On Sat, Apr 26, 2025 at 5:43 AM Tom Lane wrote:
> Andres Freund writes:
> > It's kinda sad to not have any test that tests a larger
> > io_combine_limit/io_max_combine_limit - as evidenced by this bug that'd be
> > good. However, not all platforms have PG_IOV_MAX > 16, so it seems like it'd
> > b
Andres Freund writes:
> It's kinda sad to not have any test that tests a larger
> io_combine_limit/io_max_combine_limit - as evidenced by this bug that'd be
> good. However, not all platforms have PG_IOV_MAX > 16, so it seems like it'd
> be somewhat painful to test?
Maybe just skip the test if th
Hi,
On 2025-04-25 10:26:09 -0400, Tom Lane wrote:
> Andres Freund writes:
> > void
> > assign_io_max_combine_limit(int newval, void *extra)
> > {
> > io_max_combine_limit = newval;
> > io_combine_limit = Min(io_max_combine_limit, io_combine_limit_guc);
> > }
> > void
> > assign_io_combine
Andres Freund writes:
> void
> assign_io_max_combine_limit(int newval, void *extra)
> {
> io_max_combine_limit = newval;
> io_combine_limit = Min(io_max_combine_limit, io_combine_limit_guc);
> }
> void
> assign_io_combine_limit(int newval, void *extra)
> {
> io_combine_limit_guc
Hi,
On 2025-03-18 16:18:17 +1300, Thomas Munro wrote:
> Here's a new version that also adjusts the code that just landed in
> da722699:
Something isn't quite right with this code. If I just add -c
io_combine_limit=32 to the options and do a seqscan, I get odd
failures. Mostly assertion failures
Here's a new version that also adjusts the code that just landed in da722699:
- /*
-* Each IO handle can have an PG_IOV_MAX long iovec.
-*
-* XXX: Right now the amount of space available for each IO is
PG_IOV_MAX.
-* While it's tempting to use the io_combine_l
On Wed, Feb 12, 2025 at 3:24 PM Thomas Munro wrote:
> On Wed, Feb 12, 2025 at 3:22 PM Andres Freund wrote:
> > On 2025-02-12 13:59:21 +1300, Thomas Munro wrote:
> > > How about just maintaining it in a new variable
> > > effective_io_combine_limit, whenever either of them is assigned?
> >
> > Yea
Hi,
On 2025-02-14 09:32:32 +0100, Jakub Wartak wrote:
> On Wed, Feb 12, 2025 at 1:03 AM Andres Freund wrote:
> > FWIW, I see substantial performance *regressions* with *big* IO sizes using
> > fio. Just looking at cached buffered IO.
> >
> > for s in 4 8 16 32 64 128 256 512 1024 2048 4096 8192;d
On Wed, Feb 12, 2025 at 1:03 AM Andres Freund wrote:
>
> Hi,
>
> On 2025-02-11 13:12:17 +1300, Thomas Munro wrote:
> > Tomas queried[1] the limit of 256kB (or really 32 blocks) for
> > io_combine_limit. Yeah, I think we should increase it and allow
> > experimentation with larger numbers. Note t
Hi,
On 2025-02-12 15:24:21 +1300, Thomas Munro wrote:
> On Wed, Feb 12, 2025 at 3:22 PM Andres Freund wrote:
> > On 2025-02-12 13:59:21 +1300, Thomas Munro wrote:
> > > How about just maintaining it in a new variable
> > > effective_io_combine_limit, whenever either of them is assigned?
> >
> > Y
On Wed, Feb 12, 2025 at 3:22 PM Andres Freund wrote:
> On 2025-02-12 13:59:21 +1300, Thomas Munro wrote:
> > How about just maintaining it in a new variable
> > effective_io_combine_limit, whenever either of them is assigned?
>
> Yea, that's probably the least bad way.
>
> I wonder if we should ju
Hi,
On 2025-02-12 13:59:21 +1300, Thomas Munro wrote:
> How about just maintaining it in a new variable
> effective_io_combine_limit, whenever either of them is assigned?
Yea, that's probably the least bad way.
I wonder if we should just name that variable io_combine_limit and have the
GUC be _r
On Wed, Feb 12, 2025 at 1:03 PM Andres Freund wrote:
> On 2025-02-11 13:12:17 +1300, Thomas Munro wrote:
> > I was also
> > anticipating future code that would need to multiply that number by other
> > terms to allocate shared memory, but after some off-list discussion, that
> > seems OK: such cod
Hi,
On 2025-02-11 13:12:17 +1300, Thomas Munro wrote:
> Tomas queried[1] the limit of 256kB (or really 32 blocks) for
> io_combine_limit. Yeah, I think we should increase it and allow
> experimentation with larger numbers. Note that real hardware and
> protocols have segment and size limits that
Hi,
Tomas queried[1] the limit of 256kB (or really 32 blocks) for
io_combine_limit. Yeah, I think we should increase it and allow
experimentation with larger numbers. Note that real hardware and
protocols have segment and size limits that can force the kernel to
split your I/Os, so it's not at a
15 matches
Mail list logo
