Commit Graph

47349 Commits

Author SHA1 Message Date
Rafael Espindola 8e129296fe Misc code refactorings:
* Remove unnecessary arguments now that ForceExpAbs is a method.
* Use ForceExpAbs in EmitAbsValue.

llvm-svn: 131683
2011-05-19 21:40:34 +00:00
Eric Christopher 4014e5e208 Oddly people want to use the 'r' constraint for fp constants on x86.
Fixes rdar://9218925
Fixes PR9601

llvm-svn: 131682
2011-05-19 21:33:47 +00:00
Charles Davis de745ba396 Implement the EndProlog Win64 EH method on the base MCStreamer.
llvm-svn: 131679
2011-05-19 21:24:54 +00:00
Jason W Kim d0c937d4b2 This fixes one divergence between LLVM and binutils for ARM in the
text section.

Assume the following bit of annotated assembly:

.section	.data.rel.ro,"aw",%progbits
.align	2
.LAlpha:
.long	startval(GOTOFF)

.text
.align	2
.type	main,%function
.align	4

main: ;;; assume "main" starts at offset 0x20
0x0	push	{r11, lr}
0x4	movw	r0, :lower16:(.LAlpha-(.LBeta+8))
;;; ==> (.AddrOf(.LAlpha) - ((.AddrOf(.LBeta) - .AddrOf(".")) + 8)
;;; ==> (??? - ((16-4) + 8) = -20
0x8	movt	r0, :upper16:(.LAlpha-(.LBeta+8))
;;; ==> (.AddrOf(.LAlpha) - ((.AddrOf(.LBeta) - .AddrOf(".")) + 8)
;;; ==> (??? - ((16-8) + 8) = -16
0xc	... blah

.LBeta:
0x10	add	r0, pc, r0
0x14	... blah

.LGamma:
0x18	add	r1, pc, r1

Above snippet results in the following relocs in the .o file for the
first pair of movw/movt instructions

00000024 R_ARM_MOVW_PREL_NC .LAlpha
00000028 R_ARM_MOVT_PREL .LAlpha

And the encoded instructions in the .o file for main: must be

00000020 <main>:
20:	e92d4800 push	{fp, lr}
24:	e30f0fec movw	r0, #65516	; 0xffec i.e. -20
28:	e34f0ff0 movt	r0, #65520	; 0xfff0 i.e. -16

However, llc (prior to this commit) generates the following sequence

00000020 <main>:
20:	e92d4800 push	{fp, lr}
24:	e30f0fec movw	r0, #65516	; 0xffec - i.e. -20
28:	e34f0fff movt	r0, #65535	; 0xffff - i.e. -1

What has to happen in the ArmAsmBackend is that if the relocation is PC
relative, the 16 bits encoded as part of movw and movt must be both addends,
not addresses. It makes sense to encode addresses by right shifting the value
by 16, but the result is incorrect for PIC.
i.e., the right shift by 16 for movt is ONLY valid for the NON-PCRel case.

This change agrees with what GNU as does, and makes the PIC code run.

MC/ARM/elf-movt.s covers this case.

llvm-svn: 131674
2011-05-19 20:55:25 +00:00
Devang Patel 1407fb4bbe Reapply r131605. This time with a fix, which is to use NoFolder.
llvm-svn: 131673
2011-05-19 20:52:46 +00:00
Rafael Espindola 0fc5e89c82 ADD64ri32 sign extends its argument, so we need to use a R_X86_64_32S.
Fixes PR9934.

We really need to start tblgening the relocation info :-(

llvm-svn: 131669
2011-05-19 20:32:34 +00:00
Akira Hatanaka 9e6a8cca5d Align i64 arguments to 64 bit boundaries.
llvm-svn: 131668
2011-05-19 20:29:48 +00:00
Charles Davis 7e6e8951af Implement the Win64 EH prolog instruction methods on the base MCStreamer.
I had to change the API slightly to avoid overloading issues.

llvm-svn: 131666
2011-05-19 19:35:55 +00:00
Evan Cheng 2b9bd38678 crc32 with 64-bit output zeros upper 32-bits. rdar://9467055
llvm-svn: 131664
2011-05-19 18:57:12 +00:00
Stuart Hastings 4a4e5a2b55 Update some currently-disabled code, preparing for eventual use.
llvm-svn: 131663
2011-05-19 18:48:20 +00:00
Akira Hatanaka 05fd5aa1d0 Increase number of available registers when target is MIPS32.
llvm-svn: 131660
2011-05-19 18:25:03 +00:00
Evan Cheng dc867ae1fc Add comment.
llvm-svn: 131659
2011-05-19 18:18:39 +00:00
Akira Hatanaka 92ab6db6c8 Simplify CC_MipsO32 and merge it with CC_MipsO32_VarArgs. Patch by Sasa Stankovic.
llvm-svn: 131657
2011-05-19 18:06:05 +00:00
Joerg Sonnenberger 74ba2623ce Reapply 131644 including the missing header changes:
Introduce -fatal-assembler-warnings for the obvious purpose

llvm-svn: 131655
2011-05-19 18:00:13 +00:00
Stuart Hastings c72240bbd9 Reverting 131641 to investigate 'bot complaint.
llvm-svn: 131654
2011-05-19 17:54:42 +00:00
Eli Friedman 3e2f17499e Revert r131644; it's breaking the build.
llvm-svn: 131653
2011-05-19 17:48:09 +00:00
Charles Davis 4cd8856704 Turns out GAS does have Win64 EH directives. (It also supports WinCE EH.) Make
ours compatible with GAS.

In retrospect, I should have emailed binutils about this earlier. Thanks to
Kai Tietz for pointing out that GAS already had SEH directives.

llvm-svn: 131652
2011-05-19 17:46:39 +00:00
Jim Grosbach eca54e4e6d 80 columns.
llvm-svn: 131649
2011-05-19 17:34:53 +00:00
Joerg Sonnenberger 59b523169f Introduce -fatal-assembler-warnings for the obvious purpose
llvm-svn: 131644
2011-05-19 17:27:01 +00:00
Akira Hatanaka d1465bd68b Fix data layout string. i64 is aligned to 64 bit boundaries.
llvm-svn: 131642
2011-05-19 17:21:09 +00:00
Stuart Hastings b476b0cc9f Revise MOVSX16rr8/MOVZX16rr8 (and rm variants) to no longer be
pseudos.  rdar://problem/8614450

llvm-svn: 131641
2011-05-19 16:59:50 +00:00
Cameron Zwarich 9eb5a410bd Use the correct register class for Cell varargs spilling. This fixes all of the
verifier failures in the CodeGen/CellSPU tests.

llvm-svn: 131631
2011-05-19 04:44:19 +00:00
Mon P Wang 6d9e1c7c2e Fixed sdiv and udiv for <4 x i16>. The test from r125402 still applies for this change.
llvm-svn: 131630
2011-05-19 04:15:07 +00:00
Charles Davis b4222589bf Implement the StartChained and EndChained Win64 EH methods on MCStreamer.
llvm-svn: 131629
2011-05-19 04:04:13 +00:00
Cameron Zwarich 41025dc95b Make CodeGen/PowerPC/2007-09-11-RegCoalescerAssert.ll pass with the verifier.
llvm-svn: 131627
2011-05-19 03:11:06 +00:00
Cameron Zwarich dadd73390f Fix PR8828 by removing the explicit def in MovePCToLR as well as the pointless
piclabel operand. The operand in the tablegen definition doesn't actually turn
into an MI operand, so it just confuses anything checking the TargetInstrDesc
for the number of operands. It suffices to just have an implicit def of LR.

