diff options
author | Abseil Team <absl-team@google.com> | 2020-12-14 21:51:13 -0800 |
---|---|---|
committer | Mark Barolak <mbar@google.com> | 2020-12-15 10:08:56 -0500 |
commit | 68f1ad93251744f096036e65241493774b4e7ac0 (patch) | |
tree | 82fc5e49db344026ea67cffea1c30f3ac73ed6f6 /.github | |
parent | 52acfe6fc6b8bb9b1b7854993dd0fb33f0f3c4af (diff) |
Export of internal Abseil changes
--
f0456157cdc6cef6dbb5d2f99f9dc3ca0e213fab by Abseil Team <absl-team@google.com>:
Improve flag_benchmark reproducibility, test multiple threads
This adds multiple threads to the flags benchmark, and also uses some
linker magic to ensure that the defined flags are mapped to a consistent
alignment regardless of other code changes in the benchmark binary.
The benchmark performance in multiple threads is somewhat sensitive to
the alignment of the FlagImpl class, depending on whether the flag value
and the absl::Mutex are on the same cacheline or different ones. Making
all flags be cacheline-aligned would probably waste too much memory,
so this CL instead just makes the memory alignment consistent within the
tests.
PiperOrigin-RevId: 347536882
--
a019646d306b9497a1eba7efa5e2c4c651d9979d by Derek Mauro <dmauro@google.com>:
Fix other comments in node_hash_set.h
PiperOrigin-RevId: 347462975
--
66a4b39fe84a3e81935ad08ee76291df13d1d5e6 by Abseil Team <absl-team@google.com>:
Internal change
PiperOrigin-RevId: 347436244
--
52f4d18c2c4439d26428f64ac4de41e6183919ee by Mark Barolak <mbar@google.com>:
container: fix introduction for node_hash_set.h
Import of https://github.com/abseil/abseil-cpp/pull/861
PiperOrigin-RevId: 347432772
GitOrigin-RevId: f0456157cdc6cef6dbb5d2f99f9dc3ca0e213fab
Change-Id: I629f27e5b9584d92b0a42284c1acf708779bad3f
Diffstat (limited to '.github')
0 files changed, 0 insertions, 0 deletions