leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 26 Packages Projects Releases Wiki Activity

CUTLASS 3.5.1 (#1623)

Browse Source 
* CUTLASS 3.5.1

* updates, optimizations, fixes
This commit is contained in:
Vijay Thakkar
2024-07-29 08:46:24 -04:00
committed by GitHub
parent 56b46e2d13
commit be60a0b272
312 changed files with 19793 additions and 6775 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/unit/conv/device_3x/fprop/sm90_conv2d_fprop_implicit_gemm_f16_f16_f32_tensorop_f16.cu
+8
View File
@@ -93,6 +93,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -137,6 +138,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -181,6 +183,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -225,6 +228,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//////////////////////////////////////////////////////////////////////////////////////////////////
@@ -273,6 +277,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -317,6 +322,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -361,6 +367,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -405,6 +412,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f16
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
#endif // defined(CUTLASS_ARCH_MMA_SM90_SUPPORTED)
test/unit/conv/device_3x/fprop/sm90_conv2d_fprop_implicit_gemm_f16_f16_f32_tensorop_f32.cu
+8
View File
@@ -93,6 +93,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -137,6 +138,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -181,6 +183,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -225,6 +228,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//////////////////////////////////////////////////////////////////////////////////////////////////
@@ -273,6 +277,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -317,6 +322,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -361,6 +367,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -405,6 +412,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
#endif // defined(CUTLASS_ARCH_MMA_SM90_SUPPORTED)
test/unit/conv/device_3x/fprop/sm90_conv2d_fprop_implicit_gemm_tf32_tf32_f32_tensorop_f32.cu
+8
View File
@@ -93,6 +93,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -137,6 +138,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -181,6 +183,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -225,6 +228,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//////////////////////////////////////////////////////////////////////////////////////////////////
@@ -273,6 +277,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -317,6 +322,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -361,6 +367,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
//
@@ -405,6 +412,7 @@ TEST(SM90_device_conv2d_fprop_implicitgemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f
using Conv = cutlass::conv::device::ConvUniversalAdapter<ConvKernel>;
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>());
EXPECT_TRUE(test::conv::device::TestAllConv<Conv>(/*alpha=*/1.0, /*beta=*/1.0));
}
#endif // defined(CUTLASS_ARCH_MMA_SM90_SUPPORTED)