Skip to content

F
folly
Commit e0ec3bcd authored by Nick Terrell's avatar Nick Terrell Committed by Facebook Github Bot
Browse files
Options

Fix AccessSpreader tests and benchmarks 

Summary:
* The atomic load was being optimized out of the benchmark.
* The initializer wasn't being called for the test tags, so `AccessSpreader<ThreadLocalTag>::current()` and `AccessSpreader<PthreadSelfTag>::current()` were always returning 0.

Reviewed By: ot

Differential Revision: D10154155

fbshipit-source-id: 76be9865c9ee591d24aedc64e6faf4fd24350282
parent 52b904f0
Hide whitespace changes
Inline Side-by-side
Showing with 46 additions and 38 deletions
folly/concurrency/test/CacheLocalityBenchmark.cpp
View file @ e0ec3bcd
...@@ -41,6 +41,7 @@ using namespace folly; ...@@ -41,6 +41,7 @@ using namespace folly;
Getcpu::Func AccessSpreader<tag>::pickGetcpuFunc() { \ Getcpu::Func AccessSpreader<tag>::pickGetcpuFunc() { \
return func; \ return func; \
} \ } \
template struct AccessSpreader<tag>; \
} }
DECLARE_SPREADER_TAG( DECLARE_SPREADER_TAG(
...@@ -93,44 +94,44 @@ BENCHMARK(AccessSpreaderUse, iters) { ...@@ -93,44 +94,44 @@ BENCHMARK(AccessSpreaderUse, iters) {
// ============================================================================ // ============================================================================
// folly/test/CacheLocalityTest.cpp relative time/iter iters/s // folly/test/CacheLocalityTest.cpp relative time/iter iters/s
// ============================================================================ // ============================================================================
// AccessSpreaderUse 11.94ns 83.79M // AccessSpreaderUse 11.53ns 86.75M
// ---------------------------------------------------------------------------- // ----------------------------------------------------------------------------
// contentionAtWidthGetcpu(1_stripe_0_work) 985.75ns 1.01M // contentionAtWidthGetcpu(1_stripe_0_work) 1.05us 954.62K
// contentionAtWidthGetcpu(2_stripe_0_work) 424.02ns 2.36M // contentionAtWidthGetcpu(2_stripe_0_work) 521.93ns 1.92M
// contentionAtWidthGetcpu(4_stripe_0_work) 190.13ns 5.26M // contentionAtWidthGetcpu(4_stripe_0_work) 306.91ns 3.26M
// contentionAtWidthGetcpu(8_stripe_0_work) 91.86ns 10.89M // contentionAtWidthGetcpu(8_stripe_0_work) 150.86ns 6.63M
// contentionAtWidthGetcpu(16_stripe_0_work) 29.31ns 34.12M // contentionAtWidthGetcpu(16_stripe_0_work) 86.20ns 11.60M
// contentionAtWidthGetcpu(32_stripe_0_work) 29.53ns 33.86M // contentionAtWidthGetcpu(32_stripe_0_work) 35.90ns 27.85M
// contentionAtWidthGetcpu(64_stripe_0_work) 29.93ns 33.41M // contentionAtWidthGetcpu(64_stripe_0_work) 36.76ns 27.21M
// contentionAtWidthThreadLocal(2_stripe_0_work) 609.21ns 1.64M // contentionAtWidthThreadLocal(2_stripe_0_work) 303.88ns 3.29M
// contentionAtWidthThreadLocal(4_stripe_0_work) 303.60ns 3.29M // contentionAtWidthThreadLocal(4_stripe_0_work) 225.62ns 4.43M
// contentionAtWidthThreadLocal(8_stripe_0_work) 246.57ns 4.06M // contentionAtWidthThreadLocal(8_stripe_0_work) 106.88ns 9.36M
