RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-01 16:33:37 +01:00
Code Issues Projects Releases Wiki Activity
56,344 Commits 4 Branches 5 Tags 704 MiB
5eb0c8a2bd
Commit Graph

35442 Commits

Author SHA1 Message Date
Chris Lattner
f227cb9fe7 simplify some code. 
llvm-svn: 93619
 2010-01-16 01:45:47 +00:00
Bill Wendling
d78fcdd332 Retrying r91337: 
The CIE says that the LSDA point in the FDE section is an "sdata4". That's fine,
but we need it to actually be 4-bytes in the FDE for some platforms. Allow
individual platforms to decide for themselves.

llvm-svn: 93616
 2010-01-16 01:40:55 +00:00
Chris Lattner
ea72cfbd42 eliminate uses of mangler and simplify code. 
llvm-svn: 93615
 2010-01-16 01:40:07 +00:00
Chris Lattner
8813339969 fix inverted conditional 
llvm-svn: 93614
 2010-01-16 01:37:14 +00:00
Chris Lattner
61a95137c9 CurrentFnName is now dead, remove it. 
llvm-svn: 93612
 2010-01-16 01:24:10 +00:00
Chris Lattner
bda0bbf7f9 get pic16 off CurrentFnName 
llvm-svn: 93610
 2010-01-16 01:21:04 +00:00
Chris Lattner
bb21a20f24 remove the string form of printVisibility. 
llvm-svn: 93609
 2010-01-16 01:17:26 +00:00
Chris Lattner
ef45b87192 switch more stuff onto MCSymbols 
llvm-svn: 93608
 2010-01-16 01:12:01 +00:00
Bill Wendling
921cd93e85 Temporarily revert r93581. It was causing failures in the ExecutionEngine tests 
on the build bots.

llvm-svn: 93606
 2010-01-16 01:06:58 +00:00
Chris Lattner
ff1a508835 eliminate uses of deprecated mangler apis 
llvm-svn: 93605
 2010-01-16 01:00:27 +00:00
Chris Lattner
cbe9fe90bd switch X86 target off CurFunctionName and MCIze more. 
Note that the code wasn't calling DecorateCygMingName
when emitting the ".ascii -export" stuff at the end of
file for DLLExported functions.  I don't know if it should
or not, but I'm preserving behavior.

llvm-svn: 93603
 2010-01-16 00:51:39 +00:00
Bob Wilson
a7043345ee Treat indirect branches specially only during pre-regalloc tail duplication, 
not during the later post-alloc tail duplication.

llvm-svn: 93600
 2010-01-16 00:42:25 +00:00
Chris Lattner
0bb1df70ef MCize tis, and make it keep CurrentFnSym up to date with CurrentFnName. 
llvm-svn: 93598
 2010-01-16 00:32:38 +00:00
Bob Wilson
3386047bdb Run the pre-register allocation tail duplication pass by default. Remove 
the -pre-regalloc-taildup command-line option, and add a new
-disable-early-taildup option.

llvm-svn: 93597
 2010-01-16 00:29:50 +00:00
Chris Lattner
e76d575ca2 revert the x86 part of my last patch, cygwin is mutating CurrentFnName! 
llvm-svn: 93595
 2010-01-16 00:24:20 +00:00
Chris Lattner
40eb58664f MCize a bunch more stuff, eliminating a lot of uses of the mangler 
and CurrentFnName.

llvm-svn: 93594
 2010-01-16 00:21:18 +00:00
Chris Lattner
9103a49143 remove two write-only sets. 
llvm-svn: 93590
 2010-01-15 23:56:03 +00:00
Chris Lattner
b1b91d11db supplement CurrentFnName with CurrentFnSym, which will eventually 
replace it.  Upgrade Alpha, Blackfin, and part of CellSPU to not
use mangler anymore.  CellSPU needs more invasive surgery.

llvm-svn: 93589
 2010-01-15 23:55:16 +00:00
Chris Lattner
5f4c482b72 add a version of AsmPrinter::printVisibility that takes an MCSymbol. 
llvm-svn: 93587
 2010-01-15 23:38:51 +00:00
Chris Lattner
8511f8cc1e use MCSymbol instead of getMangledName() in all cases except one. 
llvm-svn: 93582
 2010-01-15 23:31:55 +00:00
Ted Kremenek
bb1779e022 BumpPtrAllocator: Have the DefaultSlabAllocator created at runtime, not initialization time. This removes one of the 'init_constructors' reported in <rdar://problem/7545356>. 
llvm-svn: 93581
 2010-01-15 23:29:34 +00:00
Dale Johannesen
9f30926cb7 Adjust some comments per review. 
llvm-svn: 93580
 2010-01-15 23:29:29 +00:00
Chris Lattner
000ca8fe0c mc'ize a bunch of symbol stuff, eliminating std::strings. 
llvm-svn: 93578
 2010-01-15 23:26:49 +00:00
Chris Lattner
e293db0c8a add another helper 
llvm-svn: 93577
 2010-01-15 23:25:11 +00:00
David Greene
e52529d7cb Fix PR6019. A load has more than one use if it feeds a bitconvert that 
has more than one use.

llvm-svn: 93576
 2010-01-15 23:23:41 +00:00
Chris Lattner
ea4d444561 add a AsmPrinter::GetGlobalValueSymbol and GetExternalSymbolSymbol 
helper method, use it to simplify some code.

llvm-svn: 93575
 2010-01-15 23:18:17 +00:00
Ted Kremenek
0e9c12f44a Update CMake build. 
llvm-svn: 93571
 2010-01-15 22:59:46 +00:00
Ted Kremenek
89fb514857 Update CMake build. 
llvm-svn: 93570
 2010-01-15 22:59:11 +00:00
Dale Johannesen
7a3b1c5ffc DEBUG_VALUE is now variable sized, as it has a 
target-dependent memory address representation in it.
Restore X86 printing of DEBUG_VALUE; lowering is
done in X86RegisterInfo using the normal algorithm.

llvm-svn: 93565
 2010-01-15 22:22:35 +00:00
Jim Grosbach
5a66487dfd Update Thumb1 storeRegToStackSlot() and loadRegFromStackSlot() to properly 
handle physical registers R0-R7 when described as having a non-tGPR register
class.

llvm-svn: 93564
 2010-01-15 22:21:03 +00:00
Dan Gohman
7b9900094e Fix a typo that Anton noticed. 
llvm-svn: 93563
 2010-01-15 22:18:15 +00:00
Devang Patel
dfde8cf532 Add FIXME. 
llvm-svn: 93562
 2010-01-15 22:08:16 +00:00
Bob Wilson
bdd9890c7d Fix a comment typo. 
llvm-svn: 93560
 2010-01-15 21:55:02 +00:00
Eric Christopher
3c5d9dc0e3 Remove the InlineHint attribute. There are no current or planned 
users.

