Large differences in __memcpy_sse2_unaligned CPU overhead across x86_64 uarchs |
|
Issue descriptionFrom CWP data on Chrome OS stable channel, Periodic collection. Silvermont ~2.5% Airmont/SandyBridge/IvyBridge ~1.5% Haswell ~1.2% Broadwell ~0.84% I'm not sure if this is due to a uarch-specific quirk. Maybe we should ask Intel about it? |
|
►
Sign in to add a comment |
|