testsuite: Bump bytes allocated for T5837
authorBen Gamari <ben@smart-cactus.org>
Wed, 8 Feb 2017 02:30:15 +0000 (21:30 -0500)
committerBen Gamari <ben@smart-cactus.org>
Wed, 8 Feb 2017 03:38:34 +0000 (22:38 -0500)
Simon decreased this earlier today but Harbormaster doesn't reproduce his
number. I've done two things here:

 1. increased the allocations number to the Harbormaster value
 2. increased the acceptance threshold from 5% to 7%, since Simon saw a 6.6%
    change in his environment.

testsuite/tests/perf/compiler/all.T

index a9464fc..1df1eae 100644 (file)
@@ -608,7 +608,7 @@ test('T5837',
              # 2014-12-08: 115905208  Constraint solver perf improvements (esp kick-out)
              # 2016-04-06: 24199320  (x86/Linux, 64-bit machine) TypeInType
 
-           (wordsize(64), 50253880, 5)])
+           (wordsize(64), 53592736, 7)])
              # sample: 3926235424 (amd64/Linux, 15/2/2012)
              # 2012-10-02 81879216
              # 2012-09-20 87254264 amd64/Linux
@@ -638,6 +638,9 @@ test('T5837',
              #                      all the gain here is from this patch, but I think it
              #                      just pushed it over the edge, so I'm re-centreing, and
              #                      changing to 5% tolerance
+             # 2017-02-07 53592736  amd64/Linux Simon's earlier decrease appears
+             #                      to be environmentally-dependent.
+             #                      Also bumped acceptance threshold to 7%.
       ],
       compile, ['-freduction-depth=50'])