Re-center perf numbers for haddock.compiler
authorThomas Miedema <thomasmiedema@gmail.com>
Wed, 3 Jun 2015 23:42:29 +0000 (01:42 +0200)
committerThomas Miedema <thomasmiedema@gmail.com>
Thu, 4 Jun 2015 21:52:41 +0000 (23:52 +0200)
Summary:
Phabricator/Harbormaster measured the following allocation numbers for
haddock.compiler:

Expected:     33562468736
June 2nd [1]: 36740649320
Upper bound:  36918715610
June 3rd [2]: 36956620504

So although this test didn't start failing until June 2nd/3rd, the
biggest increase in allocation must have occured sometime before that.

[1] 2f0011aca137055f139bed484302679c10238d55
[2] 942cfa4e3257347dfc4644ce1a8a28db1fb0aee0

Test Plan: validate

Reviewers: austin

Subscribers: bgamari, thomie

Differential Revision: https://phabricator.haskell.org/D948

testsuite/tests/perf/haddock/all.T

index 5d1e6a0..aaa7c55 100644 (file)
@@ -86,7 +86,7 @@ test('haddock.Cabal',
 test('haddock.compiler',
      [unless(in_tree_compiler(), skip), req_haddock
      ,stats_num_field('bytes allocated',
-          [(wordsize(64), 33562468736, 10)
+          [(wordsize(64), 36740649320, 10)
             # 2012P-08-14: 26070600504 (amd64/Linux)
             # 2012-08-29: 26353100288 (amd64/Linux, new CG)
             # 2012-09-18: 26882813032 (amd64/Linux)
@@ -95,6 +95,7 @@ test('haddock.compiler',
             # 2012-11-27: 28708374824 (amd64/Linux)
             # 2014-09-10: 30353349160 (amd64/Linux) post-AMP cleanup
             # 2014-11-22: 33562468736 (amd64/Linux)
+            # 2015-06-02: 36740649320 (amd64/Linux) unknown cause
 
           ,(platform('i386-unknown-mingw32'),   902576468, 10)
             # 2012-10-30:                     13773051312 (x86/Windows)