GitOrigin-RevId: 27ed93e4c1
release-1.10
@@ -12,8 +12,8 @@ | |||
#include "src/aarch64/conv_bias/fp16/algos.h" | |||
#include "src/aarch64/conv_bias/fp16/stride2_kern.h" | |||
#include "src/arm_common/conv_bias/direct/multi_thread_common.h" | |||
#include "src/arm_common/conv_bias/postprocess_helper.h" | |||
#include "src/fallback/conv_bias/direct/multi_thread_common.h" | |||
using namespace megdnn; | |||
using namespace aarch64; | |||
@@ -43,7 +43,7 @@ size_t ConvBiasImpl::AlgoF16DirectStride2::get_workspace( | |||
const NCBKernSizeParam& param) const { | |||
MIDOUT_BEGIN(megdnn_aarch64_conv_bias_stride2_conv2357_fp16, 0, 1) { | |||
bool large_group = param.filter_meta.group >= param.nr_threads; | |||
auto wbundle = arm_common::MultithreadDirectConvCommon< | |||
auto wbundle = fallback::MultithreadDirectConvCommon< | |||
dt_float16, __fp16>::get_bundle_stride(param, large_group); | |||
return wbundle.total_size_in_bytes(); | |||
} | |||
@@ -83,7 +83,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16DirectStride2::get_kimpl | |||
conv = fp16::conv_stride2::do_conv_7x7_stride2; | |||
} | |||
WorkspaceBundle bundle = arm_common::MultithreadDirectConvCommon< | |||
WorkspaceBundle bundle = fallback::MultithreadDirectConvCommon< | |||
dt_float16, __fp16>::get_bundle_stride(param, large_group); | |||
SmallVector<NCBKern> ret_kerns; | |||
@@ -98,13 +98,13 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16DirectStride2::get_kimpl | |||
size_t OC = fm.ocpg; | |||
bundle.set(kern_param.workspace_ptr); | |||
for (size_t ic = 0; ic < IC; ic++) { | |||
arm_common::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, | |||
{ncb_index.thread_id, 0, ic}); | |||
} | |||
for (size_t oc = 0; oc < OC; oc++) { | |||
arm_common::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv, | |||
{ncb_index.thread_id, 0, oc}); | |||
@@ -116,7 +116,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16DirectStride2::get_kimpl | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
arm_common::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
}; | |||
@@ -125,7 +125,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16DirectStride2::get_kimpl | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
arm_common::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv, ncb_index.ndrange_id); | |||
}; | |||
@@ -11,9 +11,9 @@ | |||
#include "src/aarch64/conv_bias/fp32/algos.h" | |||
#include "src/aarch64/conv_bias/fp32/stride2_kern.h" | |||
#include "src/arm_common/conv_bias/direct/multi_thread_common.h" | |||
#include "src/arm_common/conv_bias/postprocess_helper.h" | |||
#include "src/fallback/conv_bias/common.h" | |||
#include "src/fallback/conv_bias/direct/multi_thread_common.h" | |||
#include "midout.h" | |||
@@ -42,8 +42,9 @@ size_t ConvBiasImpl::AlgoF32DirectStride2::get_workspace( | |||
const NCBKernSizeParam& param) const { | |||
MIDOUT_BEGIN(megdnn_aarch64_conv_bias_stride2_conv2357_fp32, 0, 1) { | |||
bool large_group = param.filter_meta.group >= param.nr_threads; | |||
auto wbundle = arm_common::MultithreadDirectConvCommon< | |||
float, float>::get_bundle_stride(param, large_group); | |||
auto wbundle = | |||
fallback::MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
param, large_group); | |||
return wbundle.total_size_in_bytes(); | |||
} | |||
MIDOUT_END(); | |||
@@ -82,7 +83,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32DirectStride2::get_kimpl | |||
} | |||
WorkspaceBundle bundle = | |||
arm_common::MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
fallback::MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
param, large_group); | |||
SmallVector<NCBKern> ret_kerns; | |||
@@ -97,13 +98,13 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32DirectStride2::get_kimpl | |||
size_t OC = fm.ocpg; | |||
bundle.set(kern_param.workspace_ptr); | |||
for (size_t ic = 0; ic < IC; ic++) { | |||
arm_common::MultithreadDirectConvCommon<float, float>:: | |||
fallback::MultithreadDirectConvCommon<float, float>:: | |||
copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, | |||
{ncb_index.thread_id, 0, ic}); | |||
} | |||
