derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
192 lines
11 KiB
MLIR
192 lines
11 KiB
MLIR
// RUN: mlir-translate --mlir-to-llvmir %s | FileCheck %s
|
|
|
|
// CHECK-LABEL: @nvvm_mma_sp_kind_m16n8k64_e4m3_f16
|
|
llvm.func @nvvm_mma_sp_kind_m16n8k64_e4m3_f16(
|
|
%a0: i32, %a1: i32, %a2: i32, %a3: i32,
|
|
%b0: i32, %b1: i32, %b2: i32, %b3: i32,
|
|
%c0: vector<2xf16>, %c1: vector<2xf16>,
|
|
%meta: i32) -> !llvm.struct<(vector<2xf16>, vector<2xf16>)> {
|
|
%sel = llvm.mlir.constant(0 : i32) : i32
|
|
// CHECK: call { <2 x half>, <2 x half> } @llvm.nvvm.mma.sp.ordered.metadata.m16n8k64.row.col.kind.f8f6f4.f16.e4m3.e4m3.f16(i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, <2 x half> {{%[0-9]+}}, <2 x half> {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 0)
|
|
%res = nvvm.mma.sp.sync A[%a0, %a1, %a2, %a3] B[%b0, %b1, %b2, %b3] C[%c0, %c1]
|
|
sparseMetadata[%meta] selector[%sel]
|
|
{kind = #nvvm.mma_kind<f8f6f4>,
|
|
multiplicandAPtxType = #nvvm.mma_type<e4m3>,
|
|
multiplicandBPtxType = #nvvm.mma_type<e4m3>,
|
|
orderedMetadata,
|
|
shape = #nvvm.shape<m = 16, n = 8, k = 64>}
|
|
: (i32, i32, vector<2xf16>) -> !llvm.struct<(vector<2xf16>, vector<2xf16>)>
|
|
llvm.return %res : !llvm.struct<(vector<2xf16>, vector<2xf16>)>
|
|
}
|
|
|
|
// CHECK-LABEL: @nvvm_mma_sp_kind_m16n8k64_e4m3_f32
|
|
llvm.func @nvvm_mma_sp_kind_m16n8k64_e4m3_f32(
|
|
%a0: i32, %a1: i32, %a2: i32, %a3: i32,
|
|
%b0: i32, %b1: i32, %b2: i32, %b3: i32,
|
|
%c0: f32, %c1: f32, %c2: f32, %c3: f32,
|
|
%meta: i32) -> !llvm.struct<(f32, f32, f32, f32)> {
|
|
%sel = llvm.mlir.constant(0 : i32) : i32
|
|
// CHECK: call { float, float, float, float } @llvm.nvvm.mma.sp.ordered.metadata.m16n8k64.row.col.f32.e4m3.e4m3.f32(i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 0)
|
|
%res = nvvm.mma.sp.sync A[%a0, %a1, %a2, %a3] B[%b0, %b1, %b2, %b3] C[%c0, %c1, %c2, %c3]
|
|
sparseMetadata[%meta] selector[%sel]
|
|
{kind = #nvvm.mma_kind<f8f6f4>,
|
|
multiplicandAPtxType = #nvvm.mma_type<e4m3>,
|
|
multiplicandBPtxType = #nvvm.mma_type<e4m3>,
|
|
orderedMetadata,
|
|
shape = #nvvm.shape<m = 16, n = 8, k = 64>}
|
|
: (i32, i32, f32) -> !llvm.struct<(f32, f32, f32, f32)>
|
|
llvm.return %res : !llvm.struct<(f32, f32, f32, f32)>
|
|
}
|
|
|
|
// CHECK-LABEL: @nvvm_mma_sp_kind_m16n8k64_e5m2_f16
|
|
llvm.func @nvvm_mma_sp_kind_m16n8k64_e5m2_f16(
|
|
%a0: i32, %a1: i32, %a2: i32, %a3: i32,
|
|
%b0: i32, %b1: i32, %b2: i32, %b3: i32,
|
|
%c0: vector<2xf16>, %c1: vector<2xf16>,
|
|
%meta: i32) -> !llvm.struct<(vector<2xf16>, vector<2xf16>)> {
|
|
%sel = llvm.mlir.constant(0 : i32) : i32
|
|
