v_lshrrev_b64 v[8:9], s6, v[4:5]
; GFX9-NEXT: v_lshlrev_b64 v[10:11], 1, v[6:7]
; GFX9-NEXT: s_cselect_b32 s4, 1, 0
; GFX9-NEXT: s_cmp_eq_u32 1, 0
; GFX9-NEXT: s_cselect_b32 s5, 1, 0
; GFX9-NEXT: s_and_b32 s4, 1, s4
; GFX9-NEXT: v_or_b32_e32 v10, v8, v10
; GFX9-NEXT: v_or_b32_e32 v11, v9, v11
; GFX9-NEXT: v_lshlrev_b64 v[8:9], 1, v[4:5]
; GFX9-NEXT: v_lshlrev_b64 v[4:5], s7, v[4:5]
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: s_and_b32 s4, 1, s5
; GFX9-NEXT: v_cndmask_b32_e32 v8, 0, v8, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v9, 0, v9, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v4, v4, v10, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v5, v5, v11, vcc
; GFX9-NEXT: v_cmp_ne_u32_e64 vcc, 0, s4
; GFX9-NEXT: v_cndmask_b32_e32 v4, v4, v6, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v5, v5, v7, vcc
; GFX9-NEXT: v_sub_u32_e32 v6, 64, v17
; GFX9-NEXT: v_lshrrev_b64 v[6:7], v6, v[8:9]
; GFX9-NEXT: v_lshlrev_b64 v[10:11], v17, v[4:5]
; GFX9-NEXT: v_subrev_u32_e32 v18, 64, v17
; GFX9-NEXT: v_or_b32_e32 v10, v6, v10
; GFX9-NEXT: v_or_b32_e32 v11, v7, v11
; GFX9-NEXT: v_lshlrev_b64 v[6:7], v17, v[8:9]
; GFX9-NEXT: v_lshlrev_b64 v[8:9], v18, v[8:9]
; GFX9-NEXT: v_cmp_gt_u32_e32 vcc, 64, v17
; GFX9-NEXT: v_and_b32_e32 v16, s8, v20
; GFX9-NEXT: v_cndmask_b32_e32 v18, 0, v6, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v19, 0, v7, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v6, v8, v10, vcc
; GFX9-NEXT: v_cndmask_b32_e32 v7, v9, v11, vcc
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v17
; GFX9-NEXT: v_cndmask_b32_e32 v8, v6, v4, vcc
; GFX9-NEXT: v_sub_u32_e32 v6, 64, v16
; GFX9-NEXT: v_cndmask_b32_e32 v9, v7, v5, vcc
; GFX9-NEXT: v_lshrrev_b64 v[4:5], v16, v[12:13]
; GFX9-NEXT: v_lshlrev_b64 v[6:7], v6, v[14:15]
; GFX9-NEXT: v_subrev_u32_e32 v10, 64, v16
; GFX9-NEXT: v_or_b32_e32 v11, v4, v6
; GFX9-NEXT: v_or_b32_e32 v17, v5, v7
; GFX9-NEXT: v_lshrrev_b64 v[6:7], v10, v[14:15]
; GFX9-NEXT: v_cmp_gt_u32_e32 vcc, 64, v16
; GFX9-NEXT: v_lshrrev_b64 v[4:5], v16, v[14:15]
; GFX9-NEXT: v_cndmask_b32_e32 v6, v6, v11, vcc
; GFX9-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v16
; GFX9-NEXT: v_cndmask_b32_e32 v7, v7, v17, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v6, v6, v12, s[4:5]
; GFX9-NEXT: v_cndmask_b32_e32 v10, 0, v4, vcc
; GFX9-NEXT: v_cndmask_b32_e64 v7, v7, v13, s[4:5]
; GFX9-NEXT: v_cndmask_b32_e32 v11, 0, v5, vcc
; GFX9-NEXT: v_or_b32_e32 v4, v18, v6
; GFX9-NEXT: v_or_b32_e32 v5, v19, v7
; GFX9-NEXT: v_or_b32_e32 v6, v8, v10
; GFX9-NEXT: v_or_b32_e32 v7, v9, v11
; GFX9-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_fshr_v2i128:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: s_sub_i32 s5, 64, 1
; GFX10-NEXT: s_sub_i32 s6, 1, 64
; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_lshrrev_b64 v[17:18], s5, v[0:1]
; GFX10-NEXT: v_lshlrev_b64 v[21:22], 1, v[2:3]
; GFX10-NEXT: s_cselect_b32 vcc_lo, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[23:24], 1, v[0:1]
; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: s_and_b32 s4, 1, vcc_lo
; GFX10-NEXT: v_lshlrev_b64 v[0:1], s6, v[0:1]
; GFX10-NEXT: v_or_b32_e32 v21, v17, v21
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s4
; GFX10-NEXT: v_or_b32_e32 v18, v18, v22
; GFX10-NEXT: v_xor_b32_e32 v19, -1, v16
; GFX10-NEXT: s_movk_i32 s7, 0x7f
; GFX10-NEXT: s_and_b32 s8, 1, s8
