ec030f72f3
canonicalization and signature generation faster Added benchmark for XlaCompilationCache::BuildSignature to measure time taken to build a signature for the cache. Base is this CL with just the changes to add the benchmark in xla_compilation_cache_test.cc, New is this whole CL. Run on desktop machine (40 X 2793 MHz CPUs); 2019-09-17T08:30:04.125894664-07:00 CPU: Intel Ivybridge with HyperThreading (20 cores) dL1:32KB dL2:256KB dL3:25MB Benchmark Base (ns) New (ns) Improvement ---------------------------------------------------------------------------- BM_BuildSignature/0 226 87 +61.5% BM_BuildSignature/1 337 171 +49.3% BM_BuildSignature/2 504 259 +48.6% BM_BuildSignature/5 1008 592 +41.3% BM_BuildSignature/10 1751 1238 +29.3% RELNOTES: n/a PiperOrigin-RevId: 276289188 Change-Id: Ia47343203f6ac587a921a92f86c2428dd04db2a7
81 lines
3.0 KiB
C++
81 lines
3.0 KiB
C++
/* Copyright 2018 The TensorFlow Authors. All Rights Reserved.
|
|
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
you may not use this file except in compliance with the License.
|
|
You may obtain a copy of the License at
|
|
|
|
http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
Unless required by applicable law or agreed to in writing, software
|
|
distributed under the License is distributed on an "AS IS" BASIS,
|
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
See the License for the specific language governing permissions and
|
|
limitations under the License.
|
|
==============================================================================*/
|
|
|
|
#include "tensorflow/compiler/jit/xla_compilation_cache.h"
|
|
|
|
#include "tensorflow/compiler/tf2xla/shape_util.h"
|
|
#include "tensorflow/core/platform/test.h"
|
|
#include "tensorflow/core/platform/test_benchmark.h"
|
|
|
|
namespace tensorflow {
|
|
namespace {
|
|
|
|
TEST(XlaCompilationCacheTest, SignatureEquality) {
|
|
NameAttrList fn;
|
|
fn.set_name("afunction");
|
|
std::vector<XlaCompiler::Argument> args(1);
|
|
args[0].kind = XlaCompiler::Argument::kConstant;
|
|
args[0].type = DT_INT32;
|
|
args[0].shape = TensorShape({4, 0});
|
|
args[0].constant_value = Tensor(DT_INT32, {4, 0});
|
|
TF_ASSERT_OK_AND_ASSIGN(XlaCompilationCache::Signature s1,
|
|
XlaCompilationCache::BuildSignature(fn, args));
|
|
|
|
args[0].type = DT_FLOAT;
|
|
args[0].constant_value = Tensor(DT_FLOAT, {4, 0});
|
|
TF_ASSERT_OK_AND_ASSIGN(XlaCompilationCache::Signature s2,
|
|
XlaCompilationCache::BuildSignature(fn, args));
|
|
|
|
args[0].shape = TensorShape({0, 4});
|
|
args[0].constant_value = Tensor(DT_FLOAT, {0, 4});
|
|
TF_ASSERT_OK_AND_ASSIGN(XlaCompilationCache::Signature s3,
|
|
XlaCompilationCache::BuildSignature(fn, args));
|
|
|
|
std::vector<XlaCompilationCache::Signature> signatures = {s1, s2, s3};
|
|
for (int i = 0; i < signatures.size(); ++i) {
|
|
for (int j = 0; j < signatures.size(); ++j) {
|
|
EXPECT_EQ(i == j, signatures[i] == signatures[j])
|
|
<< signatures[i].HumanString() << " " << signatures[j].HumanString();
|
|
}
|
|
}
|
|
}
|
|
|
|
static void BM_BuildSignature(int iters, int n_args) {
|
|
NameAttrList fn;
|
|
fn.set_name("afunction");
|
|
for (int i = 0; i < n_args; i++) {
|
|
(*fn.mutable_attr())[absl::StrCat("T", i)].set_type(DT_FLOAT);
|
|
}
|
|
std::vector<XlaCompiler::Argument> args(n_args);
|
|
for (int i = 0; i < n_args; i++) {
|
|
args[i].kind = (((i % 3) == 0) ? XlaCompiler::Argument::kConstant
|
|
: XlaCompiler::Argument::kParameter);
|
|
args[i].type = DT_INT32;
|
|
args[i].shape = TensorShape({4, 0});
|
|
args[i].constant_value = Tensor(DT_INT32, {4, 0});
|
|
}
|
|
|
|
while (--iters > 0) {
|
|
xla::StatusOr<XlaCompilationCache::Signature> s =
|
|
XlaCompilationCache::BuildSignature(fn, args);
|
|
CHECK(s.ok());
|
|
XlaCompilationCache::Signature sig = std::move(s.ValueOrDie());
|
|
}
|
|
}
|
|
BENCHMARK(BM_BuildSignature)->Arg(0)->Arg(1)->Arg(2)->Arg(5)->Arg(10);
|
|
|
|
} // namespace
|
|
} // namespace tensorflow
|