Skip to content

Commit

Permalink
Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…
Browse files Browse the repository at this point in the history 
… project_fft
  • Loading branch information
chenfeiyu committed Sep 17, 2021
2 parents 76401f4 + fcfb0af commit f8c2a2e
Show file tree
Hide file tree
Showing 23 changed files with 428 additions and 392 deletions.
2 changes: 1 addition & 1 deletion paddle/fluid/imperative/CMakeLists.txt
View file
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,7 @@ cc_library(amp SRCS amp_auto_cast.cc DEPS layer )
cc_library(tracer SRCS tracer.cc DEPS layer engine program_desc_tracer amp denormal)
cc_library(basic_engine SRCS basic_engine.cc DEPS layer gradient_accumulator)
cc_library(engine SRCS basic_engine.cc partial_grad_engine.cc DEPS layer gradient_accumulator)
cc_library(imperative_profiler SRCS profiler.cc)
cc_library(imperative_profiler SRCS profiler.cc DEPS flags)
if(NOT WIN32)
if(WITH_NCCL OR WITH_RCCL)
cc_library(imperative_all_reduce SRCS all_reduce.cc DEPS collective_helper device_context selected_rows tensor)
Expand Down
8 changes: 4 additions & 4 deletions paddle/fluid/imperative/flags.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -13,11 +13,11 @@
// limitations under the License.

#include "paddle/fluid/imperative/flags.h"
#include "gflags/gflags.h"
#include "paddle/fluid/platform/flags.h"

DEFINE_uint64(dygraph_debug, 0,
"Debug level of dygraph. This flag is not "
"open to users");
PADDLE_DEFINE_EXPORTED_uint64(dygraph_debug, 0,
"Debug level of dygraph. This flag is not "
"open to users");

namespace paddle {
namespace imperative {
Expand Down
4 changes: 2 additions & 2 deletions paddle/fluid/imperative/profiler.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -19,9 +19,9 @@
#endif
#include <glog/logging.h>
#include <mutex> // NOLINT
#include "gflags/gflags.h"
#include "paddle/fluid/platform/flags.h"

DEFINE_string(
PADDLE_DEFINE_EXPORTED_string(
tracer_profile_fname, "xxgperf",
"Profiler filename for imperative tracer, which generated by gperftools."
"Only valid when compiled `WITH_PROFILER=ON`. Empty if disable.");
Expand Down
2 changes: 1 addition & 1 deletion paddle/fluid/memory/allocation/CMakeLists.txt
View file
Original file line number Diff line number Diff line change
Expand Up @@ -99,7 +99,7 @@ cc_test(allocator_facade_abs_flags_test SRCS allocator_facade_abs_flags_test.cc

cc_test(allocator_facade_frac_flags_test SRCS allocator_facade_frac_flags_test.cc DEPS allocator_facade)

cc_library(auto_growth_best_fit_allocator SRCS auto_growth_best_fit_allocator.cc DEPS allocator aligned_allocator)
cc_library(auto_growth_best_fit_allocator SRCS auto_growth_best_fit_allocator.cc DEPS allocator aligned_allocator flags)
cc_test(auto_growth_best_fit_allocator_facade_test SRCS auto_growth_best_fit_allocator_facade_test.cc DEPS cpu_allocator auto_growth_best_fit_allocator)
cc_test(auto_growth_best_fit_allocator_test SRCS auto_growth_best_fit_allocator_test.cc DEPS auto_growth_best_fit_allocator)

Expand Down
9 changes: 5 additions & 4 deletions paddle/fluid/memory/allocation/allocator_facade.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -37,14 +37,15 @@
#endif
#include "paddle/fluid/platform/npu_info.h"

DEFINE_int64(
PADDLE_DEFINE_EXPORTED_int64(
gpu_allocator_retry_time, 10000,
"The retry time (milliseconds) when allocator fails "
"to allocate memory. No retry if this value is not greater than 0");

DEFINE_bool(use_system_allocator, false,
"Whether to use system allocator to allocate CPU and GPU memory. "
"Only used for unittests.");
PADDLE_DEFINE_EXPORTED_bool(
use_system_allocator, false,
"Whether to use system allocator to allocate CPU and GPU memory. "
"Only used for unittests.");

namespace paddle {
namespace memory {
Expand Down
27 changes: 15 additions & 12 deletions paddle/fluid/memory/allocation/auto_growth_best_fit_allocator.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,18 +17,21 @@
#include <algorithm>
#include <mutex> // NOLINT
#include "paddle/fluid/memory/allocation/aligned_allocator.h"

DEFINE_bool(free_idle_chunk, false,
"Whether to free idle chunk when each allocation is freed. "
"If false, all freed allocation would be cached to speed up next "
"allocation request. If true, no allocation would be cached. This "
"flag only works when FLAGS_allocator_strategy=auto_growth.");

DEFINE_bool(free_when_no_cache_hit, false,
"Whether to free idle chunks when no cache hit. If true, idle "
"chunk would be freed when no cache hit; if false, idle "
"chunk would be freed when out of memory occurs. This flag "
"only works when FLAGS_allocator_strategy=auto_growth.");
#include "paddle/fluid/platform/flags.h"

