| ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2 |
| ; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1030 < %s | FileCheck -check-prefixes=GCN,WAVE32,WAVE32-OPT %s |
| ; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1030 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GCN,WAVE64,WAVE64-OPT %s |
| |
| ; RUN: llc -O0 -mtriple=amdgcn-amd-amdpal -mcpu=gfx1030 < %s | FileCheck -check-prefixes=GCN,WAVE32,WAVE32-O0 %s |
| ; RUN: llc -O0 -mtriple=amdgcn-amd-amdpal -mcpu=gfx1030 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GCN,WAVE64,WAVE64-O0 %s |
| ; RUN: llc -O0 -mtriple=amdgcn-amd-amdpal -mcpu=gfx1030 -amdgpu-prealloc-sgpr-spill-vgprs=1 < %s | FileCheck -check-prefixes=GCN,WAVE32,WAVE32-WWM-PREALLOC %s |
| |
| declare ptr addrspace(5) @llvm.stacksave.p5() |
| declare void @llvm.stackrestore.p5(ptr addrspace(5)) |
| |
| define hidden void @stack_passed_argument([32 x i32], i32) { |
| ; GCN-LABEL: stack_passed_argument: |
| ; GCN: ; %bb.0: |
| ; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; GCN-NEXT: s_setpc_b64 s[30:31] |
| ret void |
| } |
| |
| define void @func_store_stacksave() { |
| ; WAVE32-OPT-LABEL: func_store_stacksave: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_lshr_b32 s4, s32, 5 |
| ; WAVE32-OPT-NEXT: ;;#ASMSTART |
| ; WAVE32-OPT-NEXT: ; use s4 |
| ; WAVE32-OPT-NEXT: ;;#ASMEND |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_store_stacksave: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_lshr_b32 s4, s32, 6 |
| ; WAVE64-OPT-NEXT: ;;#ASMSTART |
| ; WAVE64-OPT-NEXT: ; use s4 |
| ; WAVE64-OPT-NEXT: ;;#ASMEND |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_store_stacksave: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-O0-NEXT: s_lshr_b32 s4, s4, 5 |
| ; WAVE32-O0-NEXT: ;;#ASMSTART |
| ; WAVE32-O0-NEXT: ; use s4 |
| ; WAVE32-O0-NEXT: ;;#ASMEND |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_store_stacksave: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_mov_b32 s4, s32 |
| ; WAVE64-O0-NEXT: s_lshr_b32 s4, s4, 6 |
| ; WAVE64-O0-NEXT: ;;#ASMSTART |
| ; WAVE64-O0-NEXT: ; use s4 |
| ; WAVE64-O0-NEXT: ;;#ASMEND |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_store_stacksave: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshr_b32 s4, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMSTART |
| ; WAVE32-WWM-PREALLOC-NEXT: ; use s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMEND |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| %stacksave = call ptr addrspace(5) @llvm.stacksave.p5() |
| call void asm sideeffect "; use $0", "s"(ptr addrspace(5) %stacksave) |
| ret void |
| } |
| |
| define amdgpu_kernel void @kernel_store_stacksave() { |
| ; WAVE32-OPT-LABEL: kernel_store_stacksave: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_lshr_b32 s0, s32, 5 |
| ; WAVE32-OPT-NEXT: ;;#ASMSTART |
| ; WAVE32-OPT-NEXT: ; use s0 |
| ; WAVE32-OPT-NEXT: ;;#ASMEND |
| ; WAVE32-OPT-NEXT: s_endpgm |
| ; |
| ; WAVE64-OPT-LABEL: kernel_store_stacksave: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_lshr_b32 s0, s32, 6 |
| ; WAVE64-OPT-NEXT: ;;#ASMSTART |
| ; WAVE64-OPT-NEXT: ; use s0 |
| ; WAVE64-OPT-NEXT: ;;#ASMEND |
| ; WAVE64-OPT-NEXT: s_endpgm |
| ; |
| ; WAVE32-O0-LABEL: kernel_store_stacksave: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_mov_b32 s0, s32 |
| ; WAVE32-O0-NEXT: s_lshr_b32 s0, s0, 5 |
| ; WAVE32-O0-NEXT: ;;#ASMSTART |
| ; WAVE32-O0-NEXT: ; use s0 |
| ; WAVE32-O0-NEXT: ;;#ASMEND |
| ; WAVE32-O0-NEXT: s_endpgm |
| ; |
| ; WAVE64-O0-LABEL: kernel_store_stacksave: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_mov_b32 s0, s32 |
| ; WAVE64-O0-NEXT: s_lshr_b32 s0, s0, 6 |
| ; WAVE64-O0-NEXT: ;;#ASMSTART |
| ; WAVE64-O0-NEXT: ; use s0 |
| ; WAVE64-O0-NEXT: ;;#ASMEND |
| ; WAVE64-O0-NEXT: s_endpgm |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: kernel_store_stacksave: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s0, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshr_b32 s0, s0, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMSTART |
| ; WAVE32-WWM-PREALLOC-NEXT: ; use s0 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMEND |
| ; WAVE32-WWM-PREALLOC-NEXT: s_endpgm |
| %stacksave = call ptr addrspace(5) @llvm.stacksave.p5() |
| call void asm sideeffect "; use $0", "s"(ptr addrspace(5) %stacksave) |
| ret void |
| } |
| |
| define amdgpu_kernel void @kernel_store_stacksave_nocall() { |
| ; WAVE32-OPT-LABEL: kernel_store_stacksave_nocall: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_getpc_b64 s[12:13] |
| ; WAVE32-OPT-NEXT: s_mov_b32 s12, s0 |
| ; WAVE32-OPT-NEXT: v_mov_b32_e32 v0, 0 |
| ; WAVE32-OPT-NEXT: s_load_dwordx4 s[12:15], s[12:13], 0x0 |
| ; WAVE32-OPT-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_bitset0_b32 s15, 21 |
| ; WAVE32-OPT-NEXT: s_add_u32 s12, s12, s11 |
| ; WAVE32-OPT-NEXT: s_addc_u32 s13, s13, 0 |
| ; WAVE32-OPT-NEXT: s_lshr_b32 s0, s32, 5 |
| ; WAVE32-OPT-NEXT: v_mov_b32_e32 v1, s0 |
| ; WAVE32-OPT-NEXT: buffer_store_dword v0, v1, s[12:15], 0 offen |
| ; WAVE32-OPT-NEXT: s_endpgm |
| ; |
| ; WAVE64-OPT-LABEL: kernel_store_stacksave_nocall: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_getpc_b64 s[12:13] |
| ; WAVE64-OPT-NEXT: s_mov_b32 s12, s0 |
| ; WAVE64-OPT-NEXT: v_mov_b32_e32 v0, 0 |
| ; WAVE64-OPT-NEXT: s_load_dwordx4 s[12:15], s[12:13], 0x0 |
| ; WAVE64-OPT-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_add_u32 s12, s12, s11 |
| ; WAVE64-OPT-NEXT: s_addc_u32 s13, s13, 0 |
| ; WAVE64-OPT-NEXT: s_lshr_b32 s0, s32, 6 |
| ; WAVE64-OPT-NEXT: v_mov_b32_e32 v1, s0 |
| ; WAVE64-OPT-NEXT: buffer_store_dword v0, v1, s[12:15], 0 offen |
| ; WAVE64-OPT-NEXT: s_endpgm |
| ; |
| ; WAVE32-O0-LABEL: kernel_store_stacksave_nocall: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_getpc_b64 s[12:13] |
| ; WAVE32-O0-NEXT: s_mov_b32 s12, s0 |
| ; WAVE32-O0-NEXT: s_load_dwordx4 s[12:15], s[12:13], 0x0 |
| ; WAVE32-O0-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_bitset0_b32 s15, 21 |
| ; WAVE32-O0-NEXT: s_add_u32 s12, s12, s11 |
| ; WAVE32-O0-NEXT: s_addc_u32 s13, s13, 0 |
| ; WAVE32-O0-NEXT: s_mov_b32 s0, s32 |
| ; WAVE32-O0-NEXT: s_lshr_b32 s0, s0, 5 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v0, 0 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v1, s0 |
| ; WAVE32-O0-NEXT: buffer_store_dword v0, v1, s[12:15], 0 offen |
| ; WAVE32-O0-NEXT: s_endpgm |
| ; |
| ; WAVE64-O0-LABEL: kernel_store_stacksave_nocall: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_getpc_b64 s[12:13] |
| ; WAVE64-O0-NEXT: s_mov_b32 s12, s0 |
| ; WAVE64-O0-NEXT: s_load_dwordx4 s[12:15], s[12:13], 0x0 |
| ; WAVE64-O0-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_add_u32 s12, s12, s11 |
| ; WAVE64-O0-NEXT: s_addc_u32 s13, s13, 0 |
| ; WAVE64-O0-NEXT: s_mov_b32 s0, s32 |
| ; WAVE64-O0-NEXT: s_lshr_b32 s0, s0, 6 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v0, 0 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v1, s0 |
| ; WAVE64-O0-NEXT: buffer_store_dword v0, v1, s[12:15], 0 offen |
| ; WAVE64-O0-NEXT: s_endpgm |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: kernel_store_stacksave_nocall: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_getpc_b64 s[12:13] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s12, s0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_load_dwordx4 s[12:15], s[12:13], 0x0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_bitset0_b32 s15, 21 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_add_u32 s12, s12, s11 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_addc_u32 s13, s13, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s0, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshr_b32 s0, s0, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v0, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v1, s0 |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_store_dword v0, v1, s[12:15], 0 offen |
| ; WAVE32-WWM-PREALLOC-NEXT: s_endpgm |
| %stacksave = call ptr addrspace(5) @llvm.stacksave.p5() |