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v21, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v18, s4
; GFX10-NEXT: v_cmp_ne_u32_e64 s4, 0, s8
; GFX10-NEXT: v_and_b32_e32 v25, s7, v19
; GFX10-NEXT: v_cndmask_b32_e32 v17, 0, v23, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v18, 0, v24, vcc_lo
; GFX10-NEXT: v_and_b32_e32 v26, s7, v16
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v2, s4
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, v3, s4
; GFX10-NEXT: v_sub_nc_u32_e32 v2, 64, v25
; GFX10-NEXT: v_subrev_nc_u32_e32 v16, 64, v25
; GFX10-NEXT: v_sub_nc_u32_e32 v19, 64, v26
; GFX10-NEXT: v_lshlrev_b64 v[23:24], v25, v[17:18]
; GFX10-NEXT: v_lshlrev_b64 v[21:22], v25, v[0:1]
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v2, v[17:18]
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v25
; GFX10-NEXT: v_subrev_nc_u32_e32 v27, 64, v26
; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v26
; GFX10-NEXT: s_cmp_lt_u32 1, 64
; GFX10-NEXT: v_or_b32_e32 v21, v2, v21
; GFX10-NEXT: v_or_b32_e32 v22, v3, v22
; GFX10-NEXT: v_lshlrev_b64 v[2:3], v16, v[17:18]
; GFX10-NEXT: v_lshrrev_b64 v[16:17], v26, v[8:9]
; GFX10-NEXT: v_lshlrev_b64 v[18:19], v19, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e32 v23, 0, v23, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v24, 0, v24, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v21, v2, v21, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v22, v3, v22, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[2:3], v27, v[10:11]
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v25
; GFX10-NEXT: v_or_b32_e32 v16, v16, v18
; GFX10-NEXT: v_or_b32_e32 v17, v17, v19
; GFX10-NEXT: v_xor_b32_e32 v25, -1, v20
; GFX10-NEXT: v_cndmask_b32_e32 v18, v21, v0, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v22, v22, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v16, s4
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v26
; GFX10-NEXT: v_lshrrev_b64 v[0:1], v26, v[10:11]
; GFX10-NEXT: v_cndmask_b32_e64 v19, v3, v17, s4
; GFX10-NEXT: v_lshlrev_b64 v[10:11], 1, v[6:7]
; GFX10-NEXT: v_lshlrev_b64 v[16:17], 1, v[4:5]
; GFX10-NEXT: v_cndmask_b32_e32 v21, v2, v8, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[2:3], s5, v[4:5]
; GFX10-NEXT: s_cselect_b32 s5, 1, 0
; GFX10-NEXT: s_cmp_eq_u32 1, 0
; GFX10-NEXT: v_lshlrev_b64 v[4:5], s6, v[4:5]
; GFX10-NEXT: s_cselect_b32 s8, 1, 0
; GFX10-NEXT: s_and_b32 s6, 1, s5
; GFX10-NEXT: v_or_b32_e32 v2, v2, v10
; GFX10-NEXT: v_cmp_ne_u32_e64 s6, 0, s6
; GFX10-NEXT: v_or_b32_e32 v3, v3, v11
; GFX10-NEXT: s_and_b32 s8, 1, s8
; GFX10-NEXT: v_cndmask_b32_e64 v8, 0, v16, s5
; GFX10-NEXT: v_cndmask_b32_e32 v16, v19, v9, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v4, v2, s6
; GFX10-NEXT: v_and_b32_e32 v25, s7, v25
; GFX10-NEXT: v_cndmask_b32_e64 v4, v5, v3, s6
; GFX10-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s8
; GFX10-NEXT: v_cndmask_b32_e64 v9, 0, v17, s5
; GFX10-NEXT: v_cndmask_b32_e64 v19, 0, v0, s4
; GFX10-NEXT: v_or_b32_e32 v0, v23, v21
; GFX10-NEXT: v_and_b32_e32 v23, s7, v20
; GFX10-NEXT: v_cndmask_b32_e32 v3, v2, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v4, v4, v7, vcc_lo
; GFX10-NEXT: v_sub_nc_u32_e32 v2, 64, v25
; GFX10-NEXT: v_subrev_nc_u32_e32 v7, 64, v25
; GFX10-NEXT: v_sub_nc_u32_e32 v20, 64, v23
; GFX10-NEXT: v_cndmask_b32_e64 v26, 0, v1, s4
