This is an archive of the discontinued LLVM Phabricator instance.

Differential D3713

R600: Add mul24 intrinsics
ClosedPublic

Authored by arsenm on May 10 2014, 11:02 AM.

Details

Reviewers
arsenm

Diff Detail

Event Timeline

arsenm updated this revision to Diff 9283.May 10 2014, 11:02 AM
arsenm retitled this revision from to R600: Add mul24 intrinsics.
arsenm updated this object.
arsenm edited the test plan for this revision. (Show Details)
arsenm added a subscriber: Unknown Object (MLST).
arsenm accepted this revision.May 12 2014, 10:57 AM
arsenm added a reviewer: arsenm.

r208604

This revision is now accepted and ready to land.May 12 2014, 10:57 AM
arsenm closed this revision.May 12 2014, 10:57 AM

Revision Contents

PathSize
lib/
Target/
R600/
8 lines
2 lines
4 lines
test/
CodeGen/
R600/
14 lines
11 lines

Diff 9283

lib/Target/R600/AMDGPUISelLowering.cpp

Context not available.
return DAG.getNode(AMDGPUISD::UMIN, DL, VT, Op.getOperand(1), return DAG.getNode(AMDGPUISD::UMIN, DL, VT, Op.getOperand(1),
Op.getOperand(2)); Op.getOperand(2));
case AMDGPUIntrinsic::AMDGPU_imul24:
return DAG.getNode(AMDGPUISD::MUL_I24, DL, VT,
Op.getOperand(1), Op.getOperand(2));
case AMDGPUIntrinsic::AMDGPU_umul24:
return DAG.getNode(AMDGPUISD::MUL_U24, DL, VT,
Op.getOperand(1), Op.getOperand(2));
case AMDGPUIntrinsic::AMDGPU_bfe_i32: case AMDGPUIntrinsic::AMDGPU_bfe_i32:
return DAG.getNode(AMDGPUISD::BFE_I32, DL, VT, return DAG.getNode(AMDGPUISD::BFE_I32, DL, VT,
Op.getOperand(1), Op.getOperand(1),
Context not available.

lib/Target/R600/AMDGPUIntrinsics.td

Context not available.
def int_AMDGPU_imin : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; def int_AMDGPU_imin : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
def int_AMDGPU_umax : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; def int_AMDGPU_umax : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
def int_AMDGPU_umin : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; def int_AMDGPU_umin : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
def int_AMDGPU_imul24 : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
def int_AMDGPU_umul24 : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
def int_AMDGPU_cube : Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty], [IntrNoMem]>; def int_AMDGPU_cube : Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty], [IntrNoMem]>;
def int_AMDGPU_bfi : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; def int_AMDGPU_bfi : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
def int_AMDGPU_bfe_i32 : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; def int_AMDGPU_bfe_i32 : Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
Context not available.

lib/Target/R600/AMDILIntrinsics.td

Context not available.
BinaryIntInt; BinaryIntInt;
def int_AMDIL_mulhi_u32 : GCCBuiltin<"__amdil_umul_high">, def int_AMDIL_mulhi_u32 : GCCBuiltin<"__amdil_umul_high">,
BinaryIntInt; BinaryIntInt;
def int_AMDIL_mul24_i32 : GCCBuiltin<"__amdil_imul24">,
BinaryIntInt;
def int_AMDIL_mul24_u32 : GCCBuiltin<"__amdil_umul24">,
BinaryIntInt;
def int_AMDIL_mulhi24_i32 : GCCBuiltin<"__amdil_imul24_high">, def int_AMDIL_mulhi24_i32 : GCCBuiltin<"__amdil_imul24_high">,
BinaryIntInt; BinaryIntInt;
def int_AMDIL_mulhi24_u32 : GCCBuiltin<"__amdil_umul24_high">, def int_AMDIL_mulhi24_u32 : GCCBuiltin<"__amdil_umul24_high">,
Context not available.

test/CodeGen/R600/llvm.AMDGPU.imul24.ll

  • This file was added.
; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=CM -check-prefix=FUNC %s
declare i32 @llvm.AMDGPU.imul24(i32, i32) nounwind readnone
; FUNC-LABEL: @test_imul24
; SI: V_MUL_I32_I24
; CM: MUL_INT24
define void @test_imul24(i32 addrspace(1)* %out, i32 %src0, i32 %src1) nounwind {
%mul = call i32 @llvm.AMDGPU.imul24(i32 %src0, i32 %src1) nounwind readnone
store i32 %mul, i32 addrspace(1)* %out, align 4
ret void
}

test/CodeGen/R600/llvm.AMDGPU.umul24.ll

  • This file was added.
; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
declare i32 @llvm.AMDGPU.umul24(i32, i32) nounwind readnone
; SI-LABEL: @test_umul24
define void @test_umul24(i32 addrspace(1)* %out, i32 %src0, i32 %src1) nounwind {
%mul = call i32 @llvm.AMDGPU.umul24(i32 %src0, i32 %src1) nounwind readnone
store i32 %mul, i32 addrspace(1)* %out, align 4
ret void
}