// CHECK: call { <2 x half>, <2 x half> } @llvm.nvvm.mma.sp.ordered.metadata.m16n8k64.row.col.kind.f8f6f4.f16.e5m2.e5m2.f16(i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, <2 x half> {{%[0-9]+}}, <2 x half> {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 0)
|
|
%res = nvvm.mma.sp.sync A[%a0, %a1, %a2, %a3] B[%b0, %b1, %b2, %b3] C[%c0, %c1]
|
|
sparseMetadata[%meta] selector[%sel]
|
|
{kind = #nvvm.mma_kind<f8f6f4>,
|
|
multiplicandAPtxType = #nvvm.mma_type<e5m2>,
|
|
multiplicandBPtxType = #nvvm.mma_type<e5m2>,
|
|
orderedMetadata,
|
|
shape = #nvvm.shape<m = 16, n = 8, k = 64>}
|
|
: (i32, i32, vector<2xf16>) -> !llvm.struct<(vector<2xf16>, vector<2xf16>)>
|
|
llvm.return %res : !llvm.struct<(vector<2xf16>, vector<2xf16>)>
|
|
}
|
|
|
|
// CHECK-LABEL: @nvvm_mma_sp_kind_m16n8k64_e5m2_f32
|
|
llvm.func @nvvm_mma_sp_kind_m16n8k64_e5m2_f32(
|
|
%a0: i32, %a1: i32, %a2: i32, %a3: i32,
|
|
%b0: i32, %b1: i32, %b2: i32, %b3: i32,
|
|
%c0: f32, %c1: f32, %c2: f32, %c3: f32,
|
|
%meta: i32) -> !llvm.struct<(f32, f32, f32, f32)> {
|
|
%sel = llvm.mlir.constant(0 : i32) : i32
|
|
// CHECK: call { float, float, float, float } @llvm.nvvm.mma.sp.ordered.metadata.m16n8k64.row.col.f32.e5m2.e5m2.f32(i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 0)
|
|
%res = nvvm.mma.sp.sync A[%a0, %a1, %a2, %a3] B[%b0, %b1, %b2, %b3] C[%c0, %c1, %c2, %c3]
|
|
sparseMetadata[%meta] selector[%sel]
|
|
{kind = #nvvm.mma_kind<f8f6f4>,
|
|
multiplicandAPtxType = #nvvm.mma_type<e5m2>,
|
|
multiplicandBPtxType = #nvvm.mma_type<e5m2>,
|
|
orderedMetadata,
|
|
shape = #nvvm.shape<m = 16, n = 8, k = 64>}
|
|
: (i32, i32, f32) -> !llvm.struct<(f32, f32, f32, f32)>
|
|
llvm.return %res : !llvm.struct<(f32, f32, f32, f32)>
|
|
}
|
|
|
|
// CHECK-LABEL: @nvvm_mma_sp_kind_m16n8k64_e3m2_f16
|
|
llvm.func @nvvm_mma_sp_kind_m16n8k64_e3m2_f16(
|
|
%a0: i32, %a1: i32, %a2: i32, %a3: i32,
|
|
%b0: i32, %b1: i32, %b2: i32, %b3: i32,
|
|
%c0: vector<2xf16>, %c1: vector<2xf16>,
|
|
%meta: i32) -> !llvm.struct<(vector<2xf16>, vector<2xf16>)> {
|
|
%sel = llvm.mlir.constant(0 : i32) : i32
|
|
// CHECK: call { <2 x half>, <2 x half> } @llvm.nvvm.mma.sp.ordered.metadata.m16n8k64.row.col.kind.f8f6f4.f16.e3m2.e3m2.f16(i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, <2 x half> {{%[0-9]+}}, <2 x half> {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 0)
|
|
%res = nvvm.mma.sp.sync A[%a0, %a1, %a2, %a3] B[%b0, %b1, %b2, %b3] C[%c0, %c1]
|
|
sparseMetadata[%meta] selector[%sel]
|
|
{kind = #nvvm.mma_kind<f8f6f4>,
|
|
multiplicandAPtxType = #nvvm.mma_type<e3m2>,
|
|
multiplicandBPtxType = #nvvm.mma_type<e3m2>,
|
|
orderedMetadata,
|
|
shape = #nvvm.shape<m = 16, n = 8, k = 64>}
|
|
: (i32, i32, vector<2xf16>) -> !llvm.struct<(vector<2xf16>, vector<2xf16>)>
|
|
llvm.return %res : !llvm.struct<(vector<2xf16>, vector<2xf16>)>
|
|
}
|
|
|
|
// CHECK-LABEL: @nvvm_mma_sp_kind_m16n8k64_e3m2_f32
|
|
llvm.func @nvvm_mma_sp_kind_m16n8k64_e3m2_f32(
|
|
%a0: i32, %a1: i32, %a2: i32, %a3: i32,
|
|
%b0: i32, %b1: i32, %b2: i32, %b3: i32,
|
|
%c0: f32, %c1: f32, %c2: f32, %c3: f32,
|
|
%meta: i32) -> !llvm.struct<(f32, f32, f32, f32)> {
|
|
%sel = llvm.mlir.constant(0 : i32) : i32
|
|
// CHECK: call { float, float, float, float } @llvm.nvvm.mma.sp.ordered.metadata.m16n8k64.row.col.kind.f8f6f4.f32.e3m2.e3m2.f32(i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 0)
|
|
%res = nvvm.mma.sp.sync A[%a0, %a1, %a2, %a3] B[%b0, %b1, %b2, %b3] C[%c0, %c1, %c2, %c3]
|
|
sparseMetadata[%meta] selector[%sel]
|
|
{kind = #nvvm.mma_kind<f8f6f4>,
|
|
multiplicandAPtxType = #nvvm.mma_type<e3m2>,
|
|
multiplicandBPtxType = #nvvm.mma_type<e3m2>,
|
|
orderedMetadata,
|
|
shape = #nvvm.shape<m = 16, n = 8, k = 64>}
|
|
: (i32, i32, f32) -> !llvm.struct<(f32, f32, f32, f32)>
|
|
llvm.return %res : !llvm.struct<(f32, f32, f32, f32)>
|
|
}
|
|
|
|
// CHECK-LABEL: @nvvm_mma_sp_kind_m16n8k64_e2m3_f16
|
|
llvm.func @nvvm_mma_sp_kind_m16n8k64_e2m3_f16(
|
|
%a0: i32, %a1: i32, %a2: i32, %a3: i32,
|
|
%b0: i32, %b1: i32, %b2: i32, %b3: i32,
|
|
%c0: vector<2xf16>, %c1: vector<2xf16>,
|
|
%meta: i32) -> !llvm.struct<(vector<2xf16>, vector<2xf16>)> {
|
|
%sel = llvm.mlir.constant(0 : i32) : i32
|
|
// CHECK: call { <2 x half>, <2 x half> } @llvm.nvvm.mma.sp.ordered.metadata.m16n8k64.row.col.kind.f8f6f4.f16.e2m3.e2m3.f16(i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, <2 x half> {{%[0-9]+}}, <2 x half> {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 0)
|
|
%res = nvvm.mma.sp.sync A[%a0, %a1, %a2, %a3] B[%b0, %b1, %b2, %b3] C[%c0, %c1]
|
|
sparseMetadata[%meta] selector[%sel]
|
|
{kind = #nvvm.mma_kind<f8f6f4>,
|
|
multiplicandAPtxType = #nvvm.mma_type<e2m3>,
|
|
multiplicandBPtxType = #nvvm.mma_type<e2m3>,
|
|
orderedMetadata,
|
|
shape = #nvvm.shape<m = 16, n = 8, k = 64>}
|
|
: (i32, i32, vector<2xf16>) -> !llvm.struct<(vector<2xf16>, vector<2xf16>)>
|
|
llvm.return %res : !llvm.struct<(vector<2xf16>, vector<2xf16>)>
|
|
}
|
|
|
|
// CHECK-LABEL: @nvvm_mma_sp_kind_m16n8k64_e2m3_f32
|
|
llvm.func @nvvm_mma_sp_kind_m16n8k64_e2m3_f32(
|
|
%a0: i32, %a1: i32, %a2: i32, %a3: i32,
|
|
%b0: i32, %b1: i32, %b2: i32, %b3: i32,
|
|
%c0: f32, %c1: f32, %c2: f32, %c3: f32,
|
|
%meta: i32) -> !llvm.struct<(f32, f32, f32, f32)> {
|
|
%sel = llvm.mlir.constant(0 : i32) : i32
|
|
// CHECK: call { float, float, float, float } @llvm.nvvm.mma.sp.ordered.metadata.m16n8k64.row.col.kind.f8f6f4.f32.e2m3.e2m3.f32(i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 0)
|
|
%res = nvvm.mma.sp.sync A[%a0, %a1, %a2, %a3] B[%b0, %b1, %b2, %b3] C[%c0, %c1, %c2, %c3]
|
|
sparseMetadata[%meta] selector[%sel]
|
|
{kind = #nvvm.mma_kind<f8f6f4>,
|
|
multiplicandAPtxType = #nvvm.mma_type<e2m3>,
|
|
multiplicandBPtxType = #nvvm.mma_type<e2m3>,
|
|
orderedMetadata,
|
|
shape = #nvvm.shape<m = 16, n = 8, k = 64>}
|
|
: (i32, i32, f32) -> !llvm.struct<(f32, f32, f32, f32)>
|
|
llvm.return %res : !llvm.struct<(f32, f32, f32, f32)>
|
|
}
|
|
|
|
// CHECK-LABEL: @nvvm_mma_sp_kind_m16n8k64_e2m1_f16
|
|
llvm.func @nvvm_mma_sp_kind_m16n8k64_e2m1_f16(
|
|
%a0: i32, %a1: i32, %a2: i32, %a3: i32,
|
|
%b0: i32, %b1: i32, %b2: i32, %b3: i32,
|
|
%c0: vector<2xf16>, %c1: vector<2xf16>,
|
|
%meta: i32) -> !llvm.struct<(vector<2xf16>, vector<2xf16>)> {
|
|
%sel = llvm.mlir.constant(0 : i32) : i32
|
|
// CHECK: call { <2 x half>, <2 x half> } @llvm.nvvm.mma.sp.ordered.metadata.m16n8k64.row.col.kind.f8f6f4.f16.e2m1.e2m1.f16(i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, <2 x half> {{%[0-9]+}}, <2 x half> {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 0)
|
|
%res = nvvm.mma.sp.sync A[%a0, %a1, %a2, %a3] B[%b0, %b1, %b2, %b3] C[%c0, %c1]
|
|
sparseMetadata[%meta] selector[%sel]
|
|
{kind = #nvvm.mma_kind<f8f6f4>,
|
|
multiplicandAPtxType = #nvvm.mma_type<e2m1>,
|
|
multiplicandBPtxType = #nvvm.mma_type<e2m1>,
|
|
orderedMetadata,
|
|
shape = #nvvm.shape<m = 16, n = 8, k = 64>}
|
|
: (i32, i32, vector<2xf16>) -> !llvm.struct<(vector<2xf16>, vector<2xf16>)>
|
|
llvm.return %res : !llvm.struct<(vector<2xf16>, vector<2xf16>)>
|
|
}
|
|
|
|
// CHECK-LABEL: @nvvm_mma_sp_kind_m16n8k64_e2m1_f32
|
|
llvm.func @nvvm_mma_sp_kind_m16n8k64_e2m1_f32(
|
|
%a0: i32, %a1: i32, %a2: i32, %a3: i32,
|
|
%b0: i32, %b1: i32, %b2: i32, %b3: i32,
|
|
%c0: f32, %c1: f32, %c2: f32, %c3: f32,
|
|
%meta: i32) -> !llvm.struct<(f32, f32, f32, f32)> {
|
|
%sel = llvm.mlir.constant(0 : i32) : i32
|
|
// CHECK: call { float, float, float, float } @llvm.nvvm.mma.sp.ordered.metadata.m16n8k64.row.col.kind.f8f6f4.f32.e2m1.e2m1.f32(i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, float {{%[0-9]+}}, i32 {{%[0-9]+}}, i32 0)
|
|
%res = nvvm.mma.sp.sync A[%a0, %a1, %a2, %a3] B[%b0, %b1, %b2, %b3] C[%c0, %c1, %c2, %c3]
|
|
sparseMetadata[%meta] selector[%sel]
|
|
{kind = #nvvm.mma_kind<f8f6f4>,
|
|
multiplicandAPtxType = #nvvm.mma_type<e2m1>,
|
|
multiplicandBPtxType = #nvvm.mma_type<e2m1>,
|
|
orderedMetadata,
|
|
shape = #nvvm.shape<m = 16, n = 8, k = 64>}
|
|
: (i32, i32, f32) -> !llvm.struct<(f32, f32, f32, f32)>
|
|
llvm.return %res : !llvm.struct<(f32, f32, f32, f32)>
|
|
}
|