PADDLE_DEFINE_EXPORTED_READONLY_bool(
free_idle_chunk, false,
"Whether to free idle chunk when each allocation is freed. "
"If false, all freed allocation would be cached to speed up next "
"allocation request. If true, no allocation would be cached. This "
"flag only works when FLAGS_allocator_strategy=auto_growth.");

PADDLE_DEFINE_EXPORTED_READONLY_bool(
free_when_no_cache_hit, false,
"Whether to free idle chunks when no cache hit. If true, idle "
"chunk would be freed when no cache hit; if false, idle "
"chunk would be freed when out of memory occurs. This flag "
"only works when FLAGS_allocator_strategy=auto_growth.");

namespace paddle {
namespace memory {
Expand Down
13 changes: 7 additions & 6 deletions paddle/fluid/memory/allocation/naive_best_fit_allocator.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -34,12 +34,13 @@
#include "paddle/fluid/platform/xpu/xpu_header.h"
#endif

DEFINE_bool(init_allocated_mem, false,
"It is a mistake that the values of the memory allocated by "
"BuddyAllocator are always zeroed in some op's implementation. "
"To find this error in time, we use init_allocated_mem to indicate "
"that initializing the allocated memory with a small value "
"during unit testing.");
PADDLE_DEFINE_EXPORTED_bool(
init_allocated_mem, false,
"It is a mistake that the values of the memory allocated by "
"BuddyAllocator are always zeroed in some op's implementation. "
"To find this error in time, we use init_allocated_mem to indicate "
"that initializing the allocated memory with a small value "
"during unit testing.");
DECLARE_double(fraction_of_gpu_memory_to_use);
DECLARE_uint64(initial_gpu_memory_in_mb);
DECLARE_uint64(reallocate_gpu_memory_in_mb);
Expand Down
3 changes: 2 additions & 1 deletion paddle/fluid/operators/pscore/heter_listen_and_serv_op.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,8 @@ limitations under the License. */
#include "paddle/fluid/operators/pscore/heter_listen_and_serv_op.h"
#include "paddle/fluid/framework/op_registry.h"

DEFINE_int32(rpc_send_thread_num, 12, "number of threads for rpc send");
PADDLE_DEFINE_EXPORTED_int32(rpc_send_thread_num, 12,
"number of threads for rpc send");

namespace paddle {
namespace operators {
Expand Down
30 changes: 15 additions & 15 deletions paddle/fluid/operators/reshape_op.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -249,11 +249,11 @@ class ReshapeOp : public framework::OperatorWithKernel {
framework::OperatorWithKernel::IndicateVarDataType(ctx, "X");

#ifdef PADDLE_WITH_MKLDNN
if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
return framework::OpKernelType(input_data_type, ctx.GetPlace(),
framework::DataLayout::kMKLDNN,
framework::LibraryType::kMKLDNN);
}
// if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
// return framework::OpKernelType(input_data_type, ctx.GetPlace(),
// framework::DataLayout::kMKLDNN,
// framework::LibraryType::kMKLDNN);
// }
#endif
return framework::OpKernelType(input_data_type, ctx.GetPlace());
}
Expand Down Expand Up @@ -367,11 +367,11 @@ class ReshapeGradOp : public framework::OperatorWithKernel {
framework::OperatorWithKernel::IndicateVarDataType(ctx, "X");

#ifdef PADDLE_WITH_MKLDNN
if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
return framework::OpKernelType(input_data_type, ctx.GetPlace(),
framework::DataLayout::kMKLDNN,
framework::LibraryType::kMKLDNN);
}
// if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
// return framework::OpKernelType(input_data_type, ctx.GetPlace(),
// framework::DataLayout::kMKLDNN,
// framework::LibraryType::kMKLDNN);
// }
#endif
return framework::OpKernelType(input_data_type, ctx.GetPlace());
}
Expand Down Expand Up @@ -558,11 +558,11 @@ class Reshape2GradOp : public framework::OperatorWithKernel {
ctx, framework::GradVarName("Out"));

#ifdef PADDLE_WITH_MKLDNN
if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
return framework::OpKernelType(input_data_type, ctx.GetPlace(),
framework::DataLayout::kMKLDNN,
framework::LibraryType::kMKLDNN);
}
// if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
// return framework::OpKernelType(input_data_type, ctx.GetPlace(),
// framework::DataLayout::kMKLDNN,
// framework::LibraryType::kMKLDNN);
// }
#endif
return framework::OpKernelType(input_data_type, ctx.GetPlace());
}
Expand Down
40 changes: 20 additions & 20 deletions paddle/fluid/operators/squeeze_op.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -114,11 +114,11 @@ class SqueezeOp : public framework::OperatorWithKernel {
framework::OperatorWithKernel::IndicateVarDataType(ctx, "X");