llvm-svn: 93558
 2010-01-15 21:36:30 +00:00
Anton Korobeynikov
8d5197ab8a zext / truncate is free on msp430. Inform codegen about this. 
llvm-svn: 93556
 2010-01-15 21:19:43 +00:00
Anton Korobeynikov
f1080f2bbe Add branch relaxation pass (shamelessly stolen from PPC). 
llvm-svn: 93554
 2010-01-15 21:19:05 +00:00
Anton Korobeynikov
75c93b4ec3 Provide instruction sizes & encoding. No opcodes yet (but not needed so far). 
llvm-svn: 93553
 2010-01-15 21:18:39 +00:00
Anton Korobeynikov
719dd1465f Enable bit tests and setcc stuff. 
llvm-svn: 93552
 2010-01-15 21:18:18 +00:00
Anton Korobeynikov
aaed919671 Add micro-optimization which allows us to fold imm into cmp. This allows us to save 1 word (sometimes) and reduce register pressure. 
llvm-svn: 93551
 2010-01-15 21:18:02 +00:00
Anton Korobeynikov
0ca5fcdfaa Add instruction formats & support stuff 
llvm-svn: 93550
 2010-01-15 21:17:13 +00:00
Jakob Stoklund Olesen
65bc6159f6 Don't make changes to the MBB in MachineBasicBlock::canFallThrough(). 
This fixes the regression for -pre-regalloc-taildup in
MultiSource/Applications/lambda-0.1.3.

llvm-svn: 93541
 2010-01-15 20:00:12 +00:00
Jakob Stoklund Olesen
cf4c604899 Simplify logic. Any functional change is unintended. 
llvm-svn: 93540
 2010-01-15 19:59:57 +00:00
Chris Lattner
493580bb5c unbreak the build, grr symlinks. 
llvm-svn: 93539
 2010-01-15 19:51:05 +00:00
David Greene
d8faccbeab Add some debug routines to SelectionDAG to dump full DAGs. 
print/dumpWithDepth allows one to dump a DAG up to N levels deep.
dump/printWithFullDepth prints the whole DAG, subject to a depth limit
on 100 in the default case (to prevent infinite recursion).

Have CannotYetSelect to a dumpWithFullDepth so it is clearer exactly
what the non-matching DAG looks like.

llvm-svn: 93538
 2010-01-15 19:43:23 +00:00
Chris Lattner
d6b351119b fix a bug in range information for $42, eliminate an 
unneeded argument from ParseExpression.

llvm-svn: 93536
 2010-01-15 19:39:23 +00:00
Chris Lattner
e18762283a add range information for mem X86Operand's, now all 
X86Operand's have range info.

llvm-svn: 93535
 2010-01-15 19:33:43 +00:00
Chris Lattner
836cf5d129 extend MCAsmParser::ParseExpression and ParseParenExpression 
to return range information for subexpressions.  Use this to
provide range info for several new X86Operands.

llvm-svn: 93534
 2010-01-15 19:28:38 +00:00
Chris Lattner
fba7cf4b44 give X86Operand a ctor and start passing SMLoc's into it. 
llvm-svn: 93532
 2010-01-15 19:06:59 +00:00
Victor Hernandez
c1b5223e76 Improve llvm.dbg.declare intrinsic by referring directly to the storage in its first argument, via function-local metadata (instead of via a bitcast). 
This patch also cleans up code that expects there to be a bitcast in the first argument and testcases that call llvm.dbg.declare.
It also strips old llvm.dbg.declare intrinsics that did not pass metadata as the first argument.

llvm-svn: 93531
 2010-01-15 19:04:09 +00:00
Dale Johannesen
e3562c3fa2 Revert 93499. After discussion with Chris we agreed 
FrameIndexes should be lowered, but the same way as
everything else (target dependent) rather than in a
special hacked way.  The lowering needs to be done
for eventual purposes of Dwarf generation.

llvm-svn: 93530
 2010-01-15 18:58:14 +00:00
Chris Lattner
618f75b46e add range location info for registers, change 
X86Operand::Create* implementations to avoid
copy ctor use.

llvm-svn: 93528
 2010-01-15 18:51:29 +00:00
Nate Begeman
b6c8d75f62 Hook up llc's -filetype=obj to use MCStreamer if an MCCodeEmitter is available. 
Remove most of old Mach-O Writer support, it has been replaced by MCMachOStreamer

Further refactoring to completely remove MachOWriter and drive the object file
writer with the AsmPrinter MCInst/MCSection logic is forthcoming.

llvm-svn: 93527
 2010-01-15 18:51:18 +00:00
Chris Lattner
5d30d26a5d clean up the memory management of the operands. 
llvm-svn: 93526
 2010-01-15 18:44:13 +00:00
Chris Lattner
b032dc1768 refactor ParseRegister to avoid using X86Operand as a temporary 
datastructure when parsing a mem operand.

llvm-svn: 93521
 2010-01-15 18:27:19 +00:00
Victor Hernandez
97d7107d5e Revert r93504 because older uses of llvm.dbg.declare intrinsics need to be auto-upgraded 
llvm-svn: 93515
 2010-01-15 17:36:47 +00:00
Jay Foad
2277dd737c Fix http://llvm.org/PR6028, an assertion failure when an UndefValue of 
integer type is used.

llvm-svn: 93509
 2010-01-15 08:32:58 +00:00
Bob Wilson
b9a6da20e8 Change pre-regalloc tail duplication to only duplicate indirect branch blocks. 
The pre-regalloc pass caused some regressions in both compile time and
performance of the generated code, and it did not improve performance, except
for indirect branches.  I also moved the check for single-block loops to speed
up the common case when running the taildup pass before reg allocation.

llvm-svn: 93505
 2010-01-15 06:29:17 +00:00
Victor Hernandez
aee71b4e81 Improve llvm.dbg.declare intrinsic by referring directly to the storage in its first argument, via function-local metadata (instead of via a bitcast). 
This patch also cleans up code that expects there to be a bitcast in the first argument and testcases that call llvm.dbg.declare.

llvm-svn: 93504
 2010-01-15 03:37:48 +00:00
Dale Johannesen
12ae1f52c2 Lower FrameIndex operand of DEBUG_VALUE (specially) and 
print it as a comment on X86.

llvm-svn: 93499
 2010-01-15 01:54:55 +00:00
Anton Korobeynikov
fb1954f28d Fix cmp emission on msp430: we definitely should turn stuff like 
"icmp lhs, rhs" into "cmp rhs, lhs". This should fix PR5979.

llvm-svn: 93496
 2010-01-15 01:29:49 +00:00
Devang Patel
9c7b75ebb8 Do not use AT_specification die for static variables. It confuses gdb. 
llvm-svn: 93494
 2010-01-15 01:12:22 +00:00
Jim Grosbach
7239c4c92e fix 80-column violations 
llvm-svn: 93487
 2010-01-15 00:36:15 +00:00
Jim Grosbach
ed10ddd9c1 Fix 80 column violations and clean up whitespace 
llvm-svn: 93484
 2010-01-15 00:32:47 +00:00
Devang Patel
b41a4c332a Do not emit multiple AT_container_type attributes. 
We need to find a better way to emit this info.

llvm-svn: 93481
 2010-01-15 00:26:31 +00:00
Jim Grosbach
36c7973e76 Name change for consistency. No functional change. 
llvm-svn: 93480
 2010-01-15 00:22:18 +00:00
Jim Grosbach
ca3006c171 EmitAtomicCmpSwap() custome inserter needs to delete the MI passed in. EmitAtomicBinary() already does this. 
llvm-svn: 93479
 2010-01-15 00:18:34 +00:00
Jeffrey Yasskin
7a710b57f5 Teach PPC how to replaceMachineCodeForFunction correctly. (Fixes 
JITTest.FunctionIsRecompiledAndRelinked.)

llvm-svn: 93475
 2010-01-14 23:15:26 +00:00
Eric Christopher
9230a37bb0 Pad my commit stats by reducing indentation in this now separate 
commit.

llvm-svn: 93473
 2010-01-14 23:00:10 +00:00
Johnny Chen
a5f2e84b0c Added 16-bit Thumb Load/Store immediate instructions with encoding bits so that 
the disassembler can properly decode Load/Store register/immediate instructions.

llvm-svn: 93471
 2010-01-14 22:42:17 +00:00
Chris Lattner
882cb240a6 Split the TargetAsmParser "ParseInstruction" interface in half: 
the new ParseInstruction method just parses and returns a list of
target operands.  A new MatchInstruction interface is used to
turn the operand list into an MCInst.

