google / benchmark / 1287

92%

Pull #507

travis-ci

<a href="https://github.com/google/benchmark/commit/<a class=hub.com/google/benchmark/commit/cdd596fd8dff700c538de360fce1a860822297db">cdd596fd8<a href="https://github.com/google/benchmark/commit/cdd596fd8dff700c538de360fce1a860822297db">&quot;&gt;Use int64_t instead of size_t for items/bytes/iterations

</a><a class="double-link" href="https://github.com/google/benchmark/commit/<a class="double-link" href="https://github.com/google/benchmark/commit/7a767012f1c423b37069f6d315b97164b5850271">7a767012f</a>">7a767012f</a><a href="https://github.com/google/benchmark/commit/cdd596fd8dff700c538de360fce1a860822297db"> seemingly unintentionally changed the data types
for bytes_processed, items_processed, and iterations to size_t. On 32-bit systems, the
bytes_processed and items_processed variables easily overflow, and often report fewer
items/second for faster code. The fact that overflow chances depends on architecture
and even ABI (e.g. x64 vs x32 ABI on x86-64) makes it particularly annoying.
iterations counter is made to use uint32_t internally because it is incremented in the hot
loop, and incrementing uint64_t is slow on some 32-bit platforms (e.g. Asm.js). However,
the public interface State.iterations() is changed to return int64_t.

6 of 6 new or added lines in 2 files covered. (100.0%)

1476 of 1698 relevant lines covered (86.93%)

4671754.17 hits per line