Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
飞不过沧海
Paddle
提交
5303b66b
P
Paddle
项目概览
飞不过沧海
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
1
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
0
列表
看板
标记
里程碑
合并请求
0
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
0
Issue
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
未验证
提交
5303b66b
编写于
10月 12, 2022
作者:
L
Leo Chen
提交者:
GitHub
10月 12, 2022
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
clean code of interpretercore (#46891)
* refactor * refine code
上级
21fab90d
变更
6
隐藏空白更改
内联
并排
Showing
6 changed file
with
170 addition
and
204 deletion
+170
-204
paddle/fluid/framework/new_executor/data_transfer.cc
paddle/fluid/framework/new_executor/data_transfer.cc
+2
-2
paddle/fluid/framework/new_executor/interpretercore.cc
paddle/fluid/framework/new_executor/interpretercore.cc
+93
-108
paddle/fluid/framework/new_executor/interpretercore.h
paddle/fluid/framework/new_executor/interpretercore.h
+29
-31
paddle/fluid/framework/new_executor/interpretercore_util.cc
paddle/fluid/framework/new_executor/interpretercore_util.cc
+31
-45
paddle/fluid/framework/new_executor/interpretercore_util.h
paddle/fluid/framework/new_executor/interpretercore_util.h
+14
-17
paddle/fluid/framework/new_executor/new_executor_defs.h
paddle/fluid/framework/new_executor/new_executor_defs.h
+1
-1
未找到文件。
paddle/fluid/framework/new_executor/data_transfer.cc
浏览文件 @
5303b66b
...
...
@@ -378,7 +378,7 @@ std::shared_ptr<OperatorBase> TransferDevice(const std::string& var_name,
"Required src_place shall be different with dst_place, "
"but received same place: %s"
,
src_place
));
if
(
IsSupportedHetePlace
(
dst_place
))
{
if
(
IsSupportedHete
r
Place
(
dst_place
))
{
op_type
=
kMemcpyH2D
;
int
dst_place_type
=
platform
::
is_gpu_place
(
dst_place
)
?
0
:
platform
::
is_npu_place
(
dst_place
)
?
1
...
...
@@ -387,7 +387,7 @@ std::shared_ptr<OperatorBase> TransferDevice(const std::string& var_name,
:
platform
::
is_custom_place
(
dst_place
)
?
6
:
-
1
;
attr_map
=
{{
"dst_place_type"
,
dst_place_type
}};
}
else
if
(
IsSupportedHetePlace
(
src_place
))
{
}
else
if
(
IsSupportedHete
r
Place
(
src_place
))
{
op_type
=
kMemcpyD2H
;
int
dst_place_type
=
platform
::
is_cpu_place
(
dst_place
)
?
0
:
platform
::
is_cuda_pinned_place
(
dst_place
)
?
1
...
...
paddle/fluid/framework/new_executor/interpretercore.cc
浏览文件 @
5303b66b
...
...
@@ -57,6 +57,50 @@ constexpr const char* kTaskCompletion = "TaskCompletion";
namespace
paddle
{
namespace
framework
{
inline
void
SetDeviceId
(
const
platform
::
Place
&
place
)
{
// TODO(zhiqiu): reduce the cost
if
(
platform
::
is_gpu_place
(
place
))
{
#if !defined(PADDLE_WITH_CUDA) && !defined(PADDLE_WITH_HIP)
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Cannot run operator on place %s, please recompile paddle or "
"reinstall Paddle with CUDA support."
,
place
));
#else
auto
dev_id
=
place
.
device
;
platform
::
SetDeviceId
(
dev_id
);
#endif
}
else
if
(
platform
::
is_xpu_place
(
place
))
{
#ifndef PADDLE_WITH_XPU
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Cannot run operator on place %s, please recompile paddle or "
"reinstall Paddle with XPU support."
,
place
));
#else
auto
dev_id
=
place
.
device
;
platform
::
SetXPUDeviceId
(
dev_id
);
#endif
}
else
if
(
platform
::
is_npu_place
(
place
))
{
#ifndef PADDLE_WITH_ASCEND_CL
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Cannot run operator on place %s, please recompile paddle or "
"reinstall Paddle with NPU support."
,
place
));
#else
auto
dev_id
=
place
.
device
;
platform
::
SetNPUDeviceId
(
dev_id
);
#endif
}
else
if
(
platform
::
is_custom_place
(
place
))
{
#ifndef PADDLE_WITH_CUSTOM_DEVICE
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Cannot run operator on place %s, please recompile paddle or "
"reinstall Paddle with CustomDevice support."
,
place
));
#else
phi
::
DeviceManager
::
SetDevice
(
place
);
#endif
}
}
// TODO(Ruibia): Pass skip_gc_vars, used_for_jit, and other config messages by
// constructing an interpreter::ExecutionConfig
InterpreterCore
::
InterpreterCore
(
const
platform
::
Place
&
place
,
...
...