| store i32 0, ptr addrspace(5) %stacksave |
| ret void |
| } |
| |
| define void @func_stacksave_nonentry_block(i1 %cond) { |
| ; WAVE32-OPT-LABEL: func_stacksave_nonentry_block: |
| ; WAVE32-OPT: ; %bb.0: ; %bb0 |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: v_and_b32_e32 v0, 1, v0 |
| ; WAVE32-OPT-NEXT: s_mov_b32 s4, exec_lo |
| ; WAVE32-OPT-NEXT: v_cmpx_eq_u32_e32 1, v0 |
| ; WAVE32-OPT-NEXT: s_cbranch_execz .LBB4_2 |
| ; WAVE32-OPT-NEXT: ; %bb.1: ; %bb1 |
| ; WAVE32-OPT-NEXT: s_lshr_b32 s5, s32, 5 |
| ; WAVE32-OPT-NEXT: ;;#ASMSTART |
| ; WAVE32-OPT-NEXT: ; use s5 |
| ; WAVE32-OPT-NEXT: ;;#ASMEND |
| ; WAVE32-OPT-NEXT: .LBB4_2: ; %bb2 |
| ; WAVE32-OPT-NEXT: s_or_b32 exec_lo, exec_lo, s4 |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stacksave_nonentry_block: |
| ; WAVE64-OPT: ; %bb.0: ; %bb0 |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: v_and_b32_e32 v0, 1, v0 |
| ; WAVE64-OPT-NEXT: s_mov_b64 s[4:5], exec |
| ; WAVE64-OPT-NEXT: v_cmpx_eq_u32_e32 1, v0 |
| ; WAVE64-OPT-NEXT: s_cbranch_execz .LBB4_2 |
| ; WAVE64-OPT-NEXT: ; %bb.1: ; %bb1 |
| ; WAVE64-OPT-NEXT: s_lshr_b32 s6, s32, 6 |
| ; WAVE64-OPT-NEXT: ;;#ASMSTART |
| ; WAVE64-OPT-NEXT: ; use s6 |
| ; WAVE64-OPT-NEXT: ;;#ASMEND |
| ; WAVE64-OPT-NEXT: .LBB4_2: ; %bb2 |
| ; WAVE64-OPT-NEXT: s_or_b64 exec, exec, s[4:5] |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stacksave_nonentry_block: |
| ; WAVE32-O0: ; %bb.0: ; %bb0 |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_xor_saveexec_b32 s4, -1 |
| ; WAVE32-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill |
| ; WAVE32-O0-NEXT: s_mov_b32 exec_lo, s4 |
| ; WAVE32-O0-NEXT: v_and_b32_e64 v0, 1, v0 |
| ; WAVE32-O0-NEXT: v_cmp_eq_u32_e64 s5, v0, 1 |
| ; WAVE32-O0-NEXT: s_mov_b32 s4, exec_lo |
| ; WAVE32-O0-NEXT: ; implicit-def: $vgpr1 : SGPR spill to VGPR lane |
| ; WAVE32-O0-NEXT: v_writelane_b32 v1, s4, 0 |
| ; WAVE32-O0-NEXT: s_or_saveexec_b32 s7, -1 |
| ; WAVE32-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 ; 4-byte Folded Spill |
| ; WAVE32-O0-NEXT: s_mov_b32 exec_lo, s7 |
| ; WAVE32-O0-NEXT: s_and_b32 s4, s4, s5 |
| ; WAVE32-O0-NEXT: s_mov_b32 exec_lo, s4 |
| ; WAVE32-O0-NEXT: s_cbranch_execz .LBB4_2 |
| ; WAVE32-O0-NEXT: ; %bb.1: ; %bb1 |
| ; WAVE32-O0-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-O0-NEXT: s_lshr_b32 s4, s4, 5 |
| ; WAVE32-O0-NEXT: ;;#ASMSTART |
| ; WAVE32-O0-NEXT: ; use s4 |
| ; WAVE32-O0-NEXT: ;;#ASMEND |
| ; WAVE32-O0-NEXT: .LBB4_2: ; %bb2 |
| ; WAVE32-O0-NEXT: s_or_saveexec_b32 s7, -1 |
| ; WAVE32-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 ; 4-byte Folded Reload |
| ; WAVE32-O0-NEXT: s_mov_b32 exec_lo, s7 |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) |
| ; WAVE32-O0-NEXT: v_readlane_b32 s4, v1, 0 |
| ; WAVE32-O0-NEXT: s_or_b32 exec_lo, exec_lo, s4 |
| ; WAVE32-O0-NEXT: s_xor_saveexec_b32 s4, -1 |
| ; WAVE32-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload |
| ; WAVE32-O0-NEXT: s_mov_b32 exec_lo, s4 |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stacksave_nonentry_block: |
| ; WAVE64-O0: ; %bb.0: ; %bb0 |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1 |
| ; WAVE64-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill |
| ; WAVE64-O0-NEXT: s_mov_b64 exec, s[4:5] |
| ; WAVE64-O0-NEXT: v_and_b32_e64 v0, 1, v0 |
| ; WAVE64-O0-NEXT: v_cmp_eq_u32_e64 s[6:7], v0, 1 |
| ; WAVE64-O0-NEXT: s_mov_b64 s[4:5], exec |
| ; WAVE64-O0-NEXT: ; implicit-def: $vgpr1 : SGPR spill to VGPR lane |
| ; WAVE64-O0-NEXT: v_writelane_b32 v1, s4, 0 |
| ; WAVE64-O0-NEXT: v_writelane_b32 v1, s5, 1 |
| ; WAVE64-O0-NEXT: s_or_saveexec_b64 s[10:11], -1 |
| ; WAVE64-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 ; 4-byte Folded Spill |
| ; WAVE64-O0-NEXT: s_mov_b64 exec, s[10:11] |
| ; WAVE64-O0-NEXT: s_and_b64 s[4:5], s[4:5], s[6:7] |
| ; WAVE64-O0-NEXT: s_mov_b64 exec, s[4:5] |
| ; WAVE64-O0-NEXT: s_cbranch_execz .LBB4_2 |
| ; WAVE64-O0-NEXT: ; %bb.1: ; %bb1 |
| ; WAVE64-O0-NEXT: s_mov_b32 s4, s32 |
| ; WAVE64-O0-NEXT: s_lshr_b32 s4, s4, 6 |
| ; WAVE64-O0-NEXT: ;;#ASMSTART |
| ; WAVE64-O0-NEXT: ; use s4 |
| ; WAVE64-O0-NEXT: ;;#ASMEND |
| ; WAVE64-O0-NEXT: .LBB4_2: ; %bb2 |
| ; WAVE64-O0-NEXT: s_or_saveexec_b64 s[10:11], -1 |
| ; WAVE64-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 ; 4-byte Folded Reload |
| ; WAVE64-O0-NEXT: s_mov_b64 exec, s[10:11] |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) |
| ; WAVE64-O0-NEXT: v_readlane_b32 s4, v1, 0 |
| ; WAVE64-O0-NEXT: v_readlane_b32 s5, v1, 1 |
| ; WAVE64-O0-NEXT: s_or_b64 exec, exec, s[4:5] |
| ; WAVE64-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1 |
| ; WAVE64-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload |
| ; WAVE64-O0-NEXT: s_mov_b64 exec, s[4:5] |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stacksave_nonentry_block: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: ; %bb0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_xor_saveexec_b32 s4, -1 |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_store_dword v1, off, s[0:3], s32 ; 4-byte Folded Spill |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 exec_lo, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_and_b32_e64 v0, 1, v0 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_cmp_eq_u32_e64 s5, v0, 1 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s4, exec_lo |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $vgpr1 : SGPR spill to VGPR lane |
| ; WAVE32-WWM-PREALLOC-NEXT: v_writelane_b32 v1, s4, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_and_b32 s4, s4, s5 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 exec_lo, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_cbranch_execz .LBB4_2 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; %bb.1: ; %bb1 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshr_b32 s4, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMSTART |
| ; WAVE32-WWM-PREALLOC-NEXT: ; use s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMEND |
| ; WAVE32-WWM-PREALLOC-NEXT: .LBB4_2: ; %bb2 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_readlane_b32 s4, v1, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_or_b32 exec_lo, exec_lo, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_xor_saveexec_b32 s4, -1 |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_load_dword v1, off, s[0:3], s32 ; 4-byte Folded Reload |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 exec_lo, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| bb0: |
| br i1 %cond, label %bb1, label %bb2 |
| |
| bb1: |
| %stacksave = call ptr addrspace(5) @llvm.stacksave.p5() |
| call void asm sideeffect "; use $0", "s"(ptr addrspace(5) %stacksave) |
| br label %bb2 |
| |
| bb2: |
| ret void |
| } |
| |
| define void @func_stackrestore_poison() { |
| ; WAVE32-OPT-LABEL: func_stackrestore_poison: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_lshl_b32 s32, s4, 5 |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stackrestore_poison: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_lshl_b32 s32, s4, 6 |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stackrestore_poison: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr4 |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stackrestore_poison: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr4 |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_lshl_b32 s4, s4, 6 |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stackrestore_poison: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| call void @llvm.stackrestore.p5(ptr addrspace(5) poison) |
| ret void |
| } |
| |
| define void @func_stackrestore_null() { |
| ; WAVE32-OPT-LABEL: func_stackrestore_null: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_mov_b32 s32, 0 |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stackrestore_null: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_mov_b32 s32, 0 |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stackrestore_null: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_mov_b32 s4, 0 |
| ; WAVE32-O0-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stackrestore_null: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_mov_b32 s4, 0 |
| ; WAVE64-O0-NEXT: s_lshl_b32 s4, s4, 6 |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stackrestore_null: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s4, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| call void @llvm.stackrestore.p5(ptr addrspace(5) null) |
| ret void |
| } |
| |
| define void @func_stackrestore_neg1() { |
| ; WAVE32-OPT-LABEL: func_stackrestore_neg1: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_movk_i32 s32, 0xffe0 |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stackrestore_neg1: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_movk_i32 s32, 0xffc0 |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stackrestore_neg1: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_mov_b32 s4, -1 |
| ; WAVE32-O0-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stackrestore_neg1: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_mov_b32 s4, -1 |
| ; WAVE64-O0-NEXT: s_lshl_b32 s4, s4, 6 |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stackrestore_neg1: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s4, -1 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| call void @llvm.stackrestore.p5(ptr addrspace(5) inttoptr (i32 -1 to ptr addrspace(5))) |
| ret void |
| } |
| |
| define void @func_stackrestore_42() { |
| ; WAVE32-OPT-LABEL: func_stackrestore_42: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_movk_i32 s32, 0x540 |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stackrestore_42: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_movk_i32 s32, 0xa80 |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stackrestore_42: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_mov_b32 s4, 42 |
| ; WAVE32-O0-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stackrestore_42: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_mov_b32 s4, 42 |
| ; WAVE64-O0-NEXT: s_lshl_b32 s4, s4, 6 |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stackrestore_42: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s4, 42 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| call void @llvm.stackrestore.p5(ptr addrspace(5) inttoptr (i32 42 to ptr addrspace(5))) |
| ret void |
| } |
| |
| define void @func_stacksave_stackrestore() { |
| ; WAVE32-OPT-LABEL: func_stacksave_stackrestore: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stacksave_stackrestore: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stacksave_stackrestore: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stacksave_stackrestore: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_mov_b32 s4, s32 |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stacksave_stackrestore: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| %stacksave = call ptr addrspace(5) @llvm.stacksave.p5() |
| call void @llvm.stackrestore.p5(ptr addrspace(5) %stacksave) |
| ret void |
| } |
| |
| define void @func_stacksave_stackrestore_use() { |
| ; WAVE32-OPT-LABEL: func_stacksave_stackrestore_use: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_lshr_b32 s4, s32, 5 |
| ; WAVE32-OPT-NEXT: ;;#ASMSTART |
| ; WAVE32-OPT-NEXT: ; use s4 |
| ; WAVE32-OPT-NEXT: ;;#ASMEND |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stacksave_stackrestore_use: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_lshr_b32 s4, s32, 6 |
| ; WAVE64-OPT-NEXT: ;;#ASMSTART |
| ; WAVE64-OPT-NEXT: ; use s4 |
| ; WAVE64-OPT-NEXT: ;;#ASMEND |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stacksave_stackrestore_use: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-O0-NEXT: s_lshr_b32 s5, s4, 5 |
| ; WAVE32-O0-NEXT: ;;#ASMSTART |
| ; WAVE32-O0-NEXT: ; use s5 |
| ; WAVE32-O0-NEXT: ;;#ASMEND |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stacksave_stackrestore_use: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_mov_b32 s4, s32 |
| ; WAVE64-O0-NEXT: s_lshr_b32 s5, s4, 6 |
| ; WAVE64-O0-NEXT: ;;#ASMSTART |
| ; WAVE64-O0-NEXT: ; use s5 |
| ; WAVE64-O0-NEXT: ;;#ASMEND |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stacksave_stackrestore_use: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshr_b32 s5, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMSTART |
| ; WAVE32-WWM-PREALLOC-NEXT: ; use s5 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMEND |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| %stacksave = call ptr addrspace(5) @llvm.stacksave.p5() |
| call void asm sideeffect "; use $0", "s"(ptr addrspace(5) %stacksave) |
| call void @llvm.stackrestore.p5(ptr addrspace(5) %stacksave) |
| ret void |
| } |
| |
| define amdgpu_kernel void @kernel_stacksave_stackrestore_use() { |
| ; WAVE32-OPT-LABEL: kernel_stacksave_stackrestore_use: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_lshr_b32 s0, s32, 5 |
| ; WAVE32-OPT-NEXT: ;;#ASMSTART |
| ; WAVE32-OPT-NEXT: ; use s0 |
| ; WAVE32-OPT-NEXT: ;;#ASMEND |
| ; WAVE32-OPT-NEXT: s_endpgm |
| ; |
| ; WAVE64-OPT-LABEL: kernel_stacksave_stackrestore_use: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_lshr_b32 s0, s32, 6 |
| ; WAVE64-OPT-NEXT: ;;#ASMSTART |
| ; WAVE64-OPT-NEXT: ; use s0 |
| ; WAVE64-OPT-NEXT: ;;#ASMEND |
| ; WAVE64-OPT-NEXT: s_endpgm |
| ; |
| ; WAVE32-O0-LABEL: kernel_stacksave_stackrestore_use: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_mov_b32 s0, s32 |
| ; WAVE32-O0-NEXT: s_lshr_b32 s1, s0, 5 |
| ; WAVE32-O0-NEXT: ;;#ASMSTART |
| ; WAVE32-O0-NEXT: ; use s1 |
| ; WAVE32-O0-NEXT: ;;#ASMEND |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s0 |
| ; WAVE32-O0-NEXT: s_endpgm |
| ; |
| ; WAVE64-O0-LABEL: kernel_stacksave_stackrestore_use: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_mov_b32 s0, s32 |
| ; WAVE64-O0-NEXT: s_lshr_b32 s1, s0, 6 |
| ; WAVE64-O0-NEXT: ;;#ASMSTART |
| ; WAVE64-O0-NEXT: ; use s1 |
| ; WAVE64-O0-NEXT: ;;#ASMEND |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s0 |
| ; WAVE64-O0-NEXT: s_endpgm |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: kernel_stacksave_stackrestore_use: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s0, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshr_b32 s1, s0, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMSTART |
| ; WAVE32-WWM-PREALLOC-NEXT: ; use s1 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMEND |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_endpgm |
| %stacksave = call ptr addrspace(5) @llvm.stacksave.p5() |
| call void asm sideeffect "; use $0", "s"(ptr addrspace(5) %stacksave) |
| call void @llvm.stackrestore.p5(ptr addrspace(5) %stacksave) |
| ret void |
| } |
| |
| define void @func_stacksave_stackrestore_voffset(i32 %offset) { |
| ; WAVE32-OPT-LABEL: func_stacksave_stackrestore_voffset: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_lshr_b32 s4, s32, 5 |
| ; WAVE32-OPT-NEXT: v_add_nc_u32_e32 v0, s4, v0 |
| ; WAVE32-OPT-NEXT: v_readfirstlane_b32 s4, v0 |
| ; WAVE32-OPT-NEXT: s_lshl_b32 s32, s4, 5 |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stacksave_stackrestore_voffset: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_lshr_b32 s4, s32, 6 |
| ; WAVE64-OPT-NEXT: v_add_nc_u32_e32 v0, s4, v0 |
| ; WAVE64-OPT-NEXT: v_readfirstlane_b32 s4, v0 |
| ; WAVE64-OPT-NEXT: s_lshl_b32 s32, s4, 6 |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stacksave_stackrestore_voffset: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-O0-NEXT: s_lshr_b32 s4, s4, 5 |
| ; WAVE32-O0-NEXT: v_add_nc_u32_e64 v0, s4, v0 |
| ; WAVE32-O0-NEXT: v_readfirstlane_b32 s4, v0 |
| ; WAVE32-O0-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stacksave_stackrestore_voffset: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_mov_b32 s4, s32 |
| ; WAVE64-O0-NEXT: s_lshr_b32 s4, s4, 6 |
| ; WAVE64-O0-NEXT: v_add_nc_u32_e64 v0, s4, v0 |
| ; WAVE64-O0-NEXT: v_readfirstlane_b32 s4, v0 |