llvm-svn: 131626
2011-05-19 02:56:28 +00:00
Cameron Zwarich 58eafde58b Reuse the TargetInstrDesc.
llvm-svn: 131625
2011-05-19 02:56:23 +00:00
Charles Davis d991ec4e68 Implement the StartProc and EndProc Win64 EH methods on the base MCStreamer.
Based largely on Rafael Espindola's work on CFI. Other methods soon to follow.

llvm-svn: 131623
2011-05-19 02:49:00 +00:00
Rafael Espindola 964602d7ba revert 131605 to fix PR9946.
llvm-svn: 131620
2011-05-19 02:26:30 +00:00
Cameron Zwarich 00b780e280 Correctly constrain a register class when computing frame offsets, as the Thumb2
add instruction takes an rGPR. This fixes the last of PR8825.

llvm-svn: 131619
2011-05-19 02:18:27 +00:00
Cameron Zwarich fe5795a1f6 Revert r128961 because it didn't include a test and causes the verifier to fail
on CodeGen/X86/2007-05-07-InvokeSRet.ll. There is probably a bug here that was
fixed by r128961, but since there is no test or reference to a source file I have
to revert it.

llvm-svn: 131618
2011-05-19 01:56:19 +00:00
Eli Friedman 6efb64ea8e Make the demanded bits/elements optimizations preserve debug line information.
I'm not sure this is quite ideal, but I can't really think of any better way to do it.

llvm-svn: 131616
2011-05-19 01:20:42 +00:00
Jim Grosbach 13733a137c Avoid a Twine that referenced a tmp (which proceded to go out of scope before
the Twine was used).

llvm-svn: 131612
2011-05-19 00:45:05 +00:00
Devang Patel 3015a54813 Use IRBuilder.
llvm-svn: 131609
2011-05-19 00:13:33 +00:00
Devang Patel 31458a0002 Use IRBuilder while simplifying unreachable.
llvm-svn: 131607
2011-05-19 00:09:21 +00:00
Rafael Espindola f1bda7c90c Revert my previous patch. The cmake build had already been fixed.
llvm-svn: 131606
2011-05-19 00:02:45 +00:00
Devang Patel 4b13f39b77 Use IRBuilder while simplifying conditional branch.
llvm-svn: 131605
2011-05-18 23:59:51 +00:00
Eli Friedman 41e509a33d More instcombine cleanup, towards improving debug line info.
llvm-svn: 131604
2011-05-18 23:58:37 +00:00
Jim Grosbach 6bfca4efd8 Restore sanity to 131601.
llvm-svn: 131603
2011-05-18 23:56:43 +00:00
Rafael Espindola efbdd80d77 Fix the cmake build.
llvm-svn: 131602
2011-05-18 23:56:40 +00:00
Jim Grosbach b098471538 Objective C functions may use a magic '\1' on the name. Handle that when
dealing with them in the MCJIT.

llvm-svn: 131601
2011-05-18 23:53:21 +00:00
Eli Friedman d1b0f0f1b8 Shuffle StandardPasses.cpp into VMCore; add it to CMake.
llvm-svn: 131600
2011-05-18 23:51:11 +00:00
Devang Patel 7de6c4bf75 Use IRBuilder while simplifying branch.
llvm-svn: 131598
2011-05-18 23:18:47 +00:00
Eli Friedman 6fc94dd687 Revert unintentional commit.
llvm-svn: 131597
2011-05-18 23:13:10 +00:00
Eli Friedman 1754a25977 More instcombine simplifications towards better debug locations.
llvm-svn: 131596
2011-05-18 23:11:30 +00:00
Cameron Zwarich b9bef106c6 Add missing mayLoad / mayStore flags to instruction definitions without patterns,
which fixes all of the CodeGen/MBlaze verifier failures.

llvm-svn: 131595
2011-05-18 23:03:10 +00:00
Cameron Zwarich 9ddeceff19 Reserve the segment registers on x86 to fix verifier failures in any code that
uses them.

llvm-svn: 131591
2011-05-18 22:24:48 +00:00
Charles Davis 27117af772 Remove comments as Chris requested.
llvm-svn: 131590
2011-05-18 22:13:51 +00:00
Cameron Zwarich 8b2e426555 Reserve r29 on Alpha. This fixes all verifier failures in CodeGen/Alpha.
llvm-svn: 131587
2011-05-18 21:54:32 +00:00
Tanya Lattner 1d11720ae4 Handle perfect shuffle case that generates a vrev for vectors of floats.
Add test case.

llvm-svn: 131582
2011-05-18 21:44:54 +00:00
Eli Friedman fadb03d6cc Third pass at allowing plugins to modify default passes. This time with a tweak so that we don't depend on an uninitialized argument.
llvm-svn: 131581
2011-05-18 21:40:04 +00:00
Devang Patel dd14e0f7fa Use IRBuilder while simplifying return instruction.
llvm-svn: 131580
2011-05-18 21:33:11 +00:00
Cameron Zwarich 492f87de4e Fix an obvious typo in r131572.
llvm-svn: 131579
2011-05-18 21:29:07 +00:00
Cameron Zwarich ec645bf75d Fix more of PR8825. Now all of CodeGen/ARM passes with VerifyCoalescing turned on.
llvm-svn: 131578
2011-05-18 21:25:14 +00:00
Dan Gohman 3268e4d692 When forming an ICmpZero LSRUse, normalize the non-IV operand
of the comparison, so that the resulting expression is fully
normalized. This fixes PR9939.

llvm-svn: 131576
2011-05-18 21:02:18 +00:00
Charles Davis 77e0610d7c Add some more Win64 EH directives:
- StartChained and EndChained delimit a chained unwind area, which can contain
  additional operations to be undone if an exception occurs inside of it.
- UnwindOnly declares that this function doesn't handle any exceptions. If it
  has a handler, it's an unwind handler instead of an exception handler.
- Lsda declares the location and size of the LSDA, which in the Win64 EH
  scheme is kept inside the UNWIND_INFO struct. Windows itself ignores the
  LSDA; it's used by the Language-Specific Handler (the "Personality Function"
  from DWARF).

