mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-14 10:09:41 +00:00
* enable gfx940 * switch between intrinsic mfma routines on mi100/200 and mi300 * fix mfma_int8 on MI300 * disable 2 int8 examples on MI300 * Update cmake-ck-dev.sh * restore gitignore file * modify Jenkinsfile to the internal repo * Bump rocm-docs-core from 0.24.0 to 0.29.0 in /docs/sphinx Bumps [rocm-docs-core](https://github.com/RadeonOpenCompute/rocm-docs-core) from 0.24.0 to 0.29.0. - [Release notes](https://github.com/RadeonOpenCompute/rocm-docs-core/releases) - [Changelog](https://github.com/RadeonOpenCompute/rocm-docs-core/blob/develop/CHANGELOG.md) - [Commits](https://github.com/RadeonOpenCompute/rocm-docs-core/compare/v0.24.0...v0.29.0) --- updated-dependencies: - dependency-name: rocm-docs-core dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com> * initial enablement of gfx950 * fix clang format * disable examples 31 and 41 int8 on gfx950 * add code * fix build wip * fix xx * now can build * naming * minor fix * wip fix * fix macro for exp2; fix warpgemm a/b in transposedC * unify as tuple_array * Update the required Python version to 3.9 * Update executable name in test scripts * re-structure tuple/array to avoid spill * Merge function templates * Fix format * Add constraint to array<> ctor * Re-use function * Some minor changes * remove wrong code in store_raw() * fix compile issue in transpose * Rename enum Rename 'cood_transform_enum' to 'coord_transform_enum' * let more integral_constant->constant, and formating * make sure thread_buffer can be tuple/array * temp fix buffer_store spill * not using custom data type by default, now we can have ISA-level same code as opt_padding * fix compile error, fp8 not ready now * fix fp8 duplicated move/shift/and/or problem * Default use CK_TILE_FLOAT_TO_FP8_STOCHASTIC rounding mode * fix scratch in fp8 kernel * update some readme * fix merge from upstream * sync with upstream * sync upstream again * sync 22 * remove unused * fix clang-format * update README of ck_tile example * fix several issue * let python version to be 3.8 as minimal * remove ck_tile example from default cmake target like all/install/check * remove mistake * 1).support receipe in generate.py 2).use simplified mask type 3).change left/right to pass into karg * fix some bug in group-mode masking and codegen. update README * F8 quantization for FMHA forward (#1224) * Add SAccElementFunction, PComputeElementFunction, OAccElementFunction in pipeline * Add element function to fmha api * Adjust P elementwise function * Fix bug of elementwise op, our elementwise op is not inout * Add some elementwise op, prepare to quantization * Let generate.py can generate different elementwise function * To prevent compiler issue, remove the elementwise function we have not used. * Remove f8 pipeline, we should share the same pipeline even in f8 * Remove remove_cvref_t * Avoid warning * Fix wrong fp8 QK/KV block gemm setting * Check fp8 rounding error in check_err() * Set fp8 rounding error for check_err() * Use CK_TILE_FLOAT_TO_FP8_STANDARD as default fp8 rounding mode * 1. codgen the f8 api and kernel 2. f8 host code * prevent warning in filter mode * Remove not-in-use elementwise function kargs * Remove more not-in-use elementwise function kargs * Small refinements in C++ source files * Use conditional_t<> to simplify code * Support heterogeneous argument for binary function types * Re-use already-existing scales<> functor template * Fix wrong value produced by saturating * Generalize the composes<> template * Unify saturates<> implementation * Fix type errors in composes<> * Extend less_equal<> * Reuse the existing template less_equal<> in check_err() * Add equal<float> & equal<double> * Rename check_err() parameter * Rename check_err() parameter * Add FIXME comment for adding new macro in future * Remove unnecessary cast to void * Eliminate duplicated code * Avoid dividing api pool into more than 2 groups * Use more clear variable names * Use affirmative condition in if stmt * Remove blank lines * Donot perfect forwarding in composes<> * To fix compile error, revert generate.py back to4439cc107d* Fix bug of p element function * Add compute element op to host softmax * Remove element function in api interface * Extract user parameter * Rename pscale and oscale variable * rename f8 to fp8 * rename more f8 to fp8 * Add pipeline::operator() without element_functor * 1. Remove deprecated pipeline enum 2. Refine host code parameter * Use quantization range as input * 1. Rename max_dtype to dtype_max. 2. Rename scale to scale_s 3.Add init description * Refine description * prevent early return * unify _squant kernel name in cpp, update README * Adjust the default range. * Refine error message and bias range * Add fp8 benchmark and smoke test * fix fp8 swizzle_factor=4 case --------- Co-authored-by: Po Yen Chen <PoYen.Chen@amd.com> Co-authored-by: carlushuang <carlus.huang@amd.com> --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: illsilin <Illia.Silin@amd.com> Co-authored-by: Illia Silin <98187287+illsilin@users.noreply.github.com> Co-authored-by: Jing Zhang <jizha@amd.com> Co-authored-by: zjing14 <zhangjing14@gmail.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Po-Yen, Chen <PoYen.Chen@amd.com> Co-authored-by: rocking <ChunYu.Lai@amd.com> [ROCm/composable_kernel commit:db376dd8a4]
413 lines
7.8 KiB
C++
413 lines
7.8 KiB
C++
// SPDX-License-Identifier: MIT
|
|
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
|
|
|
|
#pragma once
|
|
|
|
// TODO: this folder does not match the single namespace rule. need to refactor in the future
|
|
namespace ck_tile {
|
|
namespace tensor_layout {
|
|
|
|
struct BaseTensorLayout
|
|
{
|
|
};
|
|
|
|
namespace gemm {
|
|
|
|
struct RowMajor : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "RowMajor";
|
|
};
|
|
|
|
struct ColumnMajor : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "ColumnMajor";
|
|
};
|
|
} // namespace gemm
|
|
|
|
namespace convolution {
|
|
|
|
// input tensor
|
|
// packed NCW/NCHW/NCDHW
|
|
struct NCW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NCW";
|
|
};
|
|
|
|
struct NCHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NCHW";
|
|
};
|
|
|
|
struct NCDHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NCDHW";
|
|
};
|
|
|
|
// packed GNCW/GNCHW/GNCDHW
|
|
struct GNCW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNCW";
|
|
};
|
|
|
|
struct GNCHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNCHW";
|
|
};
|
|
|
|
struct GNCDHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNCDHW";
|
|
};
|
|
|
|
// input tensor
|
|
// packed NWC/NHWC/NDHWC
|
|
struct NWC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NWC";
|
|
};
|
|
|
|
struct NHWC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NHWC";
|
|
};
|
|
|
|
struct NDHWC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NDHWC";
|
|
};
|
|
|
|
// input tensor
|
|
// packed GNWC/GNHWC/GNDHWC
|
|
struct GNWC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNWC";
|
|
};
|
|
|
|
struct GNHWC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNHWC";
|
|
};
|
|
|
|
struct GNDHWC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNDHWC";
|
|
};
|
|
|
|
// for input bias
|
|
struct GC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GC";
|
|
};
|
|
|
|
// input tensor
|
|
// packed NWGC/NHWGC/NDHWGC
|
|
struct NWGC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NWGC";
|
|
};
|
|
|
|
struct NHWGC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NHWGC";
|
|
};
|
|
|
|
struct NDHWGC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NDHWGC";
|
|
};
|
|
|
|
// input tensor
|
|
// strided layout
|
|
struct G_NW_C : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_NW_C";
|
|
};
|
|
|
|
struct G_NHW_C : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_NHW_C";
|
|
};
|
|
|
|
struct G_NDHW_C : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_NDHW_C";
|
|
};
|
|
|
|
// for input bias
|
|
struct G_C : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_C";
|
|
};
|
|
|
|
// weight tensor
|
|
// packed KCX/KCYX/KCZYX
|
|
struct KCX : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "KCX";
|
|
};
|
|
|
|
struct KCYX : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "KCYX";
|
|
};
|
|
|
|
struct KCZYX : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "KCZYX";
|
|
};
|
|
|
|
// weight tensor
|
|
// packed KCX/KCYX/KCZYX
|
|
struct GKCX : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GKCX";
|
|
};
|
|
|
|
struct GKCYX : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GKCYX";
|
|
};
|
|
|
|
struct GKCZYX : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GKCZYX";
|
|
};
|
|
|
|
// weight tensor
|
|
// packed KXC/KYXC/KZYXC
|
|
struct KXC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "KXC";
|