@@ -71,8 +115,6 @@ InterpreterCore::InterpreterCore(const platform::Place& place,
stream_analyzer_
(
place
)
{
VLOG
(
4
)
<<
"InterpreterCore(): "
<<
this
<<
" on "
<<
place_
;
is_build_
=
false
;
exception_notifier_
=
main_thread_blocker_
.
RegisterEvent
(
kExceptionCaught
);
completion_notifier_
=
main_thread_blocker_
.
RegisterEvent
(
kTaskCompletion
);
...
...
@@ -87,12 +129,6 @@ InterpreterCore::InterpreterCore(const platform::Place& place,
local_scope_
=
local_scope
;
}
var_scope_
.
SetLocalScope
(
local_scope_
);
// prune
// optmize graph pass
// convert to run graph
}
InterpreterCore
::~
InterpreterCore
()
{
...
...
@@ -111,11 +147,8 @@ InterpreterCore::~InterpreterCore() {
interpreter
::
CostInfo
InterpreterCore
::
DryRun
(
const
std
::
vector
<
std
::
string
>&
feed_names
,
const
std
::
vector
<
phi
::
DenseTensor
>&
feed_tensors
)
{
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
if
(
platform
::
is_gpu_place
(
place_
))
{
platform
::
SetDeviceId
(
place_
.
device
);
}
#endif
SetDeviceId
(
place_
);
Prepare
(
feed_names
,
feed_tensors
,
true
);
interpreter
::
CostInfo
cost_info
;
{
...
...
@@ -135,7 +168,7 @@ interpreter::CostInfo InterpreterCore::DryRun(
platform
::
DeviceContextPool
::
Instance
().
Get
(
place_
)
->
Wait
();
}
if
(
execution_config_
.
create_local_scope
)
{
if
(
HasLocalScope
()
)
{
ClearLoDTensorArrayInLocalScope
();
}
...
...
@@ -145,11 +178,7 @@ interpreter::CostInfo InterpreterCore::DryRun(
paddle
::
framework
::
FetchList
InterpreterCore
::
Run
(
const
std
::
vector
<
std
::
string
>&
feed_names
,
const
std
::
vector
<
phi
::
DenseTensor
>&
feed_tensors
)
{
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
if
(
platform
::
is_gpu_place
(
place_
))
{
platform
::
SetDeviceId
(
place_
.
device
);
}
#endif
SetDeviceId
(
place_
);
#ifdef PADDLE_WITH_MKLDNN
platform
::
AttachPointerHashToMKLDNNKey
(
this
,
place_
);
...
...
@@ -181,7 +210,7 @@ paddle::framework::FetchList InterpreterCore::Run(
}
#endif
}
if
(
execution_config_
.
create_local_scope
)
{
if
(
HasLocalScope
()
)
{
ClearLoDTensorArrayInLocalScope
();
}
...
...
@@ -196,11 +225,7 @@ paddle::framework::FetchList InterpreterCore::Run(
paddle
::
framework
::
FetchList
InterpreterCore
::
Run
(
const
std
::
vector
<
std
::
string
>&
feed_names
)
{
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
if
(
platform
::
is_gpu_place
(
place_
))
{
platform
::
SetDeviceId
(
place_
.
device
);
}
#endif
SetDeviceId
(
place_
);
#ifdef PADDLE_WITH_MKLDNN
platform
::
AttachPointerHashToMKLDNNKey
(
this
,
place_
);
...
...
@@ -208,17 +233,17 @@ paddle::framework::FetchList InterpreterCore::Run(
if
(
!
is_build_
)
{
LOG_FIRST_N
(
INFO
,
1
)
<<
"New Executor is Running."
;
paddle
::
framework
::
interpreter
::
build_variable_s
cope
(
block_
,
&
var_scope_
,
execution_config_
.
create_local_scope
);
paddle
::
framework
::
interpreter
::
BuildVariableS
cope
(
block_
,
&
var_scope_
,
HasLocalScope
()
);
std
::
vector
<
paddle
::
framework
::
OpFuncNode
>
op_func_nodes
;
paddle
::
framework
::
interpreter
::
build_op_func_l
ist
(
paddle
::
framework
::
interpreter
::
BuildOpFuncL
ist
(
place_
,
block_
,
execution_config_
.
skip_gc_vars
,
&
op_func_nodes
,
&
var_scope_
,
execution_config_
.
create_local_scope
,
HasLocalScope
()
,
execution_config_
.
used_for_jit
);
is_build_
=
true
;
SetFeedVarsInplaceSkip
(
feed_names
);
...
...
@@ -248,13 +273,13 @@ paddle::framework::FetchList InterpreterCore::Run(
#endif
}
if
(
execution_config_
.
create_local_scope
)
{
if
(
HasLocalScope
()
)
{
ClearLoDTensorArrayInLocalScope
();
}
// return Fetch Tensors
Scope
*
inner_scope
=
execution_config_
.
create_local_scope
?
local_scope_
:
var_scope_
.
GetMutableScope
();
Scope
*
inner_scope
=
HasLocalScope
()
?
local_scope_
:
var_scope_
.