for (size_t oc = 0; oc < OC; oc++) { | |||
arm_common::MultithreadDirectConvCommon<float, float>:: | |||
fallback::MultithreadDirectConvCommon<float, float>:: | |||
do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv, | |||
{ncb_index.thread_id, 0, oc}); | |||
@@ -115,7 +116,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32DirectStride2::get_kimpl | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
arm_common::MultithreadDirectConvCommon<float, float>:: | |||
fallback::MultithreadDirectConvCommon<float, float>:: | |||
copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
}; | |||
@@ -124,7 +125,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32DirectStride2::get_kimpl | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
arm_common::MultithreadDirectConvCommon<float, float>::do_conv_kern_stride( | |||
fallback::MultithreadDirectConvCommon<float, float>::do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv, ncb_index.ndrange_id); | |||
}; | |||
ret_kerns.push_back({do_conv, {group, N, OC}}); | |||
@@ -10,7 +10,6 @@ | |||
*/ | |||
#include "src/arm_common/conv_bias/f16/algos.h" | |||
#include "src/arm_common/conv_bias/direct/multi_thread_common.h" | |||
#include "src/arm_common/conv_bias/f16/direct.h" | |||
#include "src/arm_common/conv_bias/f16/do_conv_stride1.h" | |||
#include "src/arm_common/conv_bias/f16/strategy.h" | |||
@@ -18,6 +17,7 @@ | |||
#include "src/arm_common/conv_bias/postprocess_helper.h" | |||
#include "src/common/opr_delegate.h" | |||
#include "src/fallback/conv_bias/common.h" | |||
#include "src/fallback/conv_bias/direct/multi_thread_common.h" | |||
#if __ARM_FEATURE_FP16_VECTOR_ARITHMETIC | |||
#include "midout.h" | |||
MIDOUT_DECL(megdnn_arm_common_winograd_fp16) | |||
@@ -187,8 +187,9 @@ bool ConvBiasImpl::AlgoF16Direct::usable( | |||
size_t ConvBiasImpl::AlgoF16Direct::get_workspace(const NCBKernSizeParam& param) const { | |||
MIDOUT_BEGIN(megdnn_arm_common_conv_bias_fp16_kimpl, 0, 1) { | |||
bool large_group = param.filter_meta.group >= param.nr_threads; | |||
auto wbundle = MultithreadDirectConvCommon<dt_float16, __fp16>::get_bundle( | |||
param, large_group); | |||
auto wbundle = | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>::get_bundle( | |||
param, large_group); | |||
return wbundle.total_size_in_bytes(); | |||
} | |||
MIDOUT_END(); | |||
@@ -204,7 +205,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16Direct::get_kimpls( | |||
size_t group = fm.group; | |||
bool large_group = group >= param.nr_threads; | |||
WorkspaceBundle bundle = | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::get_bundle( | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>::get_bundle( | |||
param, large_group); | |||
SmallVector<NCBKern> ret_kerns; | |||
//! When group >= nr_threads, treat it as large_group, each thread process | |||
@@ -220,17 +221,20 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16Direct::get_kimpls( | |||
bundle.set(kern_param.workspace_ptr); | |||
if (fm.should_flip) { | |||
for (size_t oc = 0; oc < OC; oc++) { | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::weight_flip_kern( | |||
bundle, kern_param, ncb_index, | |||
{ncb_index.thread_id, 0, oc}); | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
weight_flip_kern( | |||
bundle, kern_param, ncb_index, | |||
{ncb_index.thread_id, 0, oc}); | |||
} | |||
} | |||
for (size_t ic = 0; ic < IC; ic++) { | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::copy_padding_kern( | |||
bundle, kern_param, ncb_index, {ncb_index.thread_id, 0, ic}); | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
copy_padding_kern( | |||
bundle, kern_param, ncb_index, | |||
{ncb_index.thread_id, 0, ic}); | |||
} | |||
for (size_t oc = 0; oc < OC; oc++) { | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::do_conv_kern( | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>::do_conv_kern( | |||
bundle, kern_param, ncb_index, fp16::conv_bias::kern_direct_f16, | |||
{ncb_index.thread_id, 0, oc}); | |||
} | |||
