leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 25 Packages Projects Releases Wiki Activity

CUTLASS 3.6.0 (#1850)

Browse Source 
* v3.6

* update changelog

* update readme

* fix typo

* fixing typos

* hopper gemm with weight prefetch

---------

Co-authored-by: yuzhai <yuzhai@nvidia.com>
Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
This commit is contained in:
Yujia Zhai
2024-10-09 15:33:27 -04:00
committed by GitHub
co-authored by yuzhai Haicheng Wu
parent 0837a2a00a
commit cc3c29a81a
354 changed files with 105937 additions and 8197 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/unit/conv/device_3x/fprop/sm90_conv1d_fprop_implicit_gemm_f16_f16_f32_tensorop_f16.cu
+17 -1
View File
@@ -84,7 +84,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f16, 6
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -127,7 +129,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f16, 6
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -169,8 +173,10 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f16, 6
cutlass::conv::collective::StageCountAutoCarveout<static_cast<int>(sizeof(typename CollectiveEpilogue::SharedStorage))>,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -213,7 +219,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f16, 6
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -261,7 +269,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f16, 1
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -304,7 +314,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f16, 1
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -347,7 +359,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f16, 1
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -390,7 +404,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f16, 1
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv1d_fprop_implicit_gemm_f16_f16_f32_tensorop_f32.cu
+19
View File
@@ -84,7 +84,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f32, 6
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -127,7 +129,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f32, 6
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -170,7 +174,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f32, 6
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -213,7 +219,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f32, 6
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -260,7 +268,10 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f32, 1
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -303,7 +314,10 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f32, 1
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -346,7 +360,10 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f32, 1
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -389,7 +406,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_f16nwc_f16nwc_f32nwc_tensor_op_f32, 1
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv1d_fprop_implicit_gemm_s8_s8_s32_tensorop_s32.cu
+21 -1
View File
@@ -55,6 +55,7 @@ using namespace cute;
// Cluster 1x1x1
//
TEST(SM90_device_conv1d_fprop_implicitgemm_s8nwc_s8nwc_s32nwc_tensor_op_s32, 64x64x64_1x1x1) {
using ElementAct = int8_t;
using ElementFlt = int8_t;
using ElementOut = int32_t;
@@ -84,10 +85,13 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_s8nwc_s8nwc_s32nwc_tensor_op_s32, 64x
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape = cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
>;
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
@@ -128,7 +132,10 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_s8nwc_s8nwc_s32nwc_tensor_op_s32, 64x
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape = cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -171,7 +178,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_s8nwc_s8nwc_s32nwc_tensor_op_s32, 64x
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape = cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -214,7 +223,10 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_s8nwc_s8nwc_s32nwc_tensor_op_s32, 64x
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape = cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -261,7 +273,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_s8nwc_s8nwc_s32nwc_tensor_op_s32, 128
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape = cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -304,7 +318,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_s8nwc_s8nwc_s32nwc_tensor_op_s32, 128
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape = cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -347,7 +363,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_s8nwc_s8nwc_s32nwc_tensor_op_s32, 128
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape = cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -390,7 +408,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_s8nwc_s8nwc_s32nwc_tensor_op_s32, 128
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape = cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv1d_fprop_implicit_gemm_tf32_tf32_f32_tensorop_f32.cu
+17
View File
@@ -84,7 +84,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_tf32nwc_tf32nwc_f32nwc_tensor_op_f32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -127,7 +129,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_tf32nwc_tf32nwc_f32nwc_tensor_op_f32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -170,7 +174,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_tf32nwc_tf32nwc_f32nwc_tensor_op_f32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -213,7 +219,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_tf32nwc_tf32nwc_f32nwc_tensor_op_f32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -260,7 +268,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_tf32nwc_tf32nwc_f32nwc_tensor_op_f32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -303,7 +313,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_tf32nwc_tf32nwc_f32nwc_tensor_op_f32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -346,7 +358,9 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_tf32nwc_tf32nwc_f32nwc_tensor_op_f32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -389,7 +403,10 @@ TEST(SM90_device_conv1d_fprop_implicitgemm_tf32nwc_tf32nwc_f32nwc_tensor_op_f32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv2d_fprop_implicit_gemm_f16_f16_f32_tensorop_f16.cu
+17
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -130,7 +132,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -175,7 +179,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -220,7 +226,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -269,7 +277,10 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -314,7 +325,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -359,7 +372,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -404,7 +419,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv2d_fprop_implicit_gemm_f16_f16_f32_tensorop_f32.cu
