On Tue, Feb 4, 2020 at 2:31 PM Andres Freund wrote:
> How much of the benefit here comes from the prefetching, and how much
> just from writing the code in a manner that allows for more out-of-order
> execution? Because there's no dependencies preventing execution of the
> next queued tuple anymor
HI,
On 2020-02-04 01:48:49 +1300, Thomas Munro wrote:
> On Fri, Apr 12, 2019 at 4:23 PM Thomas Munro wrote:
> > ... if we also prefetch during
> > the build phase ...
>
> Here's an experimental patch to investigate just that part. I tried
> initiating a prefetch of the bucket just before we copy
On Fri, Apr 12, 2019 at 4:23 PM Thomas Munro wrote:
> ... if we also prefetch during
> the build phase ...
Here's an experimental patch to investigate just that part. I tried
initiating a prefetch of the bucket just before we copy the tuple and
then finally insert it, but it doesn't seem to be f
On Fri, Apr 12, 2019 at 1:35 AM Robert Haas wrote:
> It would be interesting to see how this does with moderately-long text
> keys, say 32 or 64 byte strings, and with actually-long text keys, say
> several kB, and then with gigantic text keys, say several MB. At some
> point the key probably get
On Wed, Apr 10, 2019 at 2:10 AM Thomas Munro wrote:
> Here is an example of times for a trivial join on my laptop. Note
> that this is prefetching only the probing phase, not while building
> which should also be possible. I didn't get around to trying deeper
> prefetch pipelines as discussed ea
On Sun, Oct 14, 2018 at 11:11 PM Andrey Borodin wrote:
> > 14 окт. 2018 г., в 9:18, Thomas Munro
> > написал(а):
> >
> > + /* Prefetch the bucket for the next key */
> > + uint32 next_hash = hash_uint32(DatumGetInt32(keyval) + 1);
> > + uint32 next_bucke
On Mon, Oct 15, 2018 at 12:16 AM Dmitry Dolgov <9erthali...@gmail.com> wrote:
> > On Sun, 14 Oct 2018 at 06:19, Thomas Munro
> > wrote:
> > Cache-oblivious hash joins cause a lot of TLB and cache misses.
> > ...
> > (There is another class of cache-aware hash join algorithms that partition
> > ca
> On Sun, 14 Oct 2018 at 06:19, Thomas Munro
> wrote:
>
> Cache-oblivious hash joins cause a lot of TLB and cache misses.
> ...
> (There is another class of cache-aware hash join algorithms that partition
> carefully up front to avoid them; that's not us.)
Just out of curiosity, can you please e
Hi, Thomas!
> 14 окт. 2018 г., в 9:18, Thomas Munro
> написал(а):
>
> + /* Prefetch the bucket for the next key */
> + uint32 next_hash = hash_uint32(DatumGetInt32(keyval) + 1);
> + uint32 next_bucket = next_hash % hashtable->nbuckets;
> +
