absl/crc/internal · 9c2582eaee3dd785da65e71b04fd8036c8ee79ac · open / abseil-cpp

Optimize CRC32 for Ampere Siryn · ac364eb9

Siryn's crc32 instruction seems to have latency 3 and throughput 1, which makes the optimal ratio of pmull and crc streams close to that of tested x86 machines. Up to +120% faster for large inputs.

PiperOrigin-RevId: 568645559
Change-Id: I86b85b1b2a5d4fb3680c516c4c9044238b20fe61

committed Sep 26, 2023

ac364eb9

Name	Last commit	Last Update
..
cpu_detect.cc		Loading commit data...
cpu_detect.h		Loading commit data...
crc.cc		Loading commit data...
crc.h		Loading commit data...
crc32_x86_arm_combined_simd.h		Loading commit data...
crc32c.h		Loading commit data...
crc32c_inline.h		Loading commit data...
crc_cord_state.cc		Loading commit data...
crc_cord_state.h		Loading commit data...
crc_cord_state_test.cc		Loading commit data...
crc_internal.h		Loading commit data...
crc_memcpy.h		Loading commit data...
crc_memcpy_fallback.cc		Loading commit data...
crc_memcpy_test.cc		Loading commit data...
crc_memcpy_x86_arm_combined.cc		Loading commit data...
crc_non_temporal_memcpy.cc		Loading commit data...
crc_x86_arm_combined.cc		Loading commit data...
non_temporal_arm_intrinsics.h		Loading commit data...
non_temporal_memcpy.h		Loading commit data...
non_temporal_memcpy_test.cc		Loading commit data...