Export of internal Abseil changes.
-- da04b8cd21f6225d71397471474d34a77df0efd6 by Jon Cohen <cohenjon@google.com>: Don't use std::any, std::optional, std::variant, and friends on MacOS versions older than 10.14. Although Xcode 10 includes those headers and makes the types available to use, according to https://developer.apple.com/documentation/xcode_release_notes/xcode_10_release_notes, on MacOS 10.13 and earlier use of any functions (std::get, for example) results in an error message to upgrade to MacOS 10.14. This fixes https://github.com/abseil/abseil-cpp/issues/207. See that issue for more information on the error generated. PiperOrigin-RevId: 221844618 -- 1d99f77b4c60c5b0d7984f46e8ed63a3f969c635 by Jon Cohen <cohenjon@google.com>: raw_hash_set_test is still flaky under gcc 4.8. Since we now have the probe_test, we don't need the PerfectRatio tests. Just remove them. PiperOrigin-RevId: 221843042 -- 135cbb2a5d90963256518b3b59fe6710815e5dfa by Abseil Team <absl-team@google.com>: Update absl/algorithm/CMakeLists.txt to use new functions i.e. absl_cc_(library|test) PiperOrigin-RevId: 221828348 -- 1a5abde4f17f998ae89d87155d59f982a70202d8 by Jon Cohen <cohenjon@google.com>: Internal change PiperOrigin-RevId: 221708245 -- e03e031d4de39275989f695c768b0940cce1ff16 by Matt Armstrong <marmstrong@google.com>: Log to FATAL in throw_delegate.h ABSL_RAW_LOG(FATAL, ...) is guaranteed to abort. Previously, the code was logging to ERROR and calling abort() explicitly, which defeated any integration with absl::raw_logging_internal::AbortHook(). These changes are limited to Abseil internal APIs. PiperOrigin-RevId: 221696513 -- d13691523a3f9a5367fd1194cf9604bf4a969029 by Shahriar Rouf <nafi@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 221694877 -- f4044c56d44ba0ac2a9f218ed55f1b1f9e985eae by Abseil Team <absl-team@google.com>: Update absl/base/CMakeLists.txt to use new functions i.e. absl_cc_(library|test) PiperOrigin-RevId: 221676669 GitOrigin-RevId: da04b8cd21f6225d71397471474d34a77df0efd6 Change-Id: If6621e10d096a39b6a056a072c2727a0df0b0620
This commit is contained in:
parent
a06c4a1d90
commit
f6ae816808
9 changed files with 432 additions and 529 deletions
|
@ -14,7 +14,6 @@
|
|||
|
||||
#include "absl/container/internal/raw_hash_set.h"
|
||||
|
||||
#include <array>
|
||||
#include <cmath>
|
||||
#include <cstdint>
|
||||
#include <deque>
|
||||
|
@ -1782,143 +1781,6 @@ TEST(Table, IterationOrderChangesForSmallTables) {
|
|||
FAIL() << "Iteration order remained the same across many attempts.";
|
||||
}
|
||||
|
||||
// Fill the table to 3 different load factors (min, median, max) and evaluate
|
||||
// the percentage of perfect hits using the debug API.
|
||||
template <class Table, class AddFn>
|
||||
std::vector<double> CollectPerfectRatios(Table, AddFn add) {
|
||||
std::vector<double> results(3);
|
||||
|
||||
constexpr size_t kNumTrials = 10;
|
||||
std::vector<Table> tables(kNumTrials);
|
||||
|
||||
for (Table& t : tables) {
|
||||
using Key = typename Table::key_type;
|
||||
|
||||
// First, fill enough to have a good distribution.
|
||||
constexpr size_t kMinSize = 10000;
|
||||
std::vector<Key> keys;
|
||||
while (t.size() < kMinSize) keys.push_back(add(t));
|
||||
// Then, insert until we reach min load factor.
|
||||
double lf = t.load_factor();
|
||||
while (lf <= t.load_factor()) keys.push_back(add(t));
|
||||
|
||||
// We are now at min load factor. Take a snapshot.
|
||||
size_t perfect = 0;
|
||||
auto update_perfect = [&](Key k) {
|
||||
perfect += GetHashtableDebugNumProbes(t, k) == 0;
|
||||
};
|
||||
for (const auto& k : keys) update_perfect(k);
|
||||
|
||||
std::vector<double> perfect_ratios;
|
||||
// Keep going until we hit max load factor.
