On 12/30/2014 09:40 AM, Abhijit Menon-Sen wrote:
> Hi.
>
> I'm re-attaching the two patches as produced by format-patch. I haven't
> listed any reviewers. It's either just Andres, or maybe a lot of people.
>
> Is anyone in a position to try out the patches on MSVC and see if they
> build and work sensibly, please? (Otherwise it may be better to remove
> those bits from the patch for now.)
A couple of quick comments:
bswap32 is unused on on little-endian systems. That will give a compiler
warning.
pg_comp_crc32c_sse processes one byte at a time, until the pointer is
4-bytes aligned. Then it processes 8 bytes at a time. So it fetches the
8-byte chunks from only 4-byte aligned addresses. Is that intentional?
If unaligned access performs well, why bother with the initial
byte-at-a-time processing at all?
- Heikki