GetMutableScope
();
auto
*
fetch_var
=
inner_scope
->
FindVar
(
interpreter
::
kFetchVarName
);
if
(
fetch_var
)
{
return
std
::
move
(
*
fetch_var
->
GetMutable
<
framework
::
FetchList
>
());
...
...
@@ -327,9 +352,8 @@ std::shared_ptr<interpreter::AsyncWorkQueue> InterpreterCore::GetWorkQueue() {
}
void
InterpreterCore
::
BuildAndCacheInstructionCtx
(
Instruction
*
instr_node
)
{
Scope
*
inner_scope
=
execution_config_
.
create_local_scope
?
local_scope_
:
var_scope_
.
GetMutableScope
();
Scope
*
inner_scope
=
HasLocalScope
()
?
local_scope_
:
var_scope_
.
GetMutableScope
();
VariableValueMap
ins_map
;
for
(
auto
&
var_name_item
:
instr_node
->
Inputs
())
{
std
::
vector
<
Variable
*>
input_vars
;
...
...
@@ -355,9 +379,8 @@ void InterpreterCore::BuildAndCacheInstructionCtx(Instruction* instr_node) {
// set runtime_ctx and infershape_ctx_
if
(
instr_node
->
OpBase
()
->
Type
()
==
"cinn_launch"
)
{
// OP use scope in
// kernel
Scope
*
local_scope
=
execution_config_
.
create_local_scope
?
var_scope_
.
GetMutableLocalScope
()
:
var_scope_
.
GetMutableScope
();
Scope
*
local_scope
=
HasLocalScope
()
?
var_scope_
.
GetMutableLocalScope
()
:
var_scope_
.
GetMutableScope
();
instr_node
->
ResetContextWithScope
(
ins_map
,
outs_map
,
*
local_scope
);
}
else
{
instr_node
->
ResetContext
(
ins_map
,
outs_map
);
...
...
@@ -387,9 +410,8 @@ void InterpreterCore::BuildInplace() {
}
}
Scope
*
local_scope
=
execution_config_
.
create_local_scope
?
var_scope_
.
GetMutableLocalScope
()
:
var_scope_
.
GetMutableScope
();
Scope
*
local_scope
=
HasLocalScope
()
?
var_scope_
.
GetMutableLocalScope
()
:
var_scope_
.
GetMutableScope
();
std
::
vector
<
std
::
vector
<
size_t
>>
input_var2op
(
var_scope_
.
VarSize
());
for
(
Instruction
&
instr
:
vec_instruction_
)
{
for
(
auto
&
item
:
instr
.
Inputs
())
{
...
...
@@ -524,9 +546,8 @@ void InterpreterCore::Convert(
}
for
(
auto
var_id
:
gc_check_vars
)
{
Scope
*
inner_scope
=
execution_config_
.
create_local_scope
?
local_scope_
:
var_scope_
.
GetMutableScope
();
Scope
*
inner_scope
=
HasLocalScope
()
?
local_scope_
:
var_scope_
.
GetMutableScope
();
paddle
::
framework
::
Variable
*
var
=
inner_scope
->
FindVar
(
var_scope_
.
GetNameById
(
var_id
));
if
(
var
->
IsType
<
LoDTensor
>
()
||
var
->
IsType
<
phi
::
SelectedRows
>
()
||
...
...
@@ -629,56 +650,11 @@ void InterpreterCore::BuildSkipShareLoDInfo() {
}
}
inline
void
SetDeviceId
(
const
platform
::
Place
&
place
)
{
// TODO(zhiqiu): reduce the cost
if
(
platform
::
is_gpu_place
(
place
))
{
#if !defined(PADDLE_WITH_CUDA) && !defined(PADDLE_WITH_HIP)
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Cannot run operator on place %s, please recompile paddle or "
"reinstall Paddle with CUDA support."
,
place
));
#else
auto
dev_id
=
place
.
device
;
platform
::
SetDeviceId
(
dev_id
);
#endif
}
else
if
(
platform
::
is_xpu_place
(
place
))
{
#ifndef PADDLE_WITH_XPU
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Cannot run operator on place %s, please recompile paddle or "
"reinstall Paddle with XPU support."
,
place
));
#else
auto
dev_id
=
place
.
device
;
platform
::
SetXPUDeviceId
(
dev_id
);
#endif
}
else
if
(
platform
::
is_npu_place
(
place
))
{
#ifndef PADDLE_WITH_ASCEND_CL
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Cannot run operator on place %s, please recompile paddle or "
"reinstall Paddle with NPU support."
,
place
));
#else
auto
dev_id
=
place
.
device
;
platform
::
SetNPUDeviceId
(
dev_id
);
#endif
}
else
if
(
platform
::
is_custom_place
(
place
))
{
#ifndef PADDLE_WITH_CUSTOM_DEVICE
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Cannot run operator on place %s, please recompile paddle or "
"reinstall Paddle with CustomDevice support."