+16
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -130,7 +132,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -175,7 +179,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -220,7 +226,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -269,7 +277,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -314,7 +324,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -359,7 +371,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -404,7 +418,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv2d_fprop_implicit_gemm_s8_s8_s32_tensorop_s32.cu
+16
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_s8nhwc_s8nhwc_s32nhwc_tensor_op_s32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -129,7 +131,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_s8nhwc_s8nhwc_s32nhwc_tensor_op_s32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -173,7 +177,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_s8nhwc_s8nhwc_s32nhwc_tensor_op_s32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -217,7 +223,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_s8nhwc_s8nhwc_s32nhwc_tensor_op_s32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -265,7 +273,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_s8nhwc_s8nhwc_s32nhwc_tensor_op_s32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -309,7 +319,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_s8nhwc_s8nhwc_s32nhwc_tensor_op_s32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -353,7 +365,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_s8nhwc_s8nhwc_s32nhwc_tensor_op_s32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -397,7 +411,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_s8nhwc_s8nhwc_s32nhwc_tensor_op_s32,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv2d_fprop_implicit_gemm_tf32_tf32_f32_tensorop_f32.cu
+16
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -130,7 +132,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -175,7 +179,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -220,7 +226,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -269,7 +277,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -314,7 +324,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -359,7 +371,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -404,7 +418,9 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv3d_fprop_implicit_gemm_f16_f16_f32_tensorop_f16.cu
+17 -1
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -128,8 +130,10 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::StageCountAutoCarveout<static_cast<int>(sizeof(typename CollectiveEpilogue::SharedStorage))>,
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -173,7 +177,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -217,7 +223,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -265,7 +273,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -309,7 +319,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -353,7 +365,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -397,7 +411,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv3d_fprop_implicit_gemm_f16_f16_f32_tensorop_f32.cu
+16
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -129,7 +131,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -173,7 +177,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -217,7 +223,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -265,7 +273,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -309,7 +319,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -353,7 +365,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -397,7 +411,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_f16ndhwc_f16ndhwc_f32ndhwc_tensor_op_
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv3d_fprop_implicit_gemm_s8_s8_s32_tensorop_s32.cu
+16
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_s8ndhwc_s8ndhwc_s32ndhwc_tensor_op_s3
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -129,7 +131,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_s8ndhwc_s8ndhwc_s32ndhwc_tensor_op_s3
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -173,7 +177,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_s8ndhwc_s8ndhwc_s32ndhwc_tensor_op_s3
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -217,7 +223,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_s8ndhwc_s8ndhwc_s32ndhwc_tensor_op_s3
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -265,7 +273,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_s8ndhwc_s8ndhwc_s32ndhwc_tensor_op_s3
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -309,7 +319,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_s8ndhwc_s8ndhwc_s32ndhwc_tensor_op_s3
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -353,7 +365,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_s8ndhwc_s8ndhwc_s32ndhwc_tensor_op_s3
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -397,7 +411,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_s8ndhwc_s8ndhwc_s32ndhwc_tensor_op_s3
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
test/unit/conv/device_3x/fprop/sm90_conv3d_fprop_implicit_gemm_tf32_tf32_f32_tensorop_f32.cu
+16
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_tf32ndhwc_tf32ndhwc_f32ndhwc_tensor_o
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -129,7 +131,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_tf32ndhwc_tf32ndhwc_f32ndhwc_tensor_o
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -173,7 +177,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_tf32ndhwc_tf32ndhwc_f32ndhwc_tensor_o
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -217,7 +223,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_tf32ndhwc_tf32ndhwc_f32ndhwc_tensor_o
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -265,7 +273,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_tf32ndhwc_tf32ndhwc_f32ndhwc_tensor_o
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -309,7 +319,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_tf32ndhwc_tf32ndhwc_f32ndhwc_tensor_o
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -353,7 +365,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_tf32ndhwc_tf32ndhwc_f32ndhwc_tensor_o
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;
@@ -397,7 +411,9 @@ TEST(SM90_device_conv3d_fprop_implicitgemm_tf32ndhwc_tf32ndhwc_f32ndhwc_tensor_o
cutlass::conv::collective::KernelScheduleAuto
>::CollectiveOp;
using ProblemShape=cutlass::conv::ConvProblemShape<CollectiveMainloop::DispatchPolicy::ConvOp, CollectiveMainloop::DispatchPolicy::NumSpatialDimensions>;
using ConvKernel = cutlass::conv::kernel::ConvUniversal<
ProblemShape,
CollectiveMainloop,
CollectiveEpilogue
>;