RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-04-11 20:22:02 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
2,440 Commits 6 Branches 6 Tags
Commit Graph

361 Commits

Author SHA1 Message Date
Chris Lattner
172488eb68 Add ability to transform (x - (y - z)) into (x + (z - y)) 
llvm-svn: 2566
 2002-05-09 01:29:19 +00:00
Chris Lattner
34d5891f0a * Combine: A-(-B) -> A + B 
* Bugfix:  A + -B and -A + B

llvm-svn: 2561
 2002-05-08 22:46:53 +00:00
Chris Lattner
c38d2de2b5 Initial checkin of expression reassociation pass 
llvm-svn: 2559
 2002-05-08 22:19:27 +00:00
Chris Lattner
ce6632281f Give the unified exit node a name 
llvm-svn: 2550
 2002-05-07 22:14:45 +00:00
Chris Lattner
bbd8a94a9b Fix bug: test/Regression/Transforms/ADCE/2002-01-31-UseStuckAround.ll 
Cleanup code a lot

llvm-svn: 2547
 2002-05-07 22:11:39 +00:00
Chris Lattner
2a43c7c03d Merge all individual .h files into a single Scalar.h file 
llvm-svn: 2537
 2002-05-07 20:03:00 +00:00
Chris Lattner
e615c275c5 This doesn't use DCE! 
llvm-svn: 2536
 2002-05-07 20:02:44 +00:00
Chris Lattner
f7806b0212 Move UnifyFunctionExitNodes to Utils library: final resting place this time 
llvm-svn: 2531
 2002-05-07 19:18:48 +00:00
Chris Lattner
276dd3d721 fix comments and documentation in file 
llvm-svn: 2530
 2002-05-07 19:04:39 +00:00
Chris Lattner
e125e808a8 Split ChangeAllocations.cpp into Raise & LowerAllocations.cpp 
llvm-svn: 2529
 2002-05-07 19:02:48 +00:00
Chris Lattner
6ccbd5804a Cleanup implementation a bit 
llvm-svn: 2526
 2002-05-07 18:51:25 +00:00
Chris Lattner
9e1c79995e Updates to move some header files out of include/llvm/Transforms into 
the Scalar and Utils subdirectories

llvm-svn: 2523
 2002-05-07 18:36:35 +00:00
Chris Lattner
c5a6d978b8 Reduce dependance on TransformInternals.h, instead using the TransformUtils library 
llvm-svn: 2518
 2002-05-07 18:12:18 +00:00
Chris Lattner
a35a291f7b Add #include no longer included by llvm/Transforms/Scalar/ConstantProp.h 
llvm-svn: 2517
 2002-05-07 18:11:30 +00:00
Chris Lattner
c099f62190 Factor code out to the TransformUtils library 
llvm-svn: 2516
 2002-05-07 18:10:55 +00:00
Chris Lattner
7b20305ef9 Factor code out of TransformInternals into the Transform Utils library 
llvm-svn: 2515
 2002-05-07 18:09:15 +00:00
Chris Lattner
e817245564 Build Transforms Utils library 
llvm-svn: 2514
 2002-05-07 18:08:50 +00:00
Chris Lattner
d5477af051 Add code pulled out of TransformInternals.cpp, ConstProp.cpp, and DCE.cpp 
llvm-svn: 2513
 2002-05-07 18:07:59 +00:00
Chris Lattner
394781b660 * Convert InstWorkList to vector instead of set, because on big programs it 
is empirically faster by a noticable margin, even though duplicates can
  happen.

llvm-svn: 2511
 2002-05-07 04:29:32 +00:00
Chris Lattner
4a830e1a4a * Remove all cfg simplification stuff for a new cfg simplify pass (todo) 
* Convert to worklist instead of iterative algorithm

llvm-svn: 2510
 2002-05-07 04:24:11 +00:00
Chris Lattner
ca538273d4 Only do masking for unsigned values! 
llvm-svn: 2504
 2002-05-06 18:54:59 +00:00
Chris Lattner
66ff017174 * Eliminate dead code that should have been removed in last revision 
* Convert main constant propogation pass to be worklist driven instead of
  iterative.
* -constprop pass no longer "constant propogates" terminator instructions
   - CFG is now preserved!

llvm-svn: 2502
 2002-05-06 18:21:31 +00:00
Chris Lattner
24ebec1273 Handle X = phi Y --> X = Y 
llvm-svn: 2501
 2002-05-06 18:06:38 +00:00
Chris Lattner
910ca80941 * Use simplified interface to constant propogation stuff. 
* Remove dead PHI case (which could not work due to getNumOperands rather than
  getNumIncomingValues.  This really belongs in InstCombine, anyway so we'll
  move it there.