#ifdef PADDLE_WITH_MKLDNN
if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
return framework::OpKernelType(input_data_type, ctx.GetPlace(),
framework::DataLayout::kMKLDNN,
framework::LibraryType::kMKLDNN);
}
// if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
// return framework::OpKernelType(input_data_type, ctx.GetPlace(),
// framework::DataLayout::kMKLDNN,
// framework::LibraryType::kMKLDNN);
// }
#endif
return framework::OpKernelType(input_data_type, ctx.GetPlace());
}
Expand All @@ -141,11 +141,11 @@ class SqueezeGradOp : public framework::OperatorWithKernel {
ctx, framework::GradVarName("Out"));

#ifdef PADDLE_WITH_MKLDNN
if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
return framework::OpKernelType(input_data_type, ctx.GetPlace(),
framework::DataLayout::kMKLDNN,
framework::LibraryType::kMKLDNN);
}
// if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
// return framework::OpKernelType(input_data_type, ctx.GetPlace(),
// framework::DataLayout::kMKLDNN,
// framework::LibraryType::kMKLDNN);
// }
#endif
return framework::OpKernelType(input_data_type, ctx.GetPlace());
}
Expand Down Expand Up @@ -242,11 +242,11 @@ class Squeeze2Op : public framework::OperatorWithKernel {
framework::OperatorWithKernel::IndicateVarDataType(ctx, "X");

#ifdef PADDLE_WITH_MKLDNN
if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
return framework::OpKernelType(input_data_type, ctx.GetPlace(),
framework::DataLayout::kMKLDNN,
framework::LibraryType::kMKLDNN);
}
// if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
// return framework::OpKernelType(input_data_type, ctx.GetPlace(),
// framework::DataLayout::kMKLDNN,
// framework::LibraryType::kMKLDNN);
// }
#endif
return framework::OpKernelType(input_data_type, ctx.GetPlace());
}
Expand Down Expand Up @@ -288,11 +288,11 @@ class Squeeze2GradOp : public framework::OperatorWithKernel {
ctx, framework::GradVarName("Out"));

#ifdef PADDLE_WITH_MKLDNN
if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
return framework::OpKernelType(input_data_type, ctx.GetPlace(),
framework::DataLayout::kMKLDNN,
framework::LibraryType::kMKLDNN);
}
// if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
// return framework::OpKernelType(input_data_type, ctx.GetPlace(),
// framework::DataLayout::kMKLDNN,
// framework::LibraryType::kMKLDNN);
// }
#endif
return framework::OpKernelType(input_data_type, ctx.GetPlace());
}
Expand Down
4 changes: 2 additions & 2 deletions paddle/fluid/platform/CMakeLists.txt
View file
Original file line number Diff line number Diff line change
Expand Up @@ -37,13 +37,13 @@ if (WITH_PYTHON)
endif(NOT WIN32)
endif()

cc_library(flags SRCS flags.cc DEPS gflags)
cc_library(flags SRCS flags.cc DEPS gflags boost)
cc_library(denormal SRCS denormal.cc DEPS)

cc_library(errors SRCS errors.cc DEPS error_codes_proto)
cc_test(errors_test SRCS errors_test.cc DEPS errors enforce)

set(enforce_deps flags errors boost)
set(enforce_deps flags errors boost flags)
if(WITH_GPU)
set(enforce_deps ${enforce_deps} external_error_proto)
endif()
Expand Down
9 changes: 6 additions & 3 deletions paddle/fluid/platform/cpu_info.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@ limitations under the License. */
#endif // _WIN32

#include <algorithm>
#include "gflags/gflags.h"
#include "paddle/fluid/platform/flags.h"

DECLARE_double(fraction_of_cpu_memory_to_use);
DECLARE_uint64(initial_cpu_memory_in_mb);
Expand All @@ -42,7 +42,8 @@ DECLARE_double(fraction_of_cuda_pinned_memory_to_use);
// between host and device. Allocates too much would reduce the amount
// of memory available to the system for paging. So, by default, we
// should set false to use_pinned_memory.
DEFINE_bool(use_pinned_memory, true, "If set, allocate cpu pinned memory.");
PADDLE_DEFINE_EXPORTED_bool(use_pinned_memory, true,
"If set, allocate cpu pinned memory.");

namespace paddle {
namespace platform {
Expand All @@ -54,7 +55,9 @@ size_t CpuTotalPhysicalMemory() {
mib[1] = HW_MEMSIZE;
int64_t size = 0;
size_t len = sizeof(size);
if (sysctl(mib, 2, &size, &len, NULL, 0) == 0) return (size_t)size;
if (sysctl(mib, 2, &size, &len, NULL, 0) == 0) {
return static_cast<size_t>(size);
}
return 0L;
#elif defined(_WIN32)
MEMORYSTATUSEX sMeminfo;
Expand Down
1 change: 1 addition & 0 deletions paddle/fluid/platform/enforce.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -101,6 +101,7 @@ limitations under the License. */
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
#include "paddle/fluid/platform/type_defs.h"
#endif
#include "paddle/fluid/platform/flags.h"

namespace paddle {
namespace platform {
Expand Down
Loading

1 comment on commit f8c2a2e

@paddle-bot-old
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

✅Congratulation! Your pull request passed all required CI. You could ask reviewer(s) to approve and merge. 🎉

Please sign in to comment.