@@ -242,8 +246,9 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16Direct::get_kimpls( | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::weight_flip_kern( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
weight_flip_kern( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
}; | |||
ret_kerns.push_back({weight_flip, {group, 1_z, OC}}); | |||
} | |||
@@ -251,15 +256,16 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16Direct::get_kimpls( | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::copy_padding_kern( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
copy_padding_kern( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
}; | |||
ret_kerns.push_back({copy_padding, {group, N, IC}}); | |||
auto do_conv = [bundle]( | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::do_conv_kern( | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>::do_conv_kern( | |||
bundle, kern_param, ncb_index, fp16::conv_bias::kern_direct_f16, | |||
ncb_index.ndrange_id); | |||
}; | |||
@@ -324,9 +330,8 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16DirectStride1::get_kimpl | |||
} | |||
SWITCH_KERN(); | |||
WorkspaceBundle bundle = | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::get_bundle_stride( | |||
param, large_group); | |||
WorkspaceBundle bundle = fallback::MultithreadDirectConvCommon< | |||
dt_float16, __fp16>::get_bundle_stride(param, large_group); | |||
SmallVector<NCBKern> ret_kerns; | |||
//! When group >= nr_threads, treat it as large_group, each thread process | |||
//! one group for better performance | |||
@@ -340,15 +345,16 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16DirectStride1::get_kimpl | |||
size_t OC = fm.ocpg; | |||
bundle.set(kern_param.workspace_ptr); | |||
for (size_t ic = 0; ic < IC; ic++) { | |||
MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, | |||
{ncb_index.thread_id, 0, ic}); | |||
} | |||
for (size_t oc = 0; oc < OC; oc++) { | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv_kern_function, | |||
{ncb_index.thread_id, 0, oc}); | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv_kern_function, | |||
{ncb_index.thread_id, 0, oc}); | |||
} | |||
}; | |||
ret_kerns.push_back({exec_one_group, {group, N, 1_z}}); | |||
@@ -357,17 +363,19 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF16DirectStride1::get_kimpl | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
}; | |||
ret_kerns.push_back({copy_padding, {group, N, IC}}); | |||
auto do_conv = [bundle, conv_kern_function]( | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv_kern_function, | |||
ncb_index.ndrange_id); | |||
fallback::MultithreadDirectConvCommon<dt_float16, __fp16>:: | |||
do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv_kern_function, | |||
ncb_index.ndrange_id); | |||
}; | |||
ret_kerns.push_back({do_conv, {group, N, OC}}); | |||
} | |||
@@ -378,9 +386,8 @@ size_t ConvBiasImpl::AlgoF16DirectStride1::get_workspace( | |||
const NCBKernSizeParam& param) const { | |||
MIDOUT_BEGIN(megdnn_arm_common_conv_bias_fp16_kimpl, 1, 1) { | |||
bool large_group = param.filter_meta.group >= param.nr_threads; | |||
auto bundle = | |||
MultithreadDirectConvCommon<dt_float16, __fp16>::get_bundle_stride( | |||
param, large_group); | |||
auto bundle = fallback::MultithreadDirectConvCommon< | |||
dt_float16, __fp16>::get_bundle_stride(param, large_group); | |||
return bundle.total_size_in_bytes(); | |||
} | |||
MIDOUT_END(); | |||
@@ -11,7 +11,6 @@ | |||
*/ | |||
#include "src/arm_common/conv_bias/fp32/algos.h" | |||
#include "src/arm_common/conv_bias/direct/multi_thread_common.h" | |||
#include "src/arm_common/conv_bias/fp32/direct.h" | |||
#include "src/arm_common/conv_bias/fp32/do_conv_stride1.h" | |||
#include "src/arm_common/conv_bias/fp32/do_conv_stride2.h" | |||
@@ -20,6 +19,7 @@ | |||
#include "src/arm_common/conv_bias/postprocess_helper.h" | |||
#include "src/common/opr_delegate.h" | |||
#include "src/fallback/conv_bias/common.h" | |||
#include "src/fallback/conv_bias/direct/multi_thread_common.h" | |||
#include "midout.h" | |||
@@ -343,7 +343,7 @@ bool ConvBiasImpl::AlgoF32Direct::usable( | |||
size_t ConvBiasImpl::AlgoF32Direct::get_workspace(const NCBKernSizeParam& param) const { | |||