// contentionAtWidthThreadLocal(16_stripe_0_work) 154.84ns 6.46M // contentionAtWidthThreadLocal(16_stripe_0_work) 72.82ns 13.73M
// contentionAtWidthThreadLocal(32_stripe_0_work) 24.14ns 41.43M // contentionAtWidthThreadLocal(32_stripe_0_work) 26.53ns 37.70M
// contentionAtWidthThreadLocal(64_stripe_0_work) 23.95ns 41.75M // contentionAtWidthThreadLocal(64_stripe_0_work) 25.02ns 39.97M
// contentionAtWidthPthreadSelf(2_stripe_0_work) 722.01ns 1.39M // contentionAtWidthPthreadSelf(2_stripe_0_work) 363.34ns 2.75M
// contentionAtWidthPthreadSelf(4_stripe_0_work) 501.56ns 1.99M // contentionAtWidthPthreadSelf(4_stripe_0_work) 247.74ns 4.04M
// contentionAtWidthPthreadSelf(8_stripe_0_work) 474.58ns 2.11M // contentionAtWidthPthreadSelf(8_stripe_0_work) 109.03ns 9.17M
// contentionAtWidthPthreadSelf(16_stripe_0_work) 300.90ns 3.32M // contentionAtWidthPthreadSelf(16_stripe_0_work) 93.61ns 10.68M
// contentionAtWidthPthreadSelf(32_stripe_0_work) 175.77ns 5.69M // contentionAtWidthPthreadSelf(32_stripe_0_work) 68.63ns 14.57M
// contentionAtWidthPthreadSelf(64_stripe_0_work) 174.88ns 5.72M // contentionAtWidthPthreadSelf(64_stripe_0_work) 64.98ns 15.39M
// atomicIncrBaseline(local_incr_0_work) 16.81ns 59.51M // atomicIncrBaseline(local_incr_0_work) 13.64ns 73.33M
// ---------------------------------------------------------------------------- // ----------------------------------------------------------------------------
// contentionAtWidthGetcpu(1_stripe_500_work) 1.82us 549.97K // contentionAtWidthGetcpu(1_stripe_500_work) 1.87us 534.91K
// contentionAtWidthGetcpu(2_stripe_500_work) 533.71ns 1.87M // contentionAtWidthGetcpu(2_stripe_500_work) 1.58us 632.15K
// contentionAtWidthGetcpu(4_stripe_500_work) 424.64ns 2.35M // contentionAtWidthGetcpu(4_stripe_500_work) 622.80ns 1.61M
// contentionAtWidthGetcpu(8_stripe_500_work) 451.85ns 2.21M // contentionAtWidthGetcpu(8_stripe_500_work) 501.08ns 2.00M
// contentionAtWidthGetcpu(16_stripe_500_work) 425.54ns 2.35M // contentionAtWidthGetcpu(16_stripe_500_work) 480.42ns 2.08M
// contentionAtWidthGetcpu(32_stripe_500_work) 501.66ns 1.99M // contentionAtWidthGetcpu(32_stripe_500_work) 420.10ns 2.38M
// atomicIncrBaseline(local_incr_500_work) 438.46ns 2.28M // atomicIncrBaseline(local_incr_500_work) 407.74ns 2.45M
// ---------------------------------------------------------------------------- // ----------------------------------------------------------------------------
// contentionAtWidthGetcpu(1_stripe_1000_work) 1.88us 532.20K // contentionAtWidthGetcpu(1_stripe_1000_work) 1.93us 518.44K
// contentionAtWidthGetcpu(2_stripe_1000_work) 824.62ns 1.21M // contentionAtWidthGetcpu(2_stripe_1000_work) 1.49us 669.39K
// contentionAtWidthGetcpu(4_stripe_1000_work) 803.56ns 1.24M // contentionAtWidthGetcpu(4_stripe_1000_work) 876.77ns 1.14M
// contentionAtWidthGetcpu(8_stripe_1000_work) 926.65ns 1.08M // contentionAtWidthGetcpu(8_stripe_1000_work) 749.22ns 1.33M
// contentionAtWidthGetcpu(16_stripe_1000_work) 900.10ns 1.11M // contentionAtWidthGetcpu(16_stripe_1000_work) 754.72ns 1.32M
