WebSocket frame masking: use SIMD for ARM
WebSocket frame masking has been using SIMD on X86 for GCC and Clang for
a while. Now that Chrome no longer supports disabling ARM Neon (SIMD)
instructions at runtime, we can also perform the same optimisation for
ARM.
I have verified that this change causes a 128-bit vector XOR instruction
to be used when compiling for ARM7, and a good-quality fallback when
compiling for ARM6.
BUG=539259
Review-Url: https://codereview.chromium.org/2941453002
Cr-Commit-Position: refs/heads/master@{#478966}
1 file changed
