Bye, Bye Compaction Tables. The benefit compaction tables provides doesn't
authorReid Spencer <rspencer@reidspencer.com>
Tue, 30 Jan 2007 19:36:46 +0000 (19:36 +0000)
committerReid Spencer <rspencer@reidspencer.com>
Tue, 30 Jan 2007 19:36:46 +0000 (19:36 +0000)
commitd2bb887cd193b8b47fb15fedc7a05ab66997d4c5
tree6efd9c6a02e765929fc4d7ca6d337b5d9c9187f8
parent908504347b0565c4d4817af444012be76ba4b76f
Bye, Bye Compaction Tables. The benefit compaction tables provides doesn't
outweight its computational costs. This patch removes all compaction
table handling from the bcreader and bcwriter. For the record, here's the
difference betweeen having and not having compaction tables for some tests:

Test             With       Without   Size Chg
Olden/mst       5,602         5,598      +0.1%
viterbi        18,026        17,795      +1.3%
obsequi       162,133       166,663      -2.8%
burg          224,090       228,148      -1.8%
kimwitu++   4,933,263     5,121,159      -3.8%
176.gcc     8,470,424     9,141,539      -7.3%

It seems that it is more beneficial to larger files, but even on the largest
test case we have (176.gcc) it only amounts ot an I/O saving of 7.3%.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@33661 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Bytecode/Reader/Analyzer.cpp
lib/Bytecode/Reader/Reader.cpp
lib/Bytecode/Reader/Reader.h
lib/Bytecode/Writer/SlotCalculator.cpp
lib/Bytecode/Writer/SlotCalculator.h
lib/Bytecode/Writer/Writer.cpp