This requires new/deleting all the operands, but it also gives 
targets the ability to use polymorphic operands if they want to. 

llvm-svn: 93469
 2010-01-14 22:21:20 +00:00
Anton Korobeynikov
5eb86c0b47 Remove pseudo-MI in custom inserter. 
llvm-svn: 93467
 2010-01-14 22:09:11 +00:00
Dale Johannesen
7b7109e16e Fix a comment. 
llvm-svn: 93463
 2010-01-14 21:50:17 +00:00
Eric Christopher
013342ad23 Few minor changes that were requested. No functional change. 
llvm-svn: 93462
 2010-01-14 21:48:00 +00:00
Jim Grosbach
fb10749b6d Add comment explaining the necessity of r93456 
llvm-svn: 93459
 2010-01-14 21:38:31 +00:00
Chris Lattner
ceabb11a5e prune #includes in TargetAsmParser.h 
Pass in SMLoc of instr opcode into ParseInstruction.
Make AsmToken be a class, not a struct.

llvm-svn: 93457
 2010-01-14 21:32:45 +00:00
Jim Grosbach
d3d283dc19 Dwarf EH prepare needs to be run after SjLj prepare. Otherwise, 
catch info can get misplaced when a selector ends up more than one block
removed from the parent invoke(s). This could happen when a landing pad is
shared by multiple invokes and is also a target of a normal edge from
elsewhere.

llvm-svn: 93456
 2010-01-14 21:22:16 +00:00
Chris Lattner
c0321eb1aa introduce MCParsedAsmOperand 
llvm-svn: 93455
 2010-01-14 21:21:40 +00:00
Chris Lattner
001a94a0e3 introduce the MCParsedAsmOperand class. 
llvm-svn: 93454
 2010-01-14 21:20:55 +00:00
Evan Cheng
595cdb7091 Small tweak to inline cost computation. Ext of i/fcmp results are mostly optimized away in codegen. 
llvm-svn: 93453
 2010-01-14 21:04:31 +00:00
Anton Korobeynikov
d7f6117539 Remove spurious semicolon. 
Patch by Diego Iastrubni!

llvm-svn: 93450
 2010-01-14 20:19:51 +00:00
Victor Hernandez
ff4f0aa119 In debug builds, assert that function-local metadata has only 1 parent function 
llvm-svn: 93449
 2010-01-14 20:12:34 +00:00
Eric Christopher
b2708c4ba1 Reduce the inlining cost of functions that contain calls to easily, 
and frequently optimized functions.

llvm-svn: 93448
 2010-01-14 20:12:34 +00:00
Victor Hernandez
23d77b1e51 Simplify code that chooses when to enumerate function-local metadata operands 
llvm-svn: 93446
 2010-01-14 19:54:11 +00:00
Victor Hernandez
292297e8e2 Avoid modifying ValueEnumerator's MD ValueList by choosing which function-local MD to write based on the function currently being written 
llvm-svn: 93441
 2010-01-14 19:38:44 +00:00
Jakob Stoklund Olesen
97a8d154fb ARM "l" constraint for inline asm means R0-R7, also for Thumb2. 
This is consistent with llvm-gcc's arm/constraints.md.

Certain instructions (e.g. CBZ, CBNZ) require a low register, even in Thumb2
mode.

llvm-svn: 93436
 2010-01-14 18:19:56 +00:00
Dan Gohman
7c596d2b00 Fix a codegen abort seen in 483.xalancbmk. 
llvm-svn: 93417
 2010-01-14 03:08:49 +00:00
Victor Hernandez
6878a7e7b6 In WriteFunction(), write function-local metadata before we write the instructions, so instruction's references to metadata are fully resolved by the time they get written. 
llvm-svn: 93403
 2010-01-14 01:50:08 +00:00
Victor Hernandez
36a22e479a Fix printing of function-local metadata in AsmWriter 
llvm-svn: 93402
 2010-01-14 01:47:37 +00:00
Victor Hernandez
367ef8bf4a Clean up unnecessary return and brackets 
llvm-svn: 93401
 2010-01-14 01:46:02 +00:00
Victor Hernandez
cc00507183 Add MDNode::getFunction(), which figures out the metadata's function, if it has function that it is local to. 
llvm-svn: 93400
 2010-01-14 01:45:14 +00:00
Jakob Stoklund Olesen
5fdcc970a2 Don't fold insufficiently aligned ldr/str into ldm/stm instructions. 
An unaligned ldr causes a trap, and is then emulated by the kernel with
awesome performance. The darwin kernel does not emulate unaligned ldm/stm
Thumb2 instructions, so don't generate them.

This fixes the miscompilation of Multisource/Applications/JM/lencod for Thumb2.

Generating unaligned ldr/str pairs from a 16-bit aligned memcpy is probably
also a bad idea, but that is beyond the scope of this patch.

llvm-svn: 93393
 2010-01-14 00:54:10 +00:00
Chris Lattner
72a06499b3 this is an SSE-specific issue. 
llvm-svn: 93373
 2010-01-13 23:29:11 +00:00
Chris Lattner
0aa093dbdf X86 if conversion + tail merging issues from PR6032. 
llvm-svn: 93372
 2010-01-13 23:28:40 +00:00
Bill Wendling
488a7187b4 When the visitSub method was split into visitSub and visitFSub, this xform was 
added to the FSub version. However, the original version of this xform guarded
against doing this for floating point (!Op0->getType()->isFPOrFPVector()).

This is causing LLVM to perform incorrect xforms for code like:

void func(double *rhi, double *rlo, double xh, double xl, double yh, double yl){
  double mh, ml;
  double c = 134217729.0;
  double up, u1, u2, vp, v1, v2;
        
  up = xh*c;
  u1 = (xh - up) + up;
  u2 = xh - u1;
        
  vp = yh*c;
  v1 = (yh - vp) + vp;
  v2 = yh - v1;
        
  mh = xh*yh;
  ml = (((u1*v1 - mh) + (u1*v2)) + (u2*v1)) + (u2*v2);
  ml += xh*yl + xl*yh;
        
  *rhi = mh + ml;
  *rlo = (mh - (*rhi)) + ml;
}

The last line was optimized away, but rl is intended to be the difference
between the infinitely precise result of mh + ml and after it has been rounded
to double precision.

llvm-svn: 93369
 2010-01-13 23:23:17 +00:00
Chris Lattner
4ce6464609 fix ELF section mangling stuff for weak symbols to not use 
obsolete Mangler interfaces.

