Re: use SSE2 for is_valid_ascii

On Wed, Aug 10, 2022 at 01:50:14PM +0700, John Naylor wrote:
> Here is an updated patch using the new USE_SSE2 symbol. The style is
> different from the last one in that each stanza has platform-specific
> code. I wanted to try it this way because is_valid_ascii() is already
> written in SIMD-ish style using general purpose registers and bit
> twiddling, so it seemed natural to see the two side-by-side. Sometimes
> they can share the same comment. If we think this is bad for
> readability, I can go back to one block each, but that way leads to
> duplication of code and it's difficult to see what's different for
> each platform, IMO.

This is a neat patch. I don't know that we need an entirely separate code
block for the USE_SSE2 path, but I do think that a little bit of extra
commentary would improve the readability. IMO the existing comment for the
zero accumulator has the right amount of detail.

+ /*
+ * Set all bits in each lane of the error accumulator where input
+ * bytes are zero.
+ */
+ error_cum = _mm_or_si128(error_cum,
+ _mm_cmpeq_epi8(chunk, _mm_setzero_si128()));

I wonder if reusing a zero vector (instead of creating a new one every
time) has any noticeable effect on performance.

--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com