Saleem Abdulrasool [Sun, 9 Feb 2014 01:12:21 +0000 (01:12 +0000)]
tools: explicitly cast to avoid a warning
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201032
91177308-0d34-0410-b5e6-
96231b3b80d8
Saleem Abdulrasool [Sat, 8 Feb 2014 23:17:08 +0000 (23:17 +0000)]
tools: handle out-of-line personality 0 decoding
In some cases it is possible to have a personality 0 unwinding opcodes in the
extab (such as when .handlerdata is used in the assembly). Simply decode the 3
opcodes for that case.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201030
91177308-0d34-0410-b5e6-
96231b3b80d8
Saleem Abdulrasool [Sat, 8 Feb 2014 23:17:02 +0000 (23:17 +0000)]
ARM: change attribute tests to use parsed form
This makes the tests more readable by using the -arm-attributes decoding support
in llvm-readobj since that is now available. Change the invocation commands to
be similar to other test and use a more precise triple (the tests only require
ARM EABI support).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201029
91177308-0d34-0410-b5e6-
96231b3b80d8
Arnold Schwaighofer [Sat, 8 Feb 2014 20:41:13 +0000 (20:41 +0000)]
LoopVectorizer: Keep track of conditional store basic blocks
Before conditional store vectorization/unrolling we had only one
vectorized/unrolled basic block. After adding support for conditional store
vectorization this will not only be one block but multiple basic blocks. The
last block would have the back-edge. I updated the code to use a vector of basic
blocks instead of a single basic block and fixed the users to use the last entry
in this vector. But, I forgot to add the basic blocks to this vector!
Fixes PR18724.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201028
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Sat, 8 Feb 2014 14:53:28 +0000 (14:53 +0000)]
Pass the Mangler by reference.
It is never null and it is not used in casts, so there is no reason to use a
pointer. This matches how we pass TM.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201025
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Sat, 8 Feb 2014 06:07:27 +0000 (06:07 +0000)]
Add LLVM_OVERRIDE to a few declarations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201022
91177308-0d34-0410-b5e6-
96231b3b80d8
Juergen Ributzka [Sat, 8 Feb 2014 00:20:49 +0000 (00:20 +0000)]
[Constant Hoisting] Fix insertion point for constant materialization.
The bitcast instruction during constant materialization was not placed correcly
in the presence of phi nodes. This commit fixes the insertion point to be in the
idom instead.
This fixes PR18768
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201009
91177308-0d34-0410-b5e6-
96231b3b80d8
Juergen Ributzka [Sat, 8 Feb 2014 00:20:45 +0000 (00:20 +0000)]
[Constant Hoisting] Don't update the use list while traversing it - DOH!
This fix first traverses the whole use list of the constant expression and
keeps track of the instructions that need to be updated. Then perform the
fixup afterwards.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201008
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Fri, 7 Feb 2014 23:32:41 +0000 (23:32 +0000)]
Remove dead code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201006
91177308-0d34-0410-b5e6-
96231b3b80d8
Nikola Smiljanic [Fri, 7 Feb 2014 22:57:20 +0000 (22:57 +0000)]
Improve existing visualizers by:
- Properly displaying non null terminated StringRef.
- Auto expanding pointer types.
- Displaying real type names for PointerUnions.
- Using "size" and "capacity" across all containers.
- Simplifying code where possible.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201004
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Fri, 7 Feb 2014 22:45:13 +0000 (22:45 +0000)]
Comment cleanup. Don't repeat the function name in the comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201001
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Fri, 7 Feb 2014 22:39:17 +0000 (22:39 +0000)]
Comment cleanup. Don't repeat the function name in the comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200999
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Fri, 7 Feb 2014 22:33:56 +0000 (22:33 +0000)]
Remove training whitespace.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200998
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Fri, 7 Feb 2014 21:23:18 +0000 (21:23 +0000)]
Always create a temporary symbol to use with the cfi frame.
This is a small simplification and a small step in fixing pr18743 since
private functions on MachO should be using a 'l' prefix.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200994
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Fri, 7 Feb 2014 21:11:33 +0000 (21:11 +0000)]
Use FileCheck variables to simplify this test.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200992
91177308-0d34-0410-b5e6-
96231b3b80d8
Renato Golin [Fri, 7 Feb 2014 20:32:32 +0000 (20:32 +0000)]
Fix Darwin bots from EHABI change
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200990
91177308-0d34-0410-b5e6-
96231b3b80d8
Matt Arsenault [Fri, 7 Feb 2014 20:29:40 +0000 (20:29 +0000)]
R600/SI: Add failing test for 3 x i64 vectors.
