Re: [PATCH] Hex-coding optimizations using SVE on ARM.

Nathan Bossart <nathandbossart@gmail.com>

From: Nathan Bossart <nathandbossart@gmail.com>

To: "Chiranmoy.Bhattacharya@fujitsu.com" <Chiranmoy.Bhattacharya@fujitsu.com>

Cc: John Naylor <johncnaylorls@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>, David Rowley <dgrowleyml@gmail.com>, "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>, "Devanga.Susmitha@fujitsu.com" <Devanga.Susmitha@fujitsu.com>, "Ragesh.Hajela@fujitsu.com" <Ragesh.Hajela@fujitsu.com>

Date: 2025-09-11T03:12:52Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

Optimize hex_encode() and hex_decode() using SIMD.
- ec8719ccbfcd 19 (unreleased) landed
Speed up hex_encode with bytewise lookup
- e24d77080b36 18.0 landed

Attachments

v7-0001-Optimize-hex_encode-using-SIMD.patch (text/plain)

On Thu, Sep 04, 2025 at 02:55:50PM +0000, Chiranmoy.Bhattacharya@fujitsu.com wrote:
>> I see that there was some discussion about a Neon implementation upthread,
>> but I'm not sure we concluded anything.  For popcount, we first added a
>> Neon version before adding the SVE version, which required more complicated
>> configure/runtime checks.  Presumably Neon is available on more hardware
>> than SVE, so that could be a good place to start here, too.
> 
> We have added the Neon versions of hex encode/decode.

Thanks.  I noticed that this stuff is simple enough that we can use
port/simd.h (with a few added functions).  This is especially nice because
it takes care of x86, too.  The performance gains look similar to what you
reported for v6:

                arm
    buf  | HEAD  | patch | % diff
  -------+-------+-------+--------
      16 |    13 |     6 |   54
      64 |    34 |     9 |   74
     256 |    93 |    25 |   73
    1024 |   281 |    78 |   72
    4096 |  1086 |   227 |   79
   16384 |  4382 |   927 |   79
   65536 | 17455 |  3608 |   79

                x86
    buf  | HEAD  | patch | % diff
  -------+-------+-------+--------
      16 |    10 |     7 |   30
      64 |    29 |     9 |   69
     256 |    81 |    21 |   74
    1024 |   286 |    66 |   77
    4096 |  1106 |   253 |   77
   16384 |  4383 |   980 |   78
   65536 | 17491 |  3886 |   78

I've only modified hex_encode() for now, but I'm optimistic that we can do
something similar for hex_decode().

-- 
nathan