RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-26 14:33:02 +02:00
Code Issues Projects Releases Wiki Activity
67,164 Commits 4 Branches 5 Tags 704 MiB
09c1d135be
Commit Graph

43337 Commits

Author SHA1 Message Date
Bill Wendling
09c1d135be Missed the _RET versions of LDMIA. 
llvm-svn: 119726
 2010-11-18 19:44:29 +00:00
Eric Christopher
bc6a51d63f Rewrite stack callee saved spills and restores to use push/pop instructions. 
Remove movePastCSLoadStoreOps and associated code for simple pointer
increments. Update routines that depended upon other opcodes for save/restore.

Adjust all testcases accordingly.

llvm-svn: 119725
 2010-11-18 19:40:05 +00:00
Owen Anderson
598d36b571 Fix an order-of-deallocation issue where the AttrListImpl could be deallocated before the global 
LLVMContext, causing memory errors.  Patch by Peter Collingbourne.

llvm-svn: 119721
 2010-11-18 18:59:13 +00:00
Owen Anderson
cea4068700 Use thread-safe statics to avoid a static constructor here. This isn't thread-safe on MSVC, but we don't 
support threaded LLVM there anyways.

llvm-svn: 119718
 2010-11-18 18:49:05 +00:00
Dan Gohman
3b19bfe496 Oops, missed this file when remaing ExpandPseudos to ExpandISelPseudos. 
llvm-svn: 119717
 2010-11-18 18:48:28 +00:00
Dan Gohman
3998a0430f Rename ExpandPseudos to ExpandISelPseudos to help clarify its role. 
llvm-svn: 119716
 2010-11-18 18:45:06 +00:00
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
Jim Grosbach
fd0bab72a7 ARMPseudoInst instructions should default to being considered a single 4-byte 
instruction. Any that may be expanded otherwise by MC lowering should
override this value. rdar://8683274

llvm-svn: 119713
 2010-11-18 18:01:40 +00:00
Dan Gohman
b0ffd6beca Fix typos. 
llvm-svn: 119712
 2010-11-18 17:44:17 +00:00
Dan Gohman
3f08bf5bea Bounds-check APInt's operator[]. 
llvm-svn: 119708
 2010-11-18 17:14:56 +00:00
Dan Gohman
3a630f4051 ExpandPseudos doesn't have any dependencies, so it can use the 
simple form of INITIALIZE_PASS.

llvm-svn: 119707
 2010-11-18 17:14:05 +00:00
Dan Gohman
87b78a4726 Strip trailing whitespace. 
llvm-svn: 119706
 2010-11-18 17:06:31 +00:00
Dan Gohman
99ac1c6b83 Use llvm_unreachable for "impossible" situations. 
llvm-svn: 119705
 2010-11-18 17:05:57 +00:00
Dan Gohman
ec75e876ab Add support for PHI-translating sext, zext, and trunc instructions, 
enabling more PRE. PR8586.

llvm-svn: 119704
 2010-11-18 17:05:13 +00:00
Chris Lattner
2034d275aa slightly simplify code and substantially improve comment. Instead of 
saying "it would be bad", give an example of what is going on.

llvm-svn: 119695
 2010-11-18 08:07:09 +00:00
Chris Lattner
c752718881 remove a pointless restriction from memcpyopt. It was 
refusing to optimize two memcpy's like this:

copy A <- B
copy C <- A

if it couldn't prove that noalias(B,C).  We can eliminate
the copy by producing a memmove instead of memcpy.

llvm-svn: 119694
 2010-11-18 08:00:57 +00:00
Chris Lattner
eb29c52bce remove another pointless noalias check: M is a memcpy, so the 
source and dest are known to not overlap.

llvm-svn: 119692
 2010-11-18 07:39:57 +00:00
Chris Lattner
4d08597975 use AA::isNoAlias instead of open coding it. Remove an extraneous noalias check: 
there is no need to check to see if the source and dest of a memcpy are noalias,
behavior is undefined if not.

llvm-svn: 119691
 2010-11-18 07:38:43 +00:00
Chris Lattner
f8540ee386 finish a thought. 
llvm-svn: 119690
 2010-11-18 07:32:33 +00:00
Chris Lattner
c3e29a9a68 rearrange some code, splitting memcpy/memcpy optimization 
out of processMemCpy into its own function.