llvm-svn: 131572
2011-05-18 20:54:10 +00:00
Devang Patel 583805530c Spread use of IRBuilder even more.
llvm-svn: 131571
2011-05-18 20:53:17 +00:00
Eli Friedman 65816ac54d Revert r131556; it's breaking buildbots/clang tests.
llvm-svn: 131567
2011-05-18 20:39:27 +00:00
Devang Patel a7ec47d23c Use IRBuilder while simplifying switch instruction.
llvm-svn: 131566
2011-05-18 20:35:38 +00:00
Johnny Chen 071634612d Disassembly of tBcc was wrongly adding 4 to the SignExtend'ed imm8:'0' immediate operand.
llvm-svn: 131565
2011-05-18 20:32:41 +00:00
Devang Patel 0b373dca1f Use IRBuilder while simplifying unwind.
llvm-svn: 131561
2011-05-18 20:01:18 +00:00
Chad Rosier f4e832b14e Enables vararg functions that pass all arguments via registers to be optimized into tail-calls when possible.
llvm-svn: 131560
2011-05-18 19:59:50 +00:00
Eli Friedman 49346010f8 More instcombine cleanup aimed towards improving debug line info.
llvm-svn: 131559
2011-05-18 19:57:14 +00:00
David Chisnall 13c90786e1 Second pass at allowing plugins to modify default passes. This time without bonus inter-library dependencies.
llvm-svn: 131556
2011-05-18 19:00:41 +00:00
Evan Cheng 522fbfea3b Revise r131553. Just use the type of the input node and forgo the bitcast. rdar://9449159.
llvm-svn: 131555
2011-05-18 18:59:17 +00:00
Evan Cheng 80632c91b0 Fix an ARMTargetLowering::LowerSELECT bug: legalized result must have same type as input. Sorry test cases only trigger when dag combine is disabled. rdar://9449178
llvm-svn: 131553
2011-05-18 18:47:27 +00:00
Devang Patel 2c2ea226b7 Use IRBuilder while simplifying terminator.
llvm-svn: 131552
2011-05-18 18:43:31 +00:00
Devang Patel 767f6930bc Use IRBuilder while simplifying unconditional branch.
llvm-svn: 131551
2011-05-18 18:28:48 +00:00
Devang Patel 5c810ce4a3 Use IRBuilder while folding two entry PHINode.
llvm-svn: 131548
2011-05-18 18:16:44 +00:00
Eli Friedman 2fd66441c6 Switch more inst insertion in instcombine to IRBuilder.
llvm-svn: 131547
2011-05-18 18:10:28 +00:00
Devang Patel 15ad6761da Set up IRBuilder for use during simplification.
llvm-svn: 131545
2011-05-18 18:01:27 +00:00
Eli Friedman 0b43b9ee98 Switch more inst insertion in instcombine to IRBuilder.
llvm-svn: 131544
2011-05-18 17:58:37 +00:00
Matt Beaumont-Gay 8fa6ebf975 fix typo
llvm-svn: 131543
2011-05-18 17:37:10 +00:00
Eli Friedman cde9c1628c Switch inst insertion in instcombine transform to IRBuilder.
llvm-svn: 131542
2011-05-18 17:31:55 +00:00
Devang Patel 1fabbe921b Use IRBuiler while constant folding terminator.
llvm-svn: 131541
2011-05-18 17:26:46 +00:00
Stuart Hastings 728f6260b9 Fix inelegant initialization.
llvm-svn: 131538
2011-05-18 15:54:26 +00:00
Justin Holewinski bbdcd17d44 PTX: add flag to disable mad/fma selection
Patch by Dan Bailey

llvm-svn: 131537
2011-05-18 15:42:23 +00:00
Duncan Sands 3d9407f4eb Revert commit 131534 since it seems to have broken several buildbots.
Original log entry:
Refactor getActionType and getTypeToTransformTo ; place all of the 'decision'
code in one place.

llvm-svn: 131536
2011-05-18 14:57:56 +00:00
Nadav Rotem c5c27ede55 Refactor getActionType and getTypeToTransformTo ; place all of the 'decision'
code in one place.

llvm-svn: 131534
2011-05-18 12:26:38 +00:00
Duncan Sands 27bd0df352 Now that SrcBits and DestBits always represent the primitive size, rather
than either the primitive size or the element primitive size (in the case
of vectors), simplify the vector logic.  No functionality change.  There
is some distracting churn in the patch because I lined up comments better
while there - sorry about that.

llvm-svn: 131533
2011-05-18 10:59:25 +00:00
Duncan Sands 7f64656d21 Tighten up checking of the validity of casts. (1) The IR parser would
happily accept things like "sext <2 x i32> to <999 x i64>".  It would
also accept "sext <2 x i32> to i64", though the verifier would catch
that later.  Fixed by having castIsValid check that vector lengths match
except when doing a bitcast.  (2) When creating a cast instruction, check
that the cast is valid (this was already done when creating constexpr
casts).  While there, replace getScalarSizeInBits (used to allow more
vector casts) with getPrimitiveSizeInBits in getCastOpcode and isCastable
since vector to vector casts are now handled explicitly by passing to the
element types; i.e. this bit should result in no functional change.

llvm-svn: 131532
2011-05-18 09:21:57 +00:00
Duncan Sands a8514535a4 Teach getCastOpcode about element-by-element vector casts. For example, "trunc"
can be used to turn a <4 x i64> into a <4 x i32> but getCastOpcode would assert
if you passed these types to it.  Note that this strictly extends the previous
functionality: if getCastOpcode previously accepted two vector types (i.e. didn't
assert) then it still will and returns the same opcode (BitCast).  That's because
before it would only accept vectors with the same bitwidth, and the new code only
touches vectors with the same length.  However if two vectors have both the same
bitwidth and the same length then their element types have the same bitwidth, so
the new logic will return BitCast as before.

llvm-svn: 131530
2011-05-18 07:13:41 +00:00
Tanya Lattner 48b182c3a4 In r131488 I misunderstood how VREV works. It splits the vector in half and splits each half. Therefore, the real problem was that we were using a VREV64 for a 4xi16, when we should have been using a VREV32.
Updated test case and reverted change to the PerfectShuffle Table.

llvm-svn: 131529
2011-05-18 06:42:21 +00:00
Charles Davis 8d9c99042f Implement the Win64 EH directive methods for the assembly language streamer.
GAS has no such directives (not even mingw-w64 GAS has them), so I took
creative license with their names in assembly. I prefixed them all with
"w64_" to avoid namespace collisions, for example. If I discover that GAS
has taken a different approach, I'll change ours to match.

llvm-svn: 131525
2011-05-18 04:58:05 +00:00
Jakob Stoklund Olesen 07b5c276f4 Eliminate dead dead code elimination code.
llvm-svn: 131524
2011-05-18 04:51:15 +00:00
Jakob Stoklund Olesen f3dc225972 Also use shrinkToUses after AdjustCopiesBackFrom().
The 'last use' may not be in the same basic block, and we still want a correct
live range.

llvm-svn: 131523
2011-05-18 04:51:12 +00:00
Charles Davis 384041d0dc While thinking about how to know where the functions' boundaries are for
the purposes of the Win64 EH tables, I realized we had no way to tell where
the function ends. (MASM bounds functions with PROC and ENDP keywords.)
Add a directive to delimit the end of the function, and rename the 'frame'
directive to more accurately reflect its duality with the new directive.

llvm-svn: 131522
2011-05-18 04:47:22 +00:00
Jakob Stoklund Olesen 595a47d462 Properly shrink live ranges after deleting dead copies. Clean up after all joined copies.
LiveInterval::shrinkToUses recomputes the live range from scratch instead of
removing snippets. This should avoid the problem with dangling live ranges.

Leave physreg identity copies alone. They can be created when joining a virtreg
with a physreg. They don't affect register allocation, and they will be removed
by the rewriter.

llvm-svn: 131521
2011-05-18 04:18:19 +00:00
Cameron Zwarich f9839e4257 Fix typo.
llvm-svn: 131519
2011-05-18 02:29:50 +00:00
Cameron Zwarich d7c55fe2ef Fix more of PR8825 by correctly using rGPR registers when lowering atomic
compare-and-swap intrinsics.

llvm-svn: 131518
2011-05-18 02:20:07 +00:00
Eli Friedman 96254a0d53 Start trying to make InstCombine preserve more debug info. The idea here is to set the debug location on the IRBuilder, which will be then right location in most cases. This should magically give many transformations debug locations, and fixing places which are missing a debug location will usually just means changing the code creating it to use the IRBuilder.
As an example, the change to InstCombineCalls catches a common case where a call to a bitcast of a function is rewritten.