| ; WAVE64-O0-NEXT: s_lshl_b32 s4, s4, 6 |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stacksave_stackrestore_voffset: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshr_b32 s4, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_add_nc_u32_e64 v0, s4, v0 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_readfirstlane_b32 s4, v0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| %stacksave = call ptr addrspace(5) @llvm.stacksave.p5() |
| %gep = getelementptr i8, ptr addrspace(5) %stacksave, i32 %offset |
| call void @llvm.stackrestore.p5(ptr addrspace(5) %gep) |
| ret void |
| } |
| |
| define void @func_stacksave_vgpr(ptr addrspace(5) %stack) { |
| ; WAVE32-OPT-LABEL: func_stacksave_vgpr: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: v_readfirstlane_b32 s4, v0 |
| ; WAVE32-OPT-NEXT: s_lshl_b32 s32, s4, 5 |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stacksave_vgpr: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: v_readfirstlane_b32 s4, v0 |
| ; WAVE64-OPT-NEXT: s_lshl_b32 s32, s4, 6 |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stacksave_vgpr: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: v_readfirstlane_b32 s4, v0 |
| ; WAVE32-O0-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stacksave_vgpr: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: v_readfirstlane_b32 s4, v0 |
| ; WAVE64-O0-NEXT: s_lshl_b32 s4, s4, 6 |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stacksave_vgpr: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: v_readfirstlane_b32 s4, v0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshl_b32 s4, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| call void @llvm.stackrestore.p5(ptr addrspace(5) %stack) |
| ret void |
| } |
| |
| define amdgpu_gfx void @func_stacksave_sgpr(ptr addrspace(5) inreg %stack) { |
| ; WAVE32-OPT-LABEL: func_stacksave_sgpr: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_lshl_b32 s32, s4, 5 |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stacksave_sgpr: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_lshl_b32 s32, s4, 6 |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stacksave_sgpr: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_lshl_b32 s34, s4, 5 |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s34 |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stacksave_sgpr: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_lshl_b32 s34, s4, 6 |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s34 |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stacksave_sgpr: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshl_b32 s34, s4, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s34 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| call void @llvm.stackrestore.p5(ptr addrspace(5) %stack) |
| ret void |
| } |
| |
| define amdgpu_kernel void @kernel_stacksave_sgpr(ptr addrspace(5) %stack) { |
| ; WAVE32-OPT-LABEL: kernel_stacksave_sgpr: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_load_dword s0, s[4:5], 0x0 |
| ; WAVE32-OPT-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: ;;#ASMSTART |
| ; WAVE32-OPT-NEXT: ; use s0 |
| ; WAVE32-OPT-NEXT: ;;#ASMEND |
| ; WAVE32-OPT-NEXT: s_endpgm |
| ; |
| ; WAVE64-OPT-LABEL: kernel_stacksave_sgpr: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_load_dword s0, s[4:5], 0x0 |
| ; WAVE64-OPT-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: ;;#ASMSTART |
| ; WAVE64-OPT-NEXT: ; use s0 |
| ; WAVE64-OPT-NEXT: ;;#ASMEND |
| ; WAVE64-OPT-NEXT: s_endpgm |
| ; |
| ; WAVE32-O0-LABEL: kernel_stacksave_sgpr: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_load_dword s0, s[4:5], 0x0 |
| ; WAVE32-O0-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_mov_b32 s1, s0 |
| ; WAVE32-O0-NEXT: ;;#ASMSTART |
| ; WAVE32-O0-NEXT: ; use s1 |
| ; WAVE32-O0-NEXT: ;;#ASMEND |
| ; WAVE32-O0-NEXT: s_lshl_b32 s0, s0, 5 |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s0 |
| ; WAVE32-O0-NEXT: s_endpgm |
| ; |
| ; WAVE64-O0-LABEL: kernel_stacksave_sgpr: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_load_dword s0, s[4:5], 0x0 |
| ; WAVE64-O0-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_mov_b32 s1, s0 |
| ; WAVE64-O0-NEXT: ;;#ASMSTART |
| ; WAVE64-O0-NEXT: ; use s1 |
| ; WAVE64-O0-NEXT: ;;#ASMEND |
| ; WAVE64-O0-NEXT: s_lshl_b32 s0, s0, 6 |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s0 |
| ; WAVE64-O0-NEXT: s_endpgm |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: kernel_stacksave_sgpr: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_load_dword s0, s[4:5], 0x0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s1, s0 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMSTART |
| ; WAVE32-WWM-PREALLOC-NEXT: ; use s1 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMEND |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshl_b32 s0, s0, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_endpgm |
| call void asm sideeffect "; use $0", "s"(ptr addrspace(5) %stack) |
| call void @llvm.stackrestore.p5(ptr addrspace(5) %stack) |
| ret void |
| } |
| |
| define amdgpu_kernel void @kernel_stacksave_stackrestore_call_with_stack_objects() { |
| ; WAVE32-OPT-LABEL: kernel_stacksave_stackrestore_call_with_stack_objects: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_getpc_b64 s[20:21] |
| ; WAVE32-OPT-NEXT: s_mov_b32 s20, s0 |
| ; WAVE32-OPT-NEXT: v_lshlrev_b32_e32 v2, 20, v2 |
| ; WAVE32-OPT-NEXT: s_load_dwordx4 s[20:23], s[20:21], 0x0 |
| ; WAVE32-OPT-NEXT: v_lshlrev_b32_e32 v1, 10, v1 |
| ; WAVE32-OPT-NEXT: s_movk_i32 s32, 0x1200 |
| ; WAVE32-OPT-NEXT: s_mov_b32 s13, s9 |
| ; WAVE32-OPT-NEXT: s_mov_b32 s12, s8 |
| ; WAVE32-OPT-NEXT: s_mov_b64 s[8:9], s[4:5] |
| ; WAVE32-OPT-NEXT: s_mov_b32 s4, s32 |
| ; WAVE32-OPT-NEXT: v_mov_b32_e32 v3, 42 |
| ; WAVE32-OPT-NEXT: v_mov_b32_e32 v4, 17 |
| ; WAVE32-OPT-NEXT: v_or3_b32 v31, v0, v1, v2 |
| ; WAVE32-OPT-NEXT: s_mov_b32 s14, s10 |
| ; WAVE32-OPT-NEXT: s_mov_b32 s17, stack_passed_argument@abs32@hi |
| ; WAVE32-OPT-NEXT: s_mov_b32 s16, stack_passed_argument@abs32@lo |
| ; WAVE32-OPT-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_bitset0_b32 s23, 21 |
| ; WAVE32-OPT-NEXT: s_add_u32 s20, s20, s11 |
| ; WAVE32-OPT-NEXT: s_addc_u32 s21, s21, 0 |
| ; WAVE32-OPT-NEXT: s_mov_b64 s[10:11], s[6:7] |
| ; WAVE32-OPT-NEXT: s_lshr_b32 s15, s4, 5 |
| ; WAVE32-OPT-NEXT: s_mov_b64 s[4:5], s[0:1] |
| ; WAVE32-OPT-NEXT: s_mov_b64 s[6:7], s[2:3] |
| ; WAVE32-OPT-NEXT: s_mov_b64 s[0:1], s[20:21] |
| ; WAVE32-OPT-NEXT: s_mov_b64 s[2:3], s[22:23] |
| ; WAVE32-OPT-NEXT: buffer_store_dword v3, off, s[20:23], 0 |
| ; WAVE32-OPT-NEXT: s_waitcnt_vscnt null, 0x0 |
| ; WAVE32-OPT-NEXT: buffer_store_dword v4, off, s[20:23], s32 offset:4 |
| ; WAVE32-OPT-NEXT: s_swappc_b64 s[30:31], s[16:17] |
| ; WAVE32-OPT-NEXT: ;;#ASMSTART |
| ; WAVE32-OPT-NEXT: ; use s15 |
| ; WAVE32-OPT-NEXT: ;;#ASMEND |
| ; WAVE32-OPT-NEXT: s_endpgm |
| ; |
| ; WAVE64-OPT-LABEL: kernel_stacksave_stackrestore_call_with_stack_objects: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_getpc_b64 s[20:21] |
| ; WAVE64-OPT-NEXT: s_mov_b32 s20, s0 |
| ; WAVE64-OPT-NEXT: v_lshlrev_b32_e32 v2, 20, v2 |
| ; WAVE64-OPT-NEXT: s_load_dwordx4 s[20:23], s[20:21], 0x0 |
| ; WAVE64-OPT-NEXT: v_lshlrev_b32_e32 v1, 10, v1 |
| ; WAVE64-OPT-NEXT: s_movk_i32 s32, 0x2400 |
| ; WAVE64-OPT-NEXT: s_mov_b32 s13, s9 |
| ; WAVE64-OPT-NEXT: s_mov_b32 s12, s8 |
| ; WAVE64-OPT-NEXT: s_mov_b64 s[8:9], s[4:5] |
| ; WAVE64-OPT-NEXT: s_mov_b32 s4, s32 |
| ; WAVE64-OPT-NEXT: v_mov_b32_e32 v3, 42 |
| ; WAVE64-OPT-NEXT: v_mov_b32_e32 v4, 17 |
| ; WAVE64-OPT-NEXT: v_or3_b32 v31, v0, v1, v2 |
| ; WAVE64-OPT-NEXT: s_mov_b32 s14, s10 |
| ; WAVE64-OPT-NEXT: s_mov_b32 s17, stack_passed_argument@abs32@hi |
| ; WAVE64-OPT-NEXT: s_mov_b32 s16, stack_passed_argument@abs32@lo |
| ; WAVE64-OPT-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_add_u32 s20, s20, s11 |
| ; WAVE64-OPT-NEXT: s_addc_u32 s21, s21, 0 |
| ; WAVE64-OPT-NEXT: s_mov_b64 s[10:11], s[6:7] |
| ; WAVE64-OPT-NEXT: s_lshr_b32 s15, s4, 6 |
| ; WAVE64-OPT-NEXT: s_mov_b64 s[4:5], s[0:1] |
| ; WAVE64-OPT-NEXT: s_mov_b64 s[6:7], s[2:3] |
| ; WAVE64-OPT-NEXT: s_mov_b64 s[0:1], s[20:21] |
