| ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 |
| ; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s |
| ; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 < %s | FileCheck -check-prefix=GFX11 %s |
| |
| define i16 @v_underflow_compare_fold_i16(i16 %a, i16 %b) #0 { |
| ; GFX9-LABEL: v_underflow_compare_fold_i16: |
| ; GFX9: ; %bb.0: |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX9-NEXT: v_sub_u16_e32 v1, v0, v1 |
| ; GFX9-NEXT: v_min_u16_e32 v0, v1, v0 |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; GFX11-LABEL: v_underflow_compare_fold_i16: |
| ; GFX11: ; %bb.0: |
| ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX11-NEXT: v_sub_nc_u16 v0.h, v0.l, v1.l |
| ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) |
| ; GFX11-NEXT: v_min_u16 v0.l, v0.h, v0.l |
| ; GFX11-NEXT: s_setpc_b64 s[30:31] |
| %sub = sub i16 %a, %b |
| %cond = call i16 @llvm.umin.i16(i16 %sub, i16 %a) |
| ret i16 %cond |
| } |
| |
| define i32 @v_underflow_compare_fold_i32(i32 %a, i32 %b) #0 { |
| ; GFX9-LABEL: v_underflow_compare_fold_i32: |
| ; GFX9: ; %bb.0: |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX9-NEXT: v_sub_u32_e32 v1, v0, v1 |
| ; GFX9-NEXT: v_min_u32_e32 v0, v1, v0 |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; GFX11-LABEL: v_underflow_compare_fold_i32: |
| ; GFX11: ; %bb.0: |
| ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX11-NEXT: v_sub_nc_u32_e32 v1, v0, v1 |
| ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) |
| ; GFX11-NEXT: v_min_u32_e32 v0, v1, v0 |
| ; GFX11-NEXT: s_setpc_b64 s[30:31] |
| %sub = sub i32 %a, %b |
| %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a) |
| ret i32 %cond |
| } |
| |
| define i32 @v_underflow_compare_fold_i32_commute(i32 %a, i32 %b) #0 { |
| ; GFX9-LABEL: v_underflow_compare_fold_i32_commute: |
| ; GFX9: ; %bb.0: |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX9-NEXT: v_sub_u32_e32 v1, v0, v1 |
| ; GFX9-NEXT: v_min_u32_e32 v0, v0, v1 |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; GFX11-LABEL: v_underflow_compare_fold_i32_commute: |
| ; GFX11: ; %bb.0: |
| ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX11-NEXT: v_sub_nc_u32_e32 v1, v0, v1 |
| ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) |
| ; GFX11-NEXT: v_min_u32_e32 v0, v0, v1 |
| ; GFX11-NEXT: s_setpc_b64 s[30:31] |
| %sub = sub i32 %a, %b |
| %cond = call i32 @llvm.umin.i32(i32 %a, i32 %sub) |
| ret i32 %cond |
| } |
| |
| define i32 @v_underflow_compare_fold_i32_multi_use(i32 %a, i32 %b, ptr addrspace(1) %ptr) #0 { |
| ; GFX9-LABEL: v_underflow_compare_fold_i32_multi_use: |
| ; GFX9: ; %bb.0: |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX9-NEXT: v_sub_u32_e32 v1, v0, v1 |
| ; GFX9-NEXT: v_min_u32_e32 v0, v1, v0 |
| ; GFX9-NEXT: global_store_dword v[2:3], v1, off |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; GFX11-LABEL: v_underflow_compare_fold_i32_multi_use: |
| ; GFX11: ; %bb.0: |
| ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX11-NEXT: v_sub_nc_u32_e32 v1, v0, v1 |
| ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) |
| ; GFX11-NEXT: v_min_u32_e32 v0, v1, v0 |
| ; GFX11-NEXT: global_store_b32 v[2:3], v1, off |
| ; GFX11-NEXT: s_setpc_b64 s[30:31] |
| %sub = sub i32 %a, %b |
| store i32 %sub, ptr addrspace(1) %ptr |
| %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a) |
| ret i32 %cond |
| } |
| |
| define i64 @v_underflow_compare_fold_i64(i64 %a, i64 %b) #0 { |
| ; GFX9-LABEL: v_underflow_compare_fold_i64: |
| ; GFX9: ; %bb.0: |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 |
| ; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc |
| ; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] |
| ; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc |
| ; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; GFX11-LABEL: v_underflow_compare_fold_i64: |
| ; GFX11: ; %bb.0: |
| ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 |
| ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) |
| ; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo |
| ; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] |
| ; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3 |
| ; GFX11-NEXT: s_setpc_b64 s[30:31] |
| %sub = sub i64 %a, %b |
| %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a) |
| ret i64 %cond |
| } |
| |
| define i64 @v_underflow_compare_fold_i64_commute(i64 %a, i64 %b) #0 { |
| ; GFX9-LABEL: v_underflow_compare_fold_i64_commute: |
| ; GFX9: ; %bb.0: |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 |
| ; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc |
| ; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[0:1], v[2:3] |
| ; GFX9-NEXT: v_cndmask_b32_e32 v0, v2, v0, vcc |
| ; GFX9-NEXT: v_cndmask_b32_e32 v1, v3, v1, vcc |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; GFX11-LABEL: v_underflow_compare_fold_i64_commute: |
| ; GFX11: ; %bb.0: |
| ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 |
| ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) |
| ; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo |
| ; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[0:1], v[2:3] |
| ; GFX11-NEXT: v_dual_cndmask_b32 v0, v2, v0 :: v_dual_cndmask_b32 v1, v3, v1 |
| ; GFX11-NEXT: s_setpc_b64 s[30:31] |
| %sub = sub i64 %a, %b |
| %cond = call i64 @llvm.umin.i64(i64 %a, i64 %sub) |
| ret i64 %cond |
| } |
| |
| define i64 @v_underflow_compare_fold_i64_multi_use(i64 %a, i64 %b, ptr addrspace(1) %ptr) #0 { |
| ; GFX9-LABEL: v_underflow_compare_fold_i64_multi_use: |
| ; GFX9: ; %bb.0: |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 |
| ; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc |
| ; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] |
| ; GFX9-NEXT: global_store_dwordx2 v[4:5], v[2:3], off |
| ; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc |
| ; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; GFX11-LABEL: v_underflow_compare_fold_i64_multi_use: |
| ; GFX11: ; %bb.0: |
| ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 |
| ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) |
| ; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo |
| ; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] |
| ; GFX11-NEXT: global_store_b64 v[4:5], v[2:3], off |
| ; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3 |
| ; GFX11-NEXT: s_setpc_b64 s[30:31] |
| %sub = sub i64 %a, %b |
| store i64 %sub, ptr addrspace(1) %ptr |
| %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a) |
| ret i64 %cond |
| } |
| |
| define amdgpu_ps i16 @s_underflow_compare_fold_i16(i16 inreg %a, i16 inreg %b) #0 { |
| ; GFX9-LABEL: s_underflow_compare_fold_i16: |
| ; GFX9: ; %bb.0: |
| ; GFX9-NEXT: s_sub_i32 s1, s0, s1 |
| ; GFX9-NEXT: s_and_b32 s0, 0xffff, s0 |
| ; GFX9-NEXT: s_and_b32 s1, s1, 0xffff |
| ; GFX9-NEXT: s_min_u32 s0, s1, s0 |
| ; GFX9-NEXT: ; return to shader part epilog |
| ; |
| ; GFX11-LABEL: s_underflow_compare_fold_i16: |
| ; GFX11: ; %bb.0: |
| ; GFX11-NEXT: s_sub_i32 s1, s0, s1 |
| ; GFX11-NEXT: s_and_b32 s0, 0xffff, s0 |
| ; GFX11-NEXT: s_and_b32 s1, s1, 0xffff |
| ; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) |
| ; GFX11-NEXT: s_min_u32 s0, s1, s0 |
| ; GFX11-NEXT: ; return to shader part epilog |
| %sub = sub i16 %a, %b |
| %cond = call i16 @llvm.umin.i16(i16 %sub, i16 %a) |
| ret i16 %cond |
| } |
| |
| define amdgpu_ps i32 @s_underflow_compare_fold_i32(i32 inreg %a, i32 inreg %b) #0 { |
| ; GFX9-LABEL: s_underflow_compare_fold_i32: |
| ; GFX9: ; %bb.0: |
| ; GFX9-NEXT: s_sub_i32 s1, s0, s1 |
| ; GFX9-NEXT: s_min_u32 s0, s1, s0 |
| ; GFX9-NEXT: ; return to shader part epilog |
| ; |
| ; GFX11-LABEL: s_underflow_compare_fold_i32: |
| ; GFX11: ; %bb.0: |
| ; GFX11-NEXT: s_sub_i32 s1, s0, s1 |
| ; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) |
| ; GFX11-NEXT: s_min_u32 s0, s1, s0 |
| ; GFX11-NEXT: ; return to shader part epilog |
| %sub = sub i32 %a, %b |
| %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a) |
| ret i32 %cond |
| } |
| |
| define amdgpu_ps i64 @s_underflow_compare_fold_i64(i64 inreg %a, i64 inreg %b) #0 { |
| ; GFX9-LABEL: s_underflow_compare_fold_i64: |
| ; GFX9: ; %bb.0: |
| ; GFX9-NEXT: s_sub_u32 s2, s0, s2 |
| ; GFX9-NEXT: v_mov_b32_e32 v0, s0 |
| ; GFX9-NEXT: s_subb_u32 s3, s1, s3 |
| ; GFX9-NEXT: v_mov_b32_e32 v1, s1 |
| ; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[0:1] |
| ; GFX9-NEXT: s_and_b64 s[4:5], vcc, exec |
| ; GFX9-NEXT: s_cselect_b32 s1, s3, s1 |
| ; GFX9-NEXT: s_cselect_b32 s0, s2, s0 |
| ; GFX9-NEXT: ; return to shader part epilog |
| ; |
| ; GFX11-LABEL: s_underflow_compare_fold_i64: |
| ; GFX11: ; %bb.0: |
| ; GFX11-NEXT: s_sub_u32 s2, s0, s2 |
| ; GFX11-NEXT: s_subb_u32 s3, s1, s3 |
| ; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) |
| ; GFX11-NEXT: v_cmp_lt_u64_e64 s4, s[2:3], s[0:1] |
| ; GFX11-NEXT: s_and_b32 s4, s4, exec_lo |
| ; GFX11-NEXT: s_cselect_b32 s0, s2, s0 |
| ; GFX11-NEXT: s_cselect_b32 s1, s3, s1 |
| ; GFX11-NEXT: ; return to shader part epilog |
| %sub = sub i64 %a, %b |
| %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a) |
| ret i64 %cond |
| } |
| |
| attributes #0 = { nounwind } |