|
};
|
|
|
|
struct KYXC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "KYXC";
|
|
};
|
|
|
|
struct KZYXC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "KZYXC";
|
|
};
|
|
|
|
// weight tensor
|
|
// packed GKXC/GKYXC/GKZYXC
|
|
struct GKXC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GKXC";
|
|
};
|
|
|
|
struct GKYXC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GKYXC";
|
|
};
|
|
|
|
struct GKZYXC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GKZYXC";
|
|
};
|
|
|
|
// weight tensor
|
|
// packed KXGC/KYXGC/KZYXGC
|
|
struct KXGC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "KXGC";
|
|
};
|
|
|
|
struct KYXGC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "KYXGC";
|
|
};
|
|
|
|
struct KZYXGC : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "KZYXGC";
|
|
};
|
|
|
|
// weight tensor
|
|
// strided
|
|
struct G_K_X_C : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_K_X_C";
|
|
};
|
|
|
|
struct G_K_YX_C : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_K_YX_C";
|
|
};
|
|
|
|
struct G_K_ZYX_C : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_K_ZYX_C";
|
|
};
|
|
|
|
// output tensor
|
|
// packed NKW/NKHW/NKDHW
|
|
struct NKW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NKW";
|
|
};
|
|
|
|
struct NKHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NKHW";
|
|
};
|
|
|
|
struct NKDHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NKDHW";
|
|
};
|
|
|
|
// output tensor
|
|
// packed GNKW/GNKHW/GNKDHW
|
|
struct GNKW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNKW";
|
|
};
|
|
|
|
struct GNKHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNKHW";
|
|
};
|
|
|
|
struct GNKDHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNKDHW";
|
|
};
|
|
|
|
// output tensor
|
|
// packed NWK/NHWK/NDHWK
|
|
struct NWK : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NWK";
|
|
};
|
|
|
|
struct NHWK : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NHWK";
|
|
};
|
|
|
|
struct NDHWK : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NDHWK";
|
|
};
|
|
|
|
// output tensor
|
|
// packed GNWK/GNHWK/GNDHWK
|
|
struct GNWK : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNWK";
|
|
};
|
|
|
|
struct GNHWK : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNHWK";
|
|
};
|
|
|
|
struct GNDHWK : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNDHWK";
|
|
};
|
|
|
|
// output tensor
|
|
// packed NWGK/NHWGK/NDHWGK
|
|
struct NWGK : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NWGK";
|
|
};
|
|
|
|
struct NHWGK : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NHWGK";
|
|
};
|
|
|
|
struct NDHWGK : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NDHWGK";
|
|
};
|
|
|
|
// output tensor
|
|
// strided layout
|
|
struct G_NW_K : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_NW_K";
|
|
};
|
|
|
|
struct G_NHW_K : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_NHW_K";
|
|
};
|
|
|
|
struct G_NDHW_K : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_NDHW_K";
|
|
};
|
|
|
|
// for output bias
|
|
struct G_K : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_K";
|
|
};
|
|
|
|
// K-reduced output tensor (packed)
|
|
struct GNW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNW";
|
|
};
|
|
|
|
struct GNHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNHW";
|
|
};
|
|
|
|
struct GNDHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "GNDHW";
|
|
};
|
|
|
|
// K-reduced output tensor (packed)
|
|
struct NWG : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NWG";
|
|
};
|
|
|
|
struct NHWG : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NHWG";
|
|
};
|
|
|
|
struct NDHWG : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "NDHWG";
|
|
};
|
|
|
|
// K-reduced output tensor (strided)
|
|
struct G_NW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_NW";
|
|
};
|
|
|
|
struct G_NHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_NHW";
|
|
};
|
|
|
|
struct G_NDHW : public BaseTensorLayout
|
|
{
|
|
static constexpr const char* name = "G_NDHW";
|
|
};
|
|
|
|
} // namespace convolution
|
|
|
|
template <
|
|
typename Layout,
|
|
typename std::enable_if<std::is_base_of<BaseTensorLayout, Layout>::value, bool>::type = false>
|
|
std::ostream& operator<<(std::ostream& os, const Layout&)
|
|
{
|
|
os << Layout::name;
|
|
return os;
|
|
}
|
|
|
|
} // namespace tensor_layout
|
|
} // namespace ck_tile
|