From c409f7029b48188b6cf3edff82a0646d8e6f81c3 Mon Sep 17 00:00:00 2001 From: Stanislav Mekhanoshin Date: Fri, 11 Oct 2019 22:03:36 +0000 Subject: [PATCH] [AMDGPU] link dpp pseudos and real instructions on gfx10 This defaults to zero fi operand, but we do not expose it anyway. Should we expose it later it needs to be added to the pseudo. This enables dpp combining on gfx10. Differential Revision: https://reviews.llvm.org/D68888 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@374604 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/AMDGPU/AMDGPUMCInstLower.cpp | 4 + lib/Target/AMDGPU/VOP1Instructions.td | 29 +- lib/Target/AMDGPU/VOP2Instructions.td | 29 +- .../atomic_optimizations_local_pointer.ll | 5226 ++++++++++++++++- test/CodeGen/AMDGPU/dpp_combine.ll | 53 + 5 files changed, 5036 insertions(+), 305 deletions(-) create mode 100644 test/CodeGen/AMDGPU/dpp_combine.ll diff --git a/lib/Target/AMDGPU/AMDGPUMCInstLower.cpp b/lib/Target/AMDGPU/AMDGPUMCInstLower.cpp index ae4c32c258a..3760aed87a4 100644 --- a/lib/Target/AMDGPU/AMDGPUMCInstLower.cpp +++ b/lib/Target/AMDGPU/AMDGPUMCInstLower.cpp @@ -211,6 +211,10 @@ void AMDGPUMCInstLower::lower(const MachineInstr *MI, MCInst &OutMI) const { lowerOperand(MO, MCOp); OutMI.addOperand(MCOp); } + + int FIIdx = AMDGPU::getNamedOperandIdx(MCOpcode, AMDGPU::OpName::fi); + if (FIIdx >= (int)OutMI.getNumOperands()) + OutMI.addOperand(MCOperand::createImm(0)); } bool AMDGPUAsmPrinter::lowerOperand(const MachineOperand &MO, diff --git a/lib/Target/AMDGPU/VOP1Instructions.td b/lib/Target/AMDGPU/VOP1Instructions.td index fdb996666df..0659b1c2bdf 100644 --- a/lib/Target/AMDGPU/VOP1Instructions.td +++ b/lib/Target/AMDGPU/VOP1Instructions.td @@ -441,7 +441,7 @@ let SubtargetPredicate = isGFX10Plus in { // Target-specific instruction encodings. //===----------------------------------------------------------------------===// -class VOP1_DPP op, VOP1_Pseudo ps, VOPProfile p = ps.Pfl, bit isDPP16 = 0> : +class VOP1_DPP op, VOP1_DPP_Pseudo ps, VOPProfile p = ps.Pfl, bit isDPP16 = 0> : VOP_DPP { let hasSideEffects = ps.hasSideEffects; let Defs = ps.Defs; @@ -455,8 +455,9 @@ class VOP1_DPP op, VOP1_Pseudo ps, VOPProfile p = ps.Pfl, bit isDPP16 = let Inst{31-25} = 0x3f; } -class VOP1_DPP16 op, VOP1_Pseudo ps, VOPProfile p = ps.Pfl> : - VOP1_DPP { +class VOP1_DPP16 op, VOP1_DPP_Pseudo ps, VOPProfile p = ps.Pfl> : + VOP1_DPP, + SIMCInstr { let AssemblerPredicate = !if(p.HasExt, HasDPP16, DisableInst); let SubtargetPredicate = HasDPP16; } @@ -507,7 +508,7 @@ let AssemblerPredicate = isGFX10Plus, DecoderNamespace = "GFX10" in { } multiclass VOP1_Real_dpp_gfx10 op> { foreach _ = BoolToList(NAME#"_e32").Pfl.HasExtDPP>.ret in - def _dpp_gfx10 : VOP1_DPP16(NAME#"_e32")> { + def _dpp_gfx10 : VOP1_DPP16(NAME#"_dpp")> { let DecoderNamespace = "SDWA10"; } } @@ -840,7 +841,7 @@ def V_MOVRELD_B32_V4 : V_MOVRELD_B32_pseudo; def V_MOVRELD_B32_V8 : V_MOVRELD_B32_pseudo; def V_MOVRELD_B32_V16 : V_MOVRELD_B32_pseudo; -let OtherPredicates = [isGFX8GFX9] in { +let OtherPredicates = [isGFX8Plus] in { def : GCNPat < (i32 (int_amdgcn_mov_dpp i32:$src, timm:$dpp_ctrl, timm:$row_mask, timm:$bank_mask, @@ -858,7 +859,7 @@ def : GCNPat < (as_i1imm $bound_ctrl)) >; -} // End OtherPredicates = [isGFX8GFX9] +} // End OtherPredicates = [isGFX8Plus] let OtherPredicates = [isGFX8Plus] in { def : GCNPat< @@ -916,20 +917,4 @@ def : GCNPat < (i32 (int_amdgcn_mov_dpp8 i32:$src, timm:$dpp8)), (V_MOV_B32_dpp8_gfx10 $src, $src, (as_i32imm $dpp8), (i32 DPP8Mode.FI_0)) >; - -def : GCNPat < - (i32 (int_amdgcn_mov_dpp i32:$src, timm:$dpp_ctrl, timm:$row_mask, timm:$bank_mask, - timm:$bound_ctrl)), - (V_MOV_B32_dpp_gfx10 $src, $src, (as_i32imm $dpp_ctrl), - (as_i32imm $row_mask), (as_i32imm $bank_mask), - (as_i1imm $bound_ctrl), (i32 0)) ->; - -def : GCNPat < - (i32 (int_amdgcn_update_dpp i32:$old, i32:$src, timm:$dpp_ctrl, timm:$row_mask, - timm:$bank_mask, timm:$bound_ctrl)), - (V_MOV_B32_dpp_gfx10 $old, $src, (as_i32imm $dpp_ctrl), - (as_i32imm $row_mask), (as_i32imm $bank_mask), - (as_i1imm $bound_ctrl), (i32 0)) ->; } // End OtherPredicates = [isGFX10Plus] diff --git a/lib/Target/AMDGPU/VOP2Instructions.td b/lib/Target/AMDGPU/VOP2Instructions.td index 8a5847d6f5b..c23755fe39a 100644 --- a/lib/Target/AMDGPU/VOP2Instructions.td +++ b/lib/Target/AMDGPU/VOP2Instructions.td @@ -658,14 +658,14 @@ let Constraints = "$vdst = $src2", isCommutable = 1, IsDOT = 1 in { let SubtargetPredicate = HasDot5Insts in - defm V_DOT2C_F32_F16 : VOP2Inst_e32<"v_dot2c_f32_f16", VOP_DOT_ACC_F32_V2F16>; + defm V_DOT2C_F32_F16 : VOP2Inst<"v_dot2c_f32_f16", VOP_DOT_ACC_F32_V2F16>; let SubtargetPredicate = HasDot6Insts in - defm V_DOT4C_I32_I8 : VOP2Inst_e32<"v_dot4c_i32_i8", VOP_DOT_ACC_I32_I32>; + defm V_DOT4C_I32_I8 : VOP2Inst<"v_dot4c_i32_i8", VOP_DOT_ACC_I32_I32>; let SubtargetPredicate = HasDot4Insts in - defm V_DOT2C_I32_I16 : VOP2Inst_e32<"v_dot2c_i32_i16", VOP_DOT_ACC_I32_I32>; + defm V_DOT2C_I32_I16 : VOP2Inst<"v_dot2c_i32_i16", VOP_DOT_ACC_I32_I32>; let SubtargetPredicate = HasDot3Insts in - defm V_DOT8C_I32_I4 : VOP2Inst_e32<"v_dot8c_i32_i4", VOP_DOT_ACC_I32_I32>; + defm V_DOT8C_I32_I4 : VOP2Inst<"v_dot8c_i32_i4", VOP_DOT_ACC_I32_I32>; } let AddedComplexity = 30 in { @@ -800,7 +800,7 @@ def : GCNPat< // Target-specific instruction encodings. //===----------------------------------------------------------------------===// -class VOP2_DPP op, VOP2_Pseudo ps, +class VOP2_DPP op, VOP2_DPP_Pseudo ps, string opName = ps.OpName, VOPProfile p = ps.Pfl, bit IsDPP16 = 0> : VOP_DPP { @@ -818,13 +818,18 @@ class VOP2_DPP op, VOP2_Pseudo ps, let Inst{31} = 0x0; } -class VOP2_DPP16 op, VOP2_Pseudo ps, +class Base_VOP2_DPP16 op, VOP2_DPP_Pseudo ps, string opName = ps.OpName, VOPProfile p = ps.Pfl> : VOP2_DPP { let AssemblerPredicate = !if(p.HasExt, HasDPP16, DisableInst); let SubtargetPredicate = HasDPP16; } +class VOP2_DPP16 op, VOP2_DPP_Pseudo ps, + string opName = ps.OpName, VOPProfile p = ps.Pfl> : + Base_VOP2_DPP16, + SIMCInstr ; + class VOP2_DPP8 op, VOP2_Pseudo ps, string opName = ps.OpName, VOPProfile p = ps.Pfl> : VOP_DPP8 { @@ -885,7 +890,7 @@ let AssemblerPredicate = isGFX10Plus, DecoderNamespace = "GFX10" in { } multiclass VOP2_Real_dpp_gfx10 op> { foreach _ = BoolToList(NAME#"_e32").Pfl.HasExtDPP>.ret in - def _dpp_gfx10 : VOP2_DPP16(NAME#"_e32")> { + def _dpp_gfx10 : VOP2_DPP16(NAME#"_dpp")> { let DecoderNamespace = "SDWA10"; } } @@ -929,7 +934,7 @@ let AssemblerPredicate = isGFX10Plus, DecoderNamespace = "GFX10" in { multiclass VOP2_Real_dpp_gfx10_with_name op, string opName, string asmName> { foreach _ = BoolToList(opName#"_e32").Pfl.HasExtDPP>.ret in - def _dpp_gfx10 : VOP2_DPP16(opName#"_e32")> { + def _dpp_gfx10 : VOP2_DPP16(opName#"_dpp")> { VOP2_Pseudo ps = !cast(opName#"_e32"); let AsmString = asmName # ps.Pfl.AsmDPP16; } @@ -969,7 +974,7 @@ let AssemblerPredicate = isGFX10Plus, DecoderNamespace = "GFX10" in { } foreach _ = BoolToList(opName#"_e32").Pfl.HasExtDPP>.ret in def _dpp_gfx10 : - VOP2_DPP16(opName#"_e32"), asmName> { + VOP2_DPP16(opName#"_dpp"), asmName> { string AsmDPP = !cast(opName#"_e32").Pfl.AsmDPP16; let AsmString = asmName # !subst(", vcc", "", AsmDPP); let DecoderNamespace = "SDWA10"; @@ -992,7 +997,7 @@ let AssemblerPredicate = isGFX10Plus, DecoderNamespace = "GFX10" in { let DecoderNamespace = "SDWA10"; } def _dpp_w32_gfx10 : - VOP2_DPP16(opName#"_e32"), asmName> { + Base_VOP2_DPP16(opName#"_dpp"), asmName> { string AsmDPP = !cast(opName#"_e32").Pfl.AsmDPP16; let AsmString = asmName # !subst("vcc", "vcc_lo", AsmDPP); let isAsmParserOnly = 1; @@ -1015,7 +1020,7 @@ let AssemblerPredicate = isGFX10Plus, DecoderNamespace = "GFX10" in { let DecoderNamespace = "SDWA10"; } def _dpp_w64_gfx10 : - VOP2_DPP16(opName#"_e32"), asmName> { + Base_VOP2_DPP16(opName#"_dpp"), asmName> { string AsmDPP = !cast(opName#"_e32").Pfl.AsmDPP16; let AsmString = asmName # AsmDPP; let isAsmParserOnly = 1; @@ -1513,7 +1518,7 @@ defm V_XNOR_B32 : VOP2_Real_e32e64_vi <0x3d>; } // End SubtargetPredicate = HasDLInsts multiclass VOP2_Real_DOT_ACC_gfx9 op> : VOP2_Real_e32_vi { - def _dpp : VOP2_DPP(NAME#"_e32")>; + def _dpp_vi : VOP2_DPP(NAME#"_dpp")>; } multiclass VOP2_Real_DOT_ACC_gfx10 op> : diff --git a/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll b/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll index 84a1172d1c4..de89224c1aa 100644 --- a/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll +++ b/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll @@ -1,9 +1,9 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GCN64,GFX7LESS %s -; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -mcpu=tonga -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GCN64,GFX8MORE,GFX8MORE64,DPPCOMB %s -; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -mcpu=gfx900 -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GCN64,GFX8MORE,GFX8MORE64,DPPCOMB %s -; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GCN64,GFX8MORE,GFX8MORE64,GFX1064 %s -; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GCN32,GFX8MORE,GFX8MORE32,GFX1032 %s +; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX7LESS %s +; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -mcpu=tonga -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX8 %s +; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -mcpu=gfx900 -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9 %s +; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1064 %s +; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1032 %s declare i32 @llvm.amdgcn.workitem.id.x() @@ -12,53 +12,596 @@ declare i32 @llvm.amdgcn.workitem.id.x() ; Show that what the atomic optimization pass will do for local pointers. -; GCN-LABEL: add_i32_constant: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN32: s_bcnt1_i32_b32 s[[popcount:[0-9]+]], s[[exec_lo]] -; GCN64: s_bcnt1_i32_b64 s[[popcount:[0-9]+]], s{{\[}}[[exec_lo]]:[[exec_hi]]{{\]}} -; GCN: v_mul_u32_u24{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[popcount]], 5 -; GCN: ds_add_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @add_i32_constant(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: add_i32_constant: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s4, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s5, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr1 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX7LESS-NEXT: ; mask branch BB0_2 +; GFX7LESS-NEXT: s_cbranch_execz BB0_2 +; GFX7LESS-NEXT: BB0_1: +; GFX7LESS-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: v_mul_u32_u24_e64 v2, s4, 5 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_add_rtn_u32 v1, v1, v2 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB0_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX7LESS-NEXT: v_readfirstlane_b32 s2, v1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: v_mad_u32_u24 v0, v0, 5, s2 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: add_i32_constant: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s4, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s5, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr1 +; GFX8-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX8-NEXT: ; mask branch BB0_2 +; GFX8-NEXT: s_cbranch_execz BB0_2 +; GFX8-NEXT: BB0_1: +; GFX8-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX8-NEXT: v_mul_u32_u24_e64 v1, s4, 5 +; GFX8-NEXT: v_mov_b32_e32 v2, local_var32@abs32@lo +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_add_rtn_u32 v1, v2, v1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB0_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX8-NEXT: v_readfirstlane_b32 s2, v1 +; GFX8-NEXT: v_mad_u32_u24 v0, v0, 5, s2 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: add_i32_constant: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s4, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s5, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr1 +; GFX9-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX9-NEXT: ; mask branch BB0_2 +; GFX9-NEXT: s_cbranch_execz BB0_2 +; GFX9-NEXT: BB0_1: +; GFX9-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX9-NEXT: v_mul_u32_u24_e64 v1, s4, 5 +; GFX9-NEXT: v_mov_b32_e32 v2, local_var32@abs32@lo +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_add_rtn_u32 v1, v2, v1 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB0_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX9-NEXT: v_readfirstlane_b32 s2, v1 +; GFX9-NEXT: v_mad_u32_u24 v0, v0, 5, s2 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: add_i32_constant: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: ; implicit-def: $vgpr1 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB0_2 +; GFX1064-NEXT: s_cbranch_execz BB0_2 +; GFX1064-NEXT: BB0_1: +; GFX1064-NEXT: s_bcnt1_i32_b64 s2, s[2:3] +; GFX1064-NEXT: v_mov_b32_e32 v2, local_var32@abs32@lo +; GFX1064-NEXT: v_mul_u32_u24_e64 v1, s2, 5 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_add_rtn_u32 v1, v2, v1 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB0_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s2, v1 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: v_mad_u32_u24 v0, v0, 5, s2 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: add_i32_constant: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s3, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: ; implicit-def: $vgpr1 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s3, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: s_and_saveexec_b32 s2, vcc_lo +; GFX1032-NEXT: ; mask branch BB0_2 +; GFX1032-NEXT: s_cbranch_execz BB0_2 +; GFX1032-NEXT: BB0_1: +; GFX1032-NEXT: s_bcnt1_i32_b32 s3, s3 +; GFX1032-NEXT: v_mov_b32_e32 v2, local_var32@abs32@lo +; GFX1032-NEXT: v_mul_u32_u24_e64 v1, s3, 5 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_add_rtn_u32 v1, v2, v1 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB0_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s2 +; GFX1032-NEXT: v_readfirstlane_b32 s2, v1 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: v_mad_u32_u24 v0, v0, 5, s2 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw add i32 addrspace(3)* @local_var32, i32 5 acq_rel store i32 %old, i32 addrspace(1)* %out ret void } -; GCN-LABEL: add_i32_uniform: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN32: s_bcnt1_i32_b32 s[[popcount:[0-9]+]], s[[exec_lo]] -; GCN64: s_bcnt1_i32_b64 s[[popcount:[0-9]+]], s{{\[}}[[exec_lo]]:[[exec_hi]]{{\]}} -; GCN: s_mul_i32 s[[scalar_value:[0-9]+]], s{{[0-9]+}}, s[[popcount]] -; GCN: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] -; GCN: ds_add_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @add_i32_uniform(i32 addrspace(1)* %out, i32 %additive) { +; +; +; GFX7LESS-LABEL: add_i32_uniform: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x9 +; GFX7LESS-NEXT: s_load_dword s2, s[0:1], 0xb +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s6, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s7, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr1 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[0:1], vcc +; GFX7LESS-NEXT: ; mask branch BB1_2 +; GFX7LESS-NEXT: s_cbranch_execz BB1_2 +; GFX7LESS-NEXT: BB1_1: +; GFX7LESS-NEXT: s_bcnt1_i32_b64 s3, s[6:7] +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: s_mul_i32 s3, s2, s3 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s3 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_add_rtn_u32 v1, v1, v2 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB1_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[0:1] +; GFX7LESS-NEXT: v_readfirstlane_b32 s0, v1 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX7LESS-NEXT: s_mov_b32 s7, 0xf000 +; GFX7LESS-NEXT: v_add_i32_e32 v0, vcc, s0, v0 +; GFX7LESS-NEXT: s_mov_b32 s6, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: add_i32_uniform: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x24 +; GFX8-NEXT: s_load_dword s0, s[0:1], 0x2c +; GFX8-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s7, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr1 +; GFX8-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX8-NEXT: ; mask branch BB1_2 +; GFX8-NEXT: s_cbranch_execz BB1_2 +; GFX8-NEXT: BB1_1: +; GFX8-NEXT: s_bcnt1_i32_b64 s1, s[6:7] +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: s_mul_i32 s1, s0, s1 +; GFX8-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v2, s1 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_add_rtn_u32 v1, v1, v2 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB1_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: v_mul_lo_u32 v0, s0, v0 +; GFX8-NEXT: v_readfirstlane_b32 s0, v1 +; GFX8-NEXT: s_mov_b32 s7, 0xf000 +; GFX8-NEXT: s_mov_b32 s6, -1 +; GFX8-NEXT: v_add_u32_e32 v0, vcc, s0, v0 +; GFX8-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: add_i32_uniform: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x24 +; GFX9-NEXT: s_load_dword s0, s[0:1], 0x2c +; GFX9-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s7, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr1 +; GFX9-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX9-NEXT: ; mask branch BB1_2 +; GFX9-NEXT: s_cbranch_execz BB1_2 +; GFX9-NEXT: BB1_1: +; GFX9-NEXT: s_bcnt1_i32_b64 s1, s[6:7] +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: s_mul_i32 s1, s0, s1 +; GFX9-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v2, s1 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_add_rtn_u32 v1, v1, v2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB1_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: v_mul_lo_u32 v0, s0, v0 +; GFX9-NEXT: v_readfirstlane_b32 s0, v1 +; GFX9-NEXT: s_mov_b32 s7, 0xf000 +; GFX9-NEXT: s_mov_b32 s6, -1 +; GFX9-NEXT: v_add_u32_e32 v0, s0, v0 +; GFX9-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: add_i32_uniform: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x24 +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: s_load_dword s0, s[0:1], 0x2c +; GFX1064-NEXT: ; implicit-def: $vgpr1 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: s_and_saveexec_b64 s[6:7], vcc +; GFX1064-NEXT: ; mask branch BB1_2 +; GFX1064-NEXT: s_cbranch_execz BB1_2 +; GFX1064-NEXT: BB1_1: +; GFX1064-NEXT: s_bcnt1_i32_b64 s1, s[2:3] +; GFX1064-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: s_mul_i32 s1, s0, s1 +; GFX1064-NEXT: v_mov_b32_e32 v2, s1 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_add_rtn_u32 v1, v1, v2 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB1_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[6:7] +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: v_mul_lo_u32 v0, s0, v0 +; GFX1064-NEXT: v_readfirstlane_b32 s0, v1 +; GFX1064-NEXT: s_mov_b32 s7, 0x31016000 +; GFX1064-NEXT: s_mov_b32 s6, -1 +; GFX1064-NEXT: v_add_nc_u32_e32 v0, s0, v0 +; GFX1064-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: add_i32_uniform: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x24 +; GFX1032-NEXT: s_load_dword s0, s[0:1], 0x2c +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: ; implicit-def: $vgpr1 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: s_and_saveexec_b32 s1, vcc_lo +; GFX1032-NEXT: ; mask branch BB1_2 +; GFX1032-NEXT: s_cbranch_execz BB1_2 +; GFX1032-NEXT: BB1_1: +; GFX1032-NEXT: s_bcnt1_i32_b32 s2, s2 +; GFX1032-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: s_mul_i32 s2, s0, s2 +; GFX1032-NEXT: v_mov_b32_e32 v2, s2 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_add_rtn_u32 v1, v1, v2 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB1_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: v_mul_lo_u32 v0, s0, v0 +; GFX1032-NEXT: v_readfirstlane_b32 s0, v1 +; GFX1032-NEXT: s_mov_b32 s7, 0x31016000 +; GFX1032-NEXT: s_mov_b32 s6, -1 +; GFX1032-NEXT: v_add_nc_u32_e32 v0, s0, v0 +; GFX1032-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw add i32 addrspace(3)* @local_var32, i32 %additive acq_rel store i32 %old, i32 addrspace(1)* %out ret void } -; GCN-LABEL: add_i32_varying: ; GFX7LESS-NOT: v_mbcnt_lo_u32_b32 ; GFX7LESS-NOT: v_mbcnt_hi_u32_b32 ; GFX7LESS-NOT: s_bcnt1_i32_b64 -; GFX7LESS: ds_add_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} ; DPPCOMB: v_add_u32_dpp ; DPPCOMB: v_add_u32_dpp ; GFX8MORE32: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 31 -; GFX8MORE64: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 63 ; GFX8MORE: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] ; GFX8MORE: ds_add_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @add_i32_varying(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: add_i32_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_add_rtn_u32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: add_i32_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, 0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB2_2 +; GFX8-NEXT: s_cbranch_execz BB2_2 +; GFX8-NEXT: BB2_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_add_rtn_u32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB2_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_add_u32_e32 v0, vcc, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: add_i32_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB2_2 +; GFX9-NEXT: s_cbranch_execz BB2_2 +; GFX9-NEXT: BB2_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_add_rtn_u32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB2_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_add_u32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: add_i32_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, 0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB2_2 +; GFX1064-NEXT: s_cbranch_execz BB2_2 +; GFX1064-NEXT: BB2_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_add_rtn_u32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB2_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_add_nc_u32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: add_i32_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB2_2 +; GFX1032-NEXT: s_cbranch_execz BB2_2 +; GFX1032-NEXT: BB2_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_add_rtn_u32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB2_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_add_nc_u32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw add i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -67,64 +610,241 @@ entry: } define amdgpu_kernel void @add_i32_varying_gfx1032(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: add_i32_varying_gfx1032: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_add_rtn_u32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: add_i32_varying_gfx1032: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, 0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB3_2 +; GFX8-NEXT: s_cbranch_execz BB3_2 +; GFX8-NEXT: BB3_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_add_rtn_u32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB3_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_add_u32_e32 v0, vcc, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: add_i32_varying_gfx1032: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB3_2 +; GFX9-NEXT: s_cbranch_execz BB3_2 +; GFX9-NEXT: BB3_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_add_rtn_u32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB3_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_add_u32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: add_i32_varying_gfx1032: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, 0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB3_2 +; GFX1064-NEXT: s_cbranch_execz BB3_2 +; GFX1064-NEXT: BB3_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_add_rtn_u32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB3_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_add_nc_u32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; ; GFX1032-LABEL: add_i32_varying_gfx1032: -; GFX1032: v_mov_b32_e32 v2, v0 -; GFX1032: s_or_saveexec_b32 s2, -1 -; GFX1032: s_load_dwordx2 s[0:1], s[0:1], 0x24 -; GFX1032: v_mov_b32_e32 v1, 0 -; GFX1032: s_mov_b32 exec_lo, s2 -; GFX1032: v_cmp_ne_u32_e64 s2, 1, 0 -; GFX1032: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 -; GFX1032: s_not_b32 exec_lo, exec_lo -; GFX1032: v_mov_b32_e32 v2, 0 -; GFX1032: s_not_b32 exec_lo, exec_lo -; GFX1032: s_or_saveexec_b32 s4, -1 -; GFX1032: v_mov_b32_e32 v3, v1 -; GFX1032: v_mov_b32_e32 v4, v1 -; GFX1032: s_mov_b32 s2, -1 -; GFX1032: v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf -; GFX1032: v_add_nc_u32_e32 v2, v2, v3 -; GFX1032: v_mov_b32_e32 v3, v1 -; GFX1032: v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf -; GFX1032: v_add_nc_u32_e32 v2, v2, v3 -; GFX1032: v_mov_b32_e32 v3, v1 -; GFX1032: v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf -; GFX1032: v_add_nc_u32_e32 v2, v2, v3 -; GFX1032: v_mov_b32_e32 v3, v1 -; GFX1032: v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf -; GFX1032: v_add_nc_u32_e32 v2, v2, v3 -; GFX1032: v_mov_b32_e32 v3, v2 -; GFX1032: v_permlanex16_b32 v3, v3, -1, -1 -; GFX1032: v_mov_b32_dpp v4, v3 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf -; GFX1032: v_add_nc_u32_e32 v2, v2, v4 -; GFX1032: v_readlane_b32 s3, v2, 31 -; GFX1032: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf -; GFX1032: v_readlane_b32 s5, v2, 15 -; GFX1032: v_writelane_b32 v1, s5, 16 -; GFX1032: s_mov_b32 exec_lo, s4 -; GFX1032: v_cmp_eq_u32_e32 vcc_lo, 0, v0 -; GFX1032: s_and_saveexec_b32 s4, vcc_lo -; GFX1032: s_cbranch_execz BB3_2 -; GFX1032: BB3_1: -; GFX1032: v_mov_b32_e32 v0, local_var32@abs32@lo -; GFX1032: v_mov_b32_e32 v5, s3 -; GFX1032: s_waitcnt vmcnt(0) lgkmcnt(0) -; GFX1032: s_waitcnt_vscnt null, 0x0 -; GFX1032: ds_add_rtn_u32 v0, v0, v5 -; GFX1032: s_waitcnt vmcnt(0) lgkmcnt(0) -; GFX1032: buffer_gl0_inv -; GFX1032: buffer_gl1_inv -; GFX1032: BB3_2: -; GFX1032: v_nop -; GFX1032: s_or_b32 exec_lo, exec_lo, s4 -; GFX1032: v_readfirstlane_b32 s3, v0 -; GFX1032: v_mov_b32_e32 v0, v1 -; GFX1032: v_add_nc_u32_e32 v0, s3, v0 -; GFX1032: s_mov_b32 s3, 0x31016000 -; GFX1032: s_nop 1 -; GFX1032: s_waitcnt lgkmcnt(0) -; GFX1032: buffer_store_dword v0, off, s[0:3], 0 -; GFX1032: s_endpgm +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB3_2 +; GFX1032-NEXT: s_cbranch_execz BB3_2 +; GFX1032-NEXT: BB3_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_add_rtn_u32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB3_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_add_nc_u32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw add i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -133,74 +853,241 @@ entry: } define amdgpu_kernel void @add_i32_varying_gfx1064(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: add_i32_varying_gfx1064: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_add_rtn_u32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: add_i32_varying_gfx1064: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, 0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB4_2 +; GFX8-NEXT: s_cbranch_execz BB4_2 +; GFX8-NEXT: BB4_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_add_rtn_u32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB4_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_add_u32_e32 v0, vcc, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: add_i32_varying_gfx1064: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB4_2 +; GFX9-NEXT: s_cbranch_execz BB4_2 +; GFX9-NEXT: BB4_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_add_rtn_u32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB4_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_add_u32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; ; GFX1064-LABEL: add_i32_varying_gfx1064: -; GFX1064: v_mov_b32_e32 v2, v0 -; GFX1064: s_or_saveexec_b64 s[2:3], -1 -; GFX1064: s_load_dwordx2 s[0:1], s[0:1], 0x24 -; GFX1064: v_mov_b32_e32 v1, 0 -; GFX1064: s_mov_b64 exec, s[2:3] -; GFX1064: v_cmp_ne_u32_e64 s[2:3], 1, 0 -; GFX1064: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 -; GFX1064: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 -; GFX1064: s_not_b64 exec, exec -; GFX1064: v_mov_b32_e32 v2, 0 -; GFX1064: s_not_b64 exec, exec -; GFX1064: s_or_saveexec_b64 s[4:5], -1 -; GFX1064: v_mov_b32_e32 v3, v1 -; GFX1064: v_mov_b32_e32 v4, v1 -; GFX1064: s_mov_b32 s2, -1 -; GFX1064: v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf -; GFX1064: v_add_nc_u32_e32 v2, v2, v3 -; GFX1064: v_mov_b32_e32 v3, v1 -; GFX1064: v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf -; GFX1064: v_add_nc_u32_e32 v2, v2, v3 -; GFX1064: v_mov_b32_e32 v3, v1 -; GFX1064: v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf -; GFX1064: v_add_nc_u32_e32 v2, v2, v3 -; GFX1064: v_mov_b32_e32 v3, v1 -; GFX1064: v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf -; GFX1064: v_add_nc_u32_e32 v2, v2, v3 -; GFX1064: v_mov_b32_e32 v3, v2 -; GFX1064: v_permlanex16_b32 v3, v3, -1, -1 -; GFX1064: v_mov_b32_dpp v4, v3 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf -; GFX1064: v_add_nc_u32_e32 v2, v2, v4 -; GFX1064: v_mov_b32_e32 v4, v1 -; GFX1064: v_readlane_b32 s3, v2, 31 -; GFX1064: v_mov_b32_e32 v3, s3 -; GFX1064: v_mov_b32_dpp v4, v3 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf -; GFX1064: v_add_nc_u32_e32 v2, v2, v4 -; GFX1064: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf -; GFX1064: v_readlane_b32 s3, v2, 15 -; GFX1064: v_readlane_b32 s6, v2, 31 -; GFX1064: v_writelane_b32 v1, s3, 16 -; GFX1064: v_readlane_b32 s3, v2, 63 -; GFX1064: v_writelane_b32 v1, s6, 32 -; GFX1064: v_readlane_b32 s6, v2, 47 -; GFX1064: v_writelane_b32 v1, s6, 48 -; GFX1064: s_mov_b64 exec, s[4:5] -; GFX1064: v_cmp_eq_u32_e32 vcc, 0, v0 -; GFX1064: s_and_saveexec_b64 s[4:5], vcc -; GFX1064: s_cbranch_execz BB4_2 -; GFX1064: BB4_1: -; GFX1064: v_mov_b32_e32 v0, local_var32@abs32@lo -; GFX1064: v_mov_b32_e32 v5, s3 -; GFX1064: s_waitcnt vmcnt(0) lgkmcnt(0) -; GFX1064: s_waitcnt_vscnt null, 0x0 -; GFX1064: ds_add_rtn_u32 v0, v0, v5 -; GFX1064: s_waitcnt vmcnt(0) lgkmcnt(0) -; GFX1064: buffer_gl0_inv -; GFX1064: buffer_gl1_inv -; GFX1064: BB4_2: -; GFX1064: v_nop -; GFX1064: s_or_b64 exec, exec, s[4:5] -; GFX1064: v_readfirstlane_b32 s3, v0 -; GFX1064: v_mov_b32_e32 v0, v1 -; GFX1064: v_add_nc_u32_e32 v0, s3, v0 -; GFX1064: s_mov_b32 s3, 0x31016000 -; GFX1064: s_nop 1 -; GFX1064: s_waitcnt lgkmcnt(0) -; GFX1064: buffer_store_dword v0, off, s[0:3], 0 -; GFX1064: s_endpgm +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, 0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB4_2 +; GFX1064-NEXT: s_cbranch_execz BB4_2 +; GFX1064-NEXT: BB4_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_add_rtn_u32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB4_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_add_nc_u32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: add_i32_varying_gfx1064: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB4_2 +; GFX1032-NEXT: s_cbranch_execz BB4_2 +; GFX1032-NEXT: BB4_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_add_rtn_u32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB4_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_add_nc_u32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw add i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -208,46 +1095,495 @@ entry: ret void } -; GCN-LABEL: add_i64_constant: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN32: s_bcnt1_i32_b32 s[[popcount:[0-9]+]], s[[exec_lo]] -; GCN64: s_bcnt1_i32_b64 s[[popcount:[0-9]+]], s{{\[}}[[exec_lo]]:[[exec_hi]]{{\]}} -; GCN: v_mul_hi_u32_u24{{(_e[0-9]+)?