llvm-svn: 119687
 2010-11-18 07:02:37 +00:00
Chris Lattner
6048697a30 allow eliminating an alloca that is just copied from an constant global 
if it is passed as a byval argument.  The byval argument will just be a
read, so it is safe to read from the original global instead.  This allows
us to promote away the %agg.tmp alloca in PR8582

llvm-svn: 119686
 2010-11-18 06:41:51 +00:00
Chris Lattner
791e914b1b enhance the "alloca is just a memcpy from constant global" 
to ignore calls that obviously can't modify the alloca
because they are readonly/readnone.

llvm-svn: 119683
 2010-11-18 06:26:49 +00:00
Chris Lattner
44ccd4643d fix a small oversight in the "eliminate memcpy from constant global" 
optimization.  If the alloca that is "memcpy'd from constant" also has
a memcpy from *it*, ignore it: it is a load.  We now optimize the testcase to:

define void @test2() {
  %B = alloca %T
  %a = bitcast %T* @G to i8*
  %b = bitcast %T* %B to i8*
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %b, i8* %a, i64 124, i32 4, i1 false)
  call void @bar(i8* %b)
  ret void
}

previously we would generate:

define void @test() {
  %B = alloca %T
  %b = bitcast %T* %B to i8*
  %G.0 = getelementptr inbounds %T* @G, i32 0, i32 0
  %tmp3 = load i8* %G.0, align 4
  %G.1 = getelementptr inbounds %T* @G, i32 0, i32 1
  %G.15 = bitcast [123 x i8]* %G.1 to i8*
  %1 = bitcast [123 x i8]* %G.1 to i984*
  %srcval = load i984* %1, align 1
  %B.0 = getelementptr inbounds %T* %B, i32 0, i32 0
  store i8 %tmp3, i8* %B.0, align 4
  %B.1 = getelementptr inbounds %T* %B, i32 0, i32 1
  %B.12 = bitcast [123 x i8]* %B.1 to i8*
  %2 = bitcast [123 x i8]* %B.1 to i984*
  store i984 %srcval, i984* %2, align 1
  call void @bar(i8* %b)
  ret void
}

llvm-svn: 119682
 2010-11-18 06:20:47 +00:00
Chris Lattner
c85f76f7da trivial QoI improvement. On this invalid input: 
sahf	movl	344(%rdi),%r14d
we used to produce:

t.s:2:1: error: unexpected token in argument list

^

we now produce:
t.s:1:11: error: unexpected token in argument list
sahf	movl	344(%rdi),%r14d
    	    	^

rdar://8581401

llvm-svn: 119676
 2010-11-18 02:53:02 +00:00
Rafael Espindola
93a07b464e Change CodeGen to use .loc directives. This produces a lot more readable output 
and testing is easier.  A good example is the unknown-location.ll test that
now can just look for ".loc 1 0 0".  We also don't use a DW_LNE_set_address for
every address change anymore.

llvm-svn: 119613
 2010-11-18 02:04:25 +00:00
Evan Cheng
6b2be51f7e Silence compiler warnings. 
llvm-svn: 119610
 2010-11-18 01:43:23 +00:00
Jim Grosbach
082e9f2f2c Remove trailing whitespace. 
llvm-svn: 119608
 2010-11-18 01:39:50 +00:00
Jim Grosbach
2f9a2efb3c ARM PseudoInst instructions don't need or use an assembler string. Get rid of 
the operand to the pattern.

llvm-svn: 119607
 2010-11-18 01:38:26 +00:00
Dale Johannesen
06f479d543 Do not throw away alignment when generating the DAG for 
memset; we may need it to decide between MOVAPS and MOVUPS
later.  Adjust a test that was looking for wrong code.
PR 3866 / 8675131.

llvm-svn: 119605
 2010-11-18 01:35:23 +00:00
Evan Cheng
e63f6c7422 Code clean up. 
llvm-svn: 119604
 2010-11-18 01:28:51 +00:00
Jim Grosbach
205e0345c1 Add FIXME. 
llvm-svn: 119603
 2010-11-18 01:20:48 +00:00
Jim Grosbach
a42e2b0fcb Refactor the ARM PICADD and PICLDR* instructions to really be pseudos and not 
just pretend to be.

