Owen Anderson c2db966e5e Completely rework the datastructure GVN uses to represent the value number to leader mapping. Previously, ... this was a tree of hashtables, and a query recursed into the table for the immediate dominator ad infinitum if the initial lookup failed. This led to really bad performance on tall, narrow CFGs. We can instead replace it with what is conceptually a multimap of value numbers to leaders (actually represented by a hashtable with a list of Value*'s as the value type), and then determine which leader from that set to use very cheaply thanks to the DFS numberings maintained by DominatorTree. Because there are typically few duplicates of a given value, this scan tends to be quite fast. Additionally, we use a custom linked list and BumpPtr allocation to avoid any unnecessary allocation in representing the value-side of the multimap. This change brings with it a 15% (!) improvement in the total running time of GVN on 403.gcc, which I think is pretty good considering that includes all the "real work" being done by MemDep as well. The one downside to this approach is that we can no longer use GVN to perform simple conditional progation, but that seems like an acceptable loss since we now have LVI and CorrelatedValuePropagation to pick up the slack. If you see conditional propagation that's not happening, please file bugs against LVI or CVP. llvm-svn: 119714		2010-11-18 18:32:40 +00:00
..
2007-07-25-DominatedLoop.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2007-07-25-InfiniteLoop.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
2007-07-25-Loop.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2007-07-25-NestedLoop.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2007-07-25-SinglePredecessor.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2007-07-26-InterlockingLoops.ll	Teach InstructionSimplify about phi nodes. I chose to have it simply	2010-11-14 13:30:18 +00:00
2007-07-26-NonRedundant.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2007-07-26-PhiErasure.ll	switch GVN to use SSAUpdater. Besides removing a lot of complexity	2009-10-10 23:50:30 +00:00
2007-07-30-PredIDom.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2007-07-31-NoDomInherit.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
2007-07-31-RedundantPhi.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
2008-02-12-UndefLoad.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2008-02-13-NewPHI.ll	stop using anders-aa	2010-03-01 20:24:50 +00:00
2008-02-24-NonDominatedMemcpy.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2008-02-26-MemCpySize.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2008-07-02-Unreachable.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
2008-12-09-SelfRemove.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2008-12-12-RLE-Crash.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2008-12-14-rle-reanalyze.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2008-12-15-CacheVisited.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2009-01-21-SortInvalidation.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2009-01-22-SortInvalidation.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2009-02-17-LoadPRECrash.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2009-03-10-PREOnVoid.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2009-06-17-InvalidPRE.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
2009-07-13-MemDepSortFail.ll	Correct bogus module triple specifications.	2010-08-30 10:48:29 +00:00
2009-11-12-MemDepMallocBitCast.ll	Memory dependence analysis was incorrectly stopping to scan for stores to a pointer at bitcast uses of a malloc call.	2009-10-13 01:42:53 +00:00
2010-03-31-RedundantPHIs.ll	Teach InstructionSimplify about phi nodes. I chose to have it simply	2010-11-14 13:30:18 +00:00
2010-05-08-OneBit.ll	Fix PR7052, patch by Jakub Staszak!	2010-05-08 20:01:44 +00:00
2010-11-13-Simplify.ll	Testcase to go along with commit 118923 ("Have GVN simplify instructions	2010-11-13 21:33:19 +00:00
basic.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
bitcast-of-call.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
calls-nonlocal.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
calls-readonly.ll	Check that GVN performs this transform even if the calls	2009-10-16 12:18:23 +00:00
crash-no-aa.ll	Fix PR5744, a case where we were getting the pointer size instead of the	2009-12-10 00:11:45 +00:00
crash.ll	a testcase that doesn't crash GVN but could someday.	2010-02-11 05:08:05 +00:00
dg.exp
invariant-simple.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
lifetime-simple.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
load-constant-mem.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
load-pre-align.ll	Remove arm_apcscc from the test files. It is the default and doing this	2010-06-17 15:18:27 +00:00
load-pre-licm.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
local-pre.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
lpre-call-wrap-2.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
lpre-call-wrap.ll	Eliminate uses of %prcontext.	2009-09-05 11:35:16 +00:00
mixed.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
non-local-offset.ll	Enhance GVN to do more precise alias queries for non-local memory	2010-11-10 20:37:15 +00:00
nonescaping-malloc.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
null-aliases-nothing.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
phi-translate.ll	Add support for PHI-translating sext, zext, and trunc instructions,	2010-11-18 17:05:13 +00:00
pre-basic-add.ll	Change tests from "opt %s" to "opt < %s" so that opt doesn't see the	2009-09-11 18:01:28 +00:00
pre-load.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
pre-single-pred.ll	LoadPRE was not properly checking that the load it was PRE'ing post-dominated the block it was being hoisted to.	2010-09-25 05:26:18 +00:00
rle-must-alias.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
rle-no-phi-translate.ll	this test is failing, and is expected to.	2009-11-27 06:36:28 +00:00
rle-nonlocal.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
rle-phi-translate.ll	fix an overly conservative caching issue that caused memdep to	2009-12-19 21:29:22 +00:00
rle-semidominated.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00
rle.ll	Make BasicAliasAnalysis a normal AliasAnalysis implementation which	2010-10-18 18:04:47 +00:00