From 506d4b2cb4ec053879b81843abbfd8156d36f614 Mon Sep 17 00:00:00 2001 From: Marek Olsak Date: Wed, 4 Mar 2015 17:33:45 +0000 Subject: [PATCH] R600/SI: Add an intrinsic for S_FLBIT_I32 / V_FFBH_I32 Required by OpenGL (ARB_gpu_shader5). git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@231259 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/R600/AMDGPUIntrinsics.td | 1 + lib/Target/R600/SIInstrInfo.cpp | 1 + lib/Target/R600/SIInstructions.td | 4 +++- test/CodeGen/R600/llvm.AMDGPU.flbit.i32.ll | 28 ++++++++++++++++++++++ 4 files changed, 33 insertions(+), 1 deletion(-) create mode 100644 test/CodeGen/R600/llvm.AMDGPU.flbit.i32.ll diff --git a/lib/Target/R600/AMDGPUIntrinsics.td b/lib/Target/R600/AMDGPUIntrinsics.td index eee9c29038d..ab489cd2a4a 100644 --- a/lib/Target/R600/AMDGPUIntrinsics.td +++ b/lib/Target/R600/AMDGPUIntrinsics.td @@ -68,6 +68,7 @@ let TargetPrefix = "AMDGPU", isTarget = 1 in { def int_AMDGPU_bfe_u32 : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; def int_AMDGPU_bfm : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; def int_AMDGPU_brev : Intrinsic<[llvm_i32_ty], [llvm_i32_ty], [IntrNoMem]>; + def int_AMDGPU_flbit_i32 : Intrinsic<[llvm_i32_ty], [llvm_i32_ty], [IntrNoMem]>; def int_AMDGPU_barrier_local : Intrinsic<[], [], []>; def int_AMDGPU_barrier_global : Intrinsic<[], [], []>; } diff --git a/lib/Target/R600/SIInstrInfo.cpp b/lib/Target/R600/SIInstrInfo.cpp index a4662643224..ea9ad71f7fb 100644 --- a/lib/Target/R600/SIInstrInfo.cpp +++ b/lib/Target/R600/SIInstrInfo.cpp @@ -1420,6 +1420,7 @@ unsigned SIInstrInfo::getVALUOp(const MachineInstr &MI) { case AMDGPU::S_BCNT1_I32_B32: return AMDGPU::V_BCNT_U32_B32_e64; case AMDGPU::S_FF1_I32_B32: return AMDGPU::V_FFBL_B32_e32; case AMDGPU::S_FLBIT_I32_B32: return AMDGPU::V_FFBH_U32_e32; + case AMDGPU::S_FLBIT_I32: return AMDGPU::V_FFBH_I32_e64; } } diff --git a/lib/Target/R600/SIInstructions.td b/lib/Target/R600/SIInstructions.td index 934aaf58aab..ab1f08f087f 100644 --- a/lib/Target/R600/SIInstructions.td +++ b/lib/Target/R600/SIInstructions.td @@ -153,7 +153,9 @@ defm S_FLBIT_I32_B32 : SOP1_32 , "s_flbit_i32_b32", >; defm S_FLBIT_I32_B64 : SOP1_32_64 , "s_flbit_i32_b64", []>; -defm S_FLBIT_I32 : SOP1_32 , "s_flbit_i32", []>; +defm S_FLBIT_I32 : SOP1_32 , "s_flbit_i32", + [(set i32:$dst, (int_AMDGPU_flbit_i32 i32:$src0))] +>; defm S_FLBIT_I32_I64 : SOP1_32_64 , "s_flbit_i32_i64", []>; defm S_SEXT_I32_I8 : SOP1_32 , "s_sext_i32_i8", [(set i32:$dst, (sext_inreg i32:$src0, i8))] diff --git a/test/CodeGen/R600/llvm.AMDGPU.flbit.i32.ll b/test/CodeGen/R600/llvm.AMDGPU.flbit.i32.ll new file mode 100644 index 00000000000..20c7af8ade5 --- /dev/null +++ b/test/CodeGen/R600/llvm.AMDGPU.flbit.i32.ll @@ -0,0 +1,28 @@ +; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s +; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s + +declare i32 @llvm.AMDGPU.flbit.i32(i32) nounwind readnone + +; FUNC-LABEL: {{^}}s_flbit: +; SI: s_load_dword [[VAL:s[0-9]+]], +; SI: s_flbit_i32 [[SRESULT:s[0-9]+]], [[VAL]] +; SI: v_mov_b32_e32 [[VRESULT:v[0-9]+]], [[SRESULT]] +; SI: buffer_store_dword [[VRESULT]], +; SI: s_endpgm +define void @s_flbit(i32 addrspace(1)* noalias %out, i32 %val) nounwind { + %r = call i32 @llvm.AMDGPU.flbit.i32(i32 %val) nounwind readnone + store i32 %r, i32 addrspace(1)* %out, align 4 + ret void +} + +; FUNC-LABEL: {{^}}v_flbit: +; SI: buffer_load_dword [[VAL:v[0-9]+]], +; SI: v_ffbh_i32_e32 [[RESULT:v[0-9]+]], [[VAL]] +; SI: buffer_store_dword [[RESULT]], +; SI: s_endpgm +define void @v_flbit(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind { + %val = load i32, i32 addrspace(1)* %valptr, align 4 + %r = call i32 @llvm.AMDGPU.flbit.i32(i32 %val) nounwind readnone + store i32 %r, i32 addrspace(1)* %out, align 4 + ret void +} -- 2.34.1