,
place
));
#else
phi
::
DeviceManager
::
SetDevice
(
place
);
#endif
}
}
void
InterpreterCore
::
RunInstruction
(
const
Instruction
&
instr_node
)
{
auto
*
op
=
instr_node
.
OpBase
();
auto
place
=
instr_node
.
DeviceContext
().
GetPlace
();
Scope
*
local_scope
=
execution_config_
.
create_local_scope
?
var_scope_
.
GetMutableLocalScope
()
:
var_scope_
.
GetMutableScope
();
Scope
*
local_scope
=
HasLocalScope
()
?
var_scope_
.
GetMutableLocalScope
()
:
var_scope_
.
GetMutableScope
();
VLOG
(
4
)
<<
"Start run "
<<
place
<<
" "
<<
op
->
DebugStringEx
(
local_scope_
);
SetDeviceId
(
place
);
...
...
@@ -800,8 +776,8 @@ void InterpreterCore::RunInstruction(const Instruction& instr_node) {
void
InterpreterCore
::
ExecuteInstructionList
(
const
std
::
vector
<
Instruction
>&
vec_instr
)
{
interpreter
::
ResetAtomicGuard
guard
(
&
deps_
,
&
refs_
);
unfinished_op_numer_
=
vec_instr
.
size
();
if
(
unfinished_op_numer_
==
0
)
{
unfinished_op_num
b
er_
=
vec_instr
.
size
();
if
(
unfinished_op_num
b
er_
==
0
)
{
VLOG
(
4
)
<<
"No op to run, return"
;
return
;
}
...
...
@@ -878,8 +854,12 @@ void InterpreterCore::RunNextInstructions(
[
this
,
next_id
]
{
RunInstructionAsync
(
next_id
);
});
}
}
auto
direct_run_ops
=
interpreter
::
merge_vector
(
next_instr
.
SyncRunIds
(),
next_instr
.
DirectRunIds
());
std
::
vector
<
size_t
>
direct_run_ops
=
next_instr
.
SyncRunIds
();
direct_run_ops
.
insert
(
direct_run_ops
.
end
(),
next_instr
.
DirectRunIds
().
begin
(),
next_instr
.
DirectRunIds
().
end
());
int64_t
first_op
=
-
1
;
for
(
auto
next_id
:
direct_run_ops
)
{
if
(
IsReady
(
next_id
))
{
...
...
@@ -949,9 +929,9 @@ void InterpreterCore::RunInstructionAsync(size_t instr_id) {
return
;
}
VLOG
(
4
)
<<
"unfinished_op_num
er_: "
<<
unfinished_op_num
er_
;
if
(
UNLIKELY
(
unfinished_op_num
er_
.
fetch_sub
(
1
,
std
::
memory_order_relaxed
)
==
1
))
{
VLOG
(
4
)
<<
"unfinished_op_num
ber_: "
<<
unfinished_op_numb
er_
;
if
(
UNLIKELY
(
unfinished_op_num
ber_
.
fetch_sub
(
1
,
std
::
memory_order_relaxed
)
==
1
))
{
if
(
completion_notifier_
!=
nullptr
)
{
completion_notifier_
->
NotifyEvent
();
}
...
...
@@ -961,8 +941,11 @@ void InterpreterCore::RunInstructionAsync(size_t instr_id) {
}
}
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
void
InterpreterCore
::
RecordStreamForGC
(
const
Instruction
&
instr
)
{
#if !defined(PADDLE_WITH_CUDA) && !defined(PADDLE_WITH_HIP)
PADDLE_THROW
(
platform
::
errors
::
Unimplemented
(
"RecordStreamForGC is only implemented when compiled with GPU."
));
#else
if
(
!
IsInterpretercoreFastGCEnabled
()
||
instr
.
KernelType
()
!=
OpFuncType
::
kQueueAsync
)
{
return
;
...
...
@@ -1053,8 +1036,8 @@ void InterpreterCore::RecordStreamForGC(const Instruction& instr) {
framework
::
ToTypeName
(
var
->
Type
())));
}
}
}
#endif
}
void
InterpreterCore
::
CheckGC
(
const
Instruction
&
instr
)
{
platform
::
RecordEvent
record
(
...
...
@@ -1106,17 +1089,17 @@ void InterpreterCore::Prepare(const std::vector<std::string>& feed_names,
};
if
(
!
is_build_
)
{
paddle
::
framework
::
interpreter
::
build_variable_s
cope
(
block_
,
&
var_scope_
,
execution_config_
.
create_local_scope
);
paddle
::
framework
::
interpreter
::
BuildVariableS
cope
(
block_
,
&
var_scope_
,
HasLocalScope
()
);
FeedInput
();
std
::
vector
<
paddle
::
framework
::
OpFuncNode
>
op_func_nodes
;
paddle
::
framework
::
interpreter
::
build_op_func_l
ist
(
paddle
::
framework
::
interpreter
::
BuildOpFuncL
ist
(
place_
,
block_
,
execution_config_
.
skip_gc_vars
,
&
op_func_nodes
,
&
var_scope_
,
execution_config_
.
create_local_scope
,
HasLocalScope
()
,
execution_config_
.
used_for_jit
);
is_build_
=
true
;
SetFeedVarsInplaceSkip
(
feed_names
);
...