llvm-svn: 93356
 2010-01-13 21:29:21 +00:00
Victor Hernandez
55e8dacfed Fix comment typo 
llvm-svn: 93355
 2010-01-13 21:25:04 +00:00
Chris Lattner
c073f1009a tidy 
llvm-svn: 93352
 2010-01-13 21:21:29 +00:00
Chris Lattner
4dcb71cc65 reduce duplicate mangling logic by using MCSymbol::printMangledName. 
llvm-svn: 93351
 2010-01-13 21:12:34 +00:00
Chris Lattner
9498bee8a3 expose a static function as a static method on the MCSymbol class. 
llvm-svn: 93350
 2010-01-13 21:09:59 +00:00
Johnny Chen
08c211683b Fixed a couple of places for Thumb MOV where encoding bits are underspecified. 
llvm-svn: 93349
 2010-01-13 21:00:26 +00:00
Jakob Stoklund Olesen
27e36e52f5 Fix pasto 
llvm-svn: 93342
 2010-01-13 19:54:39 +00:00
Chris Lattner
6d1ddfeff2 stop the CBE from using deprecated Mangler stuff. 
llvm-svn: 93341
 2010-01-13 19:54:07 +00:00
Victor Hernandez
d250df09f5 Write function-local metadata as a metadata subblock of a funciton block 
llvm-svn: 93339
 2010-01-13 19:37:33 +00:00
Victor Hernandez
9938929f1f Enumerate function-local metadata (and its types and operands) only during function-incorporation, global metadata continues to be enumerated during creation of ValueEnumerator 
llvm-svn: 93338
 2010-01-13 19:36:16 +00:00
Victor Hernandez
4e63ccc1c3 Parse function-local metadata inside function blocks 
llvm-svn: 93337
 2010-01-13 19:34:08 +00:00
Chris Lattner
b255aee77a Use the GV version of getNameWithPrefix in TargetLoweringObjectFileCOFF:: 
SelectSectionForGlobal, unbreaking weak globals with no-name.

llvm-svn: 93336
 2010-01-13 19:19:17 +00:00
Evan Cheng
0fa1e2d063 Commit some changes I had managed to lose last night while refactoring the code. Avoid change use of PHI instructions because it's not legal to insert any instructions before them. 
This fixes PR6027.

llvm-svn: 93335
 2010-01-13 19:16:39 +00:00
Chris Lattner
10021a1e25 just finish MCizing FnStubInfo which cleans it up and simplifies it. 
llvm-svn: 93334
 2010-01-13 19:13:16 +00:00
Chris Lattner
5c38b08b40 don't call getNameWithPrefix repeatedly and unnecesarily. 
llvm-svn: 93333
 2010-01-13 19:05:36 +00:00
Chris Lattner
6b0af1d420 properly use MCSymbol to print the strings aquired from getNameWithPrefix. 
llvm-svn: 93332
 2010-01-13 19:00:57 +00:00
Benjamin Kramer
5db856d384 Introduce Twine::toStringRef, a variant of toVector which avoids the copy if the 
twine can be represented as a single StringRef. Use the new methode to simplify
some twine users.

llvm-svn: 93317
 2010-01-13 12:45:23 +00:00
Evan Cheng
2afc417122 Re-enable extension optimization pass. 
llvm-svn: 93313
 2010-01-13 08:45:40 +00:00
Chris Lattner
d63a5fea41 upgrade and MC'ize a few uses of makeNameProper. 
llvm-svn: 93310
 2010-01-13 08:08:33 +00:00
Chris Lattner
34047fe59f MC'ize this a bit and upgrade APIs 
llvm-svn: 93309
 2010-01-13 08:04:24 +00:00
Chris Lattner
29a0ff2e70 add a fixme, ELF MCSection isn't quite right and weak unnamed globals are broken 
on linux (even though they are pointless, they shouldn't ICE).

llvm-svn: 93308
 2010-01-13 08:02:14 +00:00
Evan Cheng
98af245e5f For now, avoid issuing extract_subreg to reuse lower 8-bit, it's not safe in 32-bit. 
llvm-svn: 93307
 2010-01-13 08:01:32 +00:00
Evan Cheng
f6f274ef64 Add comment; refactor; avoid pulling in DT if it's not used. 
llvm-svn: 93306
 2010-01-13 07:59:13 +00:00
Chris Lattner
f2090cf7fd eliminate some uses of Mangler::makeNameProper. 
llvm-svn: 93305
 2010-01-13 07:56:59 +00:00
Chris Lattner
ba8b3ed4f1 don't add the \1 to the name. 
llvm-svn: 93304
 2010-01-13 07:50:21 +00:00
Chris Lattner
1b6c061cd0 remove uses of deprecated functions, this generates slightly 
different BlockAddress labels, but nothing semantically important.

Add a FIXME that BlockAddress codegen is broken if the LLVM BB has 
an empty name (e.g. strip was run).

llvm-svn: 93303
 2010-01-13 07:30:49 +00:00
Chris Lattner
086d0d1a2b use the new form of getNameWithPrefix, not makeNameProper. 
Among other things, this would do very weird things if the 
basic block name had (e.g.) a space in it on darwin:
makeNameProper would add quotes, then the mcsymbol would 
escape the quotes.

llvm-svn: 93302
 2010-01-13 07:16:53 +00:00
Chris Lattner
9c1ca33ac3 add new isSingleStringRef()/getSingleStringRef() methods to twine, 
and use them to avoid a copy of a string in getNameWithPrefix in
the common case.  It seems like Value::setName and other places 
should use this as well?

llvm-svn: 93301
 2010-01-13 07:12:06 +00:00
Chris Lattner
951cfb6b8b ugh, my last patch just sped up a method and changed all the clients 
that I want to completely eliminate.  Add fixme's so I remember this
in the future, and add the missing helper that they should be upgraded
to use instead.

llvm-svn: 93300
 2010-01-13 07:01:09 +00:00
Chris Lattner
3c2fad1fc6 change Mangler::makeNameProper to return its result in a SmallVector 
instead of returning it in an std::string.  Based on this change:

1. Change TargetLoweringObjectFileCOFF::getCOFFSection to take a StringRef
2. Change a bunch of targets to call makeNameProper with a smallstring,
   making several of them *much* more efficient.
3. Rewrite Mangler::makeNameProper to not build names and then prepend
   prefixes, not use temporary std::strings, and to avoid other crimes.

llvm-svn: 93298
 2010-01-13 06:38:18 +00:00
Chris Lattner
23127baf26 my mistake, Mangler::makeNameProper wants to take a twine, not a stringref! 
llvm-svn: 93296
 2010-01-13 05:02:57 +00:00
Chris Lattner
b8f828a848 change makeNameProper to take a stringref instead of std::string. 
llvm-svn: 93295
 2010-01-13 04:55:33 +00:00
Chris Lattner
e8e56916d1 fix assert in AsmPrinter::EmitGlobalConstantLargeInt to match reality. 
llvm-svn: 93293
 2010-01-13 04:39:46 +00:00
Chris Lattner
633c7a95c5 reduce nesting and code duplication in AsmPrinter::EmitGlobalConstantLargeInt. 
llvm-svn: 93292
 2010-01-13 04:38:16 +00:00
Chris Lattner
4e95cbcd12 reduce indentation and add a fast-path to EmitGlobalConstant for 8-byte 
integers on 64-bit systems.

