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

[Ck tile] support rmsnorm and related fusion (#1605)

Browse Source 
* Add reduce2d new api

* Prevent user use cross warp reduction

* Fix bug of std caculation

* Add rmsnorm2d

* Add rmsnorm small example

* Remove static assert to prevent compile fail

* Add script to test performance and correctness

* Add missing cmake change

* refine naming

* refine example of rmsnorm

* Fix bug of rmsnorm

* Refine naming

* Fix cmake

* clang format

* Refine pipeline name

* Add add_rmsnorm2d_rdquant kernel

* Add reduce op

* host verification

* Fix bug of one pass pipeline

* Refine tile size

* Add two pass pipeline

* Rename two pass to three pass

* Fix bug of kSaveX == false

* Add instance library

* Add test script

* Fix bug of x verification

* Add save_x to trait

* Add README

* Move reduce2d into reduce folder

* Fix bug of welford when number of m warp > 1

* remove reduncant comment

* 1. move 06_rmsnorm2d to 10_rmsnorm2d
2. move 07_add_rmsnorm2d_rdquant to 11_add_rmsnorm2d_rdquant

* clang format and add missing header

* Add host validation of add + layernorm2d + rsquant

* Revert "Add host validation of add + layernorm2d + rsquant"

This reverts commit 936cb45797.

* Remove deprecated flag
This commit is contained in:
rocking
2024-10-30 15:22:56 +08:00
committed by GitHub
parent 8632221814
commit 3d60953477
90 changed files with 4674 additions and 128 deletions
Download Patch File Download Diff File Expand all files Collapse all files

95
include/ck_tile/core/utility/reduce_operator.hpp Normal file
View File

@@ -0,0 +1,95 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck_tile/core/config.hpp"
namespace ck_tile {
namespace ReduceOp {
// y = ReduceOp(y, x);
struct Add
{
template <typename T>
CK_TILE_HOST_DEVICE static constexpr T GetIdentityValue()
{
return type_convert<T>(0.0f);
};
template <typename T,
typename = std::enable_if_t<std::is_same_v<T, float> || std::is_same_v<T, double> ||
std::is_same_v<T, int32_t> || std::is_same_v<T, int8_t>>>
CK_TILE_HOST_DEVICE constexpr T operator()(const T& y, const T x) const
{
return y + x;
}
template <typename T,
typename = std::enable_if_t<std::is_same_v<T, half_t> || std::is_same_v<T, bf16_t>>>
CK_TILE_HOST_DEVICE constexpr T operator()(T& y, T x) const
{
float y_ = type_convert<float>(y);
float x_ = type_convert<float>(x);
return type_convert<T>(y_ + x_);
}
};
struct SquareAdd
{
template <typename T>
CK_TILE_HOST_DEVICE static constexpr T GetIdentityValue()
{
return type_convert<T>(0.0f);
};
template <typename T,
typename = std::enable_if_t<std::is_same_v<T, float> || std::is_same_v<T, double> ||
std::is_same_v<T, int32_t> || std::is_same_v<T, int8_t>>>
CK_TILE_HOST_DEVICE constexpr T operator()(const T& y, const T x) const
{
return y + (x * x);
}
};
struct Max
{
template <typename T,
typename = std::enable_if_t<std::is_same_v<T, float> || std::is_same_v<T, double> ||
std::is_same_v<T, int32_t> || std::is_same_v<T, int8_t>>>
CK_TILE_HOST_DEVICE static constexpr T GetIdentityValue()
{
return numeric<T>::min();
};
template <typename T,
typename = std::enable_if_t<std::is_same_v<T, float> || std::is_same_v<T, double> ||
std::is_same_v<T, int32_t> || std::is_same_v<T, int8_t>>>
CK_TILE_HOST_DEVICE constexpr T operator()(const T& y, const T x) const
{
return max(y, x);
}
};
struct AbsMax
{
template <typename T,
typename = std::enable_if_t<std::is_same_v<T, float> || std::is_same_v<T, double> ||
std::is_same_v<T, int32_t> || std::is_same_v<T, int8_t>>>
CK_TILE_HOST_DEVICE static constexpr T GetIdentityValue()
{
return numeric<T>::min();
};
template <typename T,
typename = std::enable_if_t<std::is_same_v<T, float> || std::is_same_v<T, double> ||
std::is_same_v<T, int32_t> || std::is_same_v<T, int8_t>>>
CK_TILE_HOST_DEVICE constexpr T operator()(const T& y, const T x) const
{
return max(y, abs(x));
}
};
} // namespace ReduceOp
} // namespace ck_tile