Dodge portability issue (apparent compiler bug) in new tablesample code.

Some of the older OS X critters in the buildfarm are failing regression,
with symptoms showing that a request for 100% sampling in BERNOULLI or
SYSTEM methods actually gets only around 50% of the table.  gdb revealed
that the computation of the "cutoff" number was producing 0x7FFFFFFF
rather than the expected 0x100000000.  Inspecting the assembly code,
it looks like gcc is trying to use lrint() instead of rint() and then
fumbling the conversion from long double to uint64.  This seems like a
clear compiler bug, but assigning the intermediate result into a plain
double variable works around it, so let's just do that.  (Another idea
would be to give up one bit of hash width so that we don't need to use
a uint64 cutoff, but let's see if this is enough.)
This commit is contained in:
Tom Lane 2015-07-25 19:42:32 -04:00
parent caef94d59f
commit d9476b8380
2 changed files with 6 additions and 2 deletions

View File

@ -144,6 +144,7 @@ bernoulli_beginsamplescan(SampleScanState *node,
{
BernoulliSamplerData *sampler = (BernoulliSamplerData *) node->tsm_state;
double percent = DatumGetFloat4(params[0]);
double dcutoff;
if (percent < 0 || percent > 100 || isnan(percent))
ereport(ERROR,
@ -155,7 +156,8 @@ bernoulli_beginsamplescan(SampleScanState *node,
* store that as a uint64, of course. Note that this gives strictly
* correct behavior at the limits of zero or one probability.
*/
sampler->cutoff = rint(((double) PG_UINT32_MAX + 1) * percent / 100);
dcutoff = rint(((double) PG_UINT32_MAX + 1) * percent / 100);
sampler->cutoff = (uint64) dcutoff;
sampler->seed = seed;
sampler->lt = InvalidOffsetNumber;

View File

@ -148,6 +148,7 @@ system_beginsamplescan(SampleScanState *node,
{
SystemSamplerData *sampler = (SystemSamplerData *) node->tsm_state;
double percent = DatumGetFloat4(params[0]);
double dcutoff;
if (percent < 0 || percent > 100 || isnan(percent))
ereport(ERROR,
@ -159,7 +160,8 @@ system_beginsamplescan(SampleScanState *node,
* store that as a uint64, of course. Note that this gives strictly
* correct behavior at the limits of zero or one probability.
*/
sampler->cutoff = rint(((double) PG_UINT32_MAX + 1) * percent / 100);
dcutoff = rint(((double) PG_UINT32_MAX + 1) * percent / 100);
sampler->cutoff = (uint64) dcutoff;
sampler->seed = seed;
sampler->nextblock = 0;
sampler->lt = InvalidOffsetNumber;