llvm-svn: 93291
 2010-01-13 04:34:19 +00:00
Chris Lattner
452f5ad8f2 reduce indentation and use early exits in AsmPrinter::EmitConstantValueOnly 
llvm-svn: 93290
 2010-01-13 04:29:19 +00:00
Victor Hernandez
6de2f65f80 Revert 93270 pending investigation of how stray non-constant values end up in ValueEnumerator's ValueList during WriteConstants() 
llvm-svn: 93289
 2010-01-13 03:18:30 +00:00
Evan Cheng
973fceab0c Disable opt-ext pass to unbreak the build for now. 
llvm-svn: 93286
 2010-01-13 01:51:43 +00:00
Evan Cheng
7fcebf982c Remove debug option I accidentally left in. 
llvm-svn: 93285
 2010-01-13 01:43:20 +00:00
Ted Kremenek
2d9c86e98b Update CMake file. 
llvm-svn: 93283
 2010-01-13 01:02:47 +00:00
Jakob Stoklund Olesen
2d819b93a2 Remove the JustSP single-register regclass. 
It was only being used by instructions with the t_addrmode_sp addressing mode,
and that is pattern matched in a way that guarantees SP is used. There is
never any register allocation done from this class.

llvm-svn: 93280
 2010-01-13 00:43:06 +00:00
Evan Cheng
76db3bb18e Add a quick pass to optimize sign / zero extension instructions. For targets where the pre-extension values are available in the subreg of the result of the extension, replace the uses of the pre-extension value with the result + extract_subreg. 
For now, this pass is fairly conservative. It only perform the replacement when both the pre- and post- extension values are used in the block. It will miss cases where the post-extension values are live, but not used.

llvm-svn: 93278
 2010-01-13 00:30:23 +00:00
Dale Johannesen
1738d7d11a Further progration of metadata operands. The 
dumper doesn't really do what I want yet, but
at least it doesn't crash now.

llvm-svn: 93272
 2010-01-13 00:00:24 +00:00
Victor Hernandez
9b62aee9ad Make WriteConstants() more robust against stray values in ValueEnumerator's ValueList 
llvm-svn: 93270
 2010-01-12 23:37:59 +00:00
Bob Wilson
0ec53edad3 Fix a comment typo. 
llvm-svn: 93261
 2010-01-12 22:18:56 +00:00
Chris Lattner
21bbaf49d5 1) Use the new SimplifyInstructionsInBlock routine instead of the copy 
in JT.

2) When cloning blocks for PHI or xor conditions, use
instsimplify to simplify the code as we go.  This allows us to 
squish common cases early in JT which opens up opportunities for
subsequent iterations, and allows it to completely simplify the
testcase.

llvm-svn: 93253
 2010-01-12 20:41:47 +00:00
Chris Lattner
87f86498c3 add a helper function. 
llvm-svn: 93251
 2010-01-12 19:40:54 +00:00
Devang Patel
1d9c156df2 Use Twine, instead of StringRef, for consistency. 
llvm-svn: 93249
 2010-01-12 18:57:56 +00:00
Devang Patel
2a6c413df3 Use ilist_tratis to autoinsert and remove NamedMDNode from MDSymbolTable. 
llvm-svn: 93247
 2010-01-12 18:34:06 +00:00
Evan Cheng
bc0b06fb16 Eliminate or_not_add and just use AddedComplexity so isel tries or_is_add patterns first. 
llvm-svn: 93245
 2010-01-12 18:31:19 +00:00
Duncan Sands
395053f13a Revert commit 93204, since it causes the assembler to barf 
on x86-64 linux with messages like this:
Error: Incorrect register `%r14' used with `l' suffix

llvm-svn: 93242
 2010-01-12 17:46:16 +00:00
Duncan Sands
1c71358944 Fix typo. 
llvm-svn: 93235
 2010-01-12 08:30:46 +00:00
Duncan Sands
f558d05561 Tweak commit 91745, which changed target data for both Mingw and Cygwin, 
to not touch Cygwin: the change caused llvm-gcc build failures due to
long double getting the wrong size.  Patch by Aaron Gray.

llvm-svn: 93234
 2010-01-12 08:21:07 +00:00
Dan Gohman
51b3e804dc Reapply the MOV64r0 patch, with a fix: MOV64r0 clobbers EFLAGS. 
llvm-svn: 93229
 2010-01-12 04:42:54 +00:00
Dan Gohman
2cd1b789c7 Update a partially obsolete comment. 
llvm-svn: 93228
 2010-01-12 04:32:35 +00:00
Dan Gohman
da0bcb49b5 Fix a typo in a comment. 
llvm-svn: 93227
 2010-01-12 04:30:26 +00:00
Chris Lattner
bc0016437d tidy up 
llvm-svn: 93222
 2010-01-12 02:07:50 +00:00
Chris Lattner
774e3967ad Teach jump threading to duplicate small blocks when the branch 
condition is a xor with a phi node.  This eliminates nonsense
like this from 176.gcc in several places:

 LBB166_84:
        testl   %eax, %eax
-       setne   %al
-       xorb    %cl, %al
-       notb    %al
-       testb   $1, %al
-       je      LBB166_85
+       je      LBB166_69
+       jmp     LBB166_85

This is rdar://7391699

llvm-svn: 93221
 2010-01-12 02:07:17 +00:00
Evan Cheng
e5b545fd60 Add TargetInstrInfo::isCoalescableInstr. It returns true if the specified 
instruction is copy like where the source and destination registers can
overlap. This is to be used by the coalescable to coalesce the source and
destination registers of instructions like X86::MOVSX64rr32. Apparently
some crazy people believe the coalescer is too simple.

llvm-svn: 93210
 2010-01-12 00:09:37 +00:00
Chris Lattner
d6d8cc7b37 some cleanup, and make it obvious that ProcessJumpOnPHI only works 
on branches by renaming it and checking for a branch at the call site.

llvm-svn: 93208
 2010-01-11 23:41:09 +00:00
Evan Cheng
a93b476689 Add manual ISD::OR fastisel selection routines. TableGen is no longer autogen them after 93152 and 93191. 
llvm-svn: 93204
 2010-01-11 22:59:27 +00:00
Chris Lattner
a8cabeeecb reenable the piece that turns trunc(zext(x)) -> x even if zext has multiple uses, 
codegen has no apparent problem with the trunc version of this, because it turns
into a simple subreg idiom

llvm-svn: 93202
 2010-01-11 22:49:40 +00:00
Chris Lattner
2749cc2036 Disable folding sext(trunc(x)) -> x (and other similar cast/cast cases) when the 
trunc has multiple uses.  Codegen is not able to coalesce the subreg case 
correctly and so this leads to higher register pressure and spilling (see PR5997).

This speeds up 256.bzip2 from 8.60 -> 8.04s on my machine, ~7%.

llvm-svn: 93200
 2010-01-11 22:45:25 +00:00
Victor Hernandez
1ca202f9ea Consolidate ConvertGlobalValIDToValue, ConvertGlobalOrMetadataValIDToValue, and ConvertValIDToValue into a more powerful ConvertValIDToValue() that does all three's work 
llvm-svn: 93197
 2010-01-11 22:31:58 +00:00
Evan Cheng
bd938ebc90 Extend r93152 to work on OR r, r. If the source set bits are known not to overlap, then select as an ADD instead. 
llvm-svn: 93191
 2010-01-11 22:03:29 +00:00
Evan Cheng
bc84a42d7b Revert 93158. It's breaking quite a few x86_64 tests. 
llvm-svn: 93185
 2010-01-11 21:13:41 +00:00
Jakob Stoklund Olesen
f1c71ef6ba Avoid adding PHI arguments for a predecessor that has gone away when a BRCOND was constant folded. 
This fixes PR5980.

llvm-svn: 93184
 2010-01-11 21:02:33 +00:00
Benjamin Kramer
400ab557d3 Remove unused string functions. 
llvm-svn: 93183
 2010-01-11 20:33:12 +00:00
Evan Cheng
4548543b0b Do not turn 8-bit OR to ADD since ADD8ri is not 3-addressfiable. 
llvm-svn: 93182
 2010-01-11 20:18:04 +00:00
Mon P Wang
e8470bbcc4 Disable transformation of select of two loads to a select of address and then a load if the 
loads are not in the default address space because the transformation discards src value info.

llvm-svn: 93180
 2010-01-11 20:12:49 +00:00
Benjamin Kramer
45e333ea95 Add StrInStrNoCase, a StringRef version of CStrInCStrNoCase. 
llvm-svn: 93174
 2010-01-11 19:45:18 +00:00
Devang Patel
9874d72976 s/NextValueNo/NextMDValueNo while processing metadata. 
llvm-svn: 93165
 2010-01-11 18:52:33 +00:00
Benjamin Kramer
7ec2a31f14 Turns out llvm-gcc still uses SplitString with a vector. Add it back until I 
have a fix.

llvm-svn: 93163
 2010-01-11 18:44:35 +00:00
Benjamin Kramer
cdbc36f961 Reimplement getToken and SplitString as "StringRef helper functions" 
- getToken is modeled after StringRef::split but it can split on multiple
  separator chars and skips leading seperators.
- SplitString is a StringRef::split variant for more than 2 elements with the
  same behaviour as getToken.

llvm-svn: 93161
 2010-01-11 18:03:24 +00:00
Dan Gohman
541c4f4c5d Use a 32-bit and with implicit zero-extension instead of a 64-bit and if it 
has an immediate with at least 32 bits of leading zeros, to avoid needing to
materialize that immediate in a register first.

FileCheckize, tidy, and extend a testcase to cover this case.

This fixes rdar://7527390.

llvm-svn: 93160
 2010-01-11 17:58:34 +00:00
Dan Gohman
5b79391087 Re-instate MOV64r0 and MOV16r0, with adjustments to work with the 
new AsmPrinter. This is perhaps less elegant than describing them
in terms of MOV32r0 and subreg operations, but it allows the
current register to rematerialize them.

llvm-svn: 93158
 2010-01-11 17:37:57 +00:00
Dan Gohman
a83443605d Pattern top-level operators don't need to be restricted to a 
single user. The _su forms are intended for non-top-level nodes.

llvm-svn: 93155
 2010-01-11 17:21:05 +00:00
Dan Gohman
5042879ae5 Reword this comment to reference a more fundamental issue. 
llvm-svn: 93154
 2010-01-11 17:14:46 +00:00
Evan Cheng
ee806a0db5 Select an OR with immediate as an ADD if the input bits are known zero. This allow the instruction to be 3address-fied if needed. 
llvm-svn: 93152
 2010-01-11 17:03:47 +00:00
David Greene
b879ff4855 Implement a feature (-vector-unaligned-mem) to allow targets to 
ignore alignment requirements for SIMD memory operands.  This
is useful on architectures like the AMD 10h that do not trap on
unaligned references if a status bit is twiddled at startup time.

llvm-svn: 93151
 2010-01-11 16:29:42 +00:00
Victor Hernandez
832bd4fd2d Respond to Chris' review: 
Make InsertDbgValueIntrinsic() and get Offset take and recieve a uint64_t.
Get constness correct for getVariable() and getValue().

llvm-svn: 93149
 2010-01-11 07:45:19 +00:00
Chris Lattner
85a6f02b94 add one more bitfield optimization, allowing clang to generate 
good code on PR4216:

_test_bitfield:                                             ## @test_bitfield
	orl	$32962, %edi
	movl	$4294941946, %eax
	andq	%rdi, %rax
	ret

instead of:

_test_bitfield:
        movl    $4294941696, %ecx
        movl    %edi, %eax
        orl     $194, %edi
        orl     $32768, %eax
        andq    $250, %rdi
        andq    %rax, %rcx
        movq    %rdi, %rax
        orq     %rcx, %rax
        ret

Evan is looking into the remaining andq+imm -> andl optimization.

llvm-svn: 93147
 2010-01-11 06:55:24 +00:00
Chris Lattner
16e36659f5 Extend CanEvaluateZExtd to handle and/or/xor more aggressively in the 
BitsToClear case.  This allows it to promote expressions which have an
and/or/xor after the lshr, promoting cases like test2 (from PR4216) 
and test3 (random extample extracted from a spec benchmark).

clang now compiles the code in PR4216 into:

_test_bitfield:                                             ## @test_bitfield
	movl	%edi, %eax
	orl	$194, %eax
	movl	$4294902010, %ecx
	andq	%rax, %rcx
	orl	$32768, %edi
	andq	$39936, %rdi
	movq	%rdi, %rax
	orq	%rcx, %rax
	ret

instead of:

_test_bitfield:                                             ## @test_bitfield
	movl	%edi, %eax
	orl	$194, %eax
	movl	$4294902010, %ecx
	andq	%rax, %rcx
	shrl	$8, %edi
	orl	$128, %edi
	shlq	$8, %rdi
	andq	$39936, %rdi
	movq	%rdi, %rax
	orq	%rcx, %rax
	ret

which is still not great, but is progress.

llvm-svn: 93145
 2010-01-11 04:05:13 +00:00
Chris Lattner
f2ba85eedc Remove the dead TD argument to CanEvaluateZExtd, and add a 
new BitsToClear result which allows us to start promoting
expressions that end with a lshr-by-constant.  This is
conservatively correct and better than what we had before
(see testcases) but still needs to be extended further.

llvm-svn: 93144
 2010-01-11 03:32:00 +00:00
Chris Lattner
d7f1b97147 improve comments, remove dead TD argument to CanEvaluateSExtd. 
llvm-svn: 93143
 2010-01-11 02:43:35 +00:00
Chris Lattner
18d753e05f teach sext optimization to handle truncs from types that are not 
the dest of the sext.

llvm-svn: 93128
 2010-01-10 20:30:41 +00:00
Chris Lattner
ca53de1ab7 teach zext optimization how to deal with truncs that don't come from 
the zext dest type.  This allows us to handle test52/53 in cast.ll,
and allows llvm-gcc to generate much better code for PR4216 in -m64
mode:

_test_bitfield:                                             ## @test_bitfield
	orl	$32962, %edi
	movl	%edi, %eax
	andl	$-25350, %eax
	ret

This also fixes a bug handling vector extends, ensuring that the
mask produced is a vector constant, not an integer constant.

llvm-svn: 93127
 2010-01-10 20:25:54 +00:00
Chris Lattner
56e327711b fix a buggy assertion, CreateIntegerCast should allow 
integer vectors as well as just integers.

llvm-svn: 93126
 2010-01-10 20:21:42 +00:00
Mikhail Glushenkov
e141fc9e1c Suppress a warning on gcc 4.4. 
warning: suggest parentheses around ‘&&’ within ‘||’.
llvm-svn: 93121
 2010-01-10 18:48:49 +00:00
Chris Lattner
67e0ef4d5d simplify CanEvaluateSExtd to return a bool now that we have a 
simpler profitability predicate.

llvm-svn: 93111
 2010-01-10 07:57:20 +00:00
Chris Lattner
fa864affbc the NumCastsRemoved argument to CanEvaluateSExtd is dead, remove it. 
llvm-svn: 93110
 2010-01-10 07:42:21 +00:00
Chris Lattner
e68d6e61b1 now that the cost model has changed, we can always consider 
elimination of a sign extend to be a win, which simplifies 
the client of CanEvaluateSExtd, and allows us to eliminate
more casts (examples taken from real code).