| ; WAVE64-OPT-NEXT: s_mov_b64 s[2:3], s[22:23] |
| ; WAVE64-OPT-NEXT: buffer_store_dword v3, off, s[20:23], 0 |
| ; WAVE64-OPT-NEXT: s_waitcnt_vscnt null, 0x0 |
| ; WAVE64-OPT-NEXT: buffer_store_dword v4, off, s[20:23], s32 offset:4 |
| ; WAVE64-OPT-NEXT: s_swappc_b64 s[30:31], s[16:17] |
| ; WAVE64-OPT-NEXT: ;;#ASMSTART |
| ; WAVE64-OPT-NEXT: ; use s15 |
| ; WAVE64-OPT-NEXT: ;;#ASMEND |
| ; WAVE64-OPT-NEXT: s_endpgm |
| ; |
| ; WAVE32-O0-LABEL: kernel_stacksave_stackrestore_call_with_stack_objects: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, 0x1200 |
| ; WAVE32-O0-NEXT: s_getpc_b64 s[20:21] |
| ; WAVE32-O0-NEXT: s_mov_b32 s20, s0 |
| ; WAVE32-O0-NEXT: s_load_dwordx4 s[20:23], s[20:21], 0x0 |
| ; WAVE32-O0-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_bitset0_b32 s23, 21 |
| ; WAVE32-O0-NEXT: s_add_u32 s20, s20, s11 |
| ; WAVE32-O0-NEXT: s_addc_u32 s21, s21, 0 |
| ; WAVE32-O0-NEXT: s_mov_b32 s14, s10 |
| ; WAVE32-O0-NEXT: s_mov_b32 s13, s9 |
| ; WAVE32-O0-NEXT: s_mov_b32 s12, s8 |
| ; WAVE32-O0-NEXT: s_mov_b64 s[10:11], s[6:7] |
| ; WAVE32-O0-NEXT: s_mov_b64 s[8:9], s[4:5] |
| ; WAVE32-O0-NEXT: s_mov_b64 s[6:7], s[2:3] |
| ; WAVE32-O0-NEXT: s_mov_b64 s[4:5], s[0:1] |
| ; WAVE32-O0-NEXT: s_mov_b32 s0, s32 |
| ; WAVE32-O0-NEXT: ; implicit-def: $vgpr32 : SGPR spill to VGPR lane |
| ; WAVE32-O0-NEXT: v_writelane_b32 v32, s0, 0 |
| ; WAVE32-O0-NEXT: s_lshr_b32 s0, s0, 5 |
| ; WAVE32-O0-NEXT: v_writelane_b32 v32, s0, 1 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v3, 42 |
| ; WAVE32-O0-NEXT: buffer_store_dword v3, off, s[20:23], 0 |
| ; WAVE32-O0-NEXT: s_waitcnt_vscnt null, 0x0 |
| ; WAVE32-O0-NEXT: s_mov_b64 s[0:1], s[20:21] |
| ; WAVE32-O0-NEXT: s_mov_b64 s[2:3], s[22:23] |
| ; WAVE32-O0-NEXT: s_mov_b32 s15, s32 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v3, 17 |
| ; WAVE32-O0-NEXT: buffer_store_dword v3, off, s[20:23], s15 offset:4 |
| ; WAVE32-O0-NEXT: s_mov_b32 s15, stack_passed_argument@abs32@hi |
| ; WAVE32-O0-NEXT: s_mov_b32 s16, stack_passed_argument@abs32@lo |
| ; WAVE32-O0-NEXT: ; kill: def $sgpr16 killed $sgpr16 def $sgpr16_sgpr17 |
| ; WAVE32-O0-NEXT: s_mov_b32 s17, s15 |
| ; WAVE32-O0-NEXT: s_mov_b32 s15, 20 |
| ; WAVE32-O0-NEXT: v_lshlrev_b32_e64 v2, s15, v2 |
| ; WAVE32-O0-NEXT: s_mov_b32 s15, 10 |
| ; WAVE32-O0-NEXT: v_lshlrev_b32_e64 v1, s15, v1 |
| ; WAVE32-O0-NEXT: v_or3_b32 v31, v0, v1, v2 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr15 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v0, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v1, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v2, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v3, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v4, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v5, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v6, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v7, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v8, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v9, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v10, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v11, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v12, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v13, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v14, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v15, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v16, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v17, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v18, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v19, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v20, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v21, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v22, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v23, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v24, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v25, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v26, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v27, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v28, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v29, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v30, s18 |
| ; WAVE32-O0-NEXT: s_swappc_b64 s[30:31], s[16:17] |
| ; WAVE32-O0-NEXT: v_readlane_b32 s1, v32, 1 |
| ; WAVE32-O0-NEXT: v_readlane_b32 s0, v32, 0 |
| ; WAVE32-O0-NEXT: ;;#ASMSTART |
| ; WAVE32-O0-NEXT: ; use s1 |
| ; WAVE32-O0-NEXT: ;;#ASMEND |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s0 |
| ; WAVE32-O0-NEXT: s_endpgm |
| ; |
| ; WAVE64-O0-LABEL: kernel_stacksave_stackrestore_call_with_stack_objects: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, 0x2400 |
| ; WAVE64-O0-NEXT: s_getpc_b64 s[24:25] |
| ; WAVE64-O0-NEXT: s_mov_b32 s24, s0 |
| ; WAVE64-O0-NEXT: s_load_dwordx4 s[24:27], s[24:25], 0x0 |
| ; WAVE64-O0-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_add_u32 s24, s24, s11 |
| ; WAVE64-O0-NEXT: s_addc_u32 s25, s25, 0 |
| ; WAVE64-O0-NEXT: s_mov_b32 s14, s10 |
| ; WAVE64-O0-NEXT: s_mov_b32 s13, s9 |
| ; WAVE64-O0-NEXT: s_mov_b32 s12, s8 |
| ; WAVE64-O0-NEXT: s_mov_b64 s[10:11], s[6:7] |
| ; WAVE64-O0-NEXT: s_mov_b64 s[8:9], s[4:5] |
| ; WAVE64-O0-NEXT: s_mov_b64 s[6:7], s[2:3] |
| ; WAVE64-O0-NEXT: s_mov_b64 s[4:5], s[0:1] |
| ; WAVE64-O0-NEXT: s_mov_b32 s0, s32 |
| ; WAVE64-O0-NEXT: ; implicit-def: $vgpr32 : SGPR spill to VGPR lane |
| ; WAVE64-O0-NEXT: v_writelane_b32 v32, s0, 0 |
| ; WAVE64-O0-NEXT: s_lshr_b32 s0, s0, 6 |
| ; WAVE64-O0-NEXT: v_writelane_b32 v32, s0, 1 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v3, 42 |
| ; WAVE64-O0-NEXT: buffer_store_dword v3, off, s[24:27], 0 |
| ; WAVE64-O0-NEXT: s_waitcnt_vscnt null, 0x0 |
| ; WAVE64-O0-NEXT: s_mov_b64 s[0:1], s[24:25] |
| ; WAVE64-O0-NEXT: s_mov_b64 s[2:3], s[26:27] |
| ; WAVE64-O0-NEXT: s_mov_b32 s15, s32 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v3, 17 |
| ; WAVE64-O0-NEXT: buffer_store_dword v3, off, s[24:27], s15 offset:4 |
| ; WAVE64-O0-NEXT: s_mov_b32 s15, stack_passed_argument@abs32@hi |
| ; WAVE64-O0-NEXT: s_mov_b32 s16, stack_passed_argument@abs32@lo |
| ; WAVE64-O0-NEXT: ; kill: def $sgpr16 killed $sgpr16 def $sgpr16_sgpr17 |
| ; WAVE64-O0-NEXT: s_mov_b32 s17, s15 |
| ; WAVE64-O0-NEXT: s_mov_b32 s15, 20 |
| ; WAVE64-O0-NEXT: v_lshlrev_b32_e64 v2, s15, v2 |
| ; WAVE64-O0-NEXT: s_mov_b32 s15, 10 |
| ; WAVE64-O0-NEXT: v_lshlrev_b32_e64 v1, s15, v1 |
| ; WAVE64-O0-NEXT: v_or3_b32 v31, v0, v1, v2 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr15 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v0, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v1, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v2, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v3, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v4, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v5, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v6, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v7, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v8, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v9, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v10, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v11, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v12, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v13, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v14, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v15, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v16, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v17, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v18, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v19, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v20, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v21, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v22, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v23, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v24, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v25, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v26, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v27, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v28, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v29, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v30, s18 |
| ; WAVE64-O0-NEXT: s_swappc_b64 s[30:31], s[16:17] |
| ; WAVE64-O0-NEXT: v_readlane_b32 s1, v32, 1 |
| ; WAVE64-O0-NEXT: v_readlane_b32 s0, v32, 0 |
| ; WAVE64-O0-NEXT: ;;#ASMSTART |
| ; WAVE64-O0-NEXT: ; use s1 |
| ; WAVE64-O0-NEXT: ;;#ASMEND |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s0 |
| ; WAVE64-O0-NEXT: s_endpgm |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: kernel_stacksave_stackrestore_call_with_stack_objects: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, 0x1200 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_getpc_b64 s[20:21] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s20, s0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_load_dwordx4 s[20:23], s[20:21], 0x0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_bitset0_b32 s23, 21 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_add_u32 s20, s20, s11 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_addc_u32 s21, s21, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s14, s10 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s13, s9 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s12, s8 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b64 s[10:11], s[6:7] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b64 s[8:9], s[4:5] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b64 s[6:7], s[2:3] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b64 s[4:5], s[0:1] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s0, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $vgpr32 : SGPR spill to VGPR lane |
| ; WAVE32-WWM-PREALLOC-NEXT: v_writelane_b32 v32, s0, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshr_b32 s0, s0, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_writelane_b32 v32, s0, 1 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v3, 42 |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_store_dword v3, off, s[20:23], 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt_vscnt null, 0x0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b64 s[0:1], s[20:21] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b64 s[2:3], s[22:23] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s15, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v3, 17 |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_store_dword v3, off, s[20:23], s15 offset:4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s15, stack_passed_argument@abs32@hi |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s16, stack_passed_argument@abs32@lo |
| ; WAVE32-WWM-PREALLOC-NEXT: ; kill: def $sgpr16 killed $sgpr16 def $sgpr16_sgpr17 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s17, s15 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s15, 20 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_lshlrev_b32_e64 v2, s15, v2 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s15, 10 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_lshlrev_b32_e64 v1, s15, v1 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_or3_b32 v31, v0, v1, v2 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr15 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v0, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v1, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v2, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v3, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v4, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v5, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v6, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v7, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v8, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v9, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v10, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v11, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v12, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v13, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v14, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v15, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v16, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v17, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v18, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v19, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v20, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v21, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v22, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v23, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v24, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v25, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v26, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v27, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v28, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v29, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v30, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_swappc_b64 s[30:31], s[16:17] |
| ; WAVE32-WWM-PREALLOC-NEXT: v_readlane_b32 s1, v32, 1 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_readlane_b32 s0, v32, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMSTART |
| ; WAVE32-WWM-PREALLOC-NEXT: ; use s1 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMEND |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_endpgm |
| %alloca = alloca [32 x i32], addrspace(5) |
| %stacksave = call ptr addrspace(5) @llvm.stacksave.p5() |
| store volatile i32 42, ptr addrspace(5) %alloca |
| call void @stack_passed_argument([32 x i32] poison, i32 17) |
| call void asm sideeffect "; use $0", "s"(ptr addrspace(5) %stacksave) |
| call void @llvm.stackrestore.p5(ptr addrspace(5) %stacksave) |
| ret void |
| } |
| |
| define void @func_stacksave_stackrestore_call_with_stack_objects() { |
| ; WAVE32-OPT-LABEL: func_stacksave_stackrestore_call_with_stack_objects: |
| ; WAVE32-OPT: ; %bb.0: |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-OPT-NEXT: s_mov_b32 s20, s33 |
| ; WAVE32-OPT-NEXT: s_mov_b32 s33, s32 |
| ; WAVE32-OPT-NEXT: s_xor_saveexec_b32 s16, -1 |
| ; WAVE32-OPT-NEXT: buffer_store_dword v32, off, s[0:3], s33 offset:128 ; 4-byte Folded Spill |
| ; WAVE32-OPT-NEXT: s_mov_b32 exec_lo, s16 |
| ; WAVE32-OPT-NEXT: v_writelane_b32 v32, s30, 0 |
| ; WAVE32-OPT-NEXT: v_mov_b32_e32 v0, 42 |
| ; WAVE32-OPT-NEXT: v_mov_b32_e32 v1, 17 |
| ; WAVE32-OPT-NEXT: s_addk_i32 s32, 0x1200 |
| ; WAVE32-OPT-NEXT: s_mov_b32 s17, stack_passed_argument@abs32@hi |
| ; WAVE32-OPT-NEXT: s_mov_b32 s18, s32 |
| ; WAVE32-OPT-NEXT: s_mov_b32 s16, stack_passed_argument@abs32@lo |
| ; WAVE32-OPT-NEXT: v_writelane_b32 v32, s31, 1 |
| ; WAVE32-OPT-NEXT: s_lshr_b32 s19, s18, 5 |
| ; WAVE32-OPT-NEXT: buffer_store_dword v0, off, s[0:3], s33 |
| ; WAVE32-OPT-NEXT: s_waitcnt_vscnt null, 0x0 |
| ; WAVE32-OPT-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:4 |
| ; WAVE32-OPT-NEXT: s_swappc_b64 s[30:31], s[16:17] |
| ; WAVE32-OPT-NEXT: s_mov_b32 s32, s18 |
| ; WAVE32-OPT-NEXT: ;;#ASMSTART |
| ; WAVE32-OPT-NEXT: ; use s19 |
| ; WAVE32-OPT-NEXT: ;;#ASMEND |
| ; WAVE32-OPT-NEXT: v_readlane_b32 s31, v32, 1 |
| ; WAVE32-OPT-NEXT: v_readlane_b32 s30, v32, 0 |
| ; WAVE32-OPT-NEXT: s_mov_b32 s32, s33 |
| ; WAVE32-OPT-NEXT: s_xor_saveexec_b32 s4, -1 |
| ; WAVE32-OPT-NEXT: buffer_load_dword v32, off, s[0:3], s33 offset:128 ; 4-byte Folded Reload |
| ; WAVE32-OPT-NEXT: s_mov_b32 exec_lo, s4 |
| ; WAVE32-OPT-NEXT: s_mov_b32 s33, s20 |
| ; WAVE32-OPT-NEXT: s_waitcnt vmcnt(0) |
| ; WAVE32-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-OPT-LABEL: func_stacksave_stackrestore_call_with_stack_objects: |
| ; WAVE64-OPT: ; %bb.0: |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-OPT-NEXT: s_mov_b32 s20, s33 |
| ; WAVE64-OPT-NEXT: s_mov_b32 s33, s32 |
| ; WAVE64-OPT-NEXT: s_xor_saveexec_b64 s[16:17], -1 |
