command_buffer_thunk.cc
imports xla/tsl/platform/env.himports xla/tsl/platform/errors.h
SHA-25670b296453555c733d06b34485834c61d34c0cc49166bdcd8258e825dfc2b0a76
MaleculeMd(In₄)
Evidence
1/* Copyright 2023 The OpenXLA Authors.
2
3Licensed under the Apache License, Version 2.0 (the "License");
4you may not use this file except in compliance with the License.
5You may obtain a copy of the License at
6
7 http://www.apache.org/licenses/LICENSE-2.0
8
9Unless required by applicable law or agreed to in writing, software
10distributed under the License is …
11:42… IND, either express or implied.
12See the License for the specific language governing permissions and
13limitations under the License.
14==============================================================================*/
15
16#include "xla/backends/gpu/runtime/command_buffer_thunk.h"
17
18#include <algorithm>
19#include <cstdint>
20#include <memory>
21#include <string>
22#include <utility>
23#include <vector>
24
25#include "absl/base/thread_annotations.h"
26#include "absl/functional/function_ref.h"
27#include "absl/status/status.h"
28#include "absl/strings/str_cat.h"
29#include "absl/synchronization/mutex.h"
30#include "xla/backends/gpu/runtime/command.h"
31#include "xla/backends/gpu/runtime/command_executor.h"
32#include "xla/backends/gpu/runtime/sequential_thunk.h"
33#include "xla/backends/gpu/runtime/thunk.h"
34#include "xla/service/buffer_assignment.h"
35#include "xla/service/gpu/buffer_allocations.h"
36#include "xla/stream_executor/command_buffer.h"
37#include "xla/stream_executor/device_address.h"
38#include "xla/stream_executor/stream_executor.h"
39#include "xla/tsl/platform/env.h"
40#include "xla/tsl/platform/errors.h"
41#include "xla/tsl/platform/logging.h"
42#include "xla/tsl/platform/statusor.h"
43#include "tsl/profiler/lib/profiler_lock.h"
44#include "tsl/profiler/lib/traceme.h"
45#include "tsl/profiler/lib/traceme_encode.h"
46
47namespace xla::gpu {
⋯5 lines
86:26… ) resident in device memory only for executable that
87 // are actually used.
88 //
89 // In a perfect world higher level framework (JAX, Tensorflow, P …
90:41… unused executables, however today they
91 // all have a pretty large LRU cache for keeping O(1000) XLA executables.
92 EvictCommandBuffers();
93 TrackC …