MIDOUT_BEGIN(megdnn_arm_common_conv_bias_f32_kimpl, 0, 1) { | |||
bool large_group = param.filter_meta.group >= param.nr_threads; | |||
auto wbundle = MultithreadDirectConvCommon<float, float>::get_bundle( | |||
auto wbundle = fallback::MultithreadDirectConvCommon<float, float>::get_bundle( | |||
param, large_group); | |||
return wbundle.total_size_in_bytes(); | |||
} | |||
@@ -359,7 +359,8 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32Direct::get_kimpls( | |||
size_t group = fm.group; | |||
bool large_group = group >= param.nr_threads; | |||
WorkspaceBundle bundle = | |||
MultithreadDirectConvCommon<float, float>::get_bundle(param, large_group); | |||
fallback::MultithreadDirectConvCommon<float, float>::get_bundle( | |||
param, large_group); | |||
SmallVector<NCBKern> ret_kerns; | |||
//! When group >= nr_threads, treat it as large_group, each thread process | |||
//! one group for better performance | |||
@@ -374,17 +375,18 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32Direct::get_kimpls( | |||
bundle.set(kern_param.workspace_ptr); | |||
if (fm.should_flip) { | |||
for (size_t oc = 0; oc < OC; oc++) { | |||
MultithreadDirectConvCommon<float, float>::weight_flip_kern( | |||
bundle, kern_param, ncb_index, | |||
{ncb_index.thread_id, 0, oc}); | |||
fallback::MultithreadDirectConvCommon<float, float>:: | |||
weight_flip_kern( | |||
bundle, kern_param, ncb_index, | |||
{ncb_index.thread_id, 0, oc}); | |||
} | |||
} | |||
for (size_t ic = 0; ic < IC; ic++) { | |||
MultithreadDirectConvCommon<float, float>::copy_padding_kern( | |||
fallback::MultithreadDirectConvCommon<float, float>::copy_padding_kern( | |||
bundle, kern_param, ncb_index, {ncb_index.thread_id, 0, ic}); | |||
} | |||
for (size_t oc = 0; oc < OC; oc++) { | |||
MultithreadDirectConvCommon<float, float>::do_conv_kern( | |||
fallback::MultithreadDirectConvCommon<float, float>::do_conv_kern( | |||
bundle, kern_param, ncb_index, fp32::conv_bias::kern_direct, | |||
{ncb_index.thread_id, 0, oc}); | |||
} | |||
@@ -396,7 +398,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32Direct::get_kimpls( | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<float, float>::weight_flip_kern( | |||
fallback::MultithreadDirectConvCommon<float, float>::weight_flip_kern( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
}; | |||
ret_kerns.push_back({weight_flip, {group, 1_z, OC}}); | |||
@@ -405,7 +407,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32Direct::get_kimpls( | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<float, float>::copy_padding_kern( | |||
fallback::MultithreadDirectConvCommon<float, float>::copy_padding_kern( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
}; | |||
ret_kerns.push_back({copy_padding, {group, N, IC}}); | |||
@@ -413,7 +415,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32Direct::get_kimpls( | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<float, float>::do_conv_kern( | |||
fallback::MultithreadDirectConvCommon<float, float>::do_conv_kern( | |||
bundle, kern_param, ncb_index, fp32::conv_bias::kern_direct, | |||
ncb_index.ndrange_id); | |||
}; | |||
@@ -452,8 +454,9 @@ size_t ConvBiasImpl::AlgoF32DirectStride1::get_workspace( | |||
const NCBKernSizeParam& param) const { | |||
MIDOUT_BEGIN(megdnn_arm_common_conv_bias_f32_kimpl, 1, 1) { | |||
bool large_group = param.filter_meta.group >= param.nr_threads; | |||
auto bundle = MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
param, large_group); | |||
auto bundle = | |||
fallback::MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
param, large_group); | |||
return bundle.total_size_in_bytes(); | |||
} | |||
MIDOUT_END(); | |||
@@ -492,7 +495,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32DirectStride1::get_kimpl | |||
SWITCH_KERN_STR1(); | |||
WorkspaceBundle bundle = | |||
MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
fallback::MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
param, large_group); | |||
SmallVector<NCBKern> ret_kerns; | |||
//! When group >= nr_threads, treat it as large_group, each thread process | |||