}} v[[value_hi:[0-9]+]], s[[popcount]], 5 -; GCN: v_mul_u32_u24{{(_e[0-9]+)?}} v[[value_lo:[0-9]+]], s[[popcount]], 5 -; GCN: ds_add_rtn_u64 v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}}, v{{[0-9]+}}, v{{\[}}[[value_lo]]:[[value_hi]]{{\]}} define amdgpu_kernel void @add_i64_constant(i64 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: add_i64_constant: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s4, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s5, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX7LESS-NEXT: ; mask branch BB5_2 +; GFX7LESS-NEXT: s_cbranch_execz BB5_2 +; GFX7LESS-NEXT: BB5_1: +; GFX7LESS-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX7LESS-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX7LESS-NEXT: v_mul_hi_u32_u24_e64 v2, s4, 5 +; GFX7LESS-NEXT: v_mul_u32_u24_e64 v1, s4, 5 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_add_rtn_u64 v[1:2], v3, v[1:2] +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB5_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX7LESS-NEXT: v_readfirstlane_b32 s2, v1 +; GFX7LESS-NEXT: v_readfirstlane_b32 s4, v2 +; GFX7LESS-NEXT: v_mul_hi_u32_u24_e32 v1, 5, v0 +; GFX7LESS-NEXT: v_mul_u32_u24_e32 v0, 5, v0 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s4 +; GFX7LESS-NEXT: v_add_i32_e32 v0, vcc, s2, v0 +; GFX7LESS-NEXT: v_addc_u32_e32 v1, vcc, v2, v1, vcc +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: add_i64_constant: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s4, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s5, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX8-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX8-NEXT: ; mask branch BB5_2 +; GFX8-NEXT: s_cbranch_execz BB5_2 +; GFX8-NEXT: BB5_1: +; GFX8-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX8-NEXT: v_mul_hi_u32_u24_e64 v2, s4, 5 +; GFX8-NEXT: v_mul_u32_u24_e64 v1, s4, 5 +; GFX8-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_add_rtn_u64 v[1:2], v3, v[1:2] +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB5_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX8-NEXT: v_readfirstlane_b32 s2, v1 +; GFX8-NEXT: v_readfirstlane_b32 s3, v2 +; GFX8-NEXT: v_mad_u64_u32 v[0:1], s[2:3], v0, 5, s[2:3] +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 2 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: add_i64_constant: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s4, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s5, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX9-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX9-NEXT: ; mask branch BB5_2 +; GFX9-NEXT: s_cbranch_execz BB5_2 +; GFX9-NEXT: BB5_1: +; GFX9-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX9-NEXT: v_mul_hi_u32_u24_e64 v2, s4, 5 +; GFX9-NEXT: v_mul_u32_u24_e64 v1, s4, 5 +; GFX9-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_add_rtn_u64 v[1:2], v3, v[1:2] +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB5_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX9-NEXT: v_readfirstlane_b32 s2, v1 +; GFX9-NEXT: v_readfirstlane_b32 s3, v2 +; GFX9-NEXT: v_mad_u64_u32 v[0:1], s[2:3], v0, 5, s[2:3] +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 2 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: add_i64_constant: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB5_2 +; GFX1064-NEXT: s_cbranch_execz BB5_2 +; GFX1064-NEXT: BB5_1: +; GFX1064-NEXT: s_bcnt1_i32_b64 s2, s[2:3] +; GFX1064-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX1064-NEXT: v_mul_hi_u32_u24_e64 v2, s2, 5 +; GFX1064-NEXT: v_mul_u32_u24_e64 v1, s2, 5 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_add_rtn_u64 v[1:2], v3, v[1:2] +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB5_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s2, v1 +; GFX1064-NEXT: v_readfirstlane_b32 s3, v2 +; GFX1064-NEXT: v_mad_u64_u32 v[0:1], s[2:3], v0, 5, s[2:3] +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: s_nop 2 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: add_i64_constant: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s3, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s3, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: s_and_saveexec_b32 s2, vcc_lo +; GFX1032-NEXT: ; mask branch BB5_2 +; GFX1032-NEXT: s_cbranch_execz BB5_2 +; GFX1032-NEXT: BB5_1: +; GFX1032-NEXT: s_bcnt1_i32_b32 s3, s3 +; GFX1032-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX1032-NEXT: v_mul_hi_u32_u24_e64 v2, s3, 5 +; GFX1032-NEXT: v_mul_u32_u24_e64 v1, s3, 5 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_add_rtn_u64 v[1:2], v3, v[1:2] +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB5_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s2 +; GFX1032-NEXT: v_readfirstlane_b32 s2, v1 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v2 +; GFX1032-NEXT: v_mad_u64_u32 v[0:1], s2, v0, 5, s[2:3] +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: s_nop 2 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw add i64 addrspace(3)* @local_var64, i64 5 acq_rel store i64 %old, i64 addrspace(1)* %out ret void } -; GCN-LABEL: add_i64_uniform: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN32: s_bcnt1_i32_b32 s{{[0-9]+}}, s[[exec_lo]] -; GCN64: s_bcnt1_i32_b64 s{{[0-9]+}}, s{{\[}}[[exec_lo]]:[[exec_hi]]{{\]}} -; GCN: ds_add_rtn_u64 v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}}, v{{[0-9]+}}, v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}} define amdgpu_kernel void @add_i64_uniform(i64 addrspace(1)* %out, i64 %additive) { +; +; +; GFX7LESS-LABEL: add_i64_uniform: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x9 +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s6, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s7, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX7LESS-NEXT: ; mask branch BB6_2 +; GFX7LESS-NEXT: s_cbranch_execz BB6_2 +; GFX7LESS-NEXT: BB6_1: +; GFX7LESS-NEXT: s_bcnt1_i32_b64 s6, s[6:7] +; GFX7LESS-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: s_mul_i32 s7, s3, s6 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, s6 +; GFX7LESS-NEXT: v_mul_hi_u32 v1, s2, v1 +; GFX7LESS-NEXT: s_mul_i32 s6, s2, s6 +; GFX7LESS-NEXT: v_add_i32_e32 v2, vcc, s7, v1 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, s6 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_add_rtn_u64 v[1:2], v3, v[1:2] +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB6_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX7LESS-NEXT: s_mov_b32 s7, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s6, -1 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: s_mov_b32 s4, s0 +; GFX7LESS-NEXT: s_mov_b32 s5, s1 +; GFX7LESS-NEXT: v_readfirstlane_b32 s0, v1 +; GFX7LESS-NEXT: v_readfirstlane_b32 s1, v2 +; GFX7LESS-NEXT: v_mul_lo_u32 v1, s3, v0 +; GFX7LESS-NEXT: v_mul_hi_u32 v2, s2, v0 +; GFX7LESS-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX7LESS-NEXT: v_add_i32_e32 v1, vcc, v2, v1 +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s1 +; GFX7LESS-NEXT: v_add_i32_e32 v0, vcc, s0, v0 +; GFX7LESS-NEXT: v_addc_u32_e32 v1, vcc, v2, v1, vcc +; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: add_i64_uniform: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s7, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB6_2 +; GFX8-NEXT: s_cbranch_execz BB6_2 +; GFX8-NEXT: BB6_1: +; GFX8-NEXT: s_bcnt1_i32_b64 s6, s[6:7] +; GFX8-NEXT: v_mov_b32_e32 v1, s6 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: v_mul_hi_u32 v1, s2, v1 +; GFX8-NEXT: s_mul_i32 s7, s3, s6 +; GFX8-NEXT: s_mul_i32 s6, s2, s6 +; GFX8-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX8-NEXT: v_add_u32_e32 v2, vcc, s7, v1 +; GFX8-NEXT: v_mov_b32_e32 v1, s6 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_add_rtn_u64 v[1:2], v3, v[1:2] +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB6_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: s_mov_b32 s4, s0 +; GFX8-NEXT: v_readfirstlane_b32 s0, v1 +; GFX8-NEXT: v_mul_lo_u32 v1, s3, v0 +; GFX8-NEXT: v_mul_hi_u32 v3, s2, v0 +; GFX8-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX8-NEXT: s_mov_b32 s5, s1 +; GFX8-NEXT: v_readfirstlane_b32 s1, v2 +; GFX8-NEXT: v_add_u32_e32 v1, vcc, v3, v1 +; GFX8-NEXT: v_mov_b32_e32 v2, s1 +; GFX8-NEXT: v_add_u32_e32 v0, vcc, s0, v0 +; GFX8-NEXT: s_mov_b32 s7, 0xf000 +; GFX8-NEXT: s_mov_b32 s6, -1 +; GFX8-NEXT: v_addc_u32_e32 v1, vcc, v2, v1, vcc +; GFX8-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: add_i64_uniform: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s7, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB6_2 +; GFX9-NEXT: s_cbranch_execz BB6_2 +; GFX9-NEXT: BB6_1: +; GFX9-NEXT: s_bcnt1_i32_b64 s6, s[6:7] +; GFX9-NEXT: v_mov_b32_e32 v1, s6 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: v_mul_hi_u32 v2, s2, v1 +; GFX9-NEXT: s_mul_i32 s7, s3, s6 +; GFX9-NEXT: s_mul_i32 s6, s2, s6 +; GFX9-NEXT: v_mov_b32_e32 v1, s6 +; GFX9-NEXT: v_add_u32_e32 v2, s7, v2 +; GFX9-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_add_rtn_u64 v[1:2], v3, v[1:2] +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB6_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: v_mul_lo_u32 v3, s3, v0 +; GFX9-NEXT: v_mul_hi_u32 v4, s2, v0 +; GFX9-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s4, s0 +; GFX9-NEXT: v_readfirstlane_b32 s0, v1 +; GFX9-NEXT: s_mov_b32 s5, s1 +; GFX9-NEXT: v_readfirstlane_b32 s1, v2 +; GFX9-NEXT: v_add_u32_e32 v1, v4, v3 +; GFX9-NEXT: v_mov_b32_e32 v2, s1 +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, s0, v0 +; GFX9-NEXT: s_mov_b32 s7, 0xf000 +; GFX9-NEXT: s_mov_b32 s6, -1 +; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, v2, v1, vcc +; GFX9-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: add_i64_uniform: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX1064-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24 +; GFX1064-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s6, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s7, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB6_2 +; GFX1064-NEXT: s_cbranch_execz BB6_2 +; GFX1064-NEXT: BB6_1: +; GFX1064-NEXT: s_bcnt1_i32_b64 s6, s[6:7] +; GFX1064-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: v_mul_hi_u32 v2, s2, s6 +; GFX1064-NEXT: s_mul_i32 s7, s2, s6 +; GFX1064-NEXT: s_mul_i32 s6, s3, s6 +; GFX1064-NEXT: v_mov_b32_e32 v1, s7 +; GFX1064-NEXT: v_add_nc_u32_e32 v2, s6, v2 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_add_rtn_u64 v[1:2], v3, v[1:2] +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB6_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: v_mul_lo_u32 v3, s3, v0 +; GFX1064-NEXT: v_mul_hi_u32 v4, s2, v0 +; GFX1064-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX1064-NEXT: v_readfirstlane_b32 s4, v1 +; GFX1064-NEXT: v_readfirstlane_b32 s5, v2 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_add_nc_u32_e32 v1, v4, v3 +; GFX1064-NEXT: v_add_co_u32_e64 v0, vcc, s4, v0 +; GFX1064-NEXT: v_add_co_ci_u32_e32 v1, vcc, s5, v1, vcc +; GFX1064-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: add_i64_uniform: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s5, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s5, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB6_2 +; GFX1032-NEXT: s_cbranch_execz BB6_2 +; GFX1032-NEXT: BB6_1: +; GFX1032-NEXT: s_bcnt1_i32_b32 s5, s5 +; GFX1032-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: v_mul_hi_u32 v2, s2, s5 +; GFX1032-NEXT: s_mul_i32 s6, s2, s5 +; GFX1032-NEXT: s_mul_i32 s5, s3, s5 +; GFX1032-NEXT: v_mov_b32_e32 v1, s6 +; GFX1032-NEXT: v_add_nc_u32_e32 v2, s5, v2 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_add_rtn_u64 v[1:2], v3, v[1:2] +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB6_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: v_mul_lo_u32 v3, s3, v0 +; GFX1032-NEXT: v_mul_hi_u32 v4, s2, v0 +; GFX1032-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX1032-NEXT: v_readfirstlane_b32 s4, v1 +; GFX1032-NEXT: v_readfirstlane_b32 s5, v2 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_add_nc_u32_e32 v1, v4, v3 +; GFX1032-NEXT: v_add_co_u32_e64 v0, vcc_lo, s4, v0 +; GFX1032-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, s5, v1, vcc_lo +; GFX1032-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw add i64 addrspace(3)* @local_var64, i64 %additive acq_rel store i64 %old, i64 addrspace(1)* %out ret void } -; GCN-LABEL: add_i64_varying: ; GCN-NOT: v_mbcnt_lo_u32_b32 ; GCN-NOT: v_mbcnt_hi_u32_b32 ; GCN-NOT: s_bcnt1_i32_b64 -; GCN: ds_add_rtn_u64 v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}}, v{{[0-9]+}}, v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}} define amdgpu_kernel void @add_i64_varying(i64 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: add_i64_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0 +; GFX7LESS-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_add_rtn_u64 v[0:1], v2, v[0:1] +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: add_i64_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_add_rtn_u64 v[0:1], v2, v[0:1] +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: add_i64_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_add_rtn_u64 v[0:1], v2, v[0:1] +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: add_i64_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_add_rtn_u64 v[0:1], v2, v[0:1] +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: add_i64_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_add_rtn_u64 v[0:1], v2, v[0:1] +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %zext = zext i32 %lane to i64 @@ -256,53 +1592,601 @@ entry: ret void } -; GCN-LABEL: sub_i32_constant: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN32: s_bcnt1_i32_b32 s[[popcount:[0-9]+]], s[[exec_lo]] -; GCN64: s_bcnt1_i32_b64 s[[popcount:[0-9]+]], s{{\[}}[[exec_lo]]:[[exec_hi]]{{\]}} -; GCN: v_mul_u32_u24{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[popcount]], 5 -; GCN: ds_sub_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @sub_i32_constant(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: sub_i32_constant: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s4, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s5, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr1 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX7LESS-NEXT: ; mask branch BB8_2 +; GFX7LESS-NEXT: s_cbranch_execz BB8_2 +; GFX7LESS-NEXT: BB8_1: +; GFX7LESS-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: v_mul_u32_u24_e64 v2, s4, 5 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_sub_rtn_u32 v1, v1, v2 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB8_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX7LESS-NEXT: v_readfirstlane_b32 s2, v1 +; GFX7LESS-NEXT: v_mul_u32_u24_e32 v0, 5, v0 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: v_sub_i32_e32 v0, vcc, s2, v0 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: sub_i32_constant: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s4, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s5, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr1 +; GFX8-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX8-NEXT: ; mask branch BB8_2 +; GFX8-NEXT: s_cbranch_execz BB8_2 +; GFX8-NEXT: BB8_1: +; GFX8-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX8-NEXT: v_mul_u32_u24_e64 v1, s4, 5 +; GFX8-NEXT: v_mov_b32_e32 v2, local_var32@abs32@lo +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_sub_rtn_u32 v1, v2, v1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB8_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX8-NEXT: v_readfirstlane_b32 s2, v1 +; GFX8-NEXT: v_mul_u32_u24_e32 v0, 5, v0 +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: sub_i32_constant: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s4, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s5, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr1 +; GFX9-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX9-NEXT: ; mask branch BB8_2 +; GFX9-NEXT: s_cbranch_execz BB8_2 +; GFX9-NEXT: BB8_1: +; GFX9-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX9-NEXT: v_mul_u32_u24_e64 v1, s4, 5 +; GFX9-NEXT: v_mov_b32_e32 v2, local_var32@abs32@lo +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_sub_rtn_u32 v1, v2, v1 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB8_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX9-NEXT: v_readfirstlane_b32 s2, v1 +; GFX9-NEXT: v_mul_u32_u24_e32 v0, 5, v0 +; GFX9-NEXT: v_sub_u32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: sub_i32_constant: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: ; implicit-def: $vgpr1 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB8_2 +; GFX1064-NEXT: s_cbranch_execz BB8_2 +; GFX1064-NEXT: BB8_1: +; GFX1064-NEXT: s_bcnt1_i32_b64 s2, s[2:3] +; GFX1064-NEXT: v_mov_b32_e32 v2, local_var32@abs32@lo +; GFX1064-NEXT: v_mul_u32_u24_e64 v1, s2, 5 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_sub_rtn_u32 v1, v2, v1 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB8_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s2, v1 +; GFX1064-NEXT: v_mul_u32_u24_e32 v0, 5, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: v_sub_nc_u32_e32 v0, s2, v0 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: s_nop 0 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: sub_i32_constant: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s3, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: ; implicit-def: $vgpr1 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s3, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: s_and_saveexec_b32 s2, vcc_lo +; GFX1032-NEXT: ; mask branch BB8_2 +; GFX1032-NEXT: s_cbranch_execz BB8_2 +; GFX1032-NEXT: BB8_1: +; GFX1032-NEXT: s_bcnt1_i32_b32 s3, s3 +; GFX1032-NEXT: v_mov_b32_e32 v2, local_var32@abs32@lo +; GFX1032-NEXT: v_mul_u32_u24_e64 v1, s3, 5 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_sub_rtn_u32 v1, v2, v1 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB8_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s2 +; GFX1032-NEXT: v_readfirstlane_b32 s2, v1 +; GFX1032-NEXT: v_mul_u32_u24_e32 v0, 5, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: v_sub_nc_u32_e32 v0, s2, v0 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: s_nop 0 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw sub i32 addrspace(3)* @local_var32, i32 5 acq_rel store i32 %old, i32 addrspace(1)* %out ret void } -; GCN-LABEL: sub_i32_uniform: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN32: s_bcnt1_i32_b32 s[[popcount:[0-9]+]], s[[exec_lo]] -; GCN64: s_bcnt1_i32_b64 s[[popcount:[0-9]+]], s{{\[}}[[exec_lo]]:[[exec_hi]]{{\]}} -; GCN: s_mul_i32 s[[scalar_value:[0-9]+]], s{{[0-9]+}}, s[[popcount]] -; GCN: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] -; GCN: ds_sub_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @sub_i32_uniform(i32 addrspace(1)* %out, i32 %subitive) { +; +; +; GFX7LESS-LABEL: sub_i32_uniform: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x9 +; GFX7LESS-NEXT: s_load_dword s2, s[0:1], 0xb +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s6, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s7, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr1 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[0:1], vcc +; GFX7LESS-NEXT: ; mask branch BB9_2 +; GFX7LESS-NEXT: s_cbranch_execz BB9_2 +; GFX7LESS-NEXT: BB9_1: +; GFX7LESS-NEXT: s_bcnt1_i32_b64 s3, s[6:7] +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: s_mul_i32 s3, s2, s3 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s3 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_sub_rtn_u32 v1, v1, v2 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB9_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[0:1] +; GFX7LESS-NEXT: v_readfirstlane_b32 s0, v1 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX7LESS-NEXT: s_mov_b32 s7, 0xf000 +; GFX7LESS-NEXT: v_sub_i32_e32 v0, vcc, s0, v0 +; GFX7LESS-NEXT: s_mov_b32 s6, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: sub_i32_uniform: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x24 +; GFX8-NEXT: s_load_dword s0, s[0:1], 0x2c +; GFX8-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s7, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr1 +; GFX8-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX8-NEXT: ; mask branch BB9_2 +; GFX8-NEXT: s_cbranch_execz BB9_2 +; GFX8-NEXT: BB9_1: +; GFX8-NEXT: s_bcnt1_i32_b64 s1, s[6:7] +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: s_mul_i32 s1, s0, s1 +; GFX8-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v2, s1 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_sub_rtn_u32 v1, v1, v2 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB9_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: v_mul_lo_u32 v0, s0, v0 +; GFX8-NEXT: v_readfirstlane_b32 s0, v1 +; GFX8-NEXT: s_mov_b32 s7, 0xf000 +; GFX8-NEXT: s_mov_b32 s6, -1 +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, s0, v0 +; GFX8-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: sub_i32_uniform: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x24 +; GFX9-NEXT: s_load_dword s0, s[0:1], 0x2c +; GFX9-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s7, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr1 +; GFX9-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX9-NEXT: ; mask branch BB9_2 +; GFX9-NEXT: s_cbranch_execz BB9_2 +; GFX9-NEXT: BB9_1: +; GFX9-NEXT: s_bcnt1_i32_b64 s1, s[6:7] +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: s_mul_i32 s1, s0, s1 +; GFX9-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v2, s1 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_sub_rtn_u32 v1, v1, v2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB9_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: v_mul_lo_u32 v0, s0, v0 +; GFX9-NEXT: v_readfirstlane_b32 s0, v1 +; GFX9-NEXT: s_mov_b32 s7, 0xf000 +; GFX9-NEXT: s_mov_b32 s6, -1 +; GFX9-NEXT: v_sub_u32_e32 v0, s0, v0 +; GFX9-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: sub_i32_uniform: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x24 +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: s_load_dword s0, s[0:1], 0x2c +; GFX1064-NEXT: ; implicit-def: $vgpr1 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: s_and_saveexec_b64 s[6:7], vcc +; GFX1064-NEXT: ; mask branch BB9_2 +; GFX1064-NEXT: s_cbranch_execz BB9_2 +; GFX1064-NEXT: BB9_1: +; GFX1064-NEXT: s_bcnt1_i32_b64 s1, s[2:3] +; GFX1064-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: s_mul_i32 s1, s0, s1 +; GFX1064-NEXT: v_mov_b32_e32 v2, s1 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_sub_rtn_u32 v1, v1, v2 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB9_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[6:7] +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: v_mul_lo_u32 v0, s0, v0 +; GFX1064-NEXT: v_readfirstlane_b32 s0, v1 +; GFX1064-NEXT: s_mov_b32 s7, 0x31016000 +; GFX1064-NEXT: s_mov_b32 s6, -1 +; GFX1064-NEXT: v_sub_nc_u32_e32 v0, s0, v0 +; GFX1064-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: sub_i32_uniform: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x24 +; GFX1032-NEXT: s_load_dword s0, s[0:1], 0x2c +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: ; implicit-def: $vgpr1 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: s_and_saveexec_b32 s1, vcc_lo +; GFX1032-NEXT: ; mask branch BB9_2 +; GFX1032-NEXT: s_cbranch_execz BB9_2 +; GFX1032-NEXT: BB9_1: +; GFX1032-NEXT: s_bcnt1_i32_b32 s2, s2 +; GFX1032-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: s_mul_i32 s2, s0, s2 +; GFX1032-NEXT: v_mov_b32_e32 v2, s2 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_sub_rtn_u32 v1, v1, v2 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB9_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: v_mul_lo_u32 v0, s0, v0 +; GFX1032-NEXT: v_readfirstlane_b32 s0, v1 +; GFX1032-NEXT: s_mov_b32 s7, 0x31016000 +; GFX1032-NEXT: s_mov_b32 s6, -1 +; GFX1032-NEXT: v_sub_nc_u32_e32 v0, s0, v0 +; GFX1032-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw sub i32 addrspace(3)* @local_var32, i32 %subitive acq_rel store i32 %old, i32 addrspace(1)* %out ret void } -; GCN-LABEL: sub_i32_varying: ; GFX7LESS-NOT: v_mbcnt_lo_u32_b32 ; GFX7LESS-NOT: v_mbcnt_hi_u32_b32 ; GFX7LESS-NOT: s_bcnt1_i32_b64 -; GFX7LESS: ds_sub_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} ; DPPCOMB: v_add_u32_dpp ; DPPCOMB: v_add_u32_dpp ; GFX8MORE32: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 31 -; GFX8MORE64: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 63 ; GFX8MORE: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] ; GFX8MORE: ds_sub_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @sub_i32_varying(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: sub_i32_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_sub_rtn_u32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: sub_i32_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, 0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB10_2 +; GFX8-NEXT: s_cbranch_execz BB10_2 +; GFX8-NEXT: BB10_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_sub_rtn_u32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB10_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: sub_i32_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB10_2 +; GFX9-NEXT: s_cbranch_execz BB10_2 +; GFX9-NEXT: BB10_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_sub_rtn_u32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB10_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_sub_u32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: sub_i32_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, 0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB10_2 +; GFX1064-NEXT: s_cbranch_execz BB10_2 +; GFX1064-NEXT: BB10_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_sub_rtn_u32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB10_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_sub_nc_u32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: sub_i32_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_add_nc_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB10_2 +; GFX1032-NEXT: s_cbranch_execz BB10_2 +; GFX1032-NEXT: BB10_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_sub_rtn_u32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB10_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_sub_nc_u32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw sub i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -310,46 +2194,505 @@ entry: ret void } -; GCN-LABEL: sub_i64_constant: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN32: s_bcnt1_i32_b32 s[[popcount:[0-9]+]], s[[exec_lo]] -; GCN64: s_bcnt1_i32_b64 s[[popcount:[0-9]+]], s{{\[}}[[exec_lo]]:[[exec_hi]]{{\]}} -; GCN: v_mul_hi_u32_u24{{(_e[0-9]+)?}} v[[value_hi:[0-9]+]], s[[popcount]], 5 -; GCN: v_mul_u32_u24{{(_e[0-9]+)?}} v[[value_lo:[0-9]+]], s[[popcount]], 5 -; GCN: ds_sub_rtn_u64 v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}}, v{{[0-9]+}}, v{{\[}}[[value_lo]]:[[value_hi]]{{\]}} define amdgpu_kernel void @sub_i64_constant(i64 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: sub_i64_constant: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s4, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s5, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX7LESS-NEXT: ; mask branch BB11_2 +; GFX7LESS-NEXT: s_cbranch_execz BB11_2 +; GFX7LESS-NEXT: BB11_1: +; GFX7LESS-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX7LESS-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX7LESS-NEXT: v_mul_hi_u32_u24_e64 v2, s4, 5 +; GFX7LESS-NEXT: v_mul_u32_u24_e64 v1, s4, 5 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_sub_rtn_u64 v[1:2], v3, v[1:2] +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB11_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX7LESS-NEXT: v_readfirstlane_b32 s2, v1 +; GFX7LESS-NEXT: v_readfirstlane_b32 s4, v2 +; GFX7LESS-NEXT: v_mul_hi_u32_u24_e32 v1, 5, v0 +; GFX7LESS-NEXT: v_mul_u32_u24_e32 v0, 5, v0 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s4 +; GFX7LESS-NEXT: v_sub_i32_e32 v0, vcc, s2, v0 +; GFX7LESS-NEXT: v_subb_u32_e32 v1, vcc, v2, v1, vcc +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: sub_i64_constant: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s4, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s5, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX8-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX8-NEXT: ; mask branch BB11_2 +; GFX8-NEXT: s_cbranch_execz BB11_2 +; GFX8-NEXT: BB11_1: +; GFX8-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX8-NEXT: v_mul_hi_u32_u24_e64 v2, s4, 5 +; GFX8-NEXT: v_mul_u32_u24_e64 v1, s4, 5 +; GFX8-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_sub_rtn_u64 v[1:2], v3, v[1:2] +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB11_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX8-NEXT: v_readfirstlane_b32 s3, v2 +; GFX8-NEXT: v_readfirstlane_b32 s2, v1 +; GFX8-NEXT: v_mul_hi_u32_u24_e32 v1, 5, v0 +; GFX8-NEXT: v_mul_u32_u24_e32 v0, 5, v0 +; GFX8-NEXT: v_mov_b32_e32 v2, s3 +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, s2, v0 +; GFX8-NEXT: v_subb_u32_e32 v1, vcc, v2, v1, vcc +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: sub_i64_constant: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[4:5], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s4, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s5, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX9-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX9-NEXT: ; mask branch BB11_2 +; GFX9-NEXT: s_cbranch_execz BB11_2 +; GFX9-NEXT: BB11_1: +; GFX9-NEXT: s_bcnt1_i32_b64 s4, s[4:5] +; GFX9-NEXT: v_mul_hi_u32_u24_e64 v2, s4, 5 +; GFX9-NEXT: v_mul_u32_u24_e64 v1, s4, 5 +; GFX9-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_sub_rtn_u64 v[1:2], v3, v[1:2] +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB11_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX9-NEXT: v_readfirstlane_b32 s3, v2 +; GFX9-NEXT: v_readfirstlane_b32 s2, v1 +; GFX9-NEXT: v_mul_hi_u32_u24_e32 v1, 5, v0 +; GFX9-NEXT: v_mul_u32_u24_e32 v0, 5, v0 +; GFX9-NEXT: v_mov_b32_e32 v2, s3 +; GFX9-NEXT: v_sub_co_u32_e32 v0, vcc, s2, v0 +; GFX9-NEXT: v_subb_co_u32_e32 v1, vcc, v2, v1, vcc +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: sub_i64_constant: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB11_2 +; GFX1064-NEXT: s_cbranch_execz BB11_2 +; GFX1064-NEXT: BB11_1: +; GFX1064-NEXT: s_bcnt1_i32_b64 s2, s[2:3] +; GFX1064-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX1064-NEXT: v_mul_hi_u32_u24_e64 v2, s2, 5 +; GFX1064-NEXT: v_mul_u32_u24_e64 v1, s2, 5 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_sub_rtn_u64 v[1:2], v3, v[1:2] +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB11_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s2, v1 +; GFX1064-NEXT: v_mul_u32_u24_e32 v1, 5, v0 +; GFX1064-NEXT: v_readfirstlane_b32 s3, v2 +; GFX1064-NEXT: v_mul_hi_u32_u24_e32 v2, 5, v0 +; GFX1064-NEXT: v_sub_co_u32_e64 v0, vcc, s2, v1 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_sub_co_ci_u32_e32 v1, vcc, s3, v2, vcc +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: sub_i64_constant: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s3, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s3, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: s_and_saveexec_b32 s2, vcc_lo +; GFX1032-NEXT: ; mask branch BB11_2 +; GFX1032-NEXT: s_cbranch_execz BB11_2 +; GFX1032-NEXT: BB11_1: +; GFX1032-NEXT: s_bcnt1_i32_b32 s3, s3 +; GFX1032-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX1032-NEXT: v_mul_hi_u32_u24_e64 v2, s3, 5 +; GFX1032-NEXT: v_mul_u32_u24_e64 v1, s3, 5 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_sub_rtn_u64 v[1:2], v3, v[1:2] +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB11_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s2 +; GFX1032-NEXT: v_readfirstlane_b32 s2, v1 +; GFX1032-NEXT: v_mul_u32_u24_e32 v1, 5, v0 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v2 +; GFX1032-NEXT: v_mul_hi_u32_u24_e32 v2, 5, v0 +; GFX1032-NEXT: v_sub_co_u32_e64 v0, vcc_lo, s2, v1 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, s3, v2, vcc_lo +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw sub i64 addrspace(3)* @local_var64, i64 5 acq_rel store i64 %old, i64 addrspace(1)* %out ret void } -; GCN-LABEL: sub_i64_uniform: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN32: s_bcnt1_i32_b32 s{{[0-9]+}}, s[[exec_lo]] -; GCN64: s_bcnt1_i32_b64 s{{[0-9]+}}, s{{\[}}[[exec_lo]]:[[exec_hi]]{{\]}} -; GCN: ds_sub_rtn_u64 v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}}, v{{[0-9]+}}, v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}} define amdgpu_kernel void @sub_i64_uniform(i64 addrspace(1)* %out, i64 %subitive) { +; +; +; GFX7LESS-LABEL: sub_i64_uniform: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x9 +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s6, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s7, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX7LESS-NEXT: ; mask branch BB12_2 +; GFX7LESS-NEXT: s_cbranch_execz BB12_2 +; GFX7LESS-NEXT: BB12_1: +; GFX7LESS-NEXT: s_bcnt1_i32_b64 s6, s[6:7] +; GFX7LESS-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: s_mul_i32 s7, s3, s6 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, s6 +; GFX7LESS-NEXT: v_mul_hi_u32 v1, s2, v1 +; GFX7LESS-NEXT: s_mul_i32 s6, s2, s6 +; GFX7LESS-NEXT: v_add_i32_e32 v2, vcc, s7, v1 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, s6 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_sub_rtn_u64 v[1:2], v3, v[1:2] +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB12_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX7LESS-NEXT: s_mov_b32 s7, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s6, -1 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: s_mov_b32 s4, s0 +; GFX7LESS-NEXT: s_mov_b32 s5, s1 +; GFX7LESS-NEXT: v_readfirstlane_b32 s0, v1 +; GFX7LESS-NEXT: v_readfirstlane_b32 s1, v2 +; GFX7LESS-NEXT: v_mul_lo_u32 v1, s3, v0 +; GFX7LESS-NEXT: v_mul_hi_u32 v2, s2, v0 +; GFX7LESS-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX7LESS-NEXT: v_add_i32_e32 v1, vcc, v2, v1 +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s1 +; GFX7LESS-NEXT: v_sub_i32_e32 v0, vcc, s0, v0 +; GFX7LESS-NEXT: v_subb_u32_e32 v1, vcc, v2, v1, vcc +; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: sub_i64_uniform: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s7, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB12_2 +; GFX8-NEXT: s_cbranch_execz BB12_2 +; GFX8-NEXT: BB12_1: +; GFX8-NEXT: s_bcnt1_i32_b64 s6, s[6:7] +; GFX8-NEXT: v_mov_b32_e32 v1, s6 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: v_mul_hi_u32 v1, s2, v1 +; GFX8-NEXT: s_mul_i32 s7, s3, s6 +; GFX8-NEXT: s_mul_i32 s6, s2, s6 +; GFX8-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX8-NEXT: v_add_u32_e32 v2, vcc, s7, v1 +; GFX8-NEXT: v_mov_b32_e32 v1, s6 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_sub_rtn_u64 v[1:2], v3, v[1:2] +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB12_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: s_mov_b32 s4, s0 +; GFX8-NEXT: v_readfirstlane_b32 s0, v1 +; GFX8-NEXT: v_mul_lo_u32 v1, s3, v0 +; GFX8-NEXT: v_mul_hi_u32 v3, s2, v0 +; GFX8-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX8-NEXT: s_mov_b32 s5, s1 +; GFX8-NEXT: v_readfirstlane_b32 s1, v2 +; GFX8-NEXT: v_add_u32_e32 v1, vcc, v3, v1 +; GFX8-NEXT: v_mov_b32_e32 v2, s1 +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, s0, v0 +; GFX8-NEXT: s_mov_b32 s7, 0xf000 +; GFX8-NEXT: s_mov_b32 s6, -1 +; GFX8-NEXT: v_subb_u32_e32 v1, vcc, v2, v1, vcc +; GFX8-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: sub_i64_uniform: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s7, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB12_2 +; GFX9-NEXT: s_cbranch_execz BB12_2 +; GFX9-NEXT: BB12_1: +; GFX9-NEXT: s_bcnt1_i32_b64 s6, s[6:7] +; GFX9-NEXT: v_mov_b32_e32 v1, s6 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: v_mul_hi_u32 v2, s2, v1 +; GFX9-NEXT: s_mul_i32 s7, s3, s6 +; GFX9-NEXT: s_mul_i32 s6, s2, s6 +; GFX9-NEXT: v_mov_b32_e32 v1, s6 +; GFX9-NEXT: v_add_u32_e32 v2, s7, v2 +; GFX9-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_sub_rtn_u64 v[1:2], v3, v[1:2] +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB12_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: v_mul_lo_u32 v3, s3, v0 +; GFX9-NEXT: v_mul_hi_u32 v4, s2, v0 +; GFX9-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s4, s0 +; GFX9-NEXT: v_readfirstlane_b32 s0, v1 +; GFX9-NEXT: s_mov_b32 s5, s1 +; GFX9-NEXT: v_readfirstlane_b32 s1, v2 +; GFX9-NEXT: v_add_u32_e32 v1, v4, v3 +; GFX9-NEXT: v_mov_b32_e32 v2, s1 +; GFX9-NEXT: v_sub_co_u32_e32 v0, vcc, s0, v0 +; GFX9-NEXT: s_mov_b32 s7, 0xf000 +; GFX9-NEXT: s_mov_b32 s6, -1 +; GFX9-NEXT: v_subb_co_u32_e32 v1, vcc, v2, v1, vcc +; GFX9-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: sub_i64_uniform: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[6:7], 1, 0 +; GFX1064-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24 +; GFX1064-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s6, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s7, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB12_2 +; GFX1064-NEXT: s_cbranch_execz BB12_2 +; GFX1064-NEXT: BB12_1: +; GFX1064-NEXT: s_bcnt1_i32_b64 s6, s[6:7] +; GFX1064-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: v_mul_hi_u32 v2, s2, s6 +; GFX1064-NEXT: s_mul_i32 s7, s2, s6 +; GFX1064-NEXT: s_mul_i32 s6, s3, s6 +; GFX1064-NEXT: v_mov_b32_e32 v1, s7 +; GFX1064-NEXT: v_add_nc_u32_e32 v2, s6, v2 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_sub_rtn_u64 v[1:2], v3, v[1:2] +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB12_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: v_mul_lo_u32 v3, s3, v0 +; GFX1064-NEXT: v_mul_hi_u32 v4, s2, v0 +; GFX1064-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX1064-NEXT: v_readfirstlane_b32 s4, v1 +; GFX1064-NEXT: v_readfirstlane_b32 s5, v2 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_add_nc_u32_e32 v1, v4, v3 +; GFX1064-NEXT: v_sub_co_u32_e64 v0, vcc, s4, v0 +; GFX1064-NEXT: v_sub_co_ci_u32_e32 v1, vcc, s5, v1, vcc +; GFX1064-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: sub_i64_uniform: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s5, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: ; implicit-def: $vgpr1_vgpr2 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s5, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB12_2 +; GFX1032-NEXT: s_cbranch_execz BB12_2 +; GFX1032-NEXT: BB12_1: +; GFX1032-NEXT: s_bcnt1_i32_b32 s5, s5 +; GFX1032-NEXT: v_mov_b32_e32 v3, local_var64@abs32@lo +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: v_mul_hi_u32 v2, s2, s5 +; GFX1032-NEXT: s_mul_i32 s6, s2, s5 +; GFX1032-NEXT: s_mul_i32 s5, s3, s5 +; GFX1032-NEXT: v_mov_b32_e32 v1, s6 +; GFX1032-NEXT: v_add_nc_u32_e32 v2, s5, v2 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_sub_rtn_u64 v[1:2], v3, v[1:2] +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB12_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: v_mul_lo_u32 v3, s3, v0 +; GFX1032-NEXT: v_mul_hi_u32 v4, s2, v0 +; GFX1032-NEXT: v_mul_lo_u32 v0, s2, v0 +; GFX1032-NEXT: v_readfirstlane_b32 s4, v1 +; GFX1032-NEXT: v_readfirstlane_b32 s5, v2 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_add_nc_u32_e32 v1, v4, v3 +; GFX1032-NEXT: v_sub_co_u32_e64 v0, vcc_lo, s4, v0 +; GFX1032-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, s5, v1, vcc_lo +; GFX1032-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw sub i64 addrspace(3)* @local_var64, i64 %subitive acq_rel store i64 %old, i64 addrspace(1)* %out ret void } -; GCN-LABEL: sub_i64_varying: ; GCN-NOT: v_mbcnt_lo_u32_b32 ; GCN-NOT: v_mbcnt_hi_u32_b32 ; GCN-NOT: s_bcnt1_i32_b64 -; GCN: ds_sub_rtn_u64 v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}}, v{{[0-9]+}}, v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}} define amdgpu_kernel void @sub_i64_varying(i64 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: sub_i64_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0 +; GFX7LESS-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_sub_rtn_u64 v[0:1], v2, v[0:1] +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: sub_i64_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_sub_rtn_u64 v[0:1], v2, v[0:1] +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: sub_i64_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_sub_rtn_u64 v[0:1], v2, v[0:1] +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: sub_i64_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_sub_rtn_u64 v[0:1], v2, v[0:1] +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: sub_i64_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_sub_rtn_u64 v[0:1], v2, v[0:1] +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %zext = zext i32 %lane to i64 @@ -358,12 +2701,245 @@ entry: ret void } -; GCN-LABEL: and_i32_varying: ; GFX8MORE32: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 31 -; GFX8MORE64: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 63 ; GFX8MORE: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] ; GFX8MORE: ds_and_rtn_b32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @and_i32_varying(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: and_i32_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_and_rtn_b32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: and_i32_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v3, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v3, s3, v3 +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: v_mov_b32_e32 v1, -1 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, -1 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_and_b32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_and_b32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB14_2 +; GFX8-NEXT: s_cbranch_execz BB14_2 +; GFX8-NEXT: BB14_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_and_rtn_b32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB14_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_and_b32_e32 v0, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: and_i32_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v3, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v3, s3, v3 +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: v_mov_b32_e32 v1, -1 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, -1 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_and_b32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_and_b32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB14_2 +; GFX9-NEXT: s_cbranch_execz BB14_2 +; GFX9-NEXT: BB14_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_and_rtn_b32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB14_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_and_b32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: and_i32_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v4, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v4, s3, v4 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: v_mov_b32_e32 v1, -1 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, -1 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_and_b32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_and_b32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB14_2 +; GFX1064-NEXT: s_cbranch_execz BB14_2 +; GFX1064-NEXT: BB14_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_and_rtn_b32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB14_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_and_b32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: and_i32_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v4, s2, 0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: v_mov_b32_e32 v1, -1 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, -1 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_and_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_and_b32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v4 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB14_2 +; GFX1032-NEXT: s_cbranch_execz BB14_2 +; GFX1032-NEXT: BB14_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_and_rtn_b32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB14_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_and_b32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw and i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -371,12 +2947,245 @@ entry: ret void } -; GCN-LABEL: or_i32_varying: ; GFX8MORE32: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 31 -; GFX8MORE64: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 63 ; GFX8MORE: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] ; GFX8MORE: ds_or_rtn_b32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @or_i32_varying(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: or_i32_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_or_rtn_b32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: or_i32_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, 0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_or_b32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_or_b32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB15_2 +; GFX8-NEXT: s_cbranch_execz BB15_2 +; GFX8-NEXT: BB15_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_or_rtn_b32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB15_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_or_b32_e32 v0, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: or_i32_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_or_b32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_or_b32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB15_2 +; GFX9-NEXT: s_cbranch_execz BB15_2 +; GFX9-NEXT: BB15_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_or_rtn_b32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB15_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_or_b32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: or_i32_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, 0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_or_b32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_or_b32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB15_2 +; GFX1064-NEXT: s_cbranch_execz BB15_2 +; GFX1064-NEXT: BB15_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_or_rtn_b32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB15_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_or_b32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: or_i32_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_or_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_or_b32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB15_2 +; GFX1032-NEXT: s_cbranch_execz BB15_2 +; GFX1032-NEXT: BB15_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_or_rtn_b32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB15_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_or_b32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw or i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -384,12 +3193,245 @@ entry: ret void } -; GCN-LABEL: xor_i32_varying: ; GFX8MORE32: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 31 -; GFX8MORE64: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 63 ; GFX8MORE: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] ; GFX8MORE: ds_xor_rtn_b32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @xor_i32_varying(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: xor_i32_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_xor_rtn_b32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: xor_i32_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, 0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_xor_b32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_xor_b32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB16_2 +; GFX8-NEXT: s_cbranch_execz BB16_2 +; GFX8-NEXT: BB16_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_xor_rtn_b32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB16_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_xor_b32_e32 v0, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: xor_i32_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_xor_b32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_xor_b32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB16_2 +; GFX9-NEXT: s_cbranch_execz BB16_2 +; GFX9-NEXT: BB16_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_xor_rtn_b32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB16_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_xor_b32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: xor_i32_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, 0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_xor_b32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_xor_b32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB16_2 +; GFX1064-NEXT: s_cbranch_execz BB16_2 +; GFX1064-NEXT: BB16_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_xor_rtn_b32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB16_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_xor_b32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: xor_i32_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_xor_b32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_xor_b32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB16_2 +; GFX1032-NEXT: s_cbranch_execz BB16_2 +; GFX1032-NEXT: BB16_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_xor_rtn_b32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB16_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_xor_b32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw xor i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -397,12 +3439,245 @@ entry: ret void } -; GCN-LABEL: max_i32_varying: ; GFX8MORE32: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 31 -; GFX8MORE64: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 63 ; GFX8MORE: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] ; GFX8MORE: ds_max_rtn_i32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @max_i32_varying(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: max_i32_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_max_rtn_i32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: max_i32_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v3, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v3, s3, v3 +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: v_bfrev_b32_e32 v1, 1 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, v1 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_max_i32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_max_i32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB17_2 +; GFX8-NEXT: s_cbranch_execz BB17_2 +; GFX8-NEXT: BB17_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_max_rtn_i32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB17_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_max_i32_e32 v0, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: max_i32_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v3, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v3, s3, v3 +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: v_bfrev_b32_e32 v1, 1 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, v1 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_max_i32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_max_i32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB17_2 +; GFX9-NEXT: s_cbranch_execz BB17_2 +; GFX9-NEXT: BB17_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_max_rtn_i32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB17_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_max_i32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: max_i32_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v4, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v4, s3, v4 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: v_bfrev_b32_e32 v1, 1 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, v1 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_max_i32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_max_i32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB17_2 +; GFX1064-NEXT: s_cbranch_execz BB17_2 +; GFX1064-NEXT: BB17_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_max_rtn_i32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB17_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_max_i32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: max_i32_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v4, s2, 0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: v_bfrev_b32_e32 v1, 1 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, v1 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_max_i32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_max_i32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v4 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB17_2 +; GFX1032-NEXT: s_cbranch_execz BB17_2 +; GFX1032-NEXT: BB17_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_max_rtn_i32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB17_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_max_i32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw max i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -410,28 +3685,440 @@ entry: ret void } -; GCN-LABEL: max_i64_constant: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN: v_mov_b32{{(_e[0-9]+)?}} v[[value_lo:[0-9]+]], 5 -; GCN: v_mov_b32{{(_e[0-9]+)?}} v[[value_hi:[0-9]+]], 0 -; GCN: ds_max_rtn_i64 v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}}, v{{[0-9]+}}, v{{\[}}[[value_lo]]:[[value_hi]]{{\]}} define amdgpu_kernel void @max_i64_constant(i64 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: max_i64_constant: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s3, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX7LESS-NEXT: ; mask branch BB18_2 +; GFX7LESS-NEXT: s_cbranch_execz BB18_2 +; GFX7LESS-NEXT: BB18_1: +; GFX7LESS-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX7LESS-NEXT: v_mov_b32_e32 v0, 5 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_max_rtn_i64 v[0:1], v2, v[0:1] +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB18_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX7LESS-NEXT: v_readfirstlane_b32 s4, v0 +; GFX7LESS-NEXT: v_readfirstlane_b32 s5, v1 +; GFX7LESS-NEXT: v_bfrev_b32_e32 v1, 1 +; GFX7LESS-NEXT: v_cndmask_b32_e64 v0, 5, 0, vcc +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: v_cndmask_b32_e32 v1, 0, v1, vcc +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s5 +; GFX7LESS-NEXT: v_mov_b32_e32 v3, s4 +; GFX7LESS-NEXT: v_cmp_gt_i64_e32 vcc, s[4:5], v[0:1] +; GFX7LESS-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc +; GFX7LESS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: max_i64_constant: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX8-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX8-NEXT: ; mask branch BB18_2 +; GFX8-NEXT: s_cbranch_execz BB18_2 +; GFX8-NEXT: BB18_1: +; GFX8-NEXT: v_mov_b32_e32 v0, 5 +; GFX8-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_max_rtn_i64 v[0:1], v2, v[0:1] +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB18_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_bfrev_b32_e32 v0, 1 +; GFX8-NEXT: v_readfirstlane_b32 s3, v1 +; GFX8-NEXT: v_cndmask_b32_e32 v1, 0, v0, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v0, 5, 0, vcc +; GFX8-NEXT: v_cmp_gt_i64_e32 vcc, s[2:3], v[0:1] +; GFX8-NEXT: v_mov_b32_e32 v2, s3 +; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc +; GFX8-NEXT: v_mov_b32_e32 v2, s2 +; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: max_i64_constant: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX9-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX9-NEXT: ; mask branch BB18_2 +; GFX9-NEXT: s_cbranch_execz BB18_2 +; GFX9-NEXT: BB18_1: +; GFX9-NEXT: v_mov_b32_e32 v0, 5 +; GFX9-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_max_rtn_i64 v[0:1], v2, v[0:1] +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB18_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_bfrev_b32_e32 v0, 1 +; GFX9-NEXT: v_readfirstlane_b32 s3, v1 +; GFX9-NEXT: v_cndmask_b32_e32 v1, 0, v0, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v0, 5, 0, vcc +; GFX9-NEXT: v_cmp_gt_i64_e32 vcc, s[2:3], v[0:1] +; GFX9-NEXT: v_mov_b32_e32 v2, s3 +; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc +; GFX9-NEXT: v_mov_b32_e32 v2, s2 +; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: max_i64_constant: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX1064-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX1064-NEXT: ; mask branch BB18_2 +; GFX1064-NEXT: s_cbranch_execz BB18_2 +; GFX1064-NEXT: BB18_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, 5 +; GFX1064-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_max_rtn_i64 v[0:1], v2, v[0:1] +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB18_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX1064-NEXT: v_readfirstlane_b32 s4, v0 +; GFX1064-NEXT: v_readfirstlane_b32 s5, v1 +; GFX1064-NEXT: v_cndmask_b32_e64 v1, 0, 0x80000000, vcc +; GFX1064-NEXT: v_cndmask_b32_e64 v0, 5, 0, vcc +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_cmp_gt_i64_e32 vcc, s[4:5], v[0:1] +; GFX1064-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc +; GFX1064-NEXT: v_cndmask_b32_e64 v0, v0, s4, vcc +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: max_i64_constant: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX1032-NEXT: s_and_saveexec_b32 s2, vcc_lo +; GFX1032-NEXT: ; mask branch BB18_2 +; GFX1032-NEXT: s_cbranch_execz BB18_2 +; GFX1032-NEXT: BB18_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, 5 +; GFX1032-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_max_rtn_i64 v[0:1], v2, v[0:1] +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB18_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s2 +; GFX1032-NEXT: v_readfirstlane_b32 s4, v0 +; GFX1032-NEXT: v_readfirstlane_b32 s5, v1 +; GFX1032-NEXT: v_cndmask_b32_e64 v1, 0, 0x80000000, vcc_lo +; GFX1032-NEXT: v_cndmask_b32_e64 v0, 5, 0, vcc_lo +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_cmp_gt_i64_e32 vcc_lo, s[4:5], v[0:1] +; GFX1032-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc_lo +; GFX1032-NEXT: v_cndmask_b32_e64 v0, v0, s4, vcc_lo +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw max i64 addrspace(3)* @local_var64, i64 5 acq_rel store i64 %old, i64 addrspace(1)* %out ret void } -; GCN-LABEL: min_i32_varying: ; GFX8MORE32: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 31 -; GFX8MORE64: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 63 ; GFX8MORE: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] ; GFX8MORE: ds_min_rtn_i32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @min_i32_varying(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: min_i32_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_min_rtn_i32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: min_i32_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v3, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v3, s3, v3 +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: v_bfrev_b32_e32 v1, -2 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, v1 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_min_i32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_min_i32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB19_2 +; GFX8-NEXT: s_cbranch_execz BB19_2 +; GFX8-NEXT: BB19_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_min_rtn_i32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB19_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_min_i32_e32 v0, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: min_i32_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v3, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v3, s3, v3 +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: v_bfrev_b32_e32 v1, -2 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, v1 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_min_i32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_min_i32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB19_2 +; GFX9-NEXT: s_cbranch_execz BB19_2 +; GFX9-NEXT: BB19_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_min_rtn_i32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB19_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_min_i32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: min_i32_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v4, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v4, s3, v4 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: v_bfrev_b32_e32 v1, -2 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, v1 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_min_i32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_min_i32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB19_2 +; GFX1064-NEXT: s_cbranch_execz BB19_2 +; GFX1064-NEXT: BB19_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_min_rtn_i32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB19_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_min_i32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: min_i32_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v4, s2, 0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: v_bfrev_b32_e32 v1, -2 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, v1 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_min_i32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_min_i32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v4 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB19_2 +; GFX1032-NEXT: s_cbranch_execz BB19_2 +; GFX1032-NEXT: BB19_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_min_rtn_i32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB19_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_min_i32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw min i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -439,28 +4126,440 @@ entry: ret void } -; GCN-LABEL: min_i64_constant: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN: v_mov_b32{{(_e[0-9]+)?}} v[[value_lo:[0-9]+]], 5 -; GCN: v_mov_b32{{(_e[0-9]+)?