Skip to content

Commit 8d3435f

Browse files
arsenmarsenm
authored
AMDGPU: Add v_smfmac_f32_32x32x64_bf8_fp8 for gfx950 (llvm#117257)
1 parent 8a5c241 commit 8d3435f

File tree

12 files changed

+505
-1
lines changed

12 files changed

+505
-1
lines changed

‎clang/include/clang/Basic/BuiltinsAMDGPU.def‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -455,6 +455,7 @@ TARGET_BUILTIN(__builtin_amdgcn_smfmac_f32_16x16x128_bf8_fp8, "V4fV4iV8iV4fiIiIi
455455
TARGET_BUILTIN(__builtin_amdgcn_smfmac_f32_16x16x128_fp8_bf8, "V4fV4iV8iV4fiIiIi", "nc", "gfx950-insts")
456456
TARGET_BUILTIN(__builtin_amdgcn_smfmac_f32_16x16x128_fp8_fp8, "V4fV4iV8iV4fiIiIi", "nc", "gfx950-insts")
457457
TARGET_BUILTIN(__builtin_amdgcn_smfmac_f32_32x32x64_bf8_bf8, "V16fV4iV8iV16fiIiIi", "nc", "gfx950-insts")
458+
TARGET_BUILTIN(__builtin_amdgcn_smfmac_f32_32x32x64_bf8_fp8, "V16fV4iV8iV16fiIiIi", "nc", "gfx950-insts")
458459

459460
//===----------------------------------------------------------------------===//
460461
// GFX12+ only builtins.

‎clang/test/CodeGenOpenCL/builtins-amdgcn-mfma.cl‎

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -545,4 +545,11 @@ void test_smfmac_f32_32x32x64_bf8_bf8(global v16f* out, v4i a, v8i b, v16f c, in
545545
*out = __builtin_amdgcn_smfmac_f32_32x32x64_bf8_bf8(a, b, c, idx, 0, 0);
546546
}
547547

548+
// CHECK-GFX950-LABEL: @test_smfmac_f32_32x32x64_bf8_fp8
549+
// CHECK-GFX950: call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.fp8(<4 x i32> %a, <8 x i32> %b, <16 x float> %c, i32 %idx, i32 0, i32 0)
550+
void test_smfmac_f32_32x32x64_bf8_fp8(global v16f* out, v4i a, v8i b, v16f c, int idx)
551+
{
552+
*out = __builtin_amdgcn_smfmac_f32_32x32x64_bf8_fp8(a, b, c, idx, 0, 0);
553+
}
554+
548555
#endif

‎clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950-param.cl‎

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -130,3 +130,9 @@ void test_smfmac_f32_32x32x64_bf8_bf8(global float16* out, int4 a, int8 b, float
130130
*out = __builtin_amdgcn_smfmac_f32_32x32x64_bf8_bf8(a, b, c, idx, d, 0); // expected-error{{argument to '__builtin_amdgcn_smfmac_f32_32x32x64_bf8_bf8' must be a constant integer}}
131131
*out = __builtin_amdgcn_smfmac_f32_32x32x64_bf8_bf8(a, b, c, idx, 0, d); // expected-error{{argument to '__builtin_amdgcn_smfmac_f32_32x32x64_bf8_bf8' must be a constant integer}}
132132
}
133+
134+
void test_smfmac_f32_32x32x64_bf8_fp8(global float16* out, int4 a, int8 b, float16 c, int idx, int d)
135+
{
136+
*out = __builtin_amdgcn_smfmac_f32_32x32x64_bf8_fp8(a, b, c, idx, d, 0); // expected-error{{argument to '__builtin_amdgcn_smfmac_f32_32x32x64_bf8_fp8' must be a constant integer}}
137+
*out = __builtin_amdgcn_smfmac_f32_32x32x64_bf8_fp8(a, b, c, idx, 0, d); // expected-error{{argument to '__builtin_amdgcn_smfmac_f32_32x32x64_bf8_fp8' must be a constant integer}}
138+
}

‎clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950.cl‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -45,6 +45,7 @@ void test(__global float4* out0, half8 a0, half8 b0, float4 c0,
4545
*out12 = __builtin_amdgcn_smfmac_f32_16x16x128_fp8_bf8(a12, b12, c12, 0, 0, 0); // expected-error{{'__builtin_amdgcn_smfmac_f32_16x16x128_fp8_bf8' needs target feature gfx950-insts}}
4646
*out12 = __builtin_amdgcn_smfmac_f32_16x16x128_fp8_fp8(a12, b12, c12, 0, 0, 0); // expected-error{{'__builtin_amdgcn_smfmac_f32_16x16x128_fp8_fp8' needs target feature gfx950-insts}}
4747
*out13 = __builtin_amdgcn_smfmac_f32_32x32x64_bf8_bf8(a13, b13, c13, 0, 0, 0); // expected-error{{'__builtin_amdgcn_smfmac_f32_32x32x64_bf8_bf8' needs target feature gfx950-insts}}
48+
*out13 = __builtin_amdgcn_smfmac_f32_32x32x64_bf8_fp8(a13, b13, c13, 0, 0, 0); // expected-error{{'__builtin_amdgcn_smfmac_f32_32x32x64_bf8_fp8' needs target feature gfx950-insts}}
4849
*out14 = __builtin_amdgcn_mfma_scale_f32_16x16x128_f8f6f4(a14, b14, c14, 0, 0, 0, d14, 0, e14); // expected-error{{'__builtin_amdgcn_mfma_scale_f32_16x16x128_f8f6f4' needs target feature gfx950-insts}}
4950
*out15 = __builtin_amdgcn_mfma_scale_f32_32x32x64_f8f6f4(a15, b15, c15, 0, 0, 0, d15, 0, e15); // expected-error{{'__builtin_amdgcn_mfma_scale_f32_32x32x64_f8f6f4' needs target feature gfx950-insts}}
5051
}

‎llvm/include/llvm/IR/IntrinsicsAMDGPU.td‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3163,6 +3163,7 @@ def int_amdgcn_smfmac_f32_16x16x128_bf8_fp8 : AMDGPUMSmfmacIntrinsic<llvm_v4f32_
31633163
def int_amdgcn_smfmac_f32_16x16x128_fp8_bf8 : AMDGPUMSmfmacIntrinsic<llvm_v4f32_ty, llvm_v4i32_ty, llvm_v8i32_ty>;
31643164
def int_amdgcn_smfmac_f32_16x16x128_fp8_fp8 : AMDGPUMSmfmacIntrinsic<llvm_v4f32_ty, llvm_v4i32_ty, llvm_v8i32_ty>;
31653165
def int_amdgcn_smfmac_f32_32x32x64_bf8_bf8 : AMDGPUMSmfmacIntrinsic<llvm_v16f32_ty, llvm_v4i32_ty, llvm_v8i32_ty>;
3166+
def int_amdgcn_smfmac_f32_32x32x64_bf8_fp8 : AMDGPUMSmfmacIntrinsic<llvm_v16f32_ty, llvm_v4i32_ty, llvm_v8i32_ty>;
31663167
}
31673168

31683169
//===----------------------------------------------------------------------===//

‎llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp‎

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1101,6 +1101,7 @@ bool AMDGPUInstructionSelector::selectG_INTRINSIC(MachineInstr &I) const {
11011101
case Intrinsic::amdgcn_smfmac_f32_16x16x128_fp8_bf8:
11021102
case Intrinsic::amdgcn_smfmac_f32_16x16x128_fp8_fp8:
11031103
case Intrinsic::amdgcn_smfmac_f32_32x32x64_bf8_bf8:
1104+
case Intrinsic::amdgcn_smfmac_f32_32x32x64_bf8_fp8:
11041105
return selectSMFMACIntrin(I);
11051106
default:
11061107
return selectImpl(I, *CoverageInfo);
@@ -3555,6 +3556,9 @@ bool AMDGPUInstructionSelector::selectSMFMACIntrin(MachineInstr &MI) const {
35553556
case Intrinsic::amdgcn_smfmac_f32_32x32x64_bf8_bf8:
35563557
Opc = AMDGPU::V_SMFMAC_F32_32X32X64_BF8_BF8_e64;
35573558
break;
3559+
case Intrinsic::amdgcn_smfmac_f32_32x32x64_bf8_fp8:
3560+
Opc = AMDGPU::V_SMFMAC_F32_32X32X64_BF8_FP8_e64;
3561+
break;
35583562
default:
35593563
llvm_unreachable("unhandled smfmac intrinsic");
35603564
}

‎llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp‎

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -4815,7 +4815,8 @@ AMDGPURegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
48154815
case Intrinsic::amdgcn_smfmac_f32_16x16x128_bf8_fp8:
48164816
case Intrinsic::amdgcn_smfmac_f32_16x16x128_fp8_bf8:
48174817
case Intrinsic::amdgcn_smfmac_f32_16x16x128_fp8_fp8:
4818-
case Intrinsic::amdgcn_smfmac_f32_32x32x64_bf8_bf8: {
4818+
case Intrinsic::amdgcn_smfmac_f32_32x32x64_bf8_bf8:
4819+
case Intrinsic::amdgcn_smfmac_f32_32x32x64_bf8_fp8: {
48194820
// vdst, srcA, srcB, srcC, idx
48204821
OpdsMapping[0] = getAGPROpMapping(MI.getOperand(0).getReg(), MRI, *TRI);
48214822
OpdsMapping[2] = getVGPROpMapping(MI.getOperand(2).getReg(), MRI, *TRI);

‎llvm/lib/Target/AMDGPU/VOP3PInstructions.td‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1063,6 +1063,7 @@ defm V_SMFMAC_F32_16X16X128_BF8_FP8 : SMFMACInst<"v_smfmac_f32_16x16x128_bf8_fp8
10631063
defm V_SMFMAC_F32_16X16X128_FP8_BF8 : SMFMACInst<"v_smfmac_f32_16x16x128_fp8_bf8", "F32_16X16X128_F8", int_amdgcn_smfmac_f32_16x16x128_fp8_bf8>;
10641064
defm V_SMFMAC_F32_16X16X128_FP8_FP8 : SMFMACInst<"v_smfmac_f32_16x16x128_fp8_fp8", "F32_16X16X128_F8", int_amdgcn_smfmac_f32_16x16x128_fp8_fp8>;
10651065
defm V_SMFMAC_F32_32X32X64_BF8_BF8 : SMFMACInst<"v_smfmac_f32_32x32x64_bf8_bf8", "F32_32X32X64_F8", int_amdgcn_smfmac_f32_32x32x64_bf8_bf8>;
1066+
defm V_SMFMAC_F32_32X32X64_BF8_FP8 : SMFMACInst<"v_smfmac_f32_32x32x64_bf8_fp8", "F32_32X32X64_F8", int_amdgcn_smfmac_f32_32x32x64_bf8_fp8>;
10661067
}
10671068

10681069
def MAIInstInfoTable : GenericTable {
@@ -2166,6 +2167,7 @@ defm V_SMFMAC_F32_16X16X128_BF8_FP8 : VOP3P_Real_SMFMAC <0x3c, "v_smfmac_f32_16x
21662167
defm V_SMFMAC_F32_16X16X128_FP8_BF8 : VOP3P_Real_SMFMAC <0x3d, "v_smfmac_f32_16x16x128fp8bf8">;
21672168
defm V_SMFMAC_F32_16X16X128_FP8_FP8 : VOP3P_Real_SMFMAC <0x43, "v_smfmac_f32_16x16x128fp8fp8">;
21682169
defm V_SMFMAC_F32_32X32X64_BF8_BF8 : VOP3P_Real_SMFMAC <0x4b, "v_smfmac_f32_32x32x64bf8bf8">;
2170+
defm V_SMFMAC_F32_32X32X64_BF8_FP8 : VOP3P_Real_SMFMAC <0x4e, "v_smfmac_f32_32x32x64bf8fp8">;
21692171

21702172
defm V_PK_FMA_F32 : VOP3P_Real_vi <0x30>;
21712173
defm V_PK_MUL_F32 : VOP3P_Real_vi <0x31>;

‎llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll‎

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -413,6 +413,15 @@ define amdgpu_kernel void @smfmac_f32_32x32x64_bf8_bf8(<4 x i32> %arg0, <8 x i32
413413
ret void
414414
}
415415

416+
declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.fp8(<4 x i32>, <8 x i32>, <16 x float>, i32, i32, i32)
417+
418+
; CHECK: DIVERGENT: %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.fp8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 1, i32 2)
419+
define amdgpu_kernel void @smfmac_f32_32x32x64_bf8_fp8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
420+
%result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.fp8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 1, i32 2)
421+
store <16 x float> %result, ptr addrspace(1) %out
422+
ret void
423+
}
424+
416425
declare i32 @llvm.amdgcn.ds.swizzle(i32, i32) #1
417426
declare i32 @llvm.amdgcn.permlane16.i32(i32, i32, i32, i32, i1, i1) #1
418427
declare i32 @llvm.amdgcn.permlanex16.i32(i32, i32, i32, i32, i1, i1) #1

0 commit comments

Comments
 (0)