; GFX10-NEXT: v_lshlrev_b64 v[10:11], v25, v[3:4]
; GFX10-NEXT: v_lshrrev_b64 v[5:6], v2, v[8:9]
; GFX10-NEXT: v_or_b32_e32 v1, v24, v16
; GFX10-NEXT: v_or_b32_e32 v2, v18, v19
; GFX10-NEXT: v_lshlrev_b64 v[16:17], v25, v[8:9]
; GFX10-NEXT: v_lshrrev_b64 v[18:19], v23, v[12:13]
; GFX10-NEXT: v_lshlrev_b64 v[20:21], v20, v[14:15]
; GFX10-NEXT: v_or_b32_e32 v10, v5, v10
; GFX10-NEXT: v_subrev_nc_u32_e32 v5, 64, v23
; GFX10-NEXT: v_cmp_gt_u32_e32 vcc_lo, 64, v25
; GFX10-NEXT: v_lshlrev_b64 v[7:8], v7, v[8:9]
; GFX10-NEXT: v_or_b32_e32 v9, v6, v11
; GFX10-NEXT: v_cmp_gt_u32_e64 s4, 64, v23
; GFX10-NEXT: v_lshrrev_b64 v[5:6], v5, v[14:15]
; GFX10-NEXT: v_cndmask_b32_e32 v11, 0, v16, vcc_lo
; GFX10-NEXT: v_or_b32_e32 v16, v18, v20
; GFX10-NEXT: v_or_b32_e32 v18, v19, v21
; GFX10-NEXT: v_cndmask_b32_e32 v10, v7, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v9, v8, v9, vcc_lo
; GFX10-NEXT: v_lshrrev_b64 v[7:8], v23, v[14:15]
; GFX10-NEXT: v_cmp_eq_u32_e64 s6, 0, v25
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v16, s4
; GFX10-NEXT: v_cmp_eq_u32_e64 s5, 0, v23
; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v18, s4
; GFX10-NEXT: v_cndmask_b32_e32 v14, 0, v17, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v10, v10, v3, s6
; GFX10-NEXT: v_cndmask_b32_e64 v9, v9, v4, s6
; GFX10-NEXT: v_cndmask_b32_e64 v4, v5, v12, s5
; GFX10-NEXT: v_cndmask_b32_e64 v5, v6, v13, s5
; GFX10-NEXT: v_cndmask_b32_e64 v6, 0, v7, s4
; GFX10-NEXT: v_cndmask_b32_e64 v7, 0, v8, s4
; GFX10-NEXT: v_or_b32_e32 v3, v22, v26
; GFX10-NEXT: v_or_b32_e32 v4, v11, v4
; GFX10-NEXT: v_or_b32_e32 v5, v14, v5
; GFX10-NEXT: v_or_b32_e32 v6, v10, v6
; GFX10-NEXT: v_or_b32_e32 v7, v9, v7
; GFX10-NEXT: s_setpc_b64 s[30:31]
%result = call <2 x i128> @llvm.fshr.v2i128(<2 x i128> %lhs, <2 x i128> %rhs, <2 x i128> %amt)
ret <2 x i128> %result
}
declare i7 @llvm.fshr.i7(i7, i7, i7) #0
declare i8 @llvm.fshr.i8(i8, i8, i8) #0
declare <2 x i8> @llvm.fshr.v2i8(<2 x i8>, <2 x i8>, <2 x i8>) #0
declare <4 x i8> @llvm.fshr.v4i8(<4 x i8>, <4 x i8>, <4 x i8>) #0
declare i16 @llvm.fshr.i16(i16, i16, i16) #0
declare <2 x i16> @llvm.fshr.v2i16(<2 x i16>, <2 x i16>, <2 x i16>) #0
declare <3 x i16> @llvm.fshr.v3i16(<3 x i16>, <3 x i16>, <3 x i16>) #0
declare <4 x i16> @llvm.fshr.v4i16(<4 x i16>, <4 x i16>, <4 x i16>) #0
declare <5 x i16> @llvm.fshr.v5i16(<5 x i16>, <5 x i16>, <5 x i16>) #0
declare <6 x i16> @llvm.fshr.v6i16(<6 x i16>, <6 x i16>, <6 x i16>) #0
declare <8 x i16> @llvm.fshr.v8i16(<8 x i16>, <8 x i16>, <8 x i16>) #0
declare i24 @llvm.fshr.i24(i24, i24, i24) #0
declare <2 x i24> @llvm.fshr.v2i24(<2 x i24>, <2 x i24>, <2 x i24>) #0
declare i32 @llvm.fshr.i32(i32, i32, i32) #0
declare <2 x i32> @llvm.fshr.v2i32(<2 x i32>, <2 x i32>, <2 x i32>) #0
declare <3 x i32> @llvm.fshr.v3i32(<3 x i32>, <3 x i32>, <3 x i32>) #0
declare <4 x i32> @llvm.fshr.v4i32(<4 x i32>, <4 x i32>, <4 x i32>) #0
declare <5 x i32> @llvm.fshr.v5i32(<5 x i32>, <5 x i32>, <5 x i32>) #0
declare <16 x i32> @llvm.fshr.v16i32(<16 x i32>, <16 x i32>, <16 x i32>) #0
declare i48 @llvm.fshr.i48(i48, i48, i48) #0
declare i64 @llvm.fshr.i64(i64, i64, i64) #0
declare <2 x i64> @llvm.fshr.v2i64(<2 x i64>, <2 x i64>, <2 x i64>) #0
declare i128 @llvm.fshr.i128(i128, i128, i128) #0
declare <2 x i128> @llvm.fshr.v2i128(<2 x i128>, <2 x i128>, <2 x i128>) #0
attributes #0 = { nounwind readnone speculatable willreturn }