Chris, does this approach look reasonable?

llvm-svn: 131516
2011-05-18 01:28:27 +00:00
Eli Friedman b9ed18f2cb Use ReplaceInstUsesWith instead of replaceAllUsesWith where appropriate in instcombine.
llvm-svn: 131512
2011-05-18 00:32:01 +00:00
Devang Patel b849cd511b Preseve line numbers while simplifying CFG.
llvm-svn: 131508
2011-05-17 23:29:05 +00:00
Cameron Zwarich 33a67ddbd2 Actually, the address operand of the Thumb2 LDREX / STREX instructions *can*
take r13, so we can just make it a GPR. This fixes PR8825.

llvm-svn: 131507
2011-05-17 23:26:20 +00:00
Cameron Zwarich c5d272766f Correct a minor problem with the Thumb2 LDREX and STREX instruction encodings. They
were marked as taking a tGPR when in reality they take an rGPR.

llvm-svn: 131506
2011-05-17 23:11:12 +00:00
Bill Wendling 0671ba8448 Conditionalize the format of the GCOV files by target type. Darwin uses the 4.2
format.

llvm-svn: 131503
2011-05-17 23:05:13 +00:00
Eli Friedman e9692808b7 Make fast-isel miss counting in -stats and -fast-isel-verbose take terminators into account; since there are many fewer isel misses with recent changes, misses caused by terminators are more significant.
llvm-svn: 131502
2011-05-17 23:02:10 +00:00
Dan Gohman abffc991dc Misc. code cleanups.
llvm-svn: 131497
2011-05-17 22:22:52 +00:00
Dan Gohman 4298df6d86 Misc. code cleanups.
llvm-svn: 131495
2011-05-17 22:20:36 +00:00
Stuart Hastings 5bd18b6638 X86 pmovsx/pmovzx ignore the upper half of their inputs.
rdar://problem/6945110

llvm-svn: 131493
2011-05-17 22:13:31 +00:00
Tanya Lattner c7e291b354 vrev is incorrectly defined in the perfect shuffle table. The ordering is backwards (should be 0x3210 versus 0x1032) which exposed a bug when doing a shuffle on a 4xi16. I've attached a test case.
llvm-svn: 131488
2011-05-17 20:48:40 +00:00
Devang Patel 341b38c22a Preserve line number information.
llvm-svn: 131482
2011-05-17 20:00:02 +00:00
Devang Patel c5933f2418 Set debug loc for new load instruction.
llvm-svn: 131481
2011-05-17 19:43:38 +00:00
Devang Patel c23bcbc498 Preserve line number information.
llvm-svn: 131480
2011-05-17 19:43:06 +00:00
Mon P Wang 6f6b44d19d Enable autodetect of popcnt
llvm-svn: 131476
2011-05-17 18:33:37 +00:00
Eli Friedman 7b27942fe7 Add x86 fast-isel for calls returning first-class aggregates. rdar://9435872.
This is r131438 with a couple small fixes.

llvm-svn: 131474
2011-05-17 18:29:03 +00:00
Eli Friedman d000a2c26e Clean up the mess created by r131467+r131469.
llvm-svn: 131471
2011-05-17 18:02:22 +00:00
Stuart Hastings c65d8eda7b Revert 131467 due to buildbot complaint.
llvm-svn: 131469
2011-05-17 16:59:46 +00:00
Stuart Hastings 3cf5308890 Fix an obscure issue in X86_64 parameter passing: if a tiny byval is
passed as the fifth parameter, insure it's passed correctly (in R9).
rdar://problem/6920088

llvm-svn: 131467
2011-05-17 16:45:55 +00:00
Jakob Stoklund Olesen 52375e6a01 Tweak cross-class coalescing to be more aggressive when the target class is small.
The greedy register allocator has live range splitting and register class
inflation, so it can actually fully undo this join, including restoring the
original register classes.

We still don't want to do this for long live ranges, mostly because of the high
register pressure of there are many constrained live ranges overlapping.

llvm-svn: 131466
2011-05-17 16:38:37 +00:00
Rafael Espindola 0693182265 Don't include information about the build into the information returned by
getHostTriple. This fixes a 32 bit clang running on a 64 bit ubuntu.

llvm-svn: 131463
2011-05-17 15:26:34 +00:00
Nadav Rotem d8edb1d5cc Fix a bug in PerformEXTRACT_VECTOR_ELTCombine. The code created an ADD SDNode
with two different types, in cases where the index and the ptr had different
types.

llvm-svn: 131461
2011-05-17 08:31:57 +00:00
Eric Christopher 56a42ebf15 Update comment.
llvm-svn: 131459
2011-05-17 08:16:14 +00:00
Eric Christopher a1d9e29552 Support XOR and AND optimization with no return value.
Finishes off rdar://8470697

llvm-svn: 131458
2011-05-17 08:10:18 +00:00
Eric Christopher abfe3131e3 Couple less magic numbers.
llvm-svn: 131457
2011-05-17 07:50:41 +00:00
Eric Christopher eb47a2a1e5 Make this code a little less magic number laden.
llvm-svn: 131456
2011-05-17 07:47:55 +00:00
Chris Lattner 1e81f57bf0 add a note
llvm-svn: 131455
2011-05-17 07:22:33 +00:00
Eli Friedman 7335e8a720 Back out r131444 and r131438; they're breaking nightly tests. I'll look into
it more tomorrow.

llvm-svn: 131451
2011-05-17 02:36:59 +00:00
Eli Friedman 83ba150f3a Add x86 fast-isel for calls returning first-class aggregates. rdar://9435872.
llvm-svn: 131438
2011-05-17 00:13:47 +00:00
Owen Anderson 97f0cf32ea @llvm.lifetime.begin acts as a load, not @llvm.lifetime.end.
llvm-svn: 131437
2011-05-17 00:05:49 +00:00
Jakob Stoklund Olesen 4edf17d91f Teach LiveInterval::isZeroLength about null SlotIndexes.
When instructions are deleted, they leave tombstone SlotIndex entries.
The isZeroLength method should ignore these null indexes.

This causes RABasic to sometimes spill a callee-saved register in the
abi-isel.ll test, so don't run that test with -regalloc=basic.  Prioritizing
register allocation according to spill weight can cause more registers to be
used.

llvm-svn: 131436
2011-05-16 23:50:05 +00:00
Jim Grosbach 4e983166bc Kill some dead code.
llvm-svn: 131431
2011-05-16 22:24:07 +00:00
Dan Gohman d282f46c6b Delete unused variables.
llvm-svn: 131430
2011-05-16 22:19:54 +00:00
Dan Gohman d4d12d14b5 Trim #includes.
llvm-svn: 131429
2011-05-16 22:14:50 +00:00
Dan Gohman ae9b1685a8 Fix whitespace and 80-column violations.
llvm-svn: 131428
2011-05-16 22:09:53 +00:00
Devang Patel a0b682db62 There is no need to force DebugLoc on a PHI at this point.
llvm-svn: 131427
2011-05-16 22:05:03 +00:00
Jim Grosbach e85c0dde7a Track how many insns fast-isel successfully selects as well as how many it
misses.