// contentionAtWidthGetcpu(32_stripe_1000_work) 890.75ns 1.12M // contentionAtWidthGetcpu(32_stripe_1000_work) 591.59ns 1.69M
// atomicIncrBaseline(local_incr_1000_work) 774.47ns 1.29M // atomicIncrBaseline(local_incr_1000_work) 611.62ns 1.64M
// ============================================================================ // ============================================================================
template <template <typename> class Tag> template <template <typename> class Tag>
static void contentionAtWidth(size_t iters, size_t stripes, size_t work) { static void contentionAtWidth(size_t iters, size_t stripes, size_t work) {
...@@ -173,7 +174,8 @@ static void contentionAtWidth(size_t iters, size_t stripes, size_t work) { ...@@ -173,7 +174,8 @@ static void contentionAtWidth(size_t iters, size_t stripes, size_t work) {
for (size_t i = iters; i > 0; --i) { for (size_t i = iters; i > 0; --i) {
++*(counters[AccessSpreader<Tag>::current(stripes)]); ++*(counters[AccessSpreader<Tag>::current(stripes)]);
for (size_t j = work; j > 0; --j) { for (size_t j = work; j > 0; --j) {
localWork.load(); auto x = localWork.load();
folly::doNotOptimizeAway(x);
} }
} }
})); }));
...@@ -181,7 +183,11 @@ static void contentionAtWidth(size_t iters, size_t stripes, size_t work) { ...@@ -181,7 +183,11 @@ static void contentionAtWidth(size_t iters, size_t stripes, size_t work) {
if (threads.size() == numThreads / 15 || threads.size() == numThreads / 5) { if (threads.size() == numThreads / 15 || threads.size() == numThreads / 5) {
// create a few dummy threads to wrap back around to 0 mod numCpus // create a few dummy threads to wrap back around to 0 mod numCpus
for (size_t i = threads.size(); i != numThreads; ++i) { for (size_t i = threads.size(); i != numThreads; ++i) {
std::thread([&]() { AccessSpreader<Tag>::current(stripes); }).join(); std::thread t([&]() {
auto x = AccessSpreader<Tag>::current(stripes);
folly::doNotOptimizeAway(x);
});
t.join();
} }
} }
} }
...@@ -214,7 +220,8 @@ atomicIncrBaseline(size_t iters, size_t work, size_t numThreads = 32) { ...@@ -214,7 +220,8 @@ atomicIncrBaseline(size_t iters, size_t work, size_t numThreads = 32) {
for (size_t i = iters; i > 0; --i) { for (size_t i = iters; i > 0; --i) {
localCounter++; localCounter++;
for (size_t j = work; j > 0; --j) { for (size_t j = work; j > 0; --j) {
localWork.load(); auto x = localWork.load();
folly::doNotOptimizeAway(x);
} }
} }
})); }));
......
folly/concurrency/test/CacheLocalityTest.cpp
View file @ e0ec3bcd
...@@ -391,6 +391,7 @@ TEST(AccessSpreader, Simple) { ...@@ -391,6 +391,7 @@ TEST(AccessSpreader, Simple) {
Getcpu::Func AccessSpreader<tag>::pickGetcpuFunc() { \ Getcpu::Func AccessSpreader<tag>::pickGetcpuFunc() { \
return func; \ return func; \
} \ } \
template struct AccessSpreader<tag>; \
} }
DECLARE_SPREADER_TAG(ManualTag, CacheLocality::uniform(16), testingGetcpu) DECLARE_SPREADER_TAG(ManualTag, CacheLocality::uniform(16), testingGetcpu)
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment