forked from zheng-da/incubator-mxnet
-
Notifications
You must be signed in to change notification settings - Fork 0
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Optimize ERF with MKL math function #5
Open
TaoLv
wants to merge
12
commits into
master
Choose a base branch
from
erf-opt
base: master
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
Open
Changes from all commits
Commits
Show all changes
12 commits
Select commit
Hold shift + click to select a range
f0c7264
mkl_func test with erf&log op, build success~
pengxin99 9311777
fix lint and build issues
TaoLv a79f7db
Try to add support to sparse array
juliusshufan 015fd0a
fix build
TaoLv 495ce36
Merge branch 'master' of https://github.com/apache/incubator-mxnet in…
TaoLv 672be6a
add functions
TaoLv c69a25c
Fix review comments
juliusshufan 2c5c20c
remove unecessary code
juliusshufan b1b6355
Update test case
juliusshufan f96c34a
minor fix
juliusshufan 06c51e9
move the position of MKL_Compute
juliusshufan acd7b56
Merge pull request #6 from juliusshufan/erf
TaoLv File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,218 @@ | ||
/* | ||
* Licensed to the Apache Software Foundation (ASF) under one | ||
* or more contributor license agreements. See the NOTICE file | ||
* distributed with this work for additional information | ||
* regarding copyright ownership. The ASF licenses this file | ||
* to you under the Apache License, Version 2.0 (the | ||
* "License"); you may not use this file except in compliance | ||
* with the License. You may obtain a copy of the License at | ||
* | ||
* http://www.apache.org/licenses/LICENSE-2.0 | ||
* | ||
* Unless required by applicable law or agreed to in writing, | ||
* software distributed under the License is distributed on an | ||
* "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY | ||
* KIND, either express or implied. See the License for the | ||
* specific language governing permissions and limitations | ||
* under the License. | ||
*/ | ||
|
||
/*! | ||
* Copyright (c) 2018 by Contributors | ||
* \file mkl_functions-inl.h | ||
* \brief | ||
* \author | ||
*/ | ||
#ifndef MXNET_OPERATOR_MKL_FUNCTIONS_INL_H_ | ||
#define MXNET_OPERATOR_MKL_FUNCTIONS_INL_H_ | ||
|
||
#if MSHADOW_USE_MKL == 1 | ||
#include "mkl.h" | ||
|
||
namespace mxnet { | ||
namespace op { | ||
namespace mkl_func { | ||
|
||
MSHADOW_XINLINE | ||
static bool check_size(const size_t n) { | ||
const size_t MKL_INT_MAX = (sizeof(MKL_INT) == sizeof(int)) ? INT_MAX : LLONG_MAX; | ||
return (n <= MKL_INT_MAX); | ||
} | ||
|
||
MSHADOW_XINLINE | ||
static bool check_type(const int t) { | ||
return (t == mshadow::kFloat32 || t == mshadow::kFloat64); | ||
} | ||
|
||
#define MXNET_MKL_UNARY_MATH_FUNC(name, func) \ | ||
struct name { \ | ||
MSHADOW_XINLINE static void Vectorize(const index_t n, const float *src, float *dst) { \ | ||
vs##func(static_cast<MKL_INT>(n), src, dst); \ | ||
} \ | ||
MSHADOW_XINLINE static void Vectorize(const index_t n, const double *src, double *dst) { \ | ||
vd##func(static_cast<MKL_INT>(n), src, dst); \ | ||
} \ | ||
}; | ||
|
||
#define MXNET_MKL_BINARY_MATH_FUNC(name, func) \ | ||
struct name { \ | ||
MSHADOW_XINLINE static void Vectorize(const index_t n, \ | ||
const float *a, \ | ||
const float *b, \ | ||
float *c) { \ | ||
vs##func(static_cast<MKL_INT>(n), a, b, c); \ | ||
} \ | ||
MSHADOW_XINLINE static void Vectorize(const index_t n, \ | ||
const double *a, \ | ||
const double *b, \ | ||
double *c) { \ | ||
vd##func(static_cast<MKL_INT>(n), a, b, c); \ | ||
} \ | ||
}; | ||
|
||
MXNET_MKL_UNARY_MATH_FUNC(erf, Erf); | ||
MXNET_MKL_UNARY_MATH_FUNC(exp, Exp); | ||
MXNET_MKL_UNARY_MATH_FUNC(exp2, Exp2); | ||
MXNET_MKL_UNARY_MATH_FUNC(exp10, Exp10); | ||
MXNET_MKL_UNARY_MATH_FUNC(expm1, Expm1); | ||
MXNET_MKL_UNARY_MATH_FUNC(log, Ln); | ||
MXNET_MKL_UNARY_MATH_FUNC(log2, Log2); | ||
MXNET_MKL_UNARY_MATH_FUNC(log10, Log10); | ||
MXNET_MKL_UNARY_MATH_FUNC(log1p, Log1p); | ||
|
||
MXNET_MKL_UNARY_MATH_FUNC(sin, Sin); | ||
MXNET_MKL_UNARY_MATH_FUNC(cos, Cos); | ||
MXNET_MKL_UNARY_MATH_FUNC(tan, Tan); | ||
MXNET_MKL_UNARY_MATH_FUNC(asin, Asin); | ||
MXNET_MKL_UNARY_MATH_FUNC(acos, Acos); | ||
MXNET_MKL_UNARY_MATH_FUNC(atan, Atan); | ||
|
||
MXNET_MKL_UNARY_MATH_FUNC(sinh, Sinh); | ||
MXNET_MKL_UNARY_MATH_FUNC(cosh, Cosh); | ||
MXNET_MKL_UNARY_MATH_FUNC(tanh, Tanh); | ||
MXNET_MKL_UNARY_MATH_FUNC(asinh, Asinh); | ||
MXNET_MKL_UNARY_MATH_FUNC(acosh, Acosh); | ||
MXNET_MKL_UNARY_MATH_FUNC(atanh, Atanh); | ||
|
||
MXNET_MKL_UNARY_MATH_FUNC(sqrt, Sqrt); | ||
MXNET_MKL_UNARY_MATH_FUNC(abs, Abs); | ||
MXNET_MKL_UNARY_MATH_FUNC(cbrt, Cbrt); | ||
MXNET_MKL_UNARY_MATH_FUNC(round, Round); | ||
MXNET_MKL_UNARY_MATH_FUNC(ceil, Ceil); | ||
MXNET_MKL_UNARY_MATH_FUNC(floor, Floor); | ||
MXNET_MKL_UNARY_MATH_FUNC(trunc, Trunc); | ||
|
||
MXNET_MKL_UNARY_MATH_FUNC(lgamma, LGamma); | ||
MXNET_MKL_UNARY_MATH_FUNC(tgamma, TGamma); | ||
MXNET_MKL_UNARY_MATH_FUNC(square, Sqr); | ||
|
||
MXNET_MKL_BINARY_MATH_FUNC(add, Add); | ||
MXNET_MKL_BINARY_MATH_FUNC(sub, Sub); | ||
MXNET_MKL_BINARY_MATH_FUNC(mul, Mul); | ||
MXNET_MKL_BINARY_MATH_FUNC(pow, Pow); | ||
MXNET_MKL_BINARY_MATH_FUNC(hypot, Hypot); | ||
|
||
|
||
template <typename DType> | ||
MSHADOW_XINLINE static void sub_(index_t n, DType *in, DType b, DType *dst) { | ||
for (index_t i = 0; i < n; i++) | ||
dst[i] = in[i] - b; | ||
} | ||
|
||
template <typename DType> | ||
MSHADOW_XINLINE static void div_(index_t n, DType *in, DType b, DType *dst) { | ||
for (index_t i = 0; i < n; i++) | ||
dst[i] = in[i] / b; | ||
} | ||
|
||
template <typename DType> | ||
MSHADOW_XINLINE static void sum_(index_t n, DType *in, DType *dst) { | ||
// dst[0] = cblas_sasum(n, in, 1); | ||
DType sum = 0.0f; | ||
for (index_t i = 0; i < n; i++) | ||
sum += in[i]; | ||