Stores of <4 x i64> do work (although they do expand to 4 stores
instead of 2), but 3 x i64 vectors fail to select.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200989
91177308-0d34-0410-b5e6-
96231b3b80d8
Renato Golin [Fri, 7 Feb 2014 20:12:49 +0000 (20:12 +0000)]
Remove -arm-disable-ehabi option
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200988
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Fri, 7 Feb 2014 19:04:43 +0000 (19:04 +0000)]
Don't internalize linkonce_odr non constant variables.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200983
91177308-0d34-0410-b5e6-
96231b3b80d8
Alexander Kornienko [Fri, 7 Feb 2014 17:42:30 +0000 (17:42 +0000)]
Fix an invalid check for duplicate option categories.
An intermediate solution until the problems with analyzer plugins linking with
llvm/Support and causing assertions due to duplicate GeneralCategory are solved.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200981
91177308-0d34-0410-b5e6-
96231b3b80d8
Sasa Stankovic [Fri, 7 Feb 2014 17:16:40 +0000 (17:16 +0000)]
[mips] Forbid the use of registers t6, t7 and t8 if the target is NaCl.
Differential Revision: http://llvm-reviews.chandlerc.com/D2694
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200978
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Fri, 7 Feb 2014 16:21:30 +0000 (16:21 +0000)]
Fix a bug with .weak_def_can_be_hidden: Mutable variables cannot use it.
Thanks to John McCall for noticing it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200977
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Fri, 7 Feb 2014 16:07:11 +0000 (16:07 +0000)]
Refactor logic into a function predicate.
No functionality change.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200976
91177308-0d34-0410-b5e6-
96231b3b80d8
Benjamin Kramer [Fri, 7 Feb 2014 12:05:36 +0000 (12:05 +0000)]
Try to unbreak the mingw32 build.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200973
91177308-0d34-0410-b5e6-
96231b3b80d8
Oliver Stannard [Fri, 7 Feb 2014 11:19:53 +0000 (11:19 +0000)]
LLVM-1163: AAPCS-VFP violation when CPRC allocated to stack
According to the AAPCS, when a CPRC is allocated to the stack, all other
VFP registers should be marked as unavailable.
I have also modified the rules for allocating non-CPRCs to the stack, to make
it more explicit that all GPRs must be made unavailable. I cannot think of a
case where the old version would produce incorrect answers, so there is no test
for this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200970
91177308-0d34-0410-b5e6-
96231b3b80d8
Sasa Stankovic [Fri, 7 Feb 2014 11:16:02 +0000 (11:16 +0000)]
Changed comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200969
91177308-0d34-0410-b5e6-
96231b3b80d8
Venkatraman Govindaraju [Fri, 7 Feb 2014 09:06:52 +0000 (09:06 +0000)]
[Sparc] Add support for parsing synthetic instruction 'mov'.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200965
91177308-0d34-0410-b5e6-
96231b3b80d8
Venkatraman Govindaraju [Fri, 7 Feb 2014 07:34:49 +0000 (07:34 +0000)]
[Sparc] Emit correct encoding for atomic instructions. Also, add support for parsing CAS instructions to test the CAS encoding.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200963
91177308-0d34-0410-b5e6-
96231b3b80d8
Venkatraman Govindaraju [Fri, 7 Feb 2014 05:54:20 +0000 (05:54 +0000)]
[Sparc] Emit relocations for Thread Local Storage (TLS) when integrated assembler is used.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200962
91177308-0d34-0410-b5e6-
96231b3b80d8
Venkatraman Govindaraju [Fri, 7 Feb 2014 04:24:35 +0000 (04:24 +0000)]
[Sparc] Emit correct relocations for PIC code when integrated assembler is used.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200961
91177308-0d34-0410-b5e6-
96231b3b80d8
Venkatraman Govindaraju [Fri, 7 Feb 2014 02:36:06 +0000 (02:36 +0000)]
[Sparc] Use SparcMCExpr::VariantKind itself as MachineOperand's target flags.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200960
91177308-0d34-0410-b5e6-
96231b3b80d8
Manman Ren [Fri, 7 Feb 2014 00:38:56 +0000 (00:38 +0000)]
PGO branch weight: fix PR18752.
Fix a bug triggered in IfConverterTriangle when CvtBB has multiple predecessors
by getting the weights before removing a successor.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200958
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Fri, 7 Feb 2014 00:16:33 +0000 (00:16 +0000)]
X86: Resolve a long standing FIXME and properly isel pextr[bw].
Generalize the AArch64 .td nodes for AssertZext and AssertSext. Use
them to match the relevant pextr store instructions.
The test widen_load-2.ll requires a slight change because with the
stores gone, the remaining instructions are scheduled in a different
order.
Add test cases for SSE4 and AVX variants.
Resolves rdar://
13414672.
Patch by Adam Nemet <anemet@apple.com>.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200957
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Thu, 6 Feb 2014 23:35:22 +0000 (23:35 +0000)]
Convert test to FileCheck.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200955
91177308-0d34-0410-b5e6-
96231b3b80d8
Quentin Colombet [Thu, 6 Feb 2014 21:44:56 +0000 (21:44 +0000)]
[CodeGenPrepare] Move away sign extensions that get in the way of addressing
mode.
Basically the idea is to transform code like this:
%idx = add nsw i32 %a, 1
%sextidx = sext i32 %idx to i64
%gep = gep i8* %myArray, i64 %sextidx
load i8* %gep
Into:
%sexta = sext i32 %a to i64
%idx = add nsw i64 %sexta, 1
%gep = gep i8* %myArray, i64 %idx
load i8* %gep
That way the computation can be folded into the addressing mode.
This transformation is done as part of the addressing mode matcher.
If the matching fails (not profitable, addressing mode not legal, etc.), the
matcher will revert the related promotions.
<rdar://problem/
15519855>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200947
91177308-0d34-0410-b5e6-
96231b3b80d8
Andrew Trick [Thu, 6 Feb 2014 19:20:41 +0000 (19:20 +0000)]
Track register pressure a bit more carefully (weird corner case).
This solves a problem where a def machine operand has no uses but has
not been marked dead. In this case, the initial RP analysis was being
extra precise and determining from LiveIntervals the the register was
actually dead. This caused us to omit the register from the RP
tracker's block live out. That's all good, but the per-instruction
summary still accounted for it as a valid def. This could cause an
assertion in the tracker later when we underflow pressure.
This is from a bug report on an out-of-tree target. It is not
reproducible on well-behaved targets. I'm just making an obvious fix
without unit test.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200941
91177308-0d34-0410-b5e6-
96231b3b80d8
Evan Cheng [Thu, 6 Feb 2014 18:51:34 +0000 (18:51 +0000)]
Revert r200095 and r200152. It turns out when compiling with -arch armv7 -mcpu=cortex-m3, the triple would still set iOS as the OS so the hack is still needed. rdar://
15984891
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200937
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Thu, 6 Feb 2014 18:36:41 +0000 (18:36 +0000)]
R600/SI: Add a MUBUF store pattern for Reg+Imm offsets
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200935
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Thu, 6 Feb 2014 18:36:39 +0000 (18:36 +0000)]
R600/SI: Add a MUBUF store pattern for Imm offsets
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200934
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Thu, 6 Feb 2014 18:36:38 +0000 (18:36 +0000)]
R600/SI: Add a MUBUF load pattern for Reg+Imm offsets
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200933
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Thu, 6 Feb 2014 18:36:34 +0000 (18:36 +0000)]
R600/SI: Use immediates offsets for SMRD instructions whenever possible
There was a problem with the old pattern, so we were copying some
larger immediates into registers when we could have been encoding
them in the instruction.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200932
91177308-0d34-0410-b5e6-
96231b3b80d8
David Peixotto [Thu, 6 Feb 2014 18:19:40 +0000 (18:19 +0000)]
Remove const_cast for STI when parsing inline asm
In a previous commit (r199818) we added a const_cast to an existing
subtarget info instead of creating a new one so that we could reuse
it when creating the TargetAsmParser for parsing inline assembly.
This cast was necessary because we needed to reuse the existing STI
to avoid generating incorrect code when the inline asm contained
mode-switching directives (e.g. .code 16).
The root cause of the failure was that there was an implicit sharing
of the STI between the parser and the MCCodeEmitter. To fix a
different but related issue, we now explicitly pass the STI to the
MCCodeEmitter (see commits r200345-r200351).
The const_cast is no longer necessary and we can now create a fresh
STI for the inline asm parser to use.
Differential Revision: http://llvm-reviews.chandlerc.com/D2709
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200929
91177308-0d34-0410-b5e6-
96231b3b80d8
Tim Northover [Thu, 6 Feb 2014 18:18:36 +0000 (18:18 +0000)]
X86: add costs for 64-bit vector ext/trunc & rebalance
The most important part of this is probably adding any cost at all for
operations like zext <8 x i8> to <8 x i32>. Before they were being
recorded as extremely costly (24, I believe) which made LLVM fall back
on a 4-wide vectorisation of a loop.
It also rebalances the values for sext, zext and trunc. Lacking any
other sane metric that might work across CPU microarchitectures I went
for instructions. This seems to be in reasonable accord with the rest
of the table (sitofp, ...) though no doubt at least one value is
sub-optimal for some bizarre reason.
Finally, separate AVX and AVX2 values are provided where appropriate.
The CodeGen is quite different in many cases.
rdar://problem/
15981990
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200928
91177308-0d34-0410-b5e6-
96231b3b80d8
Eli Bendersky [Thu, 6 Feb 2014 18:01:56 +0000 (18:01 +0000)]
Add a -suppress-warnings option to bitcode linking.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200927
91177308-0d34-0410-b5e6-
96231b3b80d8
Puyan Lotfi [Thu, 6 Feb 2014 09:57:39 +0000 (09:57 +0000)]
Yet another patch to reduce compile time for small programs:
The aim in this patch is to reduce work that VirtRegRewriter needs to do when
telling MachineRegisterInfo which physregs are in use. Up until now
VirtRegRewriter::rewrite has been doing rewriting and populating def info and
then proceeding to set whether a physreg is used based this info for every
physreg that the target provides. This can be expensive when a target has an
unusually high number of supported physregs, and is a noticeable chunk of
compile time for small programs on such targets.
So to reduce compile time, this patch simply adds the use of a SparseSet to the
rewrite function that is used to flag each physreg that is encountered in a
MachineFunction. Afterward, rather than iterating over the set of all physregs
for a given target to set the physregs used in MachineRegisterInfo, the new way
is to iterate over the set of physregs that were actually encountered and set
in the SparseSet. This improves compile time because the existing rewrite
function was iterating over all MachineOperands already, and because the
iterations afterward to setPhysRegUsed is reduced by use of the SparseSet data.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200919
91177308-0d34-0410-b5e6-
96231b3b80d8
Tim Northover [Thu, 6 Feb 2014 09:54:51 +0000 (09:54 +0000)]
X86: deduplicate V[SZ]EXT_MOVL and V[SZ]EXT nodes
I believe VZEXT_MOVL means "zero all vector elements except the first" (and
should have identical input & output types) whereas VZEXT means "zero extend
each element of a vector (discarding higher elements if necessary)".
For example:
(v4i32 (vzext (v16i8 ...)))
should zero extend the low 4 bytes of the incoming vector to 32-bits,
discarding higher bytes.
However, somewhere in the past, these two concepts had become confused, even
leading to a nonsensical VSEXT_MOVL.
This re-merges the nodes where appropriate (all VSEXT_MOVL -> VSEXT, VZEXT_MOVL
-> VZEXT when it's an actual extension).
rdar://problem/
15981990
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200918
91177308-0d34-0410-b5e6-
96231b3b80d8
Puyan Lotfi [Thu, 6 Feb 2014 09:23:24 +0000 (09:23 +0000)]
The following patch' purpose is to reduce compile time for compilation of small
programs on targets with large register files. The root of the compile time
overhead was in the use of llvm::SmallVector to hold PhysRegEntries, which
resulted in slow-down from calling llvm::SmallVector::assign(N, 0). In contrast
std::vector uses the faster __platform_bzero to zero out primitive buffers when
assign is called, while SmallVector uses an iterator.
The fix for this was simply to replace the SmallVector with a dynamically
allocated buffer and to initialize or reinitialize the buffer based on the
total registers that the target architecture requires. The changes support
cases where a pass manager may be reused for different targets, and note that
the PhysRegEntries is allocated using calloc mainly for good for, and also to
quite tools like Valgrind (see comments for more info on this).
There is an rdar to track the fact that SmallVector doesn't have platform
specific speedup optimizations inside of it for things like this, and I'll
create a bugzilla entry at some point soon as well.
TL;DR: This fix replaces the expensive llvm::SmallVector<unsigned
char>::assign(N, 0) with a call to calloc for N bytes which is much faster
because SmallVector's assign uses iterators.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200917
91177308-0d34-0410-b5e6-
96231b3b80d8
Puyan Lotfi [Thu, 6 Feb 2014 08:42:01 +0000 (08:42 +0000)]
This small change reduces compile time for small programs on targets that have
large register files. The omission of Queries.clear() is perfectly safe because
LiveIntervalUnion::Query doesn't contain any data that needs freeing and
because LiveRegMatrix::runOnFunction happens to reset the OwningArrayPtr
holding Queries every time it is run, so there's no need to zero out the
queries either. Not having to do this for very large numbers of physregs
is a noticeable constant cost reduction in compilation of small programs.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200913
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Thu, 6 Feb 2014 06:29:19 +0000 (06:29 +0000)]
A memcpy out of an fresh alloca is a no-op, delete it. Patch by Patrick Walton!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200907
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Thu, 6 Feb 2014 06:27:59 +0000 (06:27 +0000)]
Delete all of the CodeGenInstructions from CodeGenTarget destructor.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200906
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Thu, 6 Feb 2014 05:17:02 +0000 (05:17 +0000)]
[PM] Fix horrible typos that somehow didn't cause a failure in a C++11
build but spectacularly changed behavior of the C++98 build. =]
This shows my one problem with not having unittests -- basic API
expectations aren't well exercised by the integration tests because they
*happen* to not come up, even though they might later. I'll probably add
a basic unittest to complement the integration testing later, but
I wanted to revive the bots.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200905
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Thu, 6 Feb 2014 04:37:03 +0000 (04:37 +0000)]
[PM] Add a new "lazy" call graph analysis pass for the new pass manager.
The primary motivation for this pass is to separate the call graph
analysis used by the new pass manager's CGSCC pass management from the
existing call graph analysis pass. That analysis pass is (somewhat
unfortunately) over-constrained by the existing CallGraphSCCPassManager
requirements. Those requirements make it *really* hard to cleanly layer
the needed functionality for the new pass manager on top of the existing
analysis.
However, there are also a bunch of things that the pass manager would
specifically benefit from doing differently from the existing call graph
analysis, and this new implementation tries to address several of them:
- Be lazy about scanning function definitions. The existing pass eagerly
scans the entire module to build the initial graph. This new pass is
significantly more lazy, and I plan to push this even further to
maximize locality during CGSCC walks.
- Don't use a single synthetic node to partition functions with an
indirect call from functions whose address is taken. This node creates
a huge choke-point which would preclude good parallelization across
the fanout of the SCC graph when we got to the point of looking at
such changes to LLVM.
- Use a memory dense and lightweight representation of the call graph
rather than value handles and tracking call instructions. This will
require explicit update calls instead of some updates working
transparently, but should end up being significantly more efficient.
The explicit update calls ended up being needed in many cases for the
existing call graph so we don't really lose anything.
- Doesn't explicitly model SCCs and thus doesn't provide an "identity"
for an SCC which is stable across updates. This is essential for the
new pass manager to work correctly.
- Only form the graph necessary for traversing all of the functions in
an SCC friendly order. This is a much simpler graph structure and
should be more memory dense. It does limit the ways in which it is
appropriate to use this analysis. I wish I had a better name than
"call graph". I've commented extensively this aspect.
This is still very much a WIP, in fact it is really just the initial
bits. But it is about the fourth version of the initial bits that I've
implemented with each of the others running into really frustrating
problms. This looks like it will actually work and I'd like to split the
actual complexity across commits for the sake of my reviewers. =] The
rest of the implementation along with lots of wiring will follow
somewhat more rapidly now that there is a good path forward.
Naturally, this doesn't impact any of the existing optimizer. This code
is specific to the new pass manager.
A bunch of thanks are deserved for the various folks that have helped
with the design of this, especially Nick Lewycky who actually sat with
me to go through the fundamentals of the final version here.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200903
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Thu, 6 Feb 2014 04:32:33 +0000 (04:32 +0000)]
[PM] Back out one hunk of the patch in r200901 that was *supposed* to go
in my next patch. Sorry for the breakage.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200902
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Thu, 6 Feb 2014 04:25:13 +0000 (04:25 +0000)]
[PM] Wire up the analysis managers in the opt driver. This isn't really
necessary until we add analyses to the driver, but I have such an
analysis ready and wanted to split this out. This is actually exercised
by the existing tests of the new pass manager as the analysis managers
are cross-checked and validated by the function and module managers.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200901
91177308-0d34-0410-b5e6-
96231b3b80d8
Juergen Ributzka [Thu, 6 Feb 2014 04:09:06 +0000 (04:09 +0000)]
[DAG] Don't pull the binary operation though the shift if the operands have opaque constants.
During DAGCombine visitShiftByConstant assumes that certain binary operations
with only constant operands can always be folded successfully. This is no longer
true when the constant is opaque. This commit fixes visitShiftByConstant by not
performing the optimization for opaque constants. Otherwise we would end up in
an infinite DAGCombine loop.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200900
91177308-0d34-0410-b5e6-
96231b3b80d8
Manman Ren [Thu, 6 Feb 2014 01:59:22 +0000 (01:59 +0000)]
Set default of inlinecold-threshold to 225.
225 is the default value of inline-threshold. This change will make sure
we have the same inlining behavior as prior to r200886.
As Chandler points out, even though we don't have code in our testing
suite that uses cold attribute, there are larger applications that do
use cold attribute.
r200886 + this commit intend to keep the same behavior as prior to r200886.
We can later on tune the inlinecold-threshold.
The main purpose of r200886 is to help performance of instrumentation based
PGO before we actually hook up inliner with analysis passes such as BPI and BFI.
For instrumentation based PGO, we try to increase inlining of hot functions and
reduce inlining of cold functions by setting inlinecold-threshold.
Another option suggested by Chandler is to use a boolean flag that controls
if we should use OptSizeThreshold for cold functions. The default value
of the boolean flag should not change the current behavior. But it gives us
less freedom in controlling inlining of cold functions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200898
91177308-0d34-0410-b5e6-
96231b3b80d8
Kevin Enderby [Thu, 6 Feb 2014 01:21:15 +0000 (01:21 +0000)]
Update the X86 assembler for .intel_syntax to accept
the << and >> bitwise operators.
rdar://
15975725
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200896
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Thu, 6 Feb 2014 01:06:31 +0000 (01:06 +0000)]
don't set HasReliableSymbolDifference for ELF.
It is only used in MachObjectWriter.cpp. Another leftover from early days
of ELF in MC.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200895
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Thu, 6 Feb 2014 00:54:53 +0000 (00:54 +0000)]
doesSectionRequireSymbols is meaningless on ELF, remove.
This is a nop. doesSectionRequireSymbols is only used from
isSymbolLinkerVisible. isSymbolLinkerVisible only use from ELF was in
if (!Asm.isSymbolLinkerVisible(Symbol) && !Symbol.isUndefined())
return false;
if (Symbol.isTemporary())
return false;
If the symbol is a temporary this code returns false and it is irrelevant if
we take the first if or not. If the symbol is not a temporary,
Asm.isSymbolLinkerVisible returns true without ever calling
doesSectionRequireSymbols.
This was an horrible leftover from when support for ELF was first added.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200894
91177308-0d34-0410-b5e6-
96231b3b80d8
Paul Robinson [Thu, 6 Feb 2014 00:07:05 +0000 (00:07 +0000)]
Disable most IR-level transform passes on functions marked 'optnone'.
Ideally only those transform passes that run at -O0 remain enabled,
in reality we get as close as we reasonably can.
Passes are responsible for disabling themselves, it's not the job of
the pass manager to do it for them.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200892
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Thu, 6 Feb 2014 00:03:15 +0000 (00:03 +0000)]
Just returning false is the default.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200890
91177308-0d34-0410-b5e6-
96231b3b80d8
Matt Arsenault [Wed, 5 Feb 2014 23:16:05 +0000 (23:16 +0000)]
Pass address space to allowsUnalignedMemoryAccesses
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200888
91177308-0d34-0410-b5e6-
96231b3b80d8
Matt Arsenault [Wed, 5 Feb 2014 23:15:53 +0000 (23:15 +0000)]
Add address space argument to allowsUnalignedMemoryAccess.
On R600, some address spaces have more strict alignment
requirements than others.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200887
91177308-0d34-0410-b5e6-
96231b3b80d8
Manman Ren [Wed, 5 Feb 2014 22:53:44 +0000 (22:53 +0000)]
Inliner uses a smaller inline threshold for callees with cold attribute.
Added command line option inlinecold-threshold to set threshold for inlining
functions with cold attribute. Listen to the cold attribute when it would
decrease the inline threshold.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200886
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Kledzik [Wed, 5 Feb 2014 22:22:56 +0000 (22:22 +0000)]
Fix layering StringRef copy using BumpPtrAllocator.
Now to copy a string into a BumpPtrAllocator and get a StringRef to the copy:
StringRef myCopy = myStr.copy(myAllocator);
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200885
91177308-0d34-0410-b5e6-
96231b3b80d8
Quentin Colombet [Wed, 5 Feb 2014 22:13:59 +0000 (22:13 +0000)]
[RegAlloc] Add a last chance recoloring mechanism when everything else failed to
find a register.
The idea is to choose a color for the variable that cannot be allocated and
recolor its interferences around. Unlike the current register allocation scheme,
it is allowed to change the color of an already assigned (but maybe not
splittable or spillable) live interval while propagating this change to its
neighbors.
In other word, there are two things that may help finding an available color:
- Already assigned variables (RS_Done) can be recolored to different color.
- The recoloring allows to catch solutions that needs to touch more that just
the neighbors of the current allocated variable.
E.g.,
vA can use {R1, R2 }
vB can use { R2, R3}
vC can use {R1 }
Where vA, vB, and vC cannot be split anymore (they are reloads for instance) and
they all interfere.
vA is assigned R1
vB is assigned R2
vC tries to evict vA but vA is already done.
=> Regular register allocation heuristic fails.
Last chance recoloring kicks in:
vC does as if vA was evicted => vC uses R1.
vC is marked as fixed.
vA needs to find a color.
None are available.
vA cannot evict vC: vC is a fixed virtual register now.
vA does as if vB was evicted => vA uses R2.
vB needs to find a color.
R3 is available.
Recoloring => vC = R1, vA = R2, vB = R3.
<rdar://problem/
15947839>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200883
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Wed, 5 Feb 2014 21:41:42 +0000 (21:41 +0000)]
[PM] Don't require analysis results to be const in the new pass manager.
I think this was just over-eagerness on my part. The analysis results
need to often be non-const because they need to (in some cases at least)
be updated by the transformation pass in order to remain correct. It
also makes lazy analyses (a common case) needlessly annoying to write in
order to make their entire state mutable.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200881
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Wed, 5 Feb 2014 18:00:21 +0000 (18:00 +0000)]
Remove support for not using .loc directives.
Clang itself was not using this. The only way to access it was via llc.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200862
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Wed, 5 Feb 2014 17:49:31 +0000 (17:49 +0000)]
Revert "Fix an invalid check for duplicate option categories."
This reverts commit r200853.
It was causing clang/Analysis/checker-plugins.c to crash.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200858
91177308-0d34-0410-b5e6-
96231b3b80d8
Petar Jovanovic [Wed, 5 Feb 2014 17:19:30 +0000 (17:19 +0000)]
[mips] Add NaCl target and forbid indexed loads and stores for it
This patch adds NaCl target for Mips. It also forbids indexed loads and
stores if the target is NaCl.
Patch by Sasa Stankovic.
Differential Revision: http://llvm-reviews.chandlerc.com/D2690
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200855
91177308-0d34-0410-b5e6-
96231b3b80d8
Alexander Kornienko [Wed, 5 Feb 2014 16:56:37 +0000 (16:56 +0000)]
Fix an invalid check for duplicate option categories.
Summary:
The check performed in the comparator is invalid, as some STL
implementations enforce strict weak ordering by calling the comparator with the
same value. This check was also in a wrong place: the assertion would only fire
when -help was used. The new check is performed each time the category is
registered (we are not going to have thousands of them, so it's fine to do it in
O(N^2)).
Reviewers: jordan_rose
Reviewed By: jordan_rose
CC: cfe-commits, alexmc
Differential Revision: http://llvm-reviews.chandlerc.com/D2699
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200853
91177308-0d34-0410-b5e6-
96231b3b80d8
Petar Jovanovic [Wed, 5 Feb 2014 16:47:59 +0000 (16:47 +0000)]
mips: XFAIL non-extern-addend-smallcodemodel test
Small code model (and default reloc model) set Reloc::PIC_ in this test,
and PIC is not yet supported in MCJIT for MIPS.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200852
91177308-0d34-0410-b5e6-
96231b3b80d8
Elena Demikhovsky [Wed, 5 Feb 2014 16:17:36 +0000 (16:17 +0000)]
AVX-512: optimized icmp -> sext -> icmp pattern
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200849
91177308-0d34-0410-b5e6-
96231b3b80d8
Alon Mishne [Wed, 5 Feb 2014 14:23:18 +0000 (14:23 +0000)]
Test commit
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200843
91177308-0d34-0410-b5e6-
96231b3b80d8
Logan Chien [Wed, 5 Feb 2014 14:15:16 +0000 (14:15 +0000)]
ARM: Resolve thumb_bl fixup in same MCFragment.
In Thumb1 mode, bl instruction might be selected for branches between
basic blocks in the function if the offset is greater than 2KB.
However, this might cause SEGV because the destination symbol
is not marked as thumb function and the execution mode will be reset
to ARM mode.
Since we are sure that these symbols are in the same data fragment, we
can simply resolve these local symbols, and don't emit any relocation
information for this bl instruction.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200842
91177308-0d34-0410-b5e6-
96231b3b80d8
Elena Demikhovsky [Wed, 5 Feb 2014 13:03:01 +0000 (13:03 +0000)]
AVX-512: fixed a bug in EVEX encoding (the bug appeared after r200624)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200837
91177308-0d34-0410-b5e6-
96231b3b80d8
Michel Danzer [Wed, 5 Feb 2014 09:48:05 +0000 (09:48 +0000)]
R600/SI: Add pattern for zero-extending i1 to i32
Fixes opencl-example if_* tests with radeonsi.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=74469
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200830
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Wed, 5 Feb 2014 09:10:40 +0000 (09:10 +0000)]
Shrink the size of CodeGenInstruction a little bit by using bitfields. 32 bools seemed excessive.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200829
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Wed, 5 Feb 2014 07:56:49 +0000 (07:56 +0000)]
Get rid of a vector copy by just making a pointer out of the reference returned by getInstructionsByEnumValue instead of assigning it to a new vector.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200828
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Wed, 5 Feb 2014 07:27:49 +0000 (07:27 +0000)]
Fix a vector that was passed by value instead of reference.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200827
91177308-0d34-0410-b5e6-
96231b3b80d8
Kai Nacke [Wed, 5 Feb 2014 07:23:09 +0000 (07:23 +0000)]
ARM: Enable use of relocation type tlsldo in debug info for tls data.
This fixes PR18554.
Reviewers: Renato Golin, Keith Walker
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200826
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Wed, 5 Feb 2014 07:21:07 +0000 (07:21 +0000)]
Fix a doxygen comment referencing the wrong method name.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200825
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Wed, 5 Feb 2014 07:09:40 +0000 (07:09 +0000)]
Move matching for x86 BMI BLSI/BLSMSK/BLSR instructions to isel patterns instead of DAG combine. This weakens the ability to fold loads with them because we aren't able to match patterns that load the same thing twice. But maybe we should fix that if we care. The peephole optimizer will be able to fold some loads in its absense.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200824
91177308-0d34-0410-b5e6-
96231b3b80d8
Elena Demikhovsky [Wed, 5 Feb 2014 07:05:03 +0000 (07:05 +0000)]
AVX-512: Added intrinsic for cvtph2ps.
Added VPTESTNM instruction.
Added a pattern to vselect (lit tests will follow).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200823
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Wed, 5 Feb 2014 05:44:28 +0000 (05:44 +0000)]
Add CheckChildInteger to ISelMatcher operations. Removes nearly 2000 bytes from X86 matcher table.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200821
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Wed, 5 Feb 2014 05:19:19 +0000 (05:19 +0000)]
Use the information provided by getFlags to unify some code in llvm-nm.
It is not clear how much we should try to expose in getFlags. For example,
should there be a SF_Object and a SF_Text?
But for information that is already being exposed, we may as well use it in
llvm-nm.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200820
91177308-0d34-0410-b5e6-
96231b3b80d8
Todd Fiala [Wed, 5 Feb 2014 05:04:36 +0000 (05:04 +0000)]
Fix configure to find arc4random via header files.
ISSUE:
On Ubuntu 12.04 LTS, arc4random is provided by libbsd.so, which