llvm-svn: 131426
2011-05-16 21:51:07 +00:00
Eli Friedman d4a3609d30 Remove dead code. Fix associated test to use FileCheck.
llvm-svn: 131424
2011-05-16 21:28:22 +00:00
Devang Patel 8e60ff11db Preserve debug info for unused zero extended boolean argument.
Radar 9422775.

llvm-svn: 131422
2011-05-16 21:24:05 +00:00
Charles Davis 2701815ed8 Add a method I forgot in the last commit. Don't worry, this one passed
self-host :).

llvm-svn: 131421
2011-05-16 21:13:58 +00:00
Eli Friedman a4d4a0162d Make fast-isel work correctly s/uadd.with.overflow intrinsics.
llvm-svn: 131420
2011-05-16 21:06:17 +00:00
Eli Friedman 4c08bb450a Fix silly typo.
llvm-svn: 131419
2011-05-16 20:34:53 +00:00
Eli Friedman 9ac944774f Basic fast-isel of extractvalue. Not too helpful on its own, given the IR clang generates for cases like this, but it should become more useful soon.
llvm-svn: 131417
2011-05-16 20:27:46 +00:00
Jason W Kim 94c31fdd31 Add a FIXME reminder to remove ForceARMElfPIC switch.
llvm-svn: 131411
2011-05-16 16:35:21 +00:00
Rafael Espindola e90c1cb221 sets bit 0 of the function address of thumb function in .symtab
("T is 1 if the target symbol S has type STT_FUNC and the
symbol addresses a Thumb instruction ;it is 0 otherwise."
from "ELF for the ARM Architecture" 4.7.1.2)

Patch by Koan-Sin Tan!

llvm-svn: 131406
2011-05-16 16:17:21 +00:00
Rafael Espindola 71f8b08a80 Extra refactoring noticed by Eli Friedman.
llvm-svn: 131405
2011-05-16 15:48:45 +00:00
Rafael Espindola 2050af838d Don't do tail calls in a function that call setjmp. The stack might be
corrupted when setjmp returns again.

llvm-svn: 131399
2011-05-16 03:05:33 +00:00
Charles Davis 2f10078ae7 Fix copy-pasto.
llvm-svn: 131384
2011-05-15 17:28:27 +00:00
Charles Davis af18d07ed6 Add stub methods to MCStreamer for emitting Win64 exception-handling
information.

llvm-svn: 131382
2011-05-15 17:20:01 +00:00
Benjamin Kramer d96205c4e5 SimplifyCFG: Use ComputeMaskedBits to prune dead cases from switch instructions.
llvm-svn: 131345
2011-05-14 15:57:25 +00:00
Stuart Hastings 66a82b966e Avoid combining GEPs that might overflow at runtime.
rdar://problem/9267970

Patch by Julien Lerouge!

llvm-svn: 131339
2011-05-14 05:55:10 +00:00
Eli Friedman 8f1e11cde9 Fix a FIXME by moving the fast-isel implementation of the objectsize intrinsic from the x86 code to the generic code.
llvm-svn: 131332
2011-05-14 00:47:51 +00:00
Rafael Espindola df9db7ed92 Don't produce a vmovntdq if we don't have AVX support.
llvm-svn: 131330
2011-05-14 00:30:01 +00:00
Eli Friedman f080a57b81 Zap useless code; this hasn't done anything useful since fast-isel switched to being bottom-up (a very long time ago).
llvm-svn: 131329
2011-05-14 00:19:32 +00:00
Jim Grosbach f6b1846f39 Be a bit more permissive about symbols we don't understand. Just skip them
rather than throwing an error.

llvm-svn: 131322
2011-05-13 23:11:30 +00:00
Dylan Noblesmith 0bd34fbd01 ExecutionEngine: move createJIT() definition (v2)
As an ExecutionEngine class function, its definition
really belongs in ExecutionEngine.cpp, not JIT.cpp.

llvm-svn: 131320
2011-05-13 22:02:53 +00:00
Dylan Noblesmith 8418fdcd59 ExecutionEngine: push TargetMachine creation into clients (v2)
In particular, into EngineBuilder. This should only impact
the private API between the EE and EB classes, not external
clients, since JITCtor and MCJITCtor are both protected members.

llvm-svn: 131317
2011-05-13 21:51:29 +00:00
Dylan Noblesmith 1756faa472 ExecutionEngine: fix JIT/MCJIT selectTarget() duplication (v2)
This prepares for making JITCtor/MCJITCtor take a
TargetMachine* directly from clients like EngineBuilder.

llvm-svn: 131316
2011-05-13 21:36:16 +00:00
Jim Grosbach 65ed49dfc7 Teach the RtDyld to tell the memory manager about how much space a function
actually takes rather than how much memory was allocated for it. This
is more accurate and should help the manager pack things more effectively.

llvm-svn: 131305
2011-05-13 20:12:14 +00:00
Rafael Espindola e53b7d1a11 Make codegen able to handle values of empty types. This is one way
to fix PR9900. I will keep it open until sable is able to comment on it.

llvm-svn: 131294
2011-05-13 15:18:06 +00:00
Julien Lerouge 7e11f9e26d Fix a source of non determinism in FindUsedTypes, use a SetVector instead of a
set.

rdar://9423996

llvm-svn: 131283
2011-05-13 05:20:42 +00:00
Andrew Trick 03957dfeb1 Convert SimplifyIVUsers into a worklist instead of a single pass over
the users.

llvm-svn: 131277
2011-05-13 01:12:21 +00:00
Stuart Hastings aa02c0847d Since I can't reproduce the failures from 131261, re-trying with a
simplified version.  <rdar://problem/9298790>

llvm-svn: 131274
2011-05-13 00:51:54 +00:00
Stuart Hastings 8d57d8ea64 Revert 131266 and 131261 due to buildbot complaints.
rdar://problem/9298790

llvm-svn: 131269
2011-05-13 00:15:17 +00:00
Stuart Hastings 89f1b47e3a Non-fast-isel followup to 129634; correctly handle branches controlled
by non-CMP expressions.  The executable test case (129821) would test
this as well, if we had an "-O0 -disable-arm-fast-isel" LLVM-GCC
tester.  Alas, the ARM assembly would be very difficult to check with
FileCheck.

The thumb2-cbnz.ll test is affected; it generates larger code (tst.w
vs. cmp #0), but I believe the new version is correct.
rdar://problem/9298790

llvm-svn: 131261
2011-05-12 23:36:41 +00:00
Evan Cheng 8a88026ae3 Update comment.
llvm-svn: 131258
2011-05-12 22:35:48 +00:00
Devang Patel 12419aee7a Doug convinced me that DW_AT_APPLE_objc_complete_type is more appropriate name.
s/DW_AT_APPLE_objc_class_extension/DW_AT_APPLE_objc_complete_type/g

llvm-svn: 131244
2011-05-12 21:29:42 +00:00
Jim Grosbach 1b2d526c8b MCJIT section loading should just skip non-text sections rather than
erroring out completely. Some modules produce sections that aren't referenced,
so it's friendlier to clients like LLDB to just skip them, at least for now.

llvm-svn: 131243
2011-05-12 21:21:16 +00:00
Evan Cheng 43054e6159 Re-enable branchfolding common code hoisting optimization. Fixed a liveness test bug and also taught it to update liveins.
llvm-svn: 131241
2011-05-12 20:30:01 +00:00
Devang Patel 2409e7843b Let Objective-C front-end identify class extension, in dwarf output, using an attribute DW_AT_APPLE_objc_class_extension.
llvm-svn: 131238
2011-05-12 19:06:16 +00:00
Evan Cheng 33469b2cae Temporarily disable the transformation. It's breaking 186.crafty in some configuration.
llvm-svn: 131235
2011-05-12 18:44:58 +00:00
Jim Grosbach 9efe39ea38 The MCJIT memory manager needs to initialize its Module member.
llvm-svn: 131234
2011-05-12 18:21:23 +00:00
Akira Hatanaka e50a3d16e9 Fix setting of isCommutable flag.
llvm-svn: 131233
2011-05-12 17:42:08 +00:00
Evan Cheng cfdf33904b Re-commit 131172 with fix. MachineInstr identity checks should check dead
markers. In some cases a register def is dead on one path, but not on
another.