llvm-svn: 2497
 2002-05-06 17:54:10 +00:00
Chris Lattner
b40c1bf6f7 Clean up ADCE by removing extraneous wrapper class 
llvm-svn: 2494
 2002-05-06 17:27:57 +00:00
Chris Lattner
c5279917bc Combine not (not X) -> X 
llvm-svn: 2493
 2002-05-06 17:03:21 +00:00
Chris Lattner
24a94a50d8 Support more cases... 
llvm-svn: 2490
 2002-05-06 16:49:18 +00:00
Chris Lattner
24ebdf3f74 Add many more instruction combination simplifications 
llvm-svn: 2484
 2002-05-06 16:14:14 +00:00
Chris Lattner
ef7f8fc474 Merge blocks like this: 
void "test3"(bool %T) {
        br bool %T, label %BB1, label %BB1
BB1:
        ret void
}

llvm-svn: 2472
 2002-05-06 03:02:02 +00:00
Chris Lattner
d12eac0b1b Implement constant propogation of shift instructions 
llvm-svn: 2471
 2002-05-06 03:01:37 +00:00
Chris Lattner
9606fd4830 Ok, really it only takes me 3 times to get this right! 
Now use #if 0 instead of #ifdef 0

llvm-svn: 2448
 2002-05-02 21:49:50 +00:00
Chris Lattner
de442f70ac Oops, disable debugging code all the way. :) 
llvm-svn: 2447
 2002-05-02 21:47:40 +00:00
Chris Lattner
7d4f3ed1f2 * Finish the implementation of isEdgeFeasible this fixes bug: 
test/Regression/Transforms/SCCP/2002-05-02-EdgeFailure.ll
* SCCP now preserves the CFG:  It leaves conditional branches the way they
  are in the program, not simplifying them.  A seperate pass should eliminate
  the potentially dead basic blocks and edges in the CFG.

llvm-svn: 2446
 2002-05-02 21:44:00 +00:00
Chris Lattner
b30088a50d * Enable SCCP debugging to be turned on with a simple change of #define 
* Change worklist to a set so that duplicates cannot happen
* Add support for the invoke instruction
* Avoid marking store and free instructions as overdefined, since they
  cannot produce a value anyway.
* Inline the OperandChangedState implementation
* Add isEdgeFeasible in preparation to fix a bug.  Right now it behaves
  exactly as before.
* Remove obsolete comment about constant pools

llvm-svn: 2445
 2002-05-02 21:18:01 +00:00
Chris Lattner
28a8acc476 * Simplify the code by not bothering to name the folded constant 
* Do not skip the instruction immediately after a folded instruction.  This
  was causing the testcase failure:
	test/Regression/Transforms/SCCP/2002-05-02-MissSecondInst.ll

llvm-svn: 2443
 2002-05-02 20:32:51 +00:00
Chris Lattner
6212dd5c08 ExprTypeConvert will already do this transformation, no need to specialize 
it here.

llvm-svn: 2431
 2002-05-02 17:38:55 +00:00
Chris Lattner
2c44e5bca3 Be a little more efficient, do not generate loads and stores with indices in them. 
llvm-svn: 2430
 2002-05-02 17:38:14 +00:00
Chris Lattner
66f5b12141 Fix bug: test/Regression/Transforms/LevelRaise/2002-05-02-BadCastElimination.ll 
llvm-svn: 2429
 2002-05-02 17:37:34 +00:00
Chris Lattner
b7df910c1a * Add ability to eliminate a bunch of different cascading cast variations 
* Allow elimination of getelementptr X, uint 0 (which is a noop)

llvm-svn: 2428
 2002-05-02 17:06:02 +00:00
Chris Lattner
8a6104c3cb Fixed bug: test/Regression/Transforms/Mem2Reg/2002-05-01-ShouldNotPromoteThisAlloca.ll 
llvm-svn: 2423
 2002-05-01 15:38:23 +00:00
Chris Lattner
5879b0928b Remove unneccesary pass 
llvm-svn: 2420
 2002-04-30 20:52:49 +00:00
Chris Lattner
a6cb49c533 Add folding rules for mul X, 0 and mul X, 2 
llvm-svn: 2417
 2002-04-29 22:24:47 +00:00
Chris Lattner
5d1ac4dbc9 Significantly clean up SCCP pass. Now the two classes are merged and in 
an anonymous namespace where they belong.

llvm-svn: 2415
 2002-04-29 21:26:08 +00:00
Chris Lattner
48e73edf57 Range insertion now returns an iterator 
llvm-svn: 2414
 2002-04-29 21:25:34 +00:00
Chris Lattner
5d17e49ad5 Remove unused files: This is the old induction varaible cannonicalization 
pass built on top of interval analysis

llvm-svn: 2411
 2002-04-29 20:18:00 +00:00
Chris Lattner
4968cc38aa Fix for problem when allocating something like this: 
malloc(100-i);

llvm-svn: 2409
 2002-04-29 20:09:21 +00:00
Chris Lattner
f5d07697e6 changes because iMemory.h no longer #includes DerivedTypes.h 
This only requires Type.h anyway

llvm-svn: 2405
 2002-04-29 18:48:30 +00:00
Chris Lattner
cfc512e714 Eliminate dead global variables 
llvm-svn: 2400
 2002-04-29 18:13:11 +00:00
Chris Lattner
96e0c48175 Eliminate duplicate or unneccesary #include's 
llvm-svn: 2397
 2002-04-29 17:42:12 +00:00
Chris Lattner
3909059b6a Remove broken assertion. 
llvm-svn: 2396
 2002-04-29 16:20:25 +00:00