}} v[[value_hi:[0-9]+]], 0 -; GCN: ds_min_rtn_i64 v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}}, v{{[0-9]+}}, v{{\[}}[[value_lo]]:[[value_hi]]{{\]}} define amdgpu_kernel void @min_i64_constant(i64 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: min_i64_constant: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s3, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX7LESS-NEXT: ; mask branch BB20_2 +; GFX7LESS-NEXT: s_cbranch_execz BB20_2 +; GFX7LESS-NEXT: BB20_1: +; GFX7LESS-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX7LESS-NEXT: v_mov_b32_e32 v0, 5 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_min_rtn_i64 v[0:1], v2, v[0:1] +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB20_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX7LESS-NEXT: v_readfirstlane_b32 s4, v0 +; GFX7LESS-NEXT: v_readfirstlane_b32 s5, v1 +; GFX7LESS-NEXT: v_bfrev_b32_e32 v1, -2 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: v_cndmask_b32_e64 v0, 5, -1, vcc +; GFX7LESS-NEXT: v_cndmask_b32_e32 v1, 0, v1, vcc +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s5 +; GFX7LESS-NEXT: v_mov_b32_e32 v3, s4 +; GFX7LESS-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[0:1] +; GFX7LESS-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc +; GFX7LESS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: min_i64_constant: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX8-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX8-NEXT: ; mask branch BB20_2 +; GFX8-NEXT: s_cbranch_execz BB20_2 +; GFX8-NEXT: BB20_1: +; GFX8-NEXT: v_mov_b32_e32 v0, 5 +; GFX8-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_min_rtn_i64 v[0:1], v2, v[0:1] +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB20_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX8-NEXT: v_readfirstlane_b32 s4, v0 +; GFX8-NEXT: v_bfrev_b32_e32 v0, -2 +; GFX8-NEXT: v_readfirstlane_b32 s5, v1 +; GFX8-NEXT: v_cndmask_b32_e32 v1, 0, v0, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v0, 5, -1, vcc +; GFX8-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[0:1] +; GFX8-NEXT: v_mov_b32_e32 v2, s5 +; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc +; GFX8-NEXT: v_mov_b32_e32 v2, s4 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: min_i64_constant: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX9-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX9-NEXT: ; mask branch BB20_2 +; GFX9-NEXT: s_cbranch_execz BB20_2 +; GFX9-NEXT: BB20_1: +; GFX9-NEXT: v_mov_b32_e32 v0, 5 +; GFX9-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_min_rtn_i64 v[0:1], v2, v[0:1] +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB20_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX9-NEXT: v_readfirstlane_b32 s4, v0 +; GFX9-NEXT: v_bfrev_b32_e32 v0, -2 +; GFX9-NEXT: v_readfirstlane_b32 s5, v1 +; GFX9-NEXT: v_cndmask_b32_e32 v1, 0, v0, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v0, 5, -1, vcc +; GFX9-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[0:1] +; GFX9-NEXT: v_mov_b32_e32 v2, s5 +; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc +; GFX9-NEXT: v_mov_b32_e32 v2, s4 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: min_i64_constant: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX1064-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX1064-NEXT: ; mask branch BB20_2 +; GFX1064-NEXT: s_cbranch_execz BB20_2 +; GFX1064-NEXT: BB20_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, 5 +; GFX1064-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_min_rtn_i64 v[0:1], v2, v[0:1] +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB20_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX1064-NEXT: v_readfirstlane_b32 s4, v0 +; GFX1064-NEXT: v_readfirstlane_b32 s5, v1 +; GFX1064-NEXT: v_cndmask_b32_e64 v1, 0, 0x7fffffff, vcc +; GFX1064-NEXT: v_cndmask_b32_e64 v0, 5, -1, vcc +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[0:1] +; GFX1064-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc +; GFX1064-NEXT: v_cndmask_b32_e64 v0, v0, s4, vcc +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: min_i64_constant: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX1032-NEXT: s_and_saveexec_b32 s2, vcc_lo +; GFX1032-NEXT: ; mask branch BB20_2 +; GFX1032-NEXT: s_cbranch_execz BB20_2 +; GFX1032-NEXT: BB20_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, 5 +; GFX1032-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_min_rtn_i64 v[0:1], v2, v[0:1] +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB20_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s2 +; GFX1032-NEXT: v_readfirstlane_b32 s4, v0 +; GFX1032-NEXT: v_readfirstlane_b32 s5, v1 +; GFX1032-NEXT: v_cndmask_b32_e64 v1, 0, 0x7fffffff, vcc_lo +; GFX1032-NEXT: v_cndmask_b32_e64 v0, 5, -1, vcc_lo +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: v_cmp_lt_i64_e32 vcc_lo, s[4:5], v[0:1] +; GFX1032-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc_lo +; GFX1032-NEXT: v_cndmask_b32_e64 v0, v0, s4, vcc_lo +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw min i64 addrspace(3)* @local_var64, i64 5 acq_rel store i64 %old, i64 addrspace(1)* %out ret void } -; GCN-LABEL: umax_i32_varying: ; GFX8MORE32: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 31 -; GFX8MORE64: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 63 ; GFX8MORE: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] ; GFX8MORE: ds_max_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @umax_i32_varying(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: umax_i32_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_max_rtn_u32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: umax_i32_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, 0 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_max_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_max_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB21_2 +; GFX8-NEXT: s_cbranch_execz BB21_2 +; GFX8-NEXT: BB21_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_max_rtn_u32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB21_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_max_u32_e32 v0, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: umax_i32_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_max_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_max_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB21_2 +; GFX9-NEXT: s_cbranch_execz BB21_2 +; GFX9-NEXT: BB21_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_max_rtn_u32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB21_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_max_u32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: umax_i32_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, 0 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_max_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_max_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB21_2 +; GFX1064-NEXT: s_cbranch_execz BB21_2 +; GFX1064-NEXT: BB21_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_max_rtn_u32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB21_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_max_u32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: umax_i32_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, 0 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_max_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:0 +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_max_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB21_2 +; GFX1032-NEXT: s_cbranch_execz BB21_2 +; GFX1032-NEXT: BB21_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_max_rtn_u32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB21_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_max_u32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw umax i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -468,28 +4567,437 @@ entry: ret void } -; GCN-LABEL: umax_i64_constant: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN: v_mov_b32{{(_e[0-9]+)?}} v[[value_lo:[0-9]+]], 5 -; GCN: v_mov_b32{{(_e[0-9]+)?}} v[[value_hi:[0-9]+]], 0 -; GCN: ds_max_rtn_u64 v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}}, v{{[0-9]+}}, v{{\[}}[[value_lo]]:[[value_hi]]{{\]}} define amdgpu_kernel void @umax_i64_constant(i64 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: umax_i64_constant: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s3, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX7LESS-NEXT: ; mask branch BB22_2 +; GFX7LESS-NEXT: s_cbranch_execz BB22_2 +; GFX7LESS-NEXT: BB22_1: +; GFX7LESS-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX7LESS-NEXT: v_mov_b32_e32 v0, 5 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_max_rtn_u64 v[0:1], v2, v[0:1] +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB22_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX7LESS-NEXT: v_readfirstlane_b32 s4, v0 +; GFX7LESS-NEXT: v_readfirstlane_b32 s5, v1 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0 +; GFX7LESS-NEXT: v_cndmask_b32_e64 v0, 5, 0, vcc +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s4 +; GFX7LESS-NEXT: v_cmp_gt_u64_e32 vcc, s[4:5], v[0:1] +; GFX7LESS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX7LESS-NEXT: v_mov_b32_e32 v1, s5 +; GFX7LESS-NEXT: v_cndmask_b32_e32 v1, 0, v1, vcc +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: umax_i64_constant: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX8-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX8-NEXT: ; mask branch BB22_2 +; GFX8-NEXT: s_cbranch_execz BB22_2 +; GFX8-NEXT: BB22_1: +; GFX8-NEXT: v_mov_b32_e32 v0, 5 +; GFX8-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_max_rtn_u64 v[0:1], v2, v[0:1] +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB22_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_readfirstlane_b32 s3, v1 +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: v_cndmask_b32_e64 v0, 5, 0, vcc +; GFX8-NEXT: v_cmp_gt_u64_e32 vcc, s[2:3], v[0:1] +; GFX8-NEXT: v_mov_b32_e32 v1, s3 +; GFX8-NEXT: v_mov_b32_e32 v2, s2 +; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX8-NEXT: v_cndmask_b32_e32 v1, 0, v1, vcc +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: umax_i64_constant: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX9-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX9-NEXT: ; mask branch BB22_2 +; GFX9-NEXT: s_cbranch_execz BB22_2 +; GFX9-NEXT: BB22_1: +; GFX9-NEXT: v_mov_b32_e32 v0, 5 +; GFX9-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_max_rtn_u64 v[0:1], v2, v[0:1] +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB22_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_readfirstlane_b32 s3, v1 +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: v_cndmask_b32_e64 v0, 5, 0, vcc +; GFX9-NEXT: v_cmp_gt_u64_e32 vcc, s[2:3], v[0:1] +; GFX9-NEXT: v_mov_b32_e32 v1, s3 +; GFX9-NEXT: v_mov_b32_e32 v2, s2 +; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX9-NEXT: v_cndmask_b32_e32 v1, 0, v1, vcc +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: umax_i64_constant: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX1064-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX1064-NEXT: ; mask branch BB22_2 +; GFX1064-NEXT: s_cbranch_execz BB22_2 +; GFX1064-NEXT: BB22_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, 5 +; GFX1064-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_max_rtn_u64 v[0:1], v2, v[0:1] +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB22_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX1064-NEXT: v_readfirstlane_b32 s4, v0 +; GFX1064-NEXT: v_readfirstlane_b32 s5, v1 +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: v_cndmask_b32_e64 v0, 5, 0, vcc +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_cmp_gt_u64_e32 vcc, s[4:5], v[0:1] +; GFX1064-NEXT: v_cndmask_b32_e64 v0, v0, s4, vcc +; GFX1064-NEXT: v_cndmask_b32_e64 v1, 0, s5, vcc +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: umax_i64_constant: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX1032-NEXT: s_and_saveexec_b32 s2, vcc_lo +; GFX1032-NEXT: ; mask branch BB22_2 +; GFX1032-NEXT: s_cbranch_execz BB22_2 +; GFX1032-NEXT: BB22_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, 5 +; GFX1032-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_max_rtn_u64 v[0:1], v2, v[0:1] +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB22_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s2 +; GFX1032-NEXT: v_readfirstlane_b32 s4, v0 +; GFX1032-NEXT: v_readfirstlane_b32 s5, v1 +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: v_cndmask_b32_e64 v0, 5, 0, vcc_lo +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[4:5], v[0:1] +; GFX1032-NEXT: v_cndmask_b32_e64 v0, v0, s4, vcc_lo +; GFX1032-NEXT: v_cndmask_b32_e64 v1, 0, s5, vcc_lo +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw umax i64 addrspace(3)* @local_var64, i64 5 acq_rel store i64 %old, i64 addrspace(1)* %out ret void } -; GCN-LABEL: umin_i32_varying: ; GFX8MORE32: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 31 -; GFX8MORE64: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 63 ; GFX8MORE: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]] ; GFX8MORE: ds_min_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v[[value]] define amdgpu_kernel void @umin_i32_varying(i32 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: umin_i32_varying: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, local_var32@abs32@lo +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_min_rtn_u32 v0, v1, v0 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: umin_i32_varying: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v3, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v3, s3, v3 +; GFX8-NEXT: v_mov_b32_e32 v2, v0 +; GFX8-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX8-NEXT: v_mov_b32_e32 v1, -1 +; GFX8-NEXT: s_mov_b64 exec, s[2:3] +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: v_mov_b32_e32 v2, -1 +; GFX8-NEXT: s_not_b64 exec, exec +; GFX8-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX8-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_min_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX8-NEXT: s_nop 1 +; GFX8-NEXT: v_min_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX8-NEXT: v_readlane_b32 s2, v2, 63 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX8-NEXT: s_mov_b64 exec, s[4:5] +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3 +; GFX8-NEXT: ; implicit-def: $vgpr0 +; GFX8-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX8-NEXT: ; mask branch BB23_2 +; GFX8-NEXT: s_cbranch_execz BB23_2 +; GFX8-NEXT: BB23_1: +; GFX8-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v3, s2 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_min_rtn_u32 v0, v0, v3 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB23_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX8-NEXT: v_readfirstlane_b32 s2, v0 +; GFX8-NEXT: v_mov_b32_e32 v0, v1 +; GFX8-NEXT: v_min_u32_e32 v0, s2, v0 +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: s_nop 0 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: umin_i32_varying: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v3, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v3, s3, v3 +; GFX9-NEXT: v_mov_b32_e32 v2, v0 +; GFX9-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX9-NEXT: v_mov_b32_e32 v1, -1 +; GFX9-NEXT: s_mov_b64 exec, s[2:3] +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: v_mov_b32_e32 v2, -1 +; GFX9-NEXT: s_not_b64 exec, exec +; GFX9-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX9-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_min_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf +; GFX9-NEXT: s_nop 1 +; GFX9-NEXT: v_min_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf +; GFX9-NEXT: v_readlane_b32 s2, v2, 63 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf +; GFX9-NEXT: s_mov_b64 exec, s[4:5] +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3 +; GFX9-NEXT: ; implicit-def: $vgpr0 +; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX9-NEXT: ; mask branch BB23_2 +; GFX9-NEXT: s_cbranch_execz BB23_2 +; GFX9-NEXT: BB23_1: +; GFX9-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v3, s2 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_min_rtn_u32 v0, v0, v3 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB23_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX9-NEXT: v_readfirstlane_b32 s2, v0 +; GFX9-NEXT: v_mov_b32_e32 v0, v1 +; GFX9-NEXT: v_min_u32_e32 v0, s2, v0 +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: s_nop 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: umin_i32_varying: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: v_mov_b32_e32 v2, v0 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v4, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v4, s3, v4 +; GFX1064-NEXT: s_or_saveexec_b64 s[2:3], -1 +; GFX1064-NEXT: v_mov_b32_e32 v1, -1 +; GFX1064-NEXT: s_mov_b64 exec, s[2:3] +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: v_mov_b32_e32 v2, -1 +; GFX1064-NEXT: s_not_b64 exec, exec +; GFX1064-NEXT: s_or_saveexec_b64 s[4:5], -1 +; GFX1064-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_mov_b32_e32 v3, v2 +; GFX1064-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1064-NEXT: v_min_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 31 +; GFX1064-NEXT: v_mov_b32_e32 v3, s2 +; GFX1064-NEXT: v_min_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s2, v2, 15 +; GFX1064-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1064-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1064-NEXT: v_readlane_b32 s6, v2, 47 +; GFX1064-NEXT: v_writelane_b32 v1, s2, 16 +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: v_writelane_b32 v1, s3, 32 +; GFX1064-NEXT: v_readlane_b32 s3, v2, 63 +; GFX1064-NEXT: v_writelane_b32 v1, s6, 48 +; GFX1064-NEXT: s_mov_b64 exec, s[4:5] +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4 +; GFX1064-NEXT: ; implicit-def: $vgpr0 +; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GFX1064-NEXT: ; mask branch BB23_2 +; GFX1064-NEXT: s_cbranch_execz BB23_2 +; GFX1064-NEXT: BB23_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v7, s3 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_min_rtn_u32 v0, v0, v7 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB23_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5] +; GFX1064-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1064-NEXT: v_mov_b32_e32 v0, v1 +; GFX1064-NEXT: v_min_u32_e32 v0, s3, v0 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: s_nop 1 +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: umin_i32_varying: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mov_b32_e32 v2, v0 +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v4, s2, 0 +; GFX1032-NEXT: s_or_saveexec_b32 s2, -1 +; GFX1032-NEXT: v_mov_b32_e32 v1, -1 +; GFX1032-NEXT: s_mov_b32 exec_lo, s2 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: v_mov_b32_e32 v2, -1 +; GFX1032-NEXT: s_not_b32 exec_lo, exec_lo +; GFX1032-NEXT: s_or_saveexec_b32 s4, -1 +; GFX1032-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_min_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_mov_b32_e32 v3, v2 +; GFX1032-NEXT: v_permlanex16_b32 v3, v3, -1, -1 +; GFX1032-NEXT: v_min_u32_dpp v2, v3, v2 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s3, v2, 31 +; GFX1032-NEXT: v_mov_b32_dpp v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf +; GFX1032-NEXT: v_readlane_b32 s5, v2, 15 +; GFX1032-NEXT: v_writelane_b32 v1, s5, 16 +; GFX1032-NEXT: s_mov_b32 exec_lo, s4 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v4 +; GFX1032-NEXT: ; implicit-def: $vgpr0 +; GFX1032-NEXT: s_and_saveexec_b32 s4, vcc_lo +; GFX1032-NEXT: ; mask branch BB23_2 +; GFX1032-NEXT: s_cbranch_execz BB23_2 +; GFX1032-NEXT: BB23_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, local_var32@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v7, s3 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_min_rtn_u32 v0, v0, v7 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB23_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s4 +; GFX1032-NEXT: v_readfirstlane_b32 s3, v0 +; GFX1032-NEXT: v_mov_b32_e32 v0, v1 +; GFX1032-NEXT: v_min_u32_e32 v0, s3, v0 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: s_nop 1 +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %lane = call i32 @llvm.amdgcn.workitem.id.x() %old = atomicrmw umin i32 addrspace(3)* @local_var32, i32 %lane acq_rel @@ -497,16 +5005,192 @@ entry: ret void } -; GCN-LABEL: umin_i64_constant: -; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0 -; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0 -; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0 -; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]] -; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]] -; GCN: v_mov_b32{{(_e[0-9]+)?}} v[[value_lo:[0-9]+]], 5 -; GCN: v_mov_b32{{(_e[0-9]+)?}} v[[value_hi:[0-9]+]], 0 -; GCN: ds_min_rtn_u64 v{{\[}}{{[0-9]+}}:{{[0-9]+}}{{\]}}, v{{[0-9]+}}, v{{\[}}[[value_lo]]:[[value_hi]]{{\]}} define amdgpu_kernel void @umin_i64_constant(i64 addrspace(1)* %out) { +; +; +; GFX7LESS-LABEL: umin_i64_constant: +; GFX7LESS: ; %bb.0: ; %entry +; GFX7LESS-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9 +; GFX7LESS-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, s3, v0 +; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX7LESS-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX7LESS-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX7LESS-NEXT: ; mask branch BB24_2 +; GFX7LESS-NEXT: s_cbranch_execz BB24_2 +; GFX7LESS-NEXT: BB24_1: +; GFX7LESS-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX7LESS-NEXT: v_mov_b32_e32 v0, 5 +; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0 +; GFX7LESS-NEXT: s_mov_b32 m0, -1 +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: ds_min_rtn_u64 v[0:1], v2, v[0:1] +; GFX7LESS-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX7LESS-NEXT: buffer_wbinvl1 +; GFX7LESS-NEXT: BB24_2: +; GFX7LESS-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX7LESS-NEXT: v_readfirstlane_b32 s4, v0 +; GFX7LESS-NEXT: v_readfirstlane_b32 s5, v1 +; GFX7LESS-NEXT: s_mov_b32 s2, -1 +; GFX7LESS-NEXT: v_cndmask_b32_e64 v1, 0, -1, vcc +; GFX7LESS-NEXT: v_cndmask_b32_e64 v0, 5, -1, vcc +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s5 +; GFX7LESS-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[0:1] +; GFX7LESS-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc +; GFX7LESS-NEXT: v_mov_b32_e32 v2, s4 +; GFX7LESS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000 +; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0) +; GFX7LESS-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX7LESS-NEXT: s_endpgm +; +; GFX8-LABEL: umin_i64_constant: +; GFX8: ; %bb.0: ; %entry +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX8-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX8-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX8-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX8-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX8-NEXT: ; mask branch BB24_2 +; GFX8-NEXT: s_cbranch_execz BB24_2 +; GFX8-NEXT: BB24_1: +; GFX8-NEXT: v_mov_b32_e32 v0, 5 +; GFX8-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX8-NEXT: v_mov_b32_e32 v1, 0 +; GFX8-NEXT: s_mov_b32 m0, -1 +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: ds_min_rtn_u64 v[0:1], v2, v[0:1] +; GFX8-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX8-NEXT: buffer_wbinvl1_vol +; GFX8-NEXT: BB24_2: +; GFX8-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX8-NEXT: v_readfirstlane_b32 s5, v1 +; GFX8-NEXT: v_readfirstlane_b32 s4, v0 +; GFX8-NEXT: v_cndmask_b32_e64 v1, 0, -1, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v0, 5, -1, vcc +; GFX8-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[0:1] +; GFX8-NEXT: v_mov_b32_e32 v2, s5 +; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc +; GFX8-NEXT: v_mov_b32_e32 v2, s4 +; GFX8-NEXT: s_mov_b32 s2, -1 +; GFX8-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX8-NEXT: s_mov_b32 s3, 0xf000 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX8-NEXT: s_endpgm +; +; GFX9-LABEL: umin_i64_constant: +; GFX9: ; %bb.0: ; %entry +; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX9-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s2, 0 +; GFX9-NEXT: v_mbcnt_hi_u32_b32 v0, s3, v0 +; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX9-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX9-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX9-NEXT: ; mask branch BB24_2 +; GFX9-NEXT: s_cbranch_execz BB24_2 +; GFX9-NEXT: BB24_1: +; GFX9-NEXT: v_mov_b32_e32 v0, 5 +; GFX9-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX9-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: ds_min_rtn_u64 v[0:1], v2, v[0:1] +; GFX9-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX9-NEXT: buffer_wbinvl1_vol +; GFX9-NEXT: BB24_2: +; GFX9-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX9-NEXT: v_readfirstlane_b32 s5, v1 +; GFX9-NEXT: v_readfirstlane_b32 s4, v0 +; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, -1, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v0, 5, -1, vcc +; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[0:1] +; GFX9-NEXT: v_mov_b32_e32 v2, s5 +; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc +; GFX9-NEXT: v_mov_b32_e32 v2, s4 +; GFX9-NEXT: s_mov_b32 s2, -1 +; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX9-NEXT: s_mov_b32 s3, 0xf000 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX9-NEXT: s_endpgm +; +; GFX1064-LABEL: umin_i64_constant: +; GFX1064: ; %bb.0: ; %entry +; GFX1064-NEXT: v_cmp_ne_u32_e64 s[2:3], 1, 0 +; GFX1064-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1064-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1064-NEXT: v_mbcnt_hi_u32_b32_e64 v0, s3, v0 +; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GFX1064-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX1064-NEXT: s_and_saveexec_b64 s[2:3], vcc +; GFX1064-NEXT: ; mask branch BB24_2 +; GFX1064-NEXT: s_cbranch_execz BB24_2 +; GFX1064-NEXT: BB24_1: +; GFX1064-NEXT: v_mov_b32_e32 v0, 5 +; GFX1064-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1064-NEXT: v_mov_b32_e32 v1, 0 +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1064-NEXT: ds_min_rtn_u64 v[0:1], v2, v[0:1] +; GFX1064-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1064-NEXT: buffer_gl0_inv +; GFX1064-NEXT: buffer_gl1_inv +; GFX1064-NEXT: BB24_2: +; GFX1064-NEXT: v_nop +; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3] +; GFX1064-NEXT: v_readfirstlane_b32 s4, v0 +; GFX1064-NEXT: v_readfirstlane_b32 s5, v1 +; GFX1064-NEXT: v_cndmask_b32_e64 v1, 0, -1, vcc +; GFX1064-NEXT: v_cndmask_b32_e64 v0, 5, -1, vcc +; GFX1064-NEXT: s_mov_b32 s2, -1 +; GFX1064-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1064-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[0:1] +; GFX1064-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc +; GFX1064-NEXT: v_cndmask_b32_e64 v0, v0, s4, vcc +; GFX1064-NEXT: s_waitcnt lgkmcnt(0) +; GFX1064-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1064-NEXT: s_endpgm +; +; GFX1032-LABEL: umin_i64_constant: +; GFX1032: ; %bb.0: ; %entry +; GFX1032-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24 +; GFX1032-NEXT: v_cmp_ne_u32_e64 s2, 1, 0 +; GFX1032-NEXT: ; implicit-def: $vcc_hi +; GFX1032-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s2, 0 +; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0 +; GFX1032-NEXT: ; implicit-def: $vgpr0_vgpr1 +; GFX1032-NEXT: s_and_saveexec_b32 s2, vcc_lo +; GFX1032-NEXT: ; mask branch BB24_2 +; GFX1032-NEXT: s_cbranch_execz BB24_2 +; GFX1032-NEXT: BB24_1: +; GFX1032-NEXT: v_mov_b32_e32 v0, 5 +; GFX1032-NEXT: v_mov_b32_e32 v2, local_var64@abs32@lo +; GFX1032-NEXT: v_mov_b32_e32 v1, 0 +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX1032-NEXT: ds_min_rtn_u64 v[0:1], v2, v[0:1] +; GFX1032-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) +; GFX1032-NEXT: buffer_gl0_inv +; GFX1032-NEXT: buffer_gl1_inv +; GFX1032-NEXT: BB24_2: +; GFX1032-NEXT: v_nop +; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s2 +; GFX1032-NEXT: v_readfirstlane_b32 s4, v0 +; GFX1032-NEXT: v_readfirstlane_b32 s5, v1 +; GFX1032-NEXT: v_cndmask_b32_e64 v1, 0, -1, vcc_lo +; GFX1032-NEXT: v_cndmask_b32_e64 v0, 5, -1, vcc_lo +; GFX1032-NEXT: s_mov_b32 s2, -1 +; GFX1032-NEXT: s_mov_b32 s3, 0x31016000 +; GFX1032-NEXT: v_cmp_lt_u64_e32 vcc_lo, s[4:5], v[0:1] +; GFX1032-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc_lo +; GFX1032-NEXT: v_cndmask_b32_e64 v0, v0, s4, vcc_lo +; GFX1032-NEXT: s_waitcnt lgkmcnt(0) +; GFX1032-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 +; GFX1032-NEXT: s_endpgm entry: %old = atomicrmw umin i64 addrspace(3)* @local_var64, i64 5 acq_rel store i64 %old, i64 addrspace(1)* %out diff --git a/test/CodeGen/AMDGPU/dpp_combine.ll b/test/CodeGen/AMDGPU/dpp_combine.ll new file mode 100644 index 00000000000..2afc5388872 --- /dev/null +++ b/test/CodeGen/AMDGPU/dpp_combine.ll @@ -0,0 +1,53 @@ +; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck %s -check-prefix=GCN +; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck %s -check-prefix=GCN + +; GCN-LABEL: {{^}}dpp_add: +; GCN: global_load_dword [[V:v[0-9]+]], +; GCN: v_add_{{(nc_)?}}u32_dpp [[V]], [[V]], [[V]] quad_perm:[1,0,0,0] row_mask:0xf bank_mask:0xf bound_ctrl:0{{$}} +define amdgpu_kernel void @dpp_add(i32 addrspace(1)* %arg) { + %id = tail call i32 @llvm.amdgcn.workitem.id.x() + %gep = getelementptr inbounds i32, i32 addrspace(1)* %arg, i32 %id + %load = load i32, i32 addrspace(1)* %gep + %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %load, i32 %load, i32 1, i32 15, i32 15, i1 1) #0 + %add = add i32 %tmp0, %load + store i32 %add, i32 addrspace(1)* %gep + ret void +} + +; GCN-LABEL: {{^}}dpp_ceil: +; GCN: global_load_dword [[V:v[0-9]+]], +; GCN: v_ceil_f32_dpp [[V]], [[V]] quad_perm:[1,0,0,0] row_mask:0xf bank_mask:0xf bound_ctrl:0{{$}} +define amdgpu_kernel void @dpp_ceil(i32 addrspace(1)* %arg) { + %id = tail call i32 @llvm.amdgcn.workitem.id.x() + %gep = getelementptr inbounds i32, i32 addrspace(1)* %arg, i32 %id + %load = load i32, i32 addrspace(1)* %gep + %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %load, i32 %load, i32 1, i32 15, i32 15, i1 1) #0 + %tmp1 = bitcast i32 %tmp0 to float + %round = tail call float @llvm.ceil.f32(float %tmp1) + %tmp2 = bitcast float %round to i32 + store i32 %tmp2, i32 addrspace(1)* %gep + ret void +} + +; GCN-LABEL: {{^}}dpp_fadd: +; GCN: global_load_dword [[V:v[0-9]+]], +; GCN: v_add_f32_dpp [[V]], [[V]], [[V]] quad_perm:[1,0,0,0] row_mask:0xf bank_mask:0xf bound_ctrl:0{{$}} +define amdgpu_kernel void @dpp_fadd(i32 addrspace(1)* %arg) { + %id = tail call i32 @llvm.amdgcn.workitem.id.x() + %gep = getelementptr inbounds i32, i32 addrspace(1)* %arg, i32 %id + %load = load i32, i32 addrspace(1)* %gep + %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %load, i32 %load, i32 1, i32 15, i32 15, i1 1) #0 + %tmp1 = bitcast i32 %tmp0 to float + %t = bitcast i32 %load to float + %add = fadd float %tmp1, %t + %tmp2 = bitcast float %add to i32 + store i32 %tmp2, i32 addrspace(1)* %gep + ret void +} + + +declare i32 @llvm.amdgcn.workitem.id.x() +declare i32 @llvm.amdgcn.update.dpp.i32(i32, i32, i32, i32, i32, i1) #0 +declare float @llvm.ceil.f32(float) + +attributes #0 = { nounwind readnone convergent } -- 2.40.0