Nigel
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!9146 fix maxpoolgrad bug 

From: @zhao_ting_v
Reviewed-by: @kisnwang,@liangchenghui
Signed-off-by: @liangchenghui
This commit is contained in:
mindspore-ci-bot 2020-11-28 16:37:07 +08:00 committed by Gitee
parent 1752964365 2f92179d6b
commit 18f58f7db4
6 changed files with 40 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
mindspore/ccsrc/backend/kernel_compiler/cpu/adam_cpu_kernel.cc
View File

@ -88,11 +88,15 @@ bool AdamCPUKernel::Launch(const std::vector<kernel::AddressPtr> &inputs,
size_t thread_num = lens < 128 * max_thread_num ? std::ceil(lens / 128.0) : max_thread_num;
MS_LOG(INFO) << "Lens=" << lens << "; use thread_num=" << thread_num << "; max_thread_num: " << max_thread_num;
std::vector<std::thread> threads;
if (thread_num < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num;
return false;
}
threads.reserve(thread_num);
size_t start = 0;
size_t once_compute_size = (lens + thread_num - 1) / thread_num;
if (thread_num < 1 || once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num << "; once_compute_size " << once_compute_size;
if (once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: once_compute_size " << once_compute_size;
return false;
}
while (start < lens) {

16
mindspore/ccsrc/backend/kernel_compiler/cpu/arithmetic_cpu_kernel.cc
View File

@ -186,9 +186,17 @@ void ArithmeticCPUKernel::LaunchLess(const std::vector<AddressPtr> &inputs, cons
size_t thread_num = lens < 128 * max_thread_num ? std::ceil(lens / 128.0) : max_thread_num;
MS_LOG(INFO) << "Lens=" << lens << "; use thread_num=" << thread_num << "; max_thread_num: " << max_thread_num;
std::vector<std::thread> threads;
if (thread_num < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num;
return;
}
threads.reserve(thread_num);
size_t start = 0;
size_t once_compute_size = (lens + thread_num - 1) / thread_num;
if (once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: once_compute_size " << once_compute_size;
return;
}
while (start < lens) {
size_t end = (start + once_compute_size) > lens ? lens : (start + once_compute_size);
threads.emplace_back(std::thread(&ArithmeticCPUKernel::Less<T>, this, input1, input2, output, start, end));
@ -214,11 +222,15 @@ void ArithmeticCPUKernel::LaunchKernel(const std::vector<AddressPtr> &inputs, co
size_t thread_num = lens < 128 * max_thread_num ? std::ceil(lens / 128.0) : max_thread_num;
MS_LOG(INFO) << "Lens=" << lens << "; use thread_num=" << thread_num << "; max_thread_num: " << max_thread_num;
std::vector<std::thread> threads;
if (thread_num < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num;
return;
}
threads.reserve(thread_num);
size_t start = 0;
size_t once_compute_size = (lens + thread_num - 1) / thread_num;
if (thread_num < 1 || once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num << "; once_compute_size " << once_compute_size;
if (once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: once_compute_size " << once_compute_size;
return;
}
while (start < lens) {

8
mindspore/ccsrc/backend/kernel_compiler/cpu/arithmetic_self_cpu_kernel.cc
View File

@ -72,11 +72,15 @@ void ArithmeticSelfCPUKernel::LaunchKernel(const std::vector<AddressPtr> &inputs
size_t thread_num = lens < 128 * max_thread_num ? std::ceil(lens / 128.0) : max_thread_num;
MS_LOG(INFO) << "Lens=" << lens << "; use thread_num=" << thread_num << "; max_thread_num: " << max_thread_num;
std::vector<std::thread> threads;
if (thread_num < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num;
return;
}
threads.reserve(thread_num);
size_t start = 0;
size_t once_compute_size = (lens + thread_num - 1) / thread_num;
if (thread_num < 1 || once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num << "; once_compute_size " << once_compute_size;
if (once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: once_compute_size " << once_compute_size;
return;
}
while (start < lens) {

8
mindspore/ccsrc/backend/kernel_compiler/cpu/cast_cpu_kernel.cc
View File

@ -40,11 +40,15 @@ void LaunchCast(const std::vector<kernel::AddressPtr> &inputs, const std::vector
size_t thread_num = lens < 128 * max_thread_num ? std::ceil(lens / 128.0) : max_thread_num;
MS_LOG(INFO) << "Lens=" << lens << "; use thread_num=" << thread_num << "; max_thread_num: " << max_thread_num;
std::vector<std::thread> threads;
if (thread_num < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num;
return;
}
threads.reserve(thread_num);
size_t start = 0;
size_t once_compute_size = (lens + thread_num - 1) / thread_num;
if (thread_num < 1 || once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num << "; once_compute_size " << once_compute_size;
if (once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: once_compute_size " << once_compute_size;
return;
}
while (start < lens) {

8
mindspore/ccsrc/backend/kernel_compiler/cpu/eltwise_grad_cpu_kernel.cc
View File

@ -146,11 +146,15 @@ void EltWiseGradCPUKernel::LaunchKernel(const std::vector<AddressPtr> &inputs, c
size_t thread_num = lens < 128 * max_thread_num ? std::ceil(lens / 128.0) : max_thread_num;
MS_LOG(INFO) << "Lens=" << lens << "; use thread_num=" << thread_num << "; max_thread_num: " << max_thread_num;
std::vector<std::thread> threads;
if (thread_num < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num;
return;
}
threads.reserve(thread_num);
size_t start = 0;
size_t once_compute_size = (lens + thread_num - 1) / thread_num;
if (thread_num < 1 || once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: thread_num " << thread_num << "; once_compute_size " << once_compute_size;
if (once_compute_size < 1) {
MS_LOG(ERROR) << "Invalid value: once_compute_size " << once_compute_size;
return;
}
while (start < lens) {

4
mindspore/ccsrc/backend/kernel_compiler/cpu/mkldnn/pooling_grad_cpu_kernel.cc
View File

@ -83,7 +83,7 @@ void PoolingGradCPUKernel::ChannelPoolingGrad(const float *input, const float *d
size_t diff_index = 0;
for (size_t h = 0; h < dst_shape_[2]; ++h) {
box[0].first = IntToSize(std::max(h_start, 0));
box[0].second = IntToSize(std::min(h_start + SizeToInt(kernel_size_[1]), src_height));
box[0].second = IntToSize(std::min(h_start + SizeToInt(kernel_size_[0]), src_height));
for (size_t w = 0; w < src_shape_[3]; ++w) {
row_max_pair[w].first = 0;
row_max_pair[w].second = 0;
@ -91,7 +91,7 @@ void PoolingGradCPUKernel::ChannelPoolingGrad(const float *input, const float *d
int w_start = -padding_l_[1];
for (size_t w = 0; w < dst_shape_[3]; ++w) {
box[1].first = IntToSize(std::max(w_start, 0));
box[1].second = IntToSize(std::min(w_start + SizeToInt(kernel_size_[0]), src_width));
box[1].second = IntToSize(std::min(w_start + SizeToInt(kernel_size_[1]), src_width));
RowPoolingGrad(input, output, diff[diff_index], box, &row_max_pair);
diff_index += 1;
w_start += stride_;