This is passing Clang self-hosting.

llvm-svn: 131214
2011-05-12 00:56:58 +00:00
Andrew Trick 81683ed232 indvars: Added SimplifyIVUsers.
Interleave IV simplifications. Currently involves EliminateComparison
and EliminateRemainder. Next I'll add EliminateExtend.

llvm-svn: 131210
2011-05-12 00:04:28 +00:00
Matt Beaumont-Gay 7c6c90b35d Remove an unused variable and move a couple others inside DEBUG.
llvm-svn: 131208
2011-05-11 23:34:51 +00:00
Jason W Kim c09e726455 Address the last bit of relocation flag related divergence betweeen
LLVM and binutils.

With this patch, there are no functional differences between the .o
produced directly from LLVM versus the .s to .o via GNU as, for relocation tags
at least, for both PIC and non-PIC modes.

Because some non-PIC reloc tags are used (legally) on PIC, so IsPCRel flag is
necessary but not sufficient to determine whether the overall codegen mode is
PIC or not. Why is this necessary? There is an incompatibility of how relocs
are emitted in the .rodata section.  Binutils PIC likes to emit certain relocs
as section relative offsets.  Non-PIC does not do this.

So I added a hidden switch on the ELFObjectwriter "-arm-elf-force-pic" which
forces the objectwriter to pretend that all relocs are for PIC mode.


Todo: Activate ForceARMElfPIC to true if -relocation-model=pic is selected
on llc.

Todo: There are probably more issues for PIC mode on ARM/MC/ELF...

Todo: Existing tests in MC/ARM/elf-reloc*.ll need to be converted over to .s 
tests as well as expanded to cover the gamut.

llvm-svn: 131205
2011-05-11 22:53:06 +00:00
Eric Christopher 2a9dbbbb12 Turn this into a table, this will make more sense shortly.
Part of rdar://8470697

llvm-svn: 131200
2011-05-11 21:44:58 +00:00
Devang Patel 34a6620748 Identify end of prologue (and beginning of function body) using DW_LNS_set_prologue_end line table opcode.
llvm-svn: 131194
2011-05-11 19:22:19 +00:00
Jakob Stoklund Olesen 31a0b5e2f0 Avoid hoisting spills when looking at a copy from another register that is also
about to be spilled.

This can only happen when two extra snippet registers are included in the spill,
and there is a copy between them. Hoisting the spill creates problems because
the hoist will mark the copy for later dead code elimination, and spilling the
second register will turn the copy into a spill.

<rdar://problem/9420853>

llvm-svn: 131192
2011-05-11 18:25:10 +00:00
Owen Anderson b745623b71 Fix encoding of Thumb BLX register instructions. Patch by Koan-Sin Tan.
llvm-svn: 131189
2011-05-11 17:00:48 +00:00
Andrew Trick 69005e5953 Typo and missing checkin from r131186.
llvm-svn: 131187
2011-05-11 16:44:08 +00:00
Andrew Trick 55aeb55aa3 Bugpoint support for miscompilations that result in a crash.
This change allows bugpoint to pinpoint the "opt" pass and bitcode
segment responsible for a crash caused by miscompilation. At least it
works well for me now, without having to create any custom execution
wrappers.

llvm-svn: 131186
2011-05-11 16:31:24 +00:00
Nadav Rotem 8a7beb80f0 Fixes a bug in the DAGCombiner. LoadSDNodes have two values (data, chain).
If there is a store after the load node, then there is a chain, which means
that there is another user. Thus, asking hasOneUser would fail. Instead we
ask hasNUsesOfValue on the 'data' value.

llvm-svn: 131183
2011-05-11 14:40:50 +00:00
Nadav Rotem 8f971c27fb Add custom lowering of X86 vector SRA/SRL/SHL when the shift amount is a splat vector.
llvm-svn: 131179
2011-05-11 08:12:09 +00:00
Rafael Espindola 2a09d65979 Revert 131172 as it is causing clang to miscompile itself. I will try
to provide a reduced testcase.

llvm-svn: 131176
2011-05-11 03:27:17 +00:00
Bill Wendling 50117f8186 Give the 'eh.sjlj.dispatchsetup' intrinsic call the value coming from the setjmp
intrinsic call. This prevents it from being reordered so that it appears
*before* the setjmp intrinsic (thus making it completely useless).
<rdar://problem/9409683>

llvm-svn: 131174
2011-05-11 01:11:55 +00:00
Evan Cheng 05fc35e275 Add a late optimization to BranchFolding that hoist common instruction sequences
at the start of basic blocks to their common predecessor. It's actually quite
common (e.g. about 50 times in JM/lencod) and has shown to be a nice code size
benefit. e.g.

        pushq   %rax
        testl   %edi, %edi
        jne     LBB0_2
## BB#1:
        xorb    %al, %al
        popq    %rdx
        ret
LBB0_2:
        xorb    %al, %al
        callq   _foo
        popq    %rdx
        ret

=>

        pushq   %rax
        xorb    %al, %al
        testl   %edi, %edi
        je      LBB0_2
## BB#1:
        callq   _foo
LBB0_2:
        popq    %rdx
        ret

rdar://9145558

llvm-svn: 131172
2011-05-11 01:03:01 +00:00
Eric Christopher 4a34e61e53 Optimize atomic lock or that doesn't use the result value.
Next up: xor and and.

Part of rdar://8470697

llvm-svn: 131171
2011-05-10 23:57:45 +00:00
Rafael Espindola dc9d9f8eeb Fix cmake again.
llvm-svn: 131164
2011-05-10 22:42:41 +00:00
Nick Lewycky 1ec5c8a8bc Remove empty file.
llvm-svn: 131162
2011-05-10 22:38:17 +00:00
Rafael Espindola de32fa163f Avoid a gcc warning.
llvm-svn: 131161
2011-05-10 22:28:35 +00:00
Rafael Espindola b361d2b2bc Fix cmake build.
llvm-svn: 131160
2011-05-10 22:19:33 +00:00
Nick Lewycky e09457b02b Revert r131155 for now. It makes VMCore depend on Analysis and Transforms
headers.

llvm-svn: 131159
2011-05-10 22:16:06 +00:00
Rafael Espindola a678098f24 Initialize moveTypeModule.
llvm-svn: 131157
2011-05-10 21:54:59 +00:00
Eli Friedman 768de0a0f8 Disable my little CopyToReg argument hack with fast-isel. rdar://problem/9413587 .
llvm-svn: 131156
2011-05-10 21:50:58 +00:00
David Chisnall c2fcb4df5e Add support for plugins add passes to the default set of passes. The standard set of passes used by front ends can now be modified by LLVM plugins, without needing to modify any front ends.
Still to do:

- Allow replacing / removing passes (infrastructure there, just needs an infrastructure exposed)
- Defining sets of passes to be added or removed as a group
- Extending the support to allow user-defined groups of optimisations
- Allow plugins to be specified for loading automatically (e.g. from plugins.conf or some similar mechanism)

Reviewed by Nick Lewycky.

llvm-svn: 131155
2011-05-10 21:36:48 +00:00
Stuart Hastings 999fa3bf1f Correctly walk through nested and adjacent CALLSEQ_START nodes. No
test case; I've only seen this on a release branch, and I can't get it
to reproduce on trunk.  rdar://problem/7662569

llvm-svn: 131152
2011-05-10 21:20:03 +00:00
Rafael Espindola 19c1a56287 Produce a __debug_frame section on darwin ARM when appropriate.
llvm-svn: 131151
2011-05-10 21:04:45 +00:00
Rafael Espindola 99f6735532 On MachO, unlike ELF, there should be no relocation to produce the CIE pointer.
llvm-svn: 131149
2011-05-10 20:59:42 +00:00
Rafael Espindola 2fe0ee1ce2 Rename DwarfRequiresRelocationForStmtList to
DwarfRequiresRelocationForSectionOffset as this is not specific to StmtList.

llvm-svn: 131148
2011-05-10 20:35:05 +00:00
Rafael Espindola d0d2354258 The EH symbols are only needed in eh_frame, not debug_frame.
llvm-svn: 131146
2011-05-10 19:51:53 +00:00
Rafael Espindola fdc3e6fab6 Use .cfi_sections to put the unwind info in .debug_frame when possible. With
this clang will use .debug_frame in, for example,
clang -g -c -m32 test.c
This matches gcc's behaviour. It looks like .debug_frame is a bit bigger
than .eh_frame, but has the big advantage of not being allocated.

llvm-svn: 131140
2011-05-10 18:39:09 +00:00
Eric Christopher e33464663f Refactor lock versions of binary operators to be a little less
cut and paste.

llvm-svn: 131139
2011-05-10 18:36:16 +00:00
Jason W Kim 62f1ab79ec First cut at getting debugging support for ARM/MC/ELF/.o
DWARF stuff also gets fixed up by ELFARMAsmBackend::ApplyFixup(),
but the offset is not guaranteed to be mod 4 == 0 as in text/data. 

llvm-svn: 131137
2011-05-10 18:07:25 +00:00
Jakob Stoklund Olesen 50215afd8a Fix PR9883. Make sure all caches are invalidated when a live range is repaired.
The previous invalidation missed the alias interference caches.

Also add a stats counter for the number of repaired ranges.

llvm-svn: 131133
2011-05-10 17:37:41 +00:00
Rafael Espindola 27390b4a0e In a debug_frame the cfi offset is to the start of the debug_frame section!
llvm-svn: 131129
2011-05-10 15:20:23 +00:00
Rafael Espindola ec53aa9485 Add CFIStartSections to the asm printer. Add an assert that at least
one of the sections is created.

llvm-svn: 131124
2011-05-10 13:39:48 +00:00
Justin Holewinski 72d74e4606 PTX: add PTX 2.3 setting in PTX sub-target.
Patch by Wei-Ren Chen

llvm-svn: 131123
2011-05-10 12:32:11 +00:00
Rafael Espindola 1ecb12fc57 Add support for producing .deubg_frame sections.
llvm-svn: 131121
2011-05-10 03:54:12 +00:00
Rafael Espindola b6444c0401 Small cleanups.
llvm-svn: 131120
2011-05-10 03:26:21 +00:00
Rafael Espindola b6089d6cac Factor some code into a new EmitFrames method.
llvm-svn: 131119
2011-05-10 03:14:15 +00:00
Rafael Espindola 0e130d1d37 Remove unused argument.
llvm-svn: 131118
2011-05-10 03:01:39 +00:00
Rafael Espindola 74b101f3e3 Parsing and plumbing for .cfi_sections.
llvm-svn: 131117
2011-05-10 01:10:18 +00:00
Devang Patel 3fd06f760b Preserve line number information.
llvm-svn: 131112
2011-05-10 00:03:11 +00:00
Devang Patel 589845d887 Do not ignore InlinedAt while walking up scope chain to find subprogram node.
llvm-svn: 131106
2011-05-09 22:14:49 +00:00
Eric Christopher 4480428474 Look through struct wrapped types for inline asm statments.
Patch by Evan Cheng.

llvm-svn: 131093
2011-05-09 20:04:43 +00:00
Dan Gohman 0daf687e1d Change a few std::maps to DenseMaps.
llvm-svn: 131088
2011-05-09 18:44:09 +00:00
Eric Christopher 5dc19f916c Fix td file comments for Mips.
Patch by Liu <proljc@gmail.com>!

llvm-svn: 131086
2011-05-09 18:16:46 +00:00
Mon P Wang 92ff16b7bb Fixed MC encoding for index_align for VLD1/VST1 (single element from one lane) for size 32
llvm-svn: 131085
2011-05-09 17:47:27 +00:00
Duncan Sands 6be291a2cd Indent properly, no functionality change.
llvm-svn: 131082
2011-05-09 08:03:33 +00:00
NAKAMURA Takumi 392445823d Eliminate an unused line to fix a warning.
llvm-svn: 131078
2011-05-08 23:19:04 +00:00
Jakob Stoklund Olesen 89bd2ae517 Remove an assertion to fix PR9872.
It can happen that a live debug variable is the last use of a sub-register, and
the register allocator will pick a larger register class for the virtual
register.  If the allocated register doesn't support the sub-register index,
just use %noreg for the debug variables instead of asserting.

In PR9872, a debug variable ends up in the sub_8bit_hi part of a GR32_ABCD
register. The register is split and one part is inflated to GR32 and assigned
%ESI because there are no more normal uses of sub_8bit_hi.

Since %ESI doesn't have that sub-register, substPhysReg asserted. Now it will
simply insert a %noreg instead, and the debug variable will be marked
unavailable in that range.

We don't currently have a way of saying: !"value" is in bits 8-15 of %ESI, I
don't know if DWARF even supports that.

llvm-svn: 131073
2011-05-08 19:21:08 +00:00
Benjamin Kramer d724a590e5 X86: Add a bunch of peeps for add and sub of SETB.
"b + ((a < b) ? 1 : 0)" compiles into
	cmpl	%esi, %edi
	adcl	$0, %esi
instead of
	cmpl	%esi, %edi
	sbbl	%eax, %eax
	andl	$1, %eax
	addl	%esi, %eax

This saves a register, a false dependency on %eax
(Intel's CPUs still don't ignore it) and it's shorter.

llvm-svn: 131070
2011-05-08 18:36:07 +00:00
Rafael Espindola ab39c63538 Don't force relaxation of AdvanceLoc instructions on OS X. gdb is happy with
the smaller encoding and this cuts 270336 bytes from a release version of
clang and 1246272 bytes from a debug build.

llvm-svn: 131067
2011-05-08 14:35:21 +00:00
Jakob Stoklund Olesen c132174169 Eliminate the ARM sub-register indexes that are not needed by the sources.
Tablegen will invent its own names for these indexes, and the register file is a
bit simpler.

llvm-svn: 131059
2011-05-07 21:22:42 +00:00
Duncan Sands af32728a57 The comparision "max(x,y)==x" is equivalent to "x>=y". Since the max is
often expressed as "x >= y ? x : y", there is a good chance we can extract
the existing "x >= y" from it and use that as a replacement for "max(x,y)==x".

llvm-svn: 131049
2011-05-07 16:56:49 +00:00
Eric Christopher 1e3db02bda Fix the non-MC encoding of pkhbt and pkhtb.
Patch by Stephen Hines.

llvm-svn: 131045
2011-05-07 04:37:27 +00:00
Jakob Stoklund Olesen 58799a35e5 Revert ExecutionEngine patches, they either failed to build or broke unit tests.
Please ensure the build is clean and tests are passing when recommitting.

llvm-svn: 131044
2011-05-07 03:12:54 +00:00
Rafael Espindola bcabcdcedd Switch Darwin to the generic CIE/FDE printer.
llvm-svn: 131031
2011-05-06 22:29:04 +00:00
Dylan Noblesmith 4c261a2831 ExecutionEngine: delete duplicated files
Forgot to `svn rm` these in revisions 131025 / 131029.

llvm-svn: 131030
2011-05-06 22:24:04 +00:00
Dylan Noblesmith 00dd3d6001 ExecutionEngine: add missing file
From revision 131025.

llvm-svn: 131029
2011-05-06 22:20:09 +00:00
Akira Hatanaka cbb7fa68ed 1. Keep lines in 80 columns.
2. Remove unused function.
3. Correct indentation.

llvm-svn: 131028
2011-05-06 22:11:29 +00:00
Dylan Noblesmith 8c0487006e ExecutionEngine: move createJIT() definition
As an ExecutionEngine class function, its definition
really belongs in ExecutionEngine.cpp, not JIT.cpp.

llvm-svn: 131027
2011-05-06 22:07:14 +00:00
Dylan Noblesmith ab6bd20da9 ExecutionEngine: push TargetMachine creation into clients
In particular, into EngineBuilder. This should only impact
the private API between the EE and EB classes, not external
clients, since JITCtor and MCJITCtor are both protected members.

llvm-svn: 131026
2011-05-06 22:06:22 +00:00
Dylan Noblesmith fed7f0b3a0 ExecutionEngine: fix JIT/MCJIT selectTarget() duplication
This prepares for making JITCtor/MCJITCtor take a
TargetMachine* directly from clients like EngineBuilder.

llvm-svn: 131025
2011-05-06 22:05:43 +00:00
Jakob Stoklund Olesen a5c889982a Emit a proper error message when register allocators run out of registers.
This can't be just an assertion, users can always write impossible inline
assembly. Such an assembly statement should be included in the error message.

llvm-svn: 131024
2011-05-06 21:58:30 +00:00
Andrew Trick 2e116a4491 Added an assertion, and updated a comment.
llvm-svn: 131022
2011-05-06 21:52:52 +00:00
Nick Lewycky 64c9284411 It's valid to take the blockaddress of a different function, so remove this
assert in the bitcode writer. No change needed because the ValueEnumerator holds
a whole-module numbering anyhow. Fixes PR9857!

llvm-svn: 131016
2011-05-06 21:09:44 +00:00
Evan Cheng d26fc5e013 80 col violations.
llvm-svn: 131015
2011-05-06 20:52:23 +00:00
Eli Friedman 2518f8376d Make the logic for determining function alignment more explicit. No functionality change.
llvm-svn: 131012
2011-05-06 20:34:06 +00:00
Eli Friedman 7a78f66145 Use array_lengthof. No functional change.
llvm-svn: 131008
2011-05-06 19:50:10 +00:00
Jakob Stoklund Olesen 57c8f58aeb Iterate backwards over debug locations when splitting them so they can be safely erased.
This should unbreak dragonegg-i386-linux and build-self-4-mingw32.

llvm-svn: 131007
2011-05-06 19:31:19 +00:00
Andrew Trick 84f9ad9bf2 Typo: Reviewed by Alistair.
llvm-svn: 131001
2011-05-06 18:14:32 +00:00
Jakob Stoklund Olesen f8da028895 Update LiveDebugVariables after live range splitting.
After a virtual register is split, update any debug user variables that resided
in the old register. This ensures that the LiveDebugVariables are still correct
after register allocation.

This may create DBG_VALUE instructions that place a user variable in a register
in parts of the function and in a stack slot in other parts. DwarfDebug
currently doesn't support that.

llvm-svn: 130998
2011-05-06 18:00:02 +00:00
Jakob Stoklund Olesen c86fe05923 Use TargetMachine hooks to properly print debug variable locations.
llvm-svn: 130997
2011-05-06 17:59:59 +00:00
Jakob Stoklund Olesen 6cc4e4ddd0 Also count identity copies.
llvm-svn: 130996
2011-05-06 17:59:57 +00:00
Andrew Trick aab77fe574 Post-RA scheduler compile time fix. Quadratic computation of DAG node depth.
The post-ra scheduler was explicitly updating the depth of a node's
successors after scheduling it, regardless of whether the successor
was ready. This is quadratic for DAGs with transitively redundant
edges. I simply removed the useless update of depth, which is lazilly
computed later.
Fixes <rdar://problem/9044332> compiler takes way too long to build TextInput.

llvm-svn: 130992
2011-05-06 17:09:08 +00:00
Devang Patel 17b532728b Move CompileUnit::getOrCreateNameSpace() and CompileUnit::addPubType() from DwarfDebug.cpp to DwarfCompileUnit.cpp
llvm-svn: 130991
2011-05-06 16:57:54 +00:00
Rafael Espindola 705d25a1f4 Nothing else uses this label.
llvm-svn: 130989
2011-05-06 15:44:29 +00:00
Rafael Espindola 6e4fa20efb Yet more dead code.
llvm-svn: 130988
2011-05-06 15:31:55 +00:00
Rafael Espindola d06c2c1b81 Update comments.
llvm-svn: 130987
2011-05-06 15:28:56 +00:00
Rafael Espindola 4bfa978ca5 More dead code elimination.
llvm-svn: 130985
2011-05-06 15:22:26 +00:00
Rafael Espindola a716096677 Dead code elimination.
llvm-svn: 130984
2011-05-06 14:56:22 +00:00
Justin Holewinski 11d70b6b32 PTX: add PTX 2.3 language target
Patch by Wei-Ren Chen

llvm-svn: 130980
2011-05-06 11:40:36 +00:00
Duncan Sands a071c82900 Fix PR9820: a read-only call differs from a load in that a load doesn't
return the pointer being dereferenced, it returns the pointee, but a call
might return the pointer itself.

llvm-svn: 130979
2011-05-06 10:30:37 +00:00
Eli Friedman 5401962643 Re-revert r130877; it's apparently causing a regression on 197.parser,
possibly related to cbnz formation.

llvm-svn: 130977
2011-05-06 05:23:07 +00:00
Nick Lewycky a7028848a1 The computation of string length is not that complicated. Fix it, again. :)
llvm-svn: 130967
2011-05-05 23:52:18 +00:00
Rafael Espindola bb237c72a6 Remove DwarfTableException.
llvm-svn: 130964
2011-05-05 23:19:54 +00:00
Rafael Espindola ec0a3a0bbf Remove the DwarfTable enum.
llvm-svn: 130959
2011-05-05 22:14:31 +00:00