ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-04 05:31:24 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_TILE] Update flatmm related kernels (#3022)

Browse Source 
---------

Co-authored-by: Ding, Yi <yi.ding@amd.com>
Co-authored-by: felix <felix.li@amd.com>
This commit is contained in:
lalala-sh
2025-10-22 22:36:11 +08:00
committed by GitHub
parent cbd1279ae6
commit 211d64e18a
39 changed files with 11183 additions and 739 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
include/ck_tile/core/numeric/vector_type.hpp
View File

@@ -11,6 +11,7 @@
#include "ck_tile/core/numeric/half.hpp"
#include "ck_tile/core/numeric/bfloat16.hpp"
#include "ck_tile/core/numeric/pk_int4.hpp"
#include "ck_tile/core/numeric/e8m0.hpp"
#include "ck_tile/core/utility/type_traits.hpp"
namespace ck_tile {
@@ -88,7 +89,12 @@ template <typename T, typename = void>
struct vector_traits
{
using scalar_type =
std::conditional_t<std::is_same_v<remove_cvref_t<T>, pk_int4_t>, int8_t, remove_cvref_t<T>>;
std::conditional_t<std::is_same_v<remove_cvref_t<T>, pk_int4_t>,
int8_t,
std::conditional_t<std::is_same_v<remove_cvref_t<T>, pk_fp4_t> ||
std::is_same_v<remove_cvref_t<T>, e8m0_t>,
uint8_t,
remove_cvref_t<T>>>;
static constexpr index_t vector_size = 1;
};
@@ -96,7 +102,12 @@ struct vector_traits
template <typename T, index_t N>
struct vector_traits<T __attribute__((ext_vector_type(N))), void>
{
using scalar_type = std::conditional_t<std::is_same_v<T, pk_int4_t>, int8_t, T>;
using scalar_type = std::conditional_t<
std::is_same_v<T, pk_int4_t>,
int8_t,
std::conditional_t<std::is_same_v<T, pk_fp4_t> || std::is_same_v<remove_cvref_t<T>, e8m0_t>,
uint8_t,
T>>;
static constexpr index_t vector_size = N;
};
@@ -237,4 +248,10 @@ using pk_int4x4_t = int8_t __attribute__((ext_vector_type(4)));
using pk_int4x8_t = int8_t __attribute__((ext_vector_type(8)));
using pk_int4x16_t = int8_t __attribute__((ext_vector_type(16)));
using pk_int4x32_t = int8_t __attribute__((ext_vector_type(32)));
using pk_fp4x2_t = uint8_t __attribute((ext_vector_type(2)));
using pk_fp4x4_t = uint8_t __attribute((ext_vector_type(4)));
using pk_fp4x8_t = uint8_t __attribute((ext_vector_type(8)));
using pk_fp4x16_t = uint8_t __attribute((ext_vector_type(16)));
using pk_fp4x32_t = uint8_t __attribute((ext_vector_type(32)));
} // namespace ck_tile