|
||||
while (t.load_factor() < .6) {
|
||||
perfect_ratios.push_back(1.0 * perfect / t.size());
|
||||
update_perfect(add(t));
|
||||
}
|
||||
while (t.load_factor() > .5) {
|
||||
perfect_ratios.push_back(1.0 * perfect / t.size());
|
||||
update_perfect(add(t));
|
||||
}
|
||||
|
||||
results[0] += perfect_ratios.front();
|
||||
results[1] += perfect_ratios[perfect_ratios.size() / 2];
|
||||
results[2] += perfect_ratios.back();
|
||||
}
|
||||
|
||||
results[0] /= kNumTrials;
|
||||
results[1] /= kNumTrials;
|
||||
results[2] /= kNumTrials;
|
||||
|
||||
return results;
|
||||
}
|
||||
|
||||
std::vector<std::pair<double, double>> StringTablePefectRatios() {
|
||||
constexpr bool kRandomizesInserts =
|
||||
#if NDEBUG
|
||||
false;
|
||||
#else // NDEBUG
|
||||
true;
|
||||
#endif // NDEBUG
|
||||
|
||||
// The effective load factor is larger in non-opt mode because we insert
|
||||
// elements out of order.
|
||||
switch (container_internal::Group::kWidth) {
|
||||
case 8:
|
||||
if (kRandomizesInserts) {
|
||||
return {{0.986, 0.02}, {0.95, 0.02}, {0.89, 0.02}};
|
||||
} else {
|
||||
return {{0.995, 0.01}, {0.97, 0.01}, {0.89, 0.02}};
|
||||
}
|
||||
case 16:
|
||||
if (kRandomizesInserts) {
|
||||
return {{0.973, 0.01}, {0.965, 0.01}, {0.92, 0.02}};
|
||||
} else {
|
||||
return {{0.995, 0.005}, {0.99, 0.005}, {0.94, 0.01}};
|
||||
}
|
||||
}
|
||||
ABSL_RAW_LOG(FATAL, "%s", "Unknown Group width");
|
||||
return {};
|
||||
}
|
||||
|
||||
// This is almost a change detector, but it allows us to know how we are
|
||||
// affecting the probe distribution.
|
||||
TEST(Table, EffectiveLoadFactorStrings) {
|
||||
std::vector<double> perfect_ratios =
|
||||
CollectPerfectRatios(StringTable(), [](StringTable& t) {
|
||||
return t.emplace(std::to_string(t.size()), "").first->first;
|
||||
});
|
||||
|
||||
auto ratios = StringTablePefectRatios();
|
||||
if (ratios.empty()) return;
|
||||
EXPECT_THAT(perfect_ratios,
|
||||
ElementsAre(DoubleNear(ratios[0].first, ratios[0].second),
|
||||
DoubleNear(ratios[1].first, ratios[1].second),
|
||||
DoubleNear(ratios[2].first, ratios[2].second)));
|
||||
}
|
||||
|
||||
std::vector<std::pair<double, double>> IntTablePefectRatios() {
|
||||
constexpr bool kRandomizesInserts =
|
||||
#ifdef NDEBUG
|
||||
false;
|
||||
#else // NDEBUG
|
||||
true;
|
||||
#endif // NDEBUG
|
||||
|
||||
// The effective load factor is larger in non-opt mode because we insert
|
||||
// elements out of order.
|
||||
switch (container_internal::Group::kWidth) {
|
||||
case 8:
|
||||
if (kRandomizesInserts) {
|
||||
return {{0.99, 0.02}, {0.985, 0.02}, {0.95, 0.05}};
|
||||
} else {
|
||||
return {{0.99, 0.01}, {0.99, 0.01}, {0.95, 0.02}};
|
||||
}
|
||||
case 16:
|
||||
if (kRandomizesInserts) {
|
||||
return {{0.98, 0.02}, {0.978, 0.02}, {0.96, 0.02}};
|
||||
} else {
|
||||
return {{0.998, 0.003}, {0.995, 0.01}, {0.975, 0.02}};
|
||||
}
|
||||
}
|
||||
ABSL_RAW_LOG(FATAL, "%s", "Unknown Group width");
|
||||
return {};
|
||||
}
|
||||
|
||||
// This is almost a change detector, but it allows us to know how we are
|
||||
// affecting the probe distribution.
|
||||
TEST(Table, EffectiveLoadFactorInts) {
|
||||
std::vector<double> perfect_ratios = CollectPerfectRatios(
|
||||
IntTable(), [](IntTable& t) { return *t.emplace(t.size()).first; });
|
||||
|
||||
auto ratios = IntTablePefectRatios();
|
||||
if (ratios.empty()) return;
|
||||
|
||||
EXPECT_THAT(perfect_ratios,
|
||||
ElementsAre(DoubleNear(ratios[0].first, ratios[0].second),
|
||||
DoubleNear(ratios[1].first, ratios[1].second),
|
||||
DoubleNear(ratios[2].first, ratios[2].second)));
|
||||
}
|
||||
|
||||
// Confirm that we assert if we try to erase() end().
|
||||
TEST(TableDeathTest, EraseOfEndAsserts) {
|
||||
// Use an assert with side-effects to figure out if they are actually enabled.
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue