1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-01 00:12:50 +01:00
Commit Graph

83428 Commits

Author SHA1 Message Date
Duncan Sands
f9eec0d373 Use the right kind of booleans: we were emitting 0/1 booleans, instead of 0/-1
booleans.  Patch by James Benton.

llvm-svn: 159739
2012-07-05 09:32:46 +00:00
Kostya Serebryany
4d76b79c2a [tsan] fix compile-time falilure found while building Chromium with tsan (tsan issue #3). A unit test will follow separately.
llvm-svn: 159736
2012-07-05 09:07:31 +00:00
Nick Lewycky
7e3f11f34d Remove ParentMap. You can just ask the domnode for its parent. No functionality
change.

Move the "Not profitable, avoid CSE!" debug message next to where we fail the
check for profitability and use a different message for avoiding CSE due to
being in different register classes.

llvm-svn: 159729
2012-07-05 06:19:21 +00:00
Jakob Stoklund Olesen
6edf66ffe8 Make X86 call and return instructions non-variadic.
Function argument and return value registers aren't part of the
encoding, so they should be implicit operands.

llvm-svn: 159728
2012-07-04 23:53:27 +00:00
Jakob Stoklund Olesen
e8399d2b3a Allow trailing physreg RegisterSDNode operands on non-variadic instructions.
Also allow trailing register mask operands on non-variadic both
MachineSDNodes and MachineInstrs.

The extra physreg RegisterSDNode operands are added to the MI as
<imp-use> operands. This makes it possible to have non-variadic call
instructions.

Call and return instructions really are non-variadic, the argument
registers should only be used implicitly - they are not part of the
encoding.

llvm-svn: 159727
2012-07-04 23:53:23 +00:00
Jakob Stoklund Olesen
2d29683988 Print SlotIndexes when available for -print-machineinstrs.
llvm-svn: 159726
2012-07-04 23:53:19 +00:00
Jakob Stoklund Olesen
bd6116e78c Allow multiple terminators to read virtual registers.
Find the kill as the last terminator to read SrcReg.

Patch by Philipp Brüschweiler!

llvm-svn: 159722
2012-07-04 19:52:05 +00:00
Jakob Stoklund Olesen
795083115c Ensure CopyToReg nodes are always glued to the call instruction.
The CopyToReg nodes that set up the argument registers before a call
must be glued to the call instruction. Otherwise, the scheduler may emit
the physreg copies long before the call, causing long live ranges for
the fixed registers.

Besides disabling good register allocation, that can also expose
problems when EmitInstrWithCustomInserter() splits a basic block during
the live range of a physreg.

llvm-svn: 159721
2012-07-04 19:28:31 +00:00
Jakob Stoklund Olesen
df04640e90 Make sure -print-machineinstrs applies to the first pass as well.
llvm-svn: 159720
2012-07-04 19:28:27 +00:00
Rafael Espindola
d06801429f Add a testcase for pr13209. It is not a great test, but it still fails if
159509 and 159479 are reverted. It would be really nice to be able to run
just the coalescer :-(

llvm-svn: 159715
2012-07-04 16:06:00 +00:00
Alexey Samsonov
a887f63924 Fix a bug in DebugInfo lib, extend a comment for one of the methods
llvm-svn: 159707
2012-07-04 09:42:54 +00:00
Stepan Dyatkovskiy
9e98d08303 Reverted r159658:
Optimized diff operation: implemented the case when LHS and RHS subsets contains single numbers only.

llvm-svn: 159704
2012-07-04 06:07:06 +00:00
Stepan Dyatkovskiy
bb30cc0d2f Reverted r156659, due to probable performance regressions, DenseMap should be used here:
IntegersSubsetMapping
  - Replaced type of Items field from std::list with std::map. In neares future I'll test it with DenseMap and do the correspond replacement
    if possible.

llvm-svn: 159703
2012-07-04 05:53:05 +00:00
Eric Christopher
8266e56fdb Reduce some code duplication.
llvm-svn: 159701
2012-07-04 02:02:18 +00:00
Matt Beaumont-Gay
723207e88f Fix some ascii art in a comment to not have trailing backslashes (inspiration
from IfConversion.cc), and fix some spelling and grammar in the surrounding
prose.

llvm-svn: 159699
2012-07-04 01:09:45 +00:00
Jakob Stoklund Olesen
79846e5c9b Add early if-conversion support to X86.
Implement the TII hooks needed by EarlyIfConversion to create cmov
instructions and estimate their latency.

Early if-conversion is still not enabled by default.

llvm-svn: 159695
2012-07-04 00:09:58 +00:00
Jakob Stoklund Olesen
db187a51eb Add an experimental early if-conversion pass, off by default.
This pass performs if-conversion on SSA form machine code by
speculatively executing both sides of the branch and using a cmov
instruction to select the result. This can help lower the number of
branch mispredictions on architectures like x86 that don't have
predicable instructions.

The current implementation is very aggressive, and causes regressions on
mosts tests. It needs good heuristics that have yet to be implemented.

llvm-svn: 159694
2012-07-04 00:09:54 +00:00
Nuno Lopes
981f2aae39 PHINode::hasConstantValue(): return undef if the PHI is fully recursive.
Thanks Duncan for the idea

llvm-svn: 159687
2012-07-03 21:15:40 +00:00
Bill Wendling
4e014100a4 Use the DebugInfo's 'print()' method to emit the comments.
These give quite a bit more information about the DebugInfo and makes it more
readable.

llvm-svn: 159680
2012-07-03 20:01:02 +00:00
Nuno Lopes
eac3a6d03c BoundsChecking: optimize out the check for offset < 0 if size is known to be >= 0 (signed).
(LLVM optimizers cannot do this optimization by themselves)

llvm-svn: 159668
2012-07-03 17:30:18 +00:00
Nuno Lopes
af1caad2cb fold PHI nodes in SizeOffsetEvaluator whenever possible.
Unfortunately this change requires the cache map to hold WeakVHs instead

llvm-svn: 159667
2012-07-03 17:13:25 +00:00
Nuno Lopes
7de83f28a2 improve PHINode::hasConstantValue() to detect recursive cases like %phi = phi(%phi,42) as constant
llvm-svn: 159666
2012-07-03 17:10:28 +00:00
Stepan Dyatkovskiy
b0319db9b4 IntegersSubsetMappin: cosmetic changes in diff operation.
llvm-svn: 159661
2012-07-03 14:29:26 +00:00
Stepan Dyatkovskiy
b02f8f2bbc Part of r159527. Splitted into series of patches and gone with fixed PR13256:
IntegersSubsetMapping
  Added new methods
  - add(self& RHS, SuccessorClass *S)
  - detachCase
  - removeCase
  - findSuccessor
  - getCases
  - getCaseSingleNumber
  - isOverlapped
  

llvm-svn: 159660
2012-07-03 14:15:36 +00:00
Stepan Dyatkovskiy
a646b05bfe Part of r159527. Splitted into series of patches and gone with fixed PR13256:
IntegersSubsetMapping
  - Replaced type of Items field from std::list with std::map. In neares future I'll test it with DenseMap and do the correspond replacement
    if possible.

llvm-svn: 159659
2012-07-03 13:46:45 +00:00
Stepan Dyatkovskiy
037528232b Part of r159527. Splitted into series of patches and gone with fixed PR13256:
Optimized diff operation: implemented the case when LHS and RHS subsets contains single numbers only.

llvm-svn: 159658
2012-07-03 13:29:14 +00:00
Peter Collingbourne
9918d151c5 Document the llvm.fabs intrinsic.
llvm-svn: 159657
2012-07-03 12:25:40 +00:00
Chandler Carruth
22eafcefb4 Micro-optimize this function a bit. This shrinks the generated code
some, and allows the routine to be inlined into common callers. The
various bits that hit this code in their hotpath seem slightly lower on
the profile, but I can't really measure a performance improvement as
everything seems to still be bottlenecked on likely cache misses. =/

llvm-svn: 159648
2012-07-03 07:16:13 +00:00
Craig Topper
4644d3577c Remove extra space.
llvm-svn: 159647
2012-07-03 06:48:58 +00:00
Craig Topper
60bbc2fde8 Change i128mem/i256mem to f128mem/f256mem on some floating point vector instructions.
llvm-svn: 159646
2012-07-03 06:11:06 +00:00
NAKAMURA Takumi
ebea37bea1 MCContext.cpp: Fixup for my odd previous commit. No functional changes.
llvm-svn: 159645
2012-07-03 06:01:27 +00:00
Craig Topper
6fcb4454a0 Add aliases for pblendvb, blendvpd, and blendvps instructions with the implicit xmm0 operand specified. Fixes PR13252.
llvm-svn: 159644
2012-07-03 05:49:45 +00:00
NAKAMURA Takumi
33e15f03ee test/CodeGen/SPARC/private.ll: Fixup. Forgot to prune old RUN lines.
llvm-svn: 159643
2012-07-03 04:29:20 +00:00
NAKAMURA Takumi
8e3879b6ce test/CodeGen/SPARC/private.ll: FileCheck-ize.
llvm-svn: 159642
2012-07-03 04:21:57 +00:00
NAKAMURA Takumi
99db282ba1 llvm/test/lit.cfg: Retweak for Win32 to fix testing.
- execute_external should be;
    - Not on Win32.
    - Using bash.
    In reverse, "execute_internal" shoud be (Win32 && !bash).

  - lit.getBashPath() behaves differently before and after tweaking $PATH.

I will add a few explanations there later.

llvm-svn: 159641
2012-07-03 03:59:34 +00:00
NAKAMURA Takumi
a96a98098e MCContext::GetDwarfFile(): Make FileName parsing tolerant of DOSish pathsep with PathV2.
It fixes failure in test/MC/MachO/gen-dwarf.s on Win32 w/o bash.

llvm-svn: 159640
2012-07-03 03:59:29 +00:00
NAKAMURA Takumi
ebfaec6c36 test/CodeGen/X86/sincos.ll: FileCheck-ize.
llvm-svn: 159639
2012-07-03 03:59:22 +00:00
NAKAMURA Takumi
2c39bfe0fe test/CodeGen/X86/fabs.ll: FileCheck-ize.
llvm-svn: 159638
2012-07-03 03:59:15 +00:00
NAKAMURA Takumi
d1831ee438 test/CodeGen/X86/2007-09-05-InvalidAsm.ll: FileCheck-ize.
llvm-svn: 159637
2012-07-03 03:59:08 +00:00
NAKAMURA Takumi
6c6ec244da test/CodeGen/X86/2004-03-30-Select-Max.ll: FileCheck-ize.
llvm-svn: 159636
2012-07-03 03:58:59 +00:00
Jack Carter
0e58c3f697 mips32 long long register inline asm constraint support.
inlineasm-cnstrnt-bad-r-1.ll is NOT supposed to fail, so it was removed.    This resulted in the removal of a negative test (inlineasm-cnstrnt-bad-r-1.ll)
    

llvm-svn: 159625
2012-07-02 23:35:23 +00:00
Eric Christopher
07e1aa6bfe Revert " mips32 long long register inline asm constraint support." as
it appears to be breaking the bots.

This reverts commit 1b055ce320fa13f6f1ac81670d11b45e01f79876.

llvm-svn: 159619
2012-07-02 23:22:25 +00:00
Eric Christopher
185293d560 Revert "IntRange:" as it appears to be breaking self hosting.
This reverts commit b2833d9dcba88c6f0520cad760619200adc0442c.

llvm-svn: 159618
2012-07-02 23:22:21 +00:00
Jack Carter
f66ba1e2a3 deleted test/CodeGen/Mips/inlineasm-cnstrnt-bad-r-1.ll
llvm-svn: 159617
2012-07-02 23:21:22 +00:00
Chandler Carruth
33def20e63 All glory to address sanitizer. ;]
It appears to have caught a use-after-free introduced as by r159567
and/or friends which call 'addPass' from many more places. The bug in
'addPass' doesn't appear to be new, and was spotted by inspection when
ASan shown a bright light of a stacktrace at these functions.

Hopefully this will fix the ASan failure -- I have no test case other
than running an ASan-built clang over the test suite.

llvm-svn: 159614
2012-07-02 22:56:41 +00:00
Evan Cheng
6196c5f5f3 Target option DisableJumpTables is a gross hack. Move it to TargetLowering instead.
llvm-svn: 159611
2012-07-02 22:39:56 +00:00
Jack Carter
64aeffc069 mips32 long long register inline asm constraint support.
inlineasm-cnstrnt-bad-r-1.ll is NOT supposed to fail, so it was removed.    This resulted in the removal of a negative test (inlineasm-cnstrnt-bad-r-1.ll)
    

llvm-svn: 159610
2012-07-02 22:39:45 +00:00
Andrew Trick
8bd81116b5 misched: allow NULL InstrItineraries.
llvm-svn: 159599
2012-07-02 21:55:12 +00:00
Chandler Carruth
4081d7f1e7 Extend the workaround from r159593 to cover a few explicit alias
targets.

llvm-svn: 159597
2012-07-02 21:45:22 +00:00
Chandler Carruth
5be5cafa65 Revert r159588, and apply a more principled fix. Place the fix for this
in the abstraction for lit test suites so that the various other layers
of abstraction pick up the same behavioral fix, and so that we still get
a complete list of dependencies for the 'check-all' target.

This should fix the follow-on issues of the same nature with various
other build targets, including Clang targets. Sorry for the churn, and
again thanks to Matt for testing and breaking this more thoroughly.

llvm-svn: 159593
2012-07-02 21:31:03 +00:00