experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

This CL introduces a GetDebugOpName function to avoid using op_name if the Chunk is not in use.

Browse Source 
PiperOrigin-RevId: 356788273
Change-Id: I692f6b8e2553d29189872085d28ef7b0413ca75f
This commit is contained in:
Tianrun Li 2021-02-10 11:45:58 -08:00 committed by TensorFlower Gardener
parent 4499fa94bc
commit e04eb6b4af
2 changed files with 19 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
tensorflow/core/common_runtime/bfc_allocator.cc
View File

@ -1032,9 +1032,9 @@ void BFCAllocator::DumpMemoryLog(size_t num_bytes) {
(c->in_use() ? "InUse" : "Free "), " at ",
strings::Hex(reinterpret_cast<uint64>(c->ptr)), " of size ", c->size);
if (ShouldRecordOpName()) {
strings::StrAppend(
&buf, " by op ", c->op_name ? string(c->op_name) : "UNKNOWN",
" action_count ", c->action_count, " step ", c->step_id);
strings::StrAppend(&buf, " by op ", c->GetDebugOpName(),
" action_count ", c->action_count, " step ",
c->step_id);
}
strings::StrAppend(&buf, " next ", c->next);
if (timing_counter_) {
@ -1125,7 +1125,7 @@ MemoryDump BFCAllocator::RecordMemoryMapInternal() {
mc->set_size(c->size);
mc->set_requested_size(c->requested_size);
mc->set_bin(c->bin_num);
mc->set_op_name(c->op_name ? string(c->op_name) : "UNKNOWN");
mc->set_op_name(c->GetDebugOpName());
mc->set_step_id(c->step_id);
mc->set_action_count(c->action_count);
if (timing_counter_) {

16
tensorflow/core/common_runtime/bfc_allocator.h
View File

@ -193,6 +193,20 @@ class BFCAllocator : public Allocator {
uint64 step_id = 0;
uint64 action_count = 0;
// Get the op name used for memory debugging.
const char* GetDebugOpName() const {
// If chunk is not in use, although the op_name pointer is not nullptr,
// the corresponding OpKernel might have already been deallocated, and the
// op_name pointer might point to invalid memory. So in this case, return
// a special op name "UNUSED";
if (!in_use())
return "UNUSED";
else if (op_name)
return op_name;
else
return "UNKNOWN";
}
string DebugString(BFCAllocator* a,
bool recurse) TF_NO_THREAD_SAFETY_ANALYSIS {
string dbg;
@ -208,7 +222,7 @@ class BFCAllocator : public Allocator {
Chunk* n = a->ChunkFromHandle(next);
strings::StrAppend(&dbg, ", next: ", n->DebugString(a, false));
}
strings::StrAppend(&dbg, ", for: ", op_name ? op_name : "UNKNOWN",
strings::StrAppend(&dbg, ", for: ", GetDebugOpName(),
", stepid: ", step_id,
", last_action: ", action_count);
return dbg;