On Sat, May 31, 2025 at 2:41 AM Eduard Stefes wrote:
>
> On Thu, 2025-05-08 at 05:23 +0700, John Naylor wrote:
>
> > This case is a bit different, since Arm can compute hardware CRC on
> > any input size. The fast path here is only guaranteed to be taken at
> > inputs of 79 bytes or bigger, with t
On Thu, 2025-05-08 at 05:23 +0700, John Naylor wrote:
> Right. That's also true of PowerPC -- perhaps that's a different team
> than yours?
indeed that's another team. I can ping them but there is not much more
I can do.
> This case is a bit different, since Arm can compute hardware CRC on
> any
On Tue, May 27, 2025 at 3:24 AM Eduard Stefes wrote:
> So I worked on the algorithm to also work on buffers between 16-64
> bytes. Then I ran the performance measurement on two
> dataset[^raw_data_1] [^raw_data_2]. And created two diagrams
> [^attachment].
>
> my findings so far:
>
> - the optimiz
Hi,
So I worked on the algorithm to also work on buffers between 16-64
bytes. Then I ran the performance measurement on two
dataset[^raw_data_1] [^raw_data_2]. And created two diagrams
[^attachment].
my findings so far:
- the optimized crc32cvx is faster
- the sb8 performance is heavily dependi
John Naylor writes:
> We prefer that 'a' and 'b' are declared as global variables, just to
> make it as realistic as possible, although it doesn't seem to make
> much difference when I tried it on Compiler Explorer. (Same for
> autoconf)
Yeah, see commit fdb5dd6331e305f797bb589747f056062c305f0b f
On Wed, May 7, 2025 at 8:15 PM Aleksander Alekseev
wrote:
>
> I didn't review the patch but wanted to point out that when it comes
> to performance improvements it's typically useful to provide some
> benchmarks.
+1 -- It's good to have concrete numbers for the commit message, and
also to verify
On Wed, May 7, 2025 at 8:04 PM Eduard Stefes wrote:
>
> Hi,
>
> Here I send a patch that adds a vectorized version of CRC32C for the
> IBM S390X hardware. I kindly ask for a review of the code and to pick
> it up in upstream postgresql.
Thanks!
> # Why this patch:
>
> We noticed that postgres ru
Hi Eduard,
> Here I send a patch that adds a vectorized version of CRC32C for the
> IBM S390X hardware. I kindly ask for a review of the code and to pick
> it up in upstream postgresql.
> [...]
> Cheers and thanks to all for their work,
Thanks for submitting this patch.
Please register it on the
Hi,
Here I send a patch that adds a vectorized version of CRC32C for the
IBM S390X hardware. I kindly ask for a review of the code and to pick
it up in upstream postgresql.
# Why this patch:
We noticed that postgres running on an S390X will spend much longer in
CRC32C as compared to other platf