|
||
dst[0] = sum; | ||
} | ||
|
||
template <typename DType> | ||
MSHADOW_XINLINE static void max_(int n, DType * __restrict__ in, DType *dst) { | ||
dst[0] = in[0]; | ||
for (int i = 1; i < n; i++) | ||
dst[0] = (dst[0] < in[i]) ? in[i] : dst[0]; | ||
} | ||
|
||
// LayerNorm on the last dimension | ||
template <typename DType> | ||
MSHADOW_XINLINE static void LayerNormLastDim(const index_t m, | ||
const index_t n, | ||
const DType *a, | ||
const DType *b, | ||
const DType *ws, | ||
const DType *gamma, | ||
const DType *beta, | ||
const DType *mean, | ||
const DType *var, | ||
const DType eps) { | ||
#pragma omp parallel for | ||
for (index_t i = 0; i < m; i++) { | ||
DType* in_offset = a + i * n; | ||
DType* out_offset = b + i * n; | ||
DType* ws_offset = ws + i * n; | ||
|
||
sum_(n, in_offset, &(mean[i])); | ||
mean[i] /= n; | ||
sub_(n, in_offset, mean[i], out_offset); | ||
square(n, out_offset, ws_offset); | ||
sum_(n, ws_offset, &(var[i])); | ||
var[i] = sqrt(var[i] / n + eps); | ||
|
||
mul(n, out_offset, gamma, out_offset); | ||
div_(n, out_offset, var[i], out_offset); | ||
add(n, out_offset, beta, out_offset); | ||
} | ||
} | ||
|
||
// softmax on the last dimension | ||
template <typename DType> | ||
MSHADOW_XINLINE static void SoftmaxLastDim(const index_t m, | ||
const index_t n, | ||
const DType *a, | ||
const DType *b) { | ||
#pragma omp paralle for | ||
for (index_t i = 0; i < m; i++) { | ||
DType* in_offset = a + i * n; | ||
DType* out_offset = b + i * n; | ||
|
||
exp(n, in_offset, out_offset); | ||
float sum = 0.0f; | ||
sum_(n, out_offset, &sum); | ||
div_(n, out_offset, sum, out_offset); | ||
} | ||
} | ||
|
||
template <typename DType> | ||
MSHADOW_XINLINE static void LogSoftmaxLastDim(const index_t m, | ||
const index_t n, | ||
const DType *a, | ||
const DType *b) { | ||
#pragma parallel for | ||
for (index_t i = 0; i < m; i++) { | ||
DType* in_offset = a + i * n; | ||
DType* out_offset = b + i * n; | ||
|
||
DType b, logsum; | ||
max_(n, in_offset, &b); | ||
sub_(n, in_offset, b, out_offset); | ||
exp(n, out_offset, out_offset); | ||
sum_(n, out_offset, &logsum); | ||
logsum = b + logf(logsum); | ||
sub_(n, in_offset, logsum, out_offset); | ||
} | ||
} | ||
|
||
} // namespace mkl_func | ||
} // namespace op | ||
} // namespace mxnet | ||
#endif // MSHADOW_USE_MKL == 1 | ||
#endif // MXNET_OPERATOR_MKL_FUNCTIONS_INL_H_ |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have not set USE_MKL before. Just curious: is blas=MKL also tested in mxnet CI?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
MSHADOW_USE_MKL
is widely used in mshadow and mxnet to indicate MKL is used as BLAS library. Yes,USE_BLAS=mkl
is built and tested in CI:https://github.com/apache/incubator-mxnet/blob/master/ci/docker/runtime_functions.sh#L375
https://github.com/apache/incubator-mxnet/blob/master/ci/docker/runtime_functions.sh#L553