@@ -507,13 +510,16 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32DirectStride1::get_kimpl | |||
size_t OC = fm.ocpg; | |||
bundle.set(kern_param.workspace_ptr); | |||
for (size_t ic = 0; ic < IC; ic++) { | |||
MultithreadDirectConvCommon<float, float>::copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, {ncb_index.thread_id, 0, ic}); | |||
fallback::MultithreadDirectConvCommon<float, float>:: | |||
copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, | |||
{ncb_index.thread_id, 0, ic}); | |||
} | |||
for (size_t oc = 0; oc < OC; oc++) { | |||
MultithreadDirectConvCommon<float, float>::do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv_kern_function, | |||
{ncb_index.thread_id, 0, oc}); | |||
fallback::MultithreadDirectConvCommon<float, float>:: | |||
do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv_kern_function, | |||
{ncb_index.thread_id, 0, oc}); | |||
} | |||
}; | |||
ret_kerns.push_back({exec_one_group, {group, N, 1_z}}); | |||
@@ -522,15 +528,16 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32DirectStride1::get_kimpl | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<float, float>::copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
fallback::MultithreadDirectConvCommon<float, float>:: | |||
copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
}; | |||
ret_kerns.push_back({copy_padding, {group, N, IC}}); | |||
auto do_conv = [bundle, conv_kern_function]( | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<float, float>::do_conv_kern_stride( | |||
fallback::MultithreadDirectConvCommon<float, float>::do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv_kern_function, | |||
ncb_index.ndrange_id); | |||
}; | |||
@@ -570,8 +577,9 @@ size_t ConvBiasImpl::AlgoF32DirectStride2::get_workspace( | |||
const NCBKernSizeParam& param) const { | |||
MIDOUT_BEGIN(megdnn_arm_common_conv_bias_f32_kimpl, 2, 1) { | |||
bool large_group = param.filter_meta.group >= param.nr_threads; | |||
auto bundle = MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
param, large_group); | |||
auto bundle = | |||
fallback::MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
param, large_group); | |||
return bundle.total_size_in_bytes(); | |||
} | |||
MIDOUT_END(); | |||
@@ -609,7 +617,7 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32DirectStride2::get_kimpl | |||
SWITCH_KERN_STR2(); | |||
WorkspaceBundle bundle = | |||
MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
fallback::MultithreadDirectConvCommon<float, float>::get_bundle_stride( | |||
param, large_group); | |||
SmallVector<NCBKern> ret_kerns; | |||
//! When group >= nr_threads, treat it as large_group, each thread process | |||
@@ -624,13 +632,16 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32DirectStride2::get_kimpl | |||
size_t OC = fm.ocpg; | |||
bundle.set(kern_param.workspace_ptr); | |||
for (size_t ic = 0; ic < IC; ic++) { | |||
MultithreadDirectConvCommon<float, float>::copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, {ncb_index.thread_id, 0, ic}); | |||
fallback::MultithreadDirectConvCommon<float, float>:: | |||
copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, | |||
{ncb_index.thread_id, 0, ic}); | |||
} | |||
for (size_t oc = 0; oc < OC; oc++) { | |||
MultithreadDirectConvCommon<float, float>::do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv_kern_function, | |||
{ncb_index.thread_id, 0, oc}); | |||
fallback::MultithreadDirectConvCommon<float, float>:: | |||
do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv_kern_function, | |||
{ncb_index.thread_id, 0, oc}); | |||
} | |||
}; | |||
ret_kerns.push_back({exec_one_group, {group, N, 1_z}}); | |||
@@ -639,15 +650,16 @@ SmallVector<ConvBiasImpl::NCBKern> ConvBiasImpl::AlgoF32DirectStride2::get_kimpl | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<float, float>::copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
fallback::MultithreadDirectConvCommon<float, float>:: | |||
copy_padding_kern_stride( | |||
bundle, kern_param, ncb_index, ncb_index.ndrange_id); | |||
}; | |||
