Permanently accept the Haddock performance number bump, and add some TODOs
authorEdward Z. Yang <ezyang@cs.stanford.edu>
Thu, 7 Aug 2014 13:15:13 +0000 (14:15 +0100)
committerEdward Z. Yang <ezyang@cs.stanford.edu>
Thu, 7 Aug 2014 13:15:13 +0000 (14:15 +0100)
commitd026e9e889f347a86425c34b61bfb443a409d1e3
tree03c18113fc33a23532ff0df0acfe44eda46627c6
parent4855be0d0d1ac90f836a6fb54f4034f478e38fd8
Permanently accept the Haddock performance number bump, and add some TODOs

I bisected the performance difference in Haddock and found it was due to
d6aec63c009c4e57181900eb03847d7dc0fc3c7c, which I accidentally picked up
when updating Haddock 00b8f8c5b378fc679639ebe81238cf42d92aa607.  The
performance regression is justified by the fact that we are now actually
processing URLs in Haddock comments that we were not previously, so
there would be more allocation.  Time use was not affected.

The TODOs simply reflect the fact that we need updated numbers for
32-bit Linux and Windows.  Please add them when you get a chance.

Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>
testsuite/tests/perf/haddock/all.T