...
@@ -1124,7 +1107,7 @@ void InterpreterCore::Prepare(const std::vector<std::string>& feed_names,
Convert
(
&
op_func_nodes
);
}
// NOTE: Because feed_tensor will be GC after
// paddle::framework::
build_op_func_l
ist, so we should
// paddle::framework::
BuildOpFuncL
ist, so we should
// call FeedInput again.
if
(
prepare_feed
)
{
FeedInput
();
...
...
@@ -1138,6 +1121,8 @@ void InterpreterCore::SetFeedVarsInplaceSkip(
}
}
bool
InterpreterCore
::
HasLocalScope
()
const
{
return
local_scope_
!=
nullptr
;
}
std
::
shared_ptr
<
InterpreterCore
>
CreateInterpreterCore
(
const
platform
::
Place
&
place
,
const
ProgramDesc
&
prog
,
...
...
@@ -1145,11 +1130,11 @@ std::shared_ptr<InterpreterCore> CreateInterpreterCore(
const
std
::
vector
<
std
::
string
>&
fetch_names
,
const
std
::
set
<
std
::
string
>&
skip_gc_vars
)
{
std
::
shared_ptr
<
InterpreterCore
>
core
=
nullptr
;
// NOTE(Aurelius84): `
add_f
etch` will modify BlockDesc, so we should copy
// NOTE(Aurelius84): `
AddF
etch` will modify BlockDesc, so we should copy
// a new program.
auto
new_prog
=
std
::
make_shared
<
framework
::
ProgramDesc
>
(
prog
);
auto
*
block
=
new_prog
->
MutableBlock
(
0
);
interpreter
::
add_f
etch
(
fetch_names
,
block
);
interpreter
::
AddF
etch
(
fetch_names
,
block
);
core
=
std
::
make_shared
<
InterpreterCore
>
(
place
,
*
block
,
skip_gc_vars
,
scope
);
core
->
SetCopyProgram
(
new_prog
);
...
...
paddle/fluid/framework/new_executor/interpretercore.h
浏览文件 @
5303b66b
...
...
@@ -68,45 +68,42 @@ class InterpreterCore {
void
reset_scope
(
Scope
*
new_scope
);
private:
bool
BuildInplaceCheckVarIsOnlyInput
(
const
std
::
vector
<
std
::
vector
<
size_t
>>&
input_var2op
,
size_t
var_index
);
std
::
shared_ptr
<
interpreter
::
AsyncWorkQueue
>
GetWorkQueue
();
// build graph
void
Convert
(
std
::
vector
<
paddle
::
framework
::
OpFuncNode
>*
op_func_nodes
);
void
BuildOperatorDependences
();
void
BuildAndCacheInstructionCtx
(
Instruction
*
instr_node
);
void
BuildSkipShareLoDInfo
();
// inplace
void
BuildInplace
();
bool
BuildInplaceCheckVarIsOnlyInput
(
const
std
::
vector
<
std
::
vector
<
size_t
>>&
input_var2op
,
size_t
var_index
);
void
SetFeedVarsInplaceSkip
(
const
std
::
vector
<
std
::
string
>&
feed_names
);
void
BuildOperatorDependences
();
void
ClearLoDTensorArrayInLocalScope
();
void
Convert
(
std
::
vector
<
paddle
::
framework
::
OpFuncNode
>*
op_func_nodes
);
void
RunInstruction
(
const
Instruction
&
instr_node
);
// execution
void
ExecuteInstructionList
(
const
std
::
vector
<
Instruction
>&
vec_instr
);
void
RunInstructionAsync
(
size_t
instr_id
);
void
RunInstruction
(
const
Instruction
&
instr_node
);
void
RunNextInstructions
(
const
Instruction
&
instr_id
,
std
::
queue
<
size_t
>*
reserved_next_ops
);
// only used when program contains no feed op
void
Prepare
(
const
std
::
vector
<
std
::
string
>&
feed_names
,
const
std
::
vector
<
phi
::
DenseTensor
>&
feed_tensors
,
bool
prepare_feed
);
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
// gc
void
RecordStreamForGC
(
const
Instruction
&
instr
);
#endif
void
CheckGC
(
const
Instruction
&
instr
);
void
ClearLoDTensorArrayInLocalScope
();
void
RunInstructionAsync
(
size_t
instr_id
);
void
RunNextInstructions
(
const
Instruction
&
instr_id
,
std
::
queue
<
size_t
>*
reserved_next_ops
);
void
BuildSkipShareLoDInfo
();
// workqueue
std
::
shared_ptr
<
interpreter
::
AsyncWorkQueue
>
GetWorkQueue
();
void
SetFeedVarsInplaceSkip
(
const
std
::
vector
<
std
::
string
>&
feed_names
);
// scope
bool
HasLocalScope
()
const
;
private:
bool
is_build_
;
bool
is_build_
{
false
}
;
platform
::
Place
place_
;
const
BlockDesc
&
block_
;
// not owned
...