ret_kerns.push_back({copy_padding, {group, N, IC}}); | |||
auto do_conv = [bundle, conv_kern_function]( | |||
const NCBKernParam& kern_param, | |||
const NCBKernIndex& ncb_index) mutable { | |||
bundle.set(kern_param.workspace_ptr); | |||
MultithreadDirectConvCommon<float, float>::do_conv_kern_stride( | |||
fallback::MultithreadDirectConvCommon<float, float>::do_conv_kern_stride( | |||
bundle, kern_param, ncb_index, conv_kern_function, | |||
ncb_index.ndrange_id); | |||
}; | |||
@@ -1,5 +1,5 @@ | |||
/** | |||
* \file dnn/src/arm_common/conv_bias/direct/multi_thread_common.cpp | |||
* \file dnn/src/fallback/conv_bias/direct/multi_thread_common.cpp | |||
* MegEngine is Licensed under the Apache License, Version 2.0 (the "License") | |||
* | |||
* Copyright (c) 2014-2021 Megvii Inc. All rights reserved. | |||
@@ -9,12 +9,14 @@ | |||
* "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | |||
*/ | |||
#include "src/arm_common/conv_bias/direct/multi_thread_common.h" | |||
#include "src/arm_common/conv_bias/postprocess_helper.h" | |||
#include "multi_thread_common.h" | |||
#include "src/fallback/matrix_mul/opr_impl.h" | |||
using namespace megdnn; | |||
using namespace arm_common; | |||
using namespace fallback; | |||
#if MEGDNN_X86 | |||
using namespace x86; | |||
#endif | |||
namespace { | |||
bool need_dst_copy(const megdnn::fallback::ConvBiasImpl::NCBKernSizeParam& param) { | |||
@@ -354,8 +356,8 @@ void MultithreadDirectConvCommon<io_ctype, compute_ctype>::do_conv_kern_stride( | |||
kern_param.nonlineMode, kern_param.bias_type, kern_param.dst_type, 1_z, 1_z, | |||
OH, OW); | |||
}; | |||
template class megdnn::arm_common::MultithreadDirectConvCommon<float, float>; | |||
template class megdnn::fallback::MultithreadDirectConvCommon<float, float>; | |||
#if __ARM_FEATURE_FP16_VECTOR_ARITHMETIC | |||
template class megdnn::arm_common::MultithreadDirectConvCommon<dt_float16, __fp16>; | |||
template class megdnn::fallback::MultithreadDirectConvCommon<dt_float16, __fp16>; | |||
#endif | |||
// vim: syntax=cpp.doxygen |
@@ -1,5 +1,5 @@ | |||
/** | |||
* \file dnn/src/arm_common/conv_bias/direct/multi_thread_common.h | |||
* \file dnn/src/fallback/conv_bias/direct/multi_thread_common.h | |||
* MegEngine is Licensed under the Apache License, Version 2.0 (the "License") | |||
* | |||
* Copyright (c) 2014-2021 Megvii Inc. All rights reserved. | |||
@@ -10,11 +10,20 @@ | |||
*/ | |||
#pragma once | |||
#include "src/arm_common/conv_bias/opr_impl.h" | |||
#include "src/fallback/conv_bias/opr_impl.h" | |||
#include "src/fallback/matrix_mul/opr_impl.h" | |||
#if MEGDNN_X86 | |||
#include "src/x86/conv_bias/postprocess_helper.h" | |||
#elif (MEGDNN_ARMV7 || MEGDNN_AARCH64) | |||
#include "src/arm_common/conv_bias/postprocess_helper.h" | |||
#else | |||
//! TODO: optimize common postprocess_helper with general intrinsic | |||
#include "src/common/postprocess_helper.h" | |||
#endif | |||
namespace megdnn { | |||
namespace arm_common { | |||
namespace fallback { | |||
template <class io_ctype, class compute_ctype> | |||
class MultithreadDirectConvCommon { | |||
@@ -53,7 +62,7 @@ public: | |||
const CpuNDRange& workspace_ids); | |||
}; | |||
} // namespace arm_common | |||
} // namespace fallback | |||
} // namespace megdnn | |||
// vim: syntax=cpp.doxygen |
@@ -42,7 +42,7 @@ GI_FLOAT32_t GiReintInt32ToFloat32(GI_INT32_t Vector) { | |||
#elif defined(GI_SSE2_INTRINSICS) | |||
return _mm_castsi128_ps(Vector); | |||
#else | |||
return (GI_FLOAT32_t)In; | |||
return (GI_FLOAT32_t)Vector; | |||
#endif | |||
} | |||
@@ -53,7 +53,7 @@ GI_FLOAT32_t GiReintUint32ToFloat32(GI_UINT32_t Vector) { | |||
#elif defined(GI_SSE2_INTRINSICS) | |||
return _mm_castsi128_ps(Vector); | |||
#else | |||
return (GI_FLOAT32_t)In; | |||
return (GI_FLOAT32_t)Vector; | |||
#endif | |||
} | |||
@@ -1,5 +1,5 @@ | |||
/** | |||
* \file dnn/src/fallback/general_intrinsic/gi_float.h | |||
* \file dnn/src/fallback/general_intrinsic/gi_int.h | |||
* MegEngine is Licensed under the Apache License, Version 2.0 (the "License") | |||
* | |||
* Copyright (c) 2014-2022 Megvii Inc. All rights reserved. | |||