| ; WAVE64-OPT-NEXT: buffer_store_dword v32, off, s[0:3], s33 offset:128 ; 4-byte Folded Spill |
| ; WAVE64-OPT-NEXT: s_mov_b64 exec, s[16:17] |
| ; WAVE64-OPT-NEXT: v_writelane_b32 v32, s30, 0 |
| ; WAVE64-OPT-NEXT: v_mov_b32_e32 v0, 42 |
| ; WAVE64-OPT-NEXT: v_mov_b32_e32 v1, 17 |
| ; WAVE64-OPT-NEXT: s_addk_i32 s32, 0x2400 |
| ; WAVE64-OPT-NEXT: s_mov_b32 s17, stack_passed_argument@abs32@hi |
| ; WAVE64-OPT-NEXT: s_mov_b32 s18, s32 |
| ; WAVE64-OPT-NEXT: s_mov_b32 s16, stack_passed_argument@abs32@lo |
| ; WAVE64-OPT-NEXT: v_writelane_b32 v32, s31, 1 |
| ; WAVE64-OPT-NEXT: s_lshr_b32 s19, s18, 6 |
| ; WAVE64-OPT-NEXT: buffer_store_dword v0, off, s[0:3], s33 |
| ; WAVE64-OPT-NEXT: s_waitcnt_vscnt null, 0x0 |
| ; WAVE64-OPT-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:4 |
| ; WAVE64-OPT-NEXT: s_swappc_b64 s[30:31], s[16:17] |
| ; WAVE64-OPT-NEXT: s_mov_b32 s32, s18 |
| ; WAVE64-OPT-NEXT: ;;#ASMSTART |
| ; WAVE64-OPT-NEXT: ; use s19 |
| ; WAVE64-OPT-NEXT: ;;#ASMEND |
| ; WAVE64-OPT-NEXT: v_readlane_b32 s31, v32, 1 |
| ; WAVE64-OPT-NEXT: v_readlane_b32 s30, v32, 0 |
| ; WAVE64-OPT-NEXT: s_mov_b32 s32, s33 |
| ; WAVE64-OPT-NEXT: s_xor_saveexec_b64 s[4:5], -1 |
| ; WAVE64-OPT-NEXT: buffer_load_dword v32, off, s[0:3], s33 offset:128 ; 4-byte Folded Reload |
| ; WAVE64-OPT-NEXT: s_mov_b64 exec, s[4:5] |
| ; WAVE64-OPT-NEXT: s_mov_b32 s33, s20 |
| ; WAVE64-OPT-NEXT: s_waitcnt vmcnt(0) |
| ; WAVE64-OPT-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-O0-LABEL: func_stacksave_stackrestore_call_with_stack_objects: |
| ; WAVE32-O0: ; %bb.0: |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-O0-NEXT: s_mov_b32 s24, s33 |
| ; WAVE32-O0-NEXT: s_mov_b32 s33, s32 |
| ; WAVE32-O0-NEXT: s_xor_saveexec_b32 s16, -1 |
| ; WAVE32-O0-NEXT: buffer_store_dword v32, off, s[0:3], s33 offset:128 ; 4-byte Folded Spill |
| ; WAVE32-O0-NEXT: buffer_store_dword v33, off, s[0:3], s33 offset:132 ; 4-byte Folded Spill |
| ; WAVE32-O0-NEXT: s_mov_b32 exec_lo, s16 |
| ; WAVE32-O0-NEXT: s_add_i32 s32, s32, 0x1200 |
| ; WAVE32-O0-NEXT: v_writelane_b32 v32, s30, 0 |
| ; WAVE32-O0-NEXT: v_writelane_b32 v32, s31, 1 |
| ; WAVE32-O0-NEXT: s_mov_b32 s16, s32 |
| ; WAVE32-O0-NEXT: ; implicit-def: $vgpr33 : SGPR spill to VGPR lane |
| ; WAVE32-O0-NEXT: v_writelane_b32 v33, s16, 0 |
| ; WAVE32-O0-NEXT: s_lshr_b32 s16, s16, 5 |
| ; WAVE32-O0-NEXT: v_writelane_b32 v33, s16, 1 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v0, 42 |
| ; WAVE32-O0-NEXT: buffer_store_dword v0, off, s[0:3], s33 |
| ; WAVE32-O0-NEXT: s_waitcnt_vscnt null, 0x0 |
| ; WAVE32-O0-NEXT: s_mov_b64 s[22:23], s[2:3] |
| ; WAVE32-O0-NEXT: s_mov_b64 s[20:21], s[0:1] |
| ; WAVE32-O0-NEXT: s_mov_b32 s16, s32 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v0, 17 |
| ; WAVE32-O0-NEXT: buffer_store_dword v0, off, s[0:3], s16 offset:4 |
| ; WAVE32-O0-NEXT: s_mov_b32 s18, stack_passed_argument@abs32@hi |
| ; WAVE32-O0-NEXT: s_mov_b32 s16, stack_passed_argument@abs32@lo |
| ; WAVE32-O0-NEXT: ; kill: def $sgpr16 killed $sgpr16 def $sgpr16_sgpr17 |
| ; WAVE32-O0-NEXT: s_mov_b32 s17, s18 |
| ; WAVE32-O0-NEXT: s_mov_b64 s[0:1], s[20:21] |
| ; WAVE32-O0-NEXT: s_mov_b64 s[2:3], s[22:23] |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v0, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v1, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v2, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v3, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v4, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v5, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v6, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v7, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v8, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v9, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v10, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v11, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v12, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v13, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v14, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v15, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v16, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v17, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v18, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v19, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v20, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v21, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v22, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v23, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v24, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v25, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v26, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v27, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v28, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v29, s18 |
| ; WAVE32-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-O0-NEXT: v_mov_b32_e32 v30, s18 |
| ; WAVE32-O0-NEXT: s_swappc_b64 s[30:31], s[16:17] |
| ; WAVE32-O0-NEXT: v_readlane_b32 s5, v33, 1 |
| ; WAVE32-O0-NEXT: v_readlane_b32 s4, v33, 0 |
| ; WAVE32-O0-NEXT: ;;#ASMSTART |
| ; WAVE32-O0-NEXT: ; use s5 |
| ; WAVE32-O0-NEXT: ;;#ASMEND |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-O0-NEXT: v_readlane_b32 s31, v32, 1 |
| ; WAVE32-O0-NEXT: v_readlane_b32 s30, v32, 0 |
| ; WAVE32-O0-NEXT: s_mov_b32 s32, s33 |
| ; WAVE32-O0-NEXT: s_xor_saveexec_b32 s4, -1 |
| ; WAVE32-O0-NEXT: buffer_load_dword v32, off, s[0:3], s33 offset:128 ; 4-byte Folded Reload |
| ; WAVE32-O0-NEXT: buffer_load_dword v33, off, s[0:3], s33 offset:132 ; 4-byte Folded Reload |
| ; WAVE32-O0-NEXT: s_mov_b32 exec_lo, s4 |
| ; WAVE32-O0-NEXT: s_mov_b32 s33, s24 |
| ; WAVE32-O0-NEXT: s_waitcnt vmcnt(0) |
| ; WAVE32-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE64-O0-LABEL: func_stacksave_stackrestore_call_with_stack_objects: |
| ; WAVE64-O0: ; %bb.0: |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE64-O0-NEXT: s_mov_b32 s19, s33 |
| ; WAVE64-O0-NEXT: s_mov_b32 s33, s32 |
| ; WAVE64-O0-NEXT: s_xor_saveexec_b64 s[16:17], -1 |
| ; WAVE64-O0-NEXT: buffer_store_dword v32, off, s[0:3], s33 offset:128 ; 4-byte Folded Spill |
| ; WAVE64-O0-NEXT: buffer_store_dword v33, off, s[0:3], s33 offset:132 ; 4-byte Folded Spill |
| ; WAVE64-O0-NEXT: s_mov_b64 exec, s[16:17] |
| ; WAVE64-O0-NEXT: s_add_i32 s32, s32, 0x2400 |
| ; WAVE64-O0-NEXT: v_writelane_b32 v32, s30, 0 |
| ; WAVE64-O0-NEXT: v_writelane_b32 v32, s31, 1 |
| ; WAVE64-O0-NEXT: s_mov_b32 s16, s32 |
| ; WAVE64-O0-NEXT: ; implicit-def: $vgpr33 : SGPR spill to VGPR lane |
| ; WAVE64-O0-NEXT: v_writelane_b32 v33, s16, 0 |
| ; WAVE64-O0-NEXT: s_lshr_b32 s16, s16, 6 |
| ; WAVE64-O0-NEXT: v_writelane_b32 v33, s16, 1 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v0, 42 |
| ; WAVE64-O0-NEXT: buffer_store_dword v0, off, s[0:3], s33 |
| ; WAVE64-O0-NEXT: s_waitcnt_vscnt null, 0x0 |
| ; WAVE64-O0-NEXT: s_mov_b64 s[22:23], s[2:3] |
| ; WAVE64-O0-NEXT: s_mov_b64 s[20:21], s[0:1] |
| ; WAVE64-O0-NEXT: s_mov_b32 s16, s32 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v0, 17 |
| ; WAVE64-O0-NEXT: buffer_store_dword v0, off, s[0:3], s16 offset:4 |
| ; WAVE64-O0-NEXT: s_mov_b32 s18, stack_passed_argument@abs32@hi |
| ; WAVE64-O0-NEXT: s_mov_b32 s16, stack_passed_argument@abs32@lo |
| ; WAVE64-O0-NEXT: ; kill: def $sgpr16 killed $sgpr16 def $sgpr16_sgpr17 |
| ; WAVE64-O0-NEXT: s_mov_b32 s17, s18 |
| ; WAVE64-O0-NEXT: s_mov_b64 s[0:1], s[20:21] |