llvm-svn: 93109
 2010-01-10 07:40:50 +00:00
Victor Hernandez
f4272b50d6 Compute isFunctionLocal in MDNode ctor or via argument in new function getWhenValsUnresolved(). 
Document PFS argument to ParseValID() and ConvertGlobalOrMetadataValIDToValue().

llvm-svn: 93108
 2010-01-10 07:14:18 +00:00
Chris Lattner
e86619ca01 change the preferred canonical form for a sign extension to be 
lshr+ashr instead of trunc+sext.  We want to avoid type 
conversions whenever possible, it is easier to codegen expressions
without truncates and extensions.

llvm-svn: 93107
 2010-01-10 07:08:30 +00:00
Chris Lattner
00b1e8508e fix indentation of switch statements, no functionality change. 
llvm-svn: 93106
 2010-01-10 06:59:55 +00:00
Chris Lattner
1332cfaea1 fix pasto that broke bootstrap. 
llvm-svn: 93105
 2010-01-10 06:50:04 +00:00
Chris Lattner
7314e76078 simplify CanEvaluateZExtd now that we don't care about the number of 
bits known clear in the result and don't care about the # casts 
eliminated.  TD is also dead but keeping it for now.

llvm-svn: 93098
 2010-01-10 02:50:04 +00:00
Chris Lattner
1106f03886 two changes: 
1) don't try to optimize a sext or zext that is only used by a trunc, let
   the trunc get optimized first.  This avoids some pointless effort in
   some common cases since instcombine scans down a block in the first pass.
2) Change the cost model for zext elimination to consider an 'and' cheaper
   than a zext.  This allows us to do it more aggressively, and for the next
   patch to simplify the code quite a bit.

llvm-svn: 93097
 2010-01-10 02:39:31 +00:00
Chris Lattner
a04ed0659f enhance CanEvaluateZExtd to handle shift left and sext, allowing 
more expressions to be promoted and casts eliminated.

llvm-svn: 93096
 2010-01-10 02:22:12 +00:00
Chris Lattner
eb3da2ab86 remove an xform subsumed by EvaluateInDifferentType. 
llvm-svn: 93095
 2010-01-10 01:35:55 +00:00
Julien Lerouge
950d1334d1 Fix nondeterministic behavior. 
llvm-svn: 93093
 2010-01-10 01:07:22 +00:00
Chris Lattner
edb60f0861 clean up this xform by using m_Trunc. 
llvm-svn: 93092
 2010-01-10 01:04:31 +00:00
Chris Lattner
3efde81a44 inline and remove the rest of commonIntCastTransforms. 
llvm-svn: 93091
 2010-01-10 01:00:46 +00:00
Chris Lattner
492e9ca5a4 Inline the expression type promotion/demotion stuff out of 
commonIntCastTransforms into the callers, eliminating a switch,
and allowing the static predicate  methods to be moved down to
live next to the corresponding function.  No functionality 
change.

llvm-svn: 93089
 2010-01-10 00:58:42 +00:00
Chris Lattner
953394de82 "In order to ease automatic bindings generation, it would be helpful if boolean values were distinguishable from integers. The attached patch introduces "typedef int LLVMBool;", and uses LLVMBool instead of int throughout the C API, wherever a boolean value is called for." 
Patch by James Y Knight!

llvm-svn: 93079
 2010-01-09 22:27:07 +00:00
Chris Lattner
8e44e4eddc fix PR5983, reject function bodies with no blocks per the grammar. 
llvm-svn: 93067
 2010-01-09 19:20:07 +00:00
Jeffrey Yasskin
53a8f3981c Fix http://llvm.org/PR5729: x86-64 tail calls were putting their targets into 
R11, and then asserting that the target was in R9.  Since R9 isn't reserved for
the target anymore, and is used as an argument, this patch changes the
assertion.

llvm-svn: 93065
 2010-01-09 18:56:43 +00:00
Dan Gohman
771144e807 Use WriteAsOperand instead of getName() to print loop header names, 
so that unnamed blocks are handled.

llvm-svn: 93059
 2010-01-09 18:17:45 +00:00
David Chisnall
edfb3fc27c Fixed linking of modules containing aliases to constant bitcasts. Existing behaviour first tried to replace the aliases with the global that they aliased (rather than the bitcast), causing a crash on an assert because the types didn't match. When this was fixed, it then did the same thing creating the new alias (creating an alias with a different type to its aliasee). 
Linking modules containing aliases to GEPs is still not quite right.  GEPs that are equivalent to bitcasts will be replaced by bitcasts, GEPs that are not will just break.  Aliases to GEPs that are not equivalent to bitcasts are horribly broken anyway (it might be worth adding an assert when creating the alias to prevent these being created; they just cause problems later).

llvm-svn: 93052
 2010-01-09 16:27:31 +00:00
Duncan Sands
8f97b49457 Suppress use of uninitialized variable warning. 
llvm-svn: 93046
 2010-01-09 08:30:33 +00:00
Chris Lattner
b142e13f84 only factor from expressions whose uses are empty and whose 
base is the right expression type.  This fixes PR5981.

llvm-svn: 93045
 2010-01-09 06:01:36 +00:00
Dan Gohman
3708af1c59 Revert an earlier change to SIGN_EXTEND_INREG for vectors. The VTSDNode 
really does need to be a vector type, because
TargetLowering::getOperationAction for SIGN_EXTEND_INREG uses that type,
and it needs to be able to distinguish between vectors and scalars.

Also, fix some more issues with legalization of vector casts.

llvm-svn: 93043
 2010-01-09 02:13:55 +00:00
Chris Lattner
8a92d699f8 clean up this code, add a fixme. 
llvm-svn: 93042
 2010-01-09 02:02:37 +00:00
Devang Patel
bd58ff3f60 Delete NamedMDSymTable while destrucing Module. 
Disable  copy ctor and operator= for NamedMDSymTable. Hide typedef that should be public.

llvm-svn: 93041
 2010-01-09 01:44:59 +00:00
Devang Patel
de49dd9087 NamedMDNode is never used so there is no need to enumerate it here. 
llvm-svn: 93039
 2010-01-09 01:24:03 +00:00
Julien Lerouge
76c75e82d0 Fix nondeterministic behavior. 
llvm-svn: 93038
 2010-01-09 01:06:49 +00:00
Devang Patel
fbb47c2300 NamedMDNode element is either null or MDNode. Otherwise this is a malformed NamedMDNode record. 
llvm-svn: 93037
 2010-01-09 01:02:22 +00:00
Devang Patel
ea208bf99c Derive NamedMDNode from Value. 
llvm-svn: 93032
 2010-01-09 00:30:14 +00:00
Evan Cheng
afa00d14db Dan pointed out checking whether a node is dead by comparing its opcode to ISD::DELETED_NODE is not safe. Use a DAGUpdateListener to remove dead nodes from work list instead. 
llvm-svn: 93031
 2010-01-09 00:21:08 +00:00
Evan Cheng
2e497d1ed4 Fix a critical bug in 64-bit atomic operation lowering for 32-bit. The results of the cmpxchg8b instructions are being thrown away when it branches back to the top of the checking loop. This means the loop always compares against the old value and this can result in a dead lock. 
llvm-svn: 93028
 2010-01-08 23:41:50 +00:00
Eric Christopher
937778e8d0 Remove unnecessary dyn_cast and add a comment. Part of a WIP. 
llvm-svn: 93026
 2010-01-08 21:37:11 +00:00
Chris Lattner
fd117b63f7 mplement a theoretical fixme. 
llvm-svn: 93024
 2010-01-08 19:28:47 +00:00
Chris Lattner
474cd30fb7 rename CanEvaluateInDifferentType -> CanEvaluateTruncated and 
simplify it now that it is only used for truncates.

llvm-svn: 93021
 2010-01-08 19:19:23 +00:00
Evan Cheng
8b248e5016 Fix comment. 
llvm-svn: 93020
 2010-01-08 19:14:57 +00:00
Chris Lattner
05ae88cc8f teach instcombine to delete sign extending shift pairs (sra(shl X, C), C) when 
the input is already sign extended.

llvm-svn: 93019
 2010-01-08 19:04:21 +00:00
Chris Lattner
0dc48180de fix PR5978 by peeling the loop so that we avoid shifting the 
result int by 8 for the first byte.  While normally harmless,
if the result is smaller than a byte, this shift is invalid.

llvm-svn: 93018
 2010-01-08 19:02:23 +00:00
Duncan Sands
c94f8aeb36 Suppress an unused variable warning when assertions are off; 
remove some trailing whitespace while there.

llvm-svn: 93008
 2010-01-08 17:51:48 +00:00
Chris Lattner
725b9ba4e5 tidy up some stuff duncan pointed out. 
llvm-svn: 93007
 2010-01-08 17:48:19 +00:00
Johnny Chen
f782ccee01 Minor change, change the order of two "let Inst{...}" stmts within multiclass 
T2I_bin_ii12rs definition.

llvm-svn: 93006
 2010-01-08 17:41:33 +00:00
Eric Christopher
067afe673c After further thought revert the patch to make fast-isel avoid 
putting relocations into the constant pool - this isn't needed
for correctness and in the rare occasion it happens would pull
us out of fast isel for the block.

If fast-isel application startup time ever becomes an issue we
can add better support for these addresses instead of bailing.

llvm-svn: 92995
 2010-01-08 08:24:49 +00:00
Evan Cheng
f96a9ec02b ReplaceAllUsesOfValueWith may delete other nodes that the one being replaced. Do not delete dead nodes again. 
llvm-svn: 92988
 2010-01-08 02:36:12 +00:00
Evan Cheng
4f25f87baa Fix what looks to me obvious instruction definition bugs. 
1. CMPXCHG8B and CMPXCHG16B did not specify implicit physical register defs and uses.
2. LCMPXCHG8B is loading 64 bit memory, not 32 bit.

llvm-svn: 92985
 2010-01-08 01:29:19 +00:00
Eric Christopher
2a3196c2e5 Remove extraneous include. 
llvm-svn: 92972
 2010-01-08 00:05:33 +00:00
Chris Lattner
944f9c4ac1 teach ComputeNumSignBits to look through PHI nodes. 
llvm-svn: 92964
 2010-01-07 23:44:37 +00:00
Chris Lattner
db8fa82914 Enhance instcombine to reason more strongly about promoting computation 
that feeds into a zext, similar to the patch I did yesterday for sext.
There is a lot of room for extension beyond this patch.

llvm-svn: 92962
 2010-01-07 23:41:00 +00:00
Chris Lattner
e0199dff81 Fix rdar://7517201, a regression introduced by r92849. 
When folding a and(any_ext(load)) both the any_ext and the
load have to have only a single use.

This removes the anyext-uses.ll testcase which started failing
because it is unreduced and unclear what it is testing.

llvm-svn: 92950
 2010-01-07 21:59:23 +00:00
Chris Lattner
f68e328a99 factor this code better and reduce nesting at the same 
time, no functionality change.

llvm-svn: 92948
 2010-01-07 21:53:27 +00:00
Evan Cheng
4523041394 APInt'fy TargetLowering::SimplifySetCC to fix PR5963. 
llvm-svn: 92943
 2010-01-07 20:58:44 +00:00
Benjamin Kramer
bbaf8cac38 Simplify code. No intended functionality/performance change. 
llvm-svn: 92938
 2010-01-07 19:46:15 +00:00
Eric Christopher
d218bc7a2f If the data requires a relocation then don't attempt to 
add it to the constant pool for fast-isel. We already
don't add it for the normal case.

llvm-svn: 92934
 2010-01-07 19:45:14 +00:00
Eric Christopher
1d452e5c87 We need to put any kind of data with a relocation into a 
not-readonly segment on darwin.

llvm-svn: 92933
 2010-01-07 19:44:05 +00:00
Devang Patel
10ee76f7f4 Use separate namespace for named metadata. 
llvm-svn: 92931
 2010-01-07 19:39:36 +00:00
Chris Lattner
320b635dfd constant materialization could be improved. 
llvm-svn: 92921
 2010-01-07 17:53:10 +00:00
Benjamin Kramer
badb9914d2 Kill dead store. 
llvm-svn: 92920
 2010-01-07 17:50:57 +00:00
Benjamin Kramer
a97aab4995 Remove dead variable. 
llvm-svn: 92919
 2010-01-07 17:29:08 +00:00
Benjamin Kramer
22ec2524fd Use pop_back_val instead of back()+pop_back. 
llvm-svn: 92918
 2010-01-07 17:27:56 +00:00
Benjamin Kramer
df696ee6b4 Use a do-while loop instead of while + boolean. 
llvm-svn: 92912
 2010-01-07 13:50:07 +00:00
Duncan Sands
8299315be7 Be less stingy as to how many selects and phi nodes we 
are prepared to look through.

llvm-svn: 92898
 2010-01-07 05:48:42 +00:00
Douglas Gregor
c8114b239b More trivial optimizations to a function well outside the critical path 
llvm-svn: 92896
 2010-01-07 02:24:06 +00:00
Chris Lattner
1a64e58fa3 handle ConstantVector while I'm in here. 
llvm-svn: 92892
 2010-01-07 01:20:20 +00:00
Chris Lattner
0b0caf0877 fix a globalopt crash on 'bullet' (handling evaluation of a store 
to an element of a vector in a static ctor) which occurs with an 
unrelated patch I'm testing.  Annoyingly, EvaluateStoreInto basically 
does exactly the same stuff as InsertElement constant folding, but it
now handles vectors, and you can't insertelement into a vector.  It
would be 'really nice' if GEP into a vector were not legal.

llvm-svn: 92889
 2010-01-07 01:16:21 +00:00
Evan Cheng
51d86260ff Fix a minor regression from my dag combiner changes. One more place which needs to look pass truncates. 
llvm-svn: 92885
 2010-01-07 00:54:06 +00:00
Douglas Gregor
b7161a0b1c Switch StringRef::edit_distance over to using raw pointers, since both 
std::vector and llvm::SmallVector have annoying performance
tradeoffs. No, I don't expect this to matter, and now it won't.

llvm-svn: 92884
 2010-01-07 00:51:54 +00:00
Jim Grosbach
40aafaec86 80 column violations 
llvm-svn: 92876
 2010-01-06 23:54:42 +00:00
Jakob Stoklund Olesen
b90023e707 Allow double defs after tail duplication. 
llvm-svn: 92874
 2010-01-06 23:52:46 +00:00
Jakob Stoklund Olesen
a63aa4e54b Add Target hook to duplicate machine instructions. 
Some instructions refer to unique labels, and so cannot be trivially cloned
with CloneMachineInstr.

llvm-svn: 92873
 2010-01-06 23:47:07 +00:00