llvm-svn: 119602
 2010-11-18 01:15:56 +00:00
Owen Anderson
b7970b2c6a Try again at providing Thumb2 encodings for basic multiplication operators. 
llvm-svn: 119601
 2010-11-18 01:08:42 +00:00
Jim Grosbach
42d103250b Refactor a few ARM load instructions to better parameterize things and re-use 
common encoding information.

llvm-svn: 119598
 2010-11-18 00:46:58 +00:00
Owen Anderson
e8906ba112 Revert r119593 while I figure out my testing disagrees with the buildbot. 
llvm-svn: 119597
 2010-11-18 00:42:51 +00:00
Dan Gohman
3213622610 Introduce memoization for ScalarEvolution dominates and properlyDominates 
queries, and SCEVExpander getRelevantLoop queries.

llvm-svn: 119595
 2010-11-18 00:34:22 +00:00
Owen Anderson
47a64ab90c Provide correct Thumb2 encodings for basic multiplication operators. 
llvm-svn: 119593
 2010-11-18 00:19:10 +00:00
John Thompson
b33f935bc3 Bug 8621 fix - pointer cast stripped from inline asm constraint argument. 
llvm-svn: 119590
 2010-11-17 23:58:47 +00:00
Jim Grosbach
5d9d8356fa Clean up LEApcrel instuction(s) a bit. It's not really a Pseudo, so don't mark 
it as such. Add some encoding information.

llvm-svn: 119588
 2010-11-17 23:33:14 +00:00
Dan Gohman
ecff21cc67 Factor out the code for purging a SCEV from all the various memoization maps. 
Some of these maps may merge in the future, but for now it's convenient to have
a utility function for them.

llvm-svn: 119587
 2010-11-17 23:28:48 +00:00
Dan Gohman
8e594dfe39 Merge the implementations of isLoopInvariant and hasComputableLoopEvolution, and 
memoize the results. This improves compile time in code which highly complex
expressions which get queried many times.

llvm-svn: 119584
 2010-11-17 23:21:44 +00:00
Dan Gohman
fb6ea18d0a Make SCEV::getType() and SCEV::print non-virtual. Move SCEV::hasOperand 
to ScalarEvolution. Delete SCEV::~SCEV. SCEV is no longer virtual.

llvm-svn: 119578
 2010-11-17 22:27:42 +00:00
Owen Anderson
ea6ac4cdff Second attempt at correct encodings for Thumb2 bitfield instructions. 
llvm-svn: 119575
 2010-11-17 22:16:31 +00:00
Jim Grosbach
16aeabf0d5 Fix comment typo. 
llvm-svn: 119573
 2010-11-17 21:57:51 +00:00
Dan Gohman
9bbb0fa515 Move SCEV::dominates and properlyDominates to ScalarEvolution. 
llvm-svn: 119570
 2010-11-17 21:41:58 +00:00
Bob Wilson
a217a6e40b Change ARMGlobalMerge to keep BSS globals in separate pools. 
This completes the fixes for Radar 8673120.

llvm-svn: 119566
 2010-11-17 21:25:39 +00:00
Bob Wilson
819660b716 Fix ARMGlobalMerge pass to check if globals are entirely within range. 
It is generally not sufficient to check if the starting offset is in range
of the maximum offset that can be efficiently used for the target.

llvm-svn: 119565
 2010-11-17 21:25:36 +00:00
Bob Wilson
49bf8702f0 Change the symbol for merged globals from "merged" to "_MergedGlobals". 
This makes it more clear that the symbol is an internal, compiler-generated
name and gives a little more description about its contents.

llvm-svn: 119564
 2010-11-17 21:25:33 +00:00
Bob Wilson
fd7399b6a6 Fix the ARMGlobalMerge pass to look at variable sizes instead of pointer sizes. 
It was mistakenly looking at the pointer type when checking for the size of
global variables.  This is a partial fix for Radar 8673120.

llvm-svn: 119563
 2010-11-17 21:25:27 +00:00
Dan Gohman
04df5af12b Move SCEV::isLoopInvariant and hasComputableLoopEvolution to be member 
functions of ScalarEvolution, in preparation for memoization and
other optimizations.

llvm-svn: 119562
 2010-11-17 21:23:15 +00:00