On Wed, 2020-03-04 at 12:06 -0800, Jeff Davis wrote:
> I tracked the problem down.
Because of the name _FORTIFY_SOURCE, I got somewhat concerned that this
change (logtape-20200303) was somehow interfering with a safety
mechanism in the compiler.
There's a mechanism in GCC called object size track
On Wed, 2020-03-04 at 11:57 +0800, Adam Lee wrote:
> Master(e537aed61d): 13342.844 ms 13195.982 ms 13271.023 ms
> With my patch(a pointer): 13020.029 ms 13008.158 ms 13063.658 ms
> With your patch(flexible array): 12870.117 ms 12814.725 ms 13119.255
> ms
I tracked the problem down.
When we change
On Tue, Mar 03, 2020 at 08:46:24AM -0800, Jeff Davis wrote:
> On Tue, 2020-03-03 at 09:49 +0800, Adam Lee wrote:
> > Master: 12729ms 12970ms 12999ms
> > With my patch(a pointer): 12965ms 13273ms 13116ms
> > With your patch(flexible array): 12906ms 12991ms 13043ms
>
> Hmm.. looks like you didn't re
On Tue, 2020-03-03 at 09:49 +0800, Adam Lee wrote:
> Master: 12729ms 12970ms 12999ms
> With my patch(a pointer): 12965ms 13273ms 13116ms
> With your patch(flexible array): 12906ms 12991ms 13043ms
Hmm.. looks like you didn't reproduce the difference I saw. What
compiler/version?
Regards,
J
On Fri, Feb 28, 2020 at 12:38:55PM -0800, Jeff Davis wrote:
> On Fri, 2020-02-28 at 14:16 +0800, Adam Lee wrote:
> > I noticed another difference, I was using palloc0(), which could be
> > one of the
> > reason, but not sure.
>
> I changed the palloc0()'s in your code to plain palloc(), and it did
On Fri, 2020-02-28 at 14:16 +0800, Adam Lee wrote:
> I noticed another difference, I was using palloc0(), which could be
> one of the
> reason, but not sure.
I changed the palloc0()'s in your code to plain palloc(), and it didn't
make any perceptible difference. Still slower than the version I pos
On Thu, Feb 27, 2020 at 01:01:08PM -0800, Jeff Davis wrote:
> Attached is a patch that exports a new function from logtape.c:
>
>extern LogicalTapeSet *LogicalTapeSetExtend(
> LogicalTapeSet *lts, int nAdditional);
>
> The purpose is to allow adding new tapes dynamically for the u
Attached is a patch that exports a new function from logtape.c:
extern LogicalTapeSet *LogicalTapeSetExtend(
LogicalTapeSet *lts, int nAdditional);
The purpose is to allow adding new tapes dynamically for the upcoming
Hash Aggregation work[0]. HashAgg doesn't know in advance how ma