...
@@ -127,11 +124,7 @@ class InterpreterCore {
std
::
vector
<
Instruction
>
vec_instruction_
;
// deconstruct before OpFuncNode
// last_live_ops_[i] contains the id of operators that last access var[i]
std
::
map
<
size_t
,
std
::
set
<
size_t
>>
last_live_ops_
;
std
::
vector
<
size_t
>
dependecy_count_
;
std
::
atomic
<
size_t
>
unfinished_op_numer_
{
0
};
std
::
atomic
<
size_t
>
unfinished_op_number_
{
0
};
VariableScope
var_scope_
;
Scope
*
local_scope_
{
nullptr
};
// not owned
...
...
@@ -145,8 +138,13 @@ class InterpreterCore {
std
::
unique_ptr
<
InterpreterCoreGarbageCollector
>
gc_
;
std
::
future
<
std
::
unique_ptr
<
AtomicVectorSizeT
>>
atomic_deps_
;
std
::
future
<
std
::
unique_ptr
<
AtomicVectorSizeT
>>
atomic_var_ref_
;
// last_live_ops_[i] contains the id of operators that last access the i-th
// var
std
::
map
<
size_t
,
std
::
set
<
size_t
>>
last_live_ops_
;
// dependecy_count_[i] contains the number of dependencies that the i-th op
// need to wait
std
::
vector
<
size_t
>
dependecy_count_
;
std
::
vector
<
std
::
shared_ptr
<
interpreter
::
OpDepInfo
>>
deps_
;
std
::
vector
<
std
::
shared_ptr
<
interpreter
::
VarRefInfo
>>
refs_
;
...
...
paddle/fluid/framework/new_executor/interpretercore_util.cc
浏览文件 @
5303b66b
...
...
@@ -122,8 +122,8 @@ bool var_can_be_deleted(const std::string& name, const BlockDesc& block) {
std
::
unordered_map
<
const
paddle
::
framework
::
OperatorBase
*
,
std
::
vector
<
std
::
string
>>
get_unused_v
ars
(
const
BlockDesc
&
block
,
const
std
::
vector
<
std
::
shared_ptr
<
OperatorBase
>>&
ops
)
{
GetUnusedV
ars
(
const
BlockDesc
&
block
,
const
std
::
vector
<
std
::
shared_ptr
<
OperatorBase
>>&
ops
)
{
std
::
unordered_map
<
std
::
string
,
size_t
>
var_op_idx_map
;
for
(
size_t
i
=
0
;
i
<
ops
.
size
();
++
i
)
{
...
...
@@ -166,17 +166,17 @@ get_unused_vars(const BlockDesc& block,
for
(
auto
&
name_op_idx_pair
:
var_op_idx_map
)
{
auto
&
name
=
name_op_idx_pair
.
first
;
size_t
op_idx
=
name_op_idx_pair
.
second
;
result
[
op
s
[
op_idx
].
get
()
].
emplace_back
(
name
);
VLOG
(
4
)
<<
op
s
[
op_idx
].
get
()
->
Type
()
<<
" "
<<
name
;
auto
op
=
ops
[
op_idx
].
get
();
result
[
op
].
emplace_back
(
name
);
VLOG
(
4
)
<<
op
->
Type
()
<<
" "
<<
name
;
}
VLOG
(
4
)
<<
"gc map size:"
<<
result
.
size
();
return
result
;
}
void
build_variable_s
cope
(
const
framework
::
BlockDesc
&
block
,
VariableScope
*
var_scope
,
bool
use_local_scope
)
{
void
BuildVariableS
cope
(
const
framework
::
BlockDesc
&
block
,
VariableScope
*
var_scope
,
bool
use_local_scope
)
{
VLOG
(
3
)
<<
"Creating Variables"
;
auto
inner_scope
=
var_scope
->
GetMutableScope
();
...
...
@@ -214,8 +214,8 @@ void build_variable_scope(const framework::BlockDesc& block,
}
}
void
create_all_o
ps
(
const
framework
::
BlockDesc
&
block
,
std
::
vector
<
std
::
unique_ptr
<
OperatorBase
>>*
ops
)
{
void
CreateAllO
ps
(
const
framework
::
BlockDesc
&
block
,
std
::
vector
<
std
::
unique_ptr
<
OperatorBase
>>*
ops
)
{
for
(
auto
&
op
:
block
.
AllOps
())
{
auto
op_type
=
op
->
Type
();
VLOG
(
8
)
<<
"CreateOp from : "
<<
op_type
;
...
...