| ; WAVE64-O0-NEXT: s_mov_b64 s[2:3], s[22:23] |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v0, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v1, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v2, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v3, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v4, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v5, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v6, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v7, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v8, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v9, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v10, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v11, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v12, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v13, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v14, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v15, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v16, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v17, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v18, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v19, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v20, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v21, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v22, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v23, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v24, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v25, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v26, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v27, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v28, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v29, s18 |
| ; WAVE64-O0-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE64-O0-NEXT: v_mov_b32_e32 v30, s18 |
| ; WAVE64-O0-NEXT: s_swappc_b64 s[30:31], s[16:17] |
| ; WAVE64-O0-NEXT: v_readlane_b32 s5, v33, 1 |
| ; WAVE64-O0-NEXT: v_readlane_b32 s4, v33, 0 |
| ; WAVE64-O0-NEXT: ;;#ASMSTART |
| ; WAVE64-O0-NEXT: ; use s5 |
| ; WAVE64-O0-NEXT: ;;#ASMEND |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s4 |
| ; WAVE64-O0-NEXT: v_readlane_b32 s31, v32, 1 |
| ; WAVE64-O0-NEXT: v_readlane_b32 s30, v32, 0 |
| ; WAVE64-O0-NEXT: s_mov_b32 s32, s33 |
| ; WAVE64-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1 |
| ; WAVE64-O0-NEXT: buffer_load_dword v32, off, s[0:3], s33 offset:128 ; 4-byte Folded Reload |
| ; WAVE64-O0-NEXT: buffer_load_dword v33, off, s[0:3], s33 offset:132 ; 4-byte Folded Reload |
| ; WAVE64-O0-NEXT: s_mov_b64 exec, s[4:5] |
| ; WAVE64-O0-NEXT: s_mov_b32 s33, s19 |
| ; WAVE64-O0-NEXT: s_waitcnt vmcnt(0) |
| ; WAVE64-O0-NEXT: s_setpc_b64 s[30:31] |
| ; |
| ; WAVE32-WWM-PREALLOC-LABEL: func_stacksave_stackrestore_call_with_stack_objects: |
| ; WAVE32-WWM-PREALLOC: ; %bb.0: |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s24, s33 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s33, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_xor_saveexec_b32 s16, -1 |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_store_dword v32, off, s[0:3], s33 offset:128 ; 4-byte Folded Spill |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_store_dword v33, off, s[0:3], s33 offset:132 ; 4-byte Folded Spill |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 exec_lo, s16 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_add_i32 s32, s32, 0x1200 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_writelane_b32 v33, s30, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_writelane_b32 v33, s31, 1 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s16, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $vgpr32 : SGPR spill to VGPR lane |
| ; WAVE32-WWM-PREALLOC-NEXT: v_writelane_b32 v32, s16, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_lshr_b32 s16, s16, 5 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_writelane_b32 v32, s16, 1 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v0, 42 |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_store_dword v0, off, s[0:3], s33 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt_vscnt null, 0x0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b64 s[22:23], s[2:3] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b64 s[20:21], s[0:1] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s16, s32 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v0, 17 |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_store_dword v0, off, s[0:3], s16 offset:4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s18, stack_passed_argument@abs32@hi |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s16, stack_passed_argument@abs32@lo |
| ; WAVE32-WWM-PREALLOC-NEXT: ; kill: def $sgpr16 killed $sgpr16 def $sgpr16_sgpr17 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s17, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b64 s[0:1], s[20:21] |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b64 s[2:3], s[22:23] |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v0, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v1, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v2, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v3, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v4, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v5, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v6, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v7, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v8, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v9, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v10, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v11, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v12, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v13, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v14, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v15, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v16, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v17, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v18, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v19, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v20, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v21, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v22, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v23, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v24, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v25, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v26, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v27, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v28, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v29, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: ; implicit-def: $sgpr18 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_mov_b32_e32 v30, s18 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_swappc_b64 s[30:31], s[16:17] |
| ; WAVE32-WWM-PREALLOC-NEXT: v_readlane_b32 s5, v32, 1 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_readlane_b32 s4, v32, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMSTART |
| ; WAVE32-WWM-PREALLOC-NEXT: ; use s5 |
| ; WAVE32-WWM-PREALLOC-NEXT: ;;#ASMEND |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_readlane_b32 s31, v33, 1 |
| ; WAVE32-WWM-PREALLOC-NEXT: v_readlane_b32 s30, v33, 0 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s32, s33 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_xor_saveexec_b32 s4, -1 |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_load_dword v32, off, s[0:3], s33 offset:128 ; 4-byte Folded Reload |
| ; WAVE32-WWM-PREALLOC-NEXT: buffer_load_dword v33, off, s[0:3], s33 offset:132 ; 4-byte Folded Reload |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 exec_lo, s4 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_mov_b32 s33, s24 |
| ; WAVE32-WWM-PREALLOC-NEXT: s_waitcnt vmcnt(0) |
| ; WAVE32-WWM-PREALLOC-NEXT: s_setpc_b64 s[30:31] |
| %alloca = alloca [32 x i32], addrspace(5) |
| %stacksave = call ptr addrspace(5) @llvm.stacksave.p5() |
| store volatile i32 42, ptr addrspace(5) %alloca |
| call void @stack_passed_argument([32 x i32] poison, i32 17) |
| call void asm sideeffect "; use $0", "s"(ptr addrspace(5) %stacksave) |
| call void @llvm.stackrestore.p5(ptr addrspace(5) %stacksave) |
| ret void |
| } |
| |
| !llvm.module.flags = !{!0} |
| !0 = !{i32 1, !"amdhsa_code_object_version", i32 500} |
| |
| ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: |
| ; WAVE32: {{.*}} |
| ; WAVE64: {{.*}} |