leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 29 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/dgrad/sm90_conv1d_dgrad_implicit_gemm_f16_f16_f32_tensorop_f16.cu
+16
View File
@@ -87,7 +87,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -130,7 +132,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -173,7 +177,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -216,7 +222,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -264,7 +272,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -307,7 +317,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -350,7 +362,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -393,7 +407,9 @@ TEST(SM90_device_conv1d_dgrad_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/dgrad/sm90_conv1d_dgrad_implicit_gemm_f16_f16_f32_tensorop_f32.cu
+16
View File
@@ -87,7 +87,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -130,7 +132,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -173,7 +177,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -216,7 +222,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -263,7 +271,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -306,7 +316,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -349,7 +361,9 @@ TEST(SM90_device_conv1d_dgrad_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
>;
@@ -392,7 +406,9 @@ TEST(SM90_device_conv1d_dgrad_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/dgrad/sm90_conv2d_dgrad_implicit_gemm_f16_f16_f32_tensorop_f16.cu
+16
View File
@@ -88,7 +88,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -132,7 +134,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -176,7 +180,9 @@ TEST(SM90_device_conv2d_dgrad_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_dgrad_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
>;
@@ -268,7 +276,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -312,7 +322,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -356,7 +368,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -400,7 +414,9 @@ TEST(SM90_device_conv2d_dgrad_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/dgrad/sm90_conv2d_dgrad_implicit_gemm_f16_f16_f32_tensorop_f32.cu
+16
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -129,7 +131,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -173,7 +177,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -217,7 +223,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -265,7 +273,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -309,7 +319,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -353,7 +365,9 @@ TEST(SM90_device_conv2d_dgrad_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
>;
@@ -397,7 +411,9 @@ TEST(SM90_device_conv2d_dgrad_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/dgrad/sm90_conv3d_dgrad_implicit_gemm_f16_f16_f32_tensorop_f16.cu
+16
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv3d_dgrad_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
>;
@@ -130,7 +132,9 @@ TEST(SM90_device_conv3d_dgrad_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
>;
@@ -174,7 +178,9 @@ TEST(SM90_device_conv3d_dgrad_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
>;
@@ -218,7 +224,9 @@ TEST(SM90_device_conv3d_dgrad_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
>;
@@ -266,7 +274,9 @@ TEST(SM90_device_conv3d_dgrad_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
>;
@@ -310,7 +320,9 @@ TEST(SM90_device_conv3d_dgrad_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
>;
@@ -354,7 +366,9 @@ TEST(SM90_device_conv3d_dgrad_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
>;
@@ -398,7 +412,9 @@ TEST(SM90_device_conv3d_dgrad_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/dgrad/sm90_conv3d_dgrad_implicit_gemm_f16_f16_f32_tensorop_f32.cu
+16
View File
@@ -85,7 +85,9 @@ TEST(SM90_device_conv3d_dgrad_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_dgrad_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_dgrad_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_dgrad_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_dgrad_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_dgrad_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_dgrad_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_dgrad_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
>;