@@ -289,9 +289,9 @@ std::tuple<VariableValueMap, VariableIdMap> BuildVariableMap(
return
std
::
make_tuple
(
name2var
,
name2id
);
}
void
apply_device_g
uard
(
const
OperatorBase
*
op_base
,
const
platform
::
Place
&
place
,
OpKernelType
*
expected_kernel_key
)
{
void
ApplyDeviceG
uard
(
const
OperatorBase
*
op_base
,
const
platform
::
Place
&
place
,
OpKernelType
*
expected_kernel_key
)
{
bool
need_change_place
=
(
op_base
->
HasAttr
(
"op_device"
)
&&
(
op_base
->
Attr
<
std
::
string
>
(
"op_device"
).
length
()
>
0
));
...
...
@@ -352,7 +352,7 @@ void apply_device_guard(const OperatorBase* op_base,
}
}
void
deal_operator_b
ase
(
const
platform
::
Place
&
place
,
void
HandleOperatorB
ase
(
const
platform
::
Place
&
place
,
const
VariableScope
*
var_scope
,
std
::
shared_ptr
<
OperatorBase
>
op_base
,
OpFuncNode
*
op_func_node
,
...
...
@@ -361,7 +361,7 @@ void deal_operator_base(const platform::Place& place,
auto
*
dev_ctx
=
pool
.
Get
(
place
);
// input, output is prepared. set the other attributes.
op_func_node
->
operator_base_
=
op_base
;
if
(
IsSupportedHetePlace
(
place
))
{
if
(
IsSupportedHete
r
Place
(
place
))
{
op_func_node
->
type_
=
OpFuncType
::
kQueueAsync
;
}
else
if
(
platform
::
is_cpu_place
(
place
))
{
op_func_node
->
type_
=
OpFuncType
::
kQueueSync
;
...
...
@@ -382,19 +382,19 @@ void deal_operator_base(const platform::Place& place,
op_func_node
->
dev_ctx_
=
dev_ctx
;
}
void
build_op_func_l
ist
(
const
platform
::
Place
&
place
,
const
framework
::
BlockDesc
&
block
,
const
std
::
set
<
std
::
string
>&
skip_gc_vars
,
std
::
vector
<
OpFuncNode
>*
vec_func_list
,
VariableScope
*
var_scope
,
bool
use_local_scope
,
bool
used_for_jit
)
{
void
BuildOpFuncL
ist
(
const
platform
::
Place
&
place
,
const
framework
::
BlockDesc
&
block
,
const
std
::
set
<
std
::
string
>&
skip_gc_vars
,
std
::
vector
<
OpFuncNode
>*
vec_func_list
,
VariableScope
*
var_scope
,
bool
use_local_scope
,
bool
used_for_jit
)
{
Scope
*
local_scope
=
use_local_scope
?
var_scope
->
GetMutableLocalScope
()
:
var_scope
->
GetMutableScope
();
std
::
vector
<
std
::
unique_ptr
<
OperatorBase
>>
ops_unique
;
// its elements will be moved to vec_func_list
// Step 1: create all ops for current block.
create_all_o
ps
(
block
,
&
ops_unique
);
CreateAllO
ps
(
block
,
&
ops_unique
);
if
(
!
used_for_jit
)
{
// If gc is enabled and block size > 1
...
...
@@ -415,7 +415,7 @@ void build_op_func_list(const platform::Place& place,
for
(
auto
&
op_unique
:
ops_unique
)
{
ops
.
emplace_back
(
std
::
move
(
op_unique
));
}
auto
unused_var_map
=
get_unused_v
ars
(
block
,
ops
);
auto
unused_var_map
=
GetUnusedV
ars
(
block
,
ops
);
bool
flag_log_is_printed
=
false
;
for
(
size_t
i
=
0
;
i
<
ops
.
size
();
++
i
)
{
...
...
@@ -485,10 +485,10 @@ void build_op_func_list(const platform::Place& place,
try
{
if
(
dynamic_cast
<
framework
::
OperatorWithKernel
*>
(
op
)
==
nullptr
)
{
VLOG
(
4
)
<<
"HandleOperatorBase"
;
// op is not a operatorwithkernel, so direcly run OperatorBase::Run()
deal_operator_b
ase
(
HandleOperatorB
ase
(
place
,
var_scope
,
ops
[
i
],
&
op_func_node
,
local_scope
);
VLOG
(
4
)
<<
"deal_operator_base"
;
}
else
{
VLOG
(
4
)
<<
"OP is not null"
;
auto
op_with_kernel
=
const_cast
<
framework
::
OperatorWithKernel
*>
(
...
...
@@ -522,7 +522,7 @@ void build_op_func_list(const platform::Place& place,
op_with_kernel
->
GetExpectedKernelType
(
exec_ctx
);
VLOG
(
4
)
<<
"get expected_kernel_key"
;
// change device by the device_guard()
apply_device_g
uard
(
op
,
place
,
&
expected_kernel_key
);
ApplyDeviceG
uard
(
op
,
place
,
&
expected_kernel_key
);
VLOG
(
4
)
<<
"expected_kernel_key : "
<<
expected_kernel_key
;
// step 2. select op kernel
...
...
@@ -565,7 +565,7 @@ void build_op_func_list(const platform::Place& place,
dev_ctx
=
pool
.
Get
(
kernel_type
.
place_
);
}
op_func_node
.
dev_ctx_
=
dev_ctx
;
if
(
IsSupportedHetePlace
(
kernel_type
.
place_
))
{
if
(
IsSupportedHete
r
Place
(
kernel_type
.
place_
))
{
op_func_node
.
type_
=
OpFuncType
::
kQueueAsync
;
}
else
if
(
platform
::
is_cpu_place
(
kernel_type
.
place_
))
{
op_func_node
.
type_
=
OpFuncType
::
kQueueSync
;
...
...
@@ -667,7 +667,7 @@ void build_op_func_list(const platform::Place& place,
vec_func_list
->
emplace_back
(
op_func_node
);
// gc---------------------------------------------
------------------------------
// gc---------------------------------------------
auto
iter
=
unused_var_map
.
find
(
op
);
if
(
iter
==
unused_var_map
.
end
())
{
interpreter
::
LogDeviceMemoryStats
(
place
);
...
...
@@ -702,8 +702,8 @@ void build_op_func_list(const platform::Place& place,
memory
::
Release
(
place
);
}
void
add_f
etch
(
const
std
::
vector
<
std
::
string
>&
fetch_names
,
framework
::
BlockDesc
*
block
)
{
void
AddF
etch
(
const
std
::
vector
<
std
::
string
>&
fetch_names
,
framework
::
BlockDesc
*
block
)
{
auto
*
fetch_holder
=
block
->
Var
(
kFetchVarName
);
fetch_holder
->
SetType
(
proto
::
VarType
::
FETCH_LIST
);
fetch_holder
->
SetPersistable
(
true
);
...
...
@@ -721,20 +721,6 @@ void add_fetch(const std::vector<std::string>& fetch_names,
}
}
std
::
vector
<
size_t
>
merge_vector
(
const
std
::
vector
<
size_t
>&
first
,
const
std
::
vector
<
size_t
>&
second
)
{
std
::
vector
<
size_t
>
out
(
first
.
size
()
+
second
.
size
());
std
::
merge
(
first
.
begin
(),
first
.
end
(),
second
.
begin
(),
second
.
end
(),
out
.
begin
());
std
::
vector
<
size_t
>::
iterator
it
;
it
=
std
::
unique
(
out
.
begin
(),
out
.
end
());
out
.
resize
(
std
::
distance
(
out
.
begin
(),
it
));
return
out
;
}
}
// namespace interpreter
}
// namespace framework
}
// namespace paddle
paddle/fluid/framework/new_executor/interpretercore_util.h
浏览文件 @
5303b66b
...
...
@@ -66,23 +66,20 @@ class AsyncWorkQueue {
void
LogDeviceMemoryStats
(
const
platform
::
Place
&
place
);
void
build_variable_scope
(
const
framework
::
BlockDesc
&
block
,
VariableScope
*
var_scope
,
bool
use_local_scope
=
true
);
void
build_op_func_list
(
const
platform
::
Place
&
place
,
const
framework
::
BlockDesc
&
block
,
const
std
::
set
<
std
::
string
>&
skip_gc_vars
,
std
::
vector
<
OpFuncNode
>*
vec_func_list
,
VariableScope
*
scope
,
bool
use_local_scope
=
true
,
bool
used_for_jit
=
false
);
void
add_fetch
(
const
std
::
vector
<
std
::
string
>&
fetch_names
,
framework
::
BlockDesc
*
block
);
std
::
vector
<
size_t
>
merge_vector
(
const
std
::
vector
<
size_t
>&
first
,
const
std
::
vector
<
size_t
>&
second
);
void
BuildVariableScope
(
const
framework
::
BlockDesc
&
block
,
VariableScope
*
var_scope
,
bool
use_local_scope
=
true
);
void
BuildOpFuncList
(
const
platform
::
Place
&
place
,
const
framework
::
BlockDesc
&
block
,
const
std
::
set
<
std
::
string
>&
skip_gc_vars
,
std
::
vector
<
OpFuncNode
>*
vec_func_list
,
VariableScope
*
scope
,
bool
use_local_scope
=
true
,
bool
used_for_jit
=
false
);
void
AddFetch
(
const
std
::
vector
<
std
::
string
>&
fetch_names
,
framework
::
BlockDesc
*
block
);
}
// namespace interpreter
}
// namespace framework
...
...
paddle/fluid/framework/new_executor/new_executor_defs.h
浏览文件 @
5303b66b
...
...
@@ -392,7 +392,7 @@ static bool IsCpuOp(const Instruction& instr) {
}
// is supported heterogeneous place
static
bool
IsSupportedHetePlace
(
const
phi
::
Place
&
place
)
{
static
bool
IsSupportedHete
r
Place
(
const
phi
::
Place
&
place
)
{
return
platform
::
is_gpu_place
(
place
)
||
platform
::
is_npu_place
(
place
)
||
platform
::
is_xpu_place
(
place
)
||
platform
::
is_ipu_place
(
place
)
||
platform
::
is_custom_place
(
place
);
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录