ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-20 06:49:15 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add the last two forward instance traits. (#3134)

Browse Source 
* Add InstanceTraits for DeviceGroupedConvFwdMultipleD_Wmma_CShuffle

* Add InstanceTraits for kernel_grouped_conv_fwd_dl_multiple_d

* A few small changes to fix broken instance traits.
This commit is contained in:
John Shumway
2025-10-31 07:52:42 -07:00
committed by GitHub
parent 1977e4b96a
commit 5ed2046bee
17 changed files with 1207 additions and 82 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_dl_multiple_d_nhwc_kyxc_nhwk.hpp
View File

@@ -24,6 +24,10 @@
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/host_utility/io.hpp"
#ifdef CK_EXPERIMENTAL_BUILDER
#include "ck_tile/builder/reflect/instance_traits_device_grouped_conv_fwd_dl_multiple_d_nhwc_kyxc_nhwk.hpp"
#endif
namespace ck {
namespace tensor_operation {
namespace device {
@@ -1048,6 +1052,19 @@ struct DeviceGroupedConvFwdDlMultipleD_NHWC_KYXC_NHWK
return str.str();
}
#ifdef CK_EXPERIMENTAL_BUILDER
std::string GetInstanceString() const override
{
static_assert(ck_tile::reflect::HasInstanceTraits<DeviceOp>,
"Specialization of instance_traits not found. Please check that a "
"specialization exists in file "
"ck_tile/builder/reflect/"
"instance_traits_device_grouped_conv_fwd_dl_multiple_d_nhwc_kyxc_nhwk.hpp "
"for the given template parameters.");
return ck_tile::reflect::instance_string<DeviceOp>();
}
#endif
};
} // namespace device

17
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_d_wmma_cshuffle.hpp
View File

@@ -24,6 +24,10 @@
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/host_utility/io.hpp"
#ifdef CK_EXPERIMENTAL_BUILDER
#include "ck_tile/builder/reflect/instance_traits_device_grouped_conv_fwd_multiple_d_wmma_cshuffle.hpp"
#endif
namespace ck {
namespace tensor_operation {
namespace device {
@@ -1003,6 +1007,19 @@ struct DeviceGroupedConvFwdMultipleD_Wmma_CShuffle
return str.str();
}
#ifdef CK_EXPERIMENTAL_BUILDER
std::string GetInstanceString() const override
{
static_assert(ck_tile::reflect::HasInstanceTraits<DeviceOp>,
"Specialization of instance_traits not found. Please check that a "
"specialization exists in file "
"ck_tile/builder/reflect/"
"instance_traits_device_grouped_conv_fwd_multiple_d_wmma_cshuffle.hpp "
"for the given template parameters.");
return ck_tile::reflect::instance_string<DeviceOp>();
}
#endif
};
} // namespace device

2
include/ck/tensor_operation/gpu/element/unary_element_wise_operation.hpp
View File

@@ -1760,6 +1760,8 @@ struct FastNumericArrayConverter<uint8_t, half_t, N>
struct DynamicUnaryOp
{
static constexpr const char* name = "DynamicUnaryOp";
__host__ __device__ DynamicUnaryOp() = delete;
__host__ __device__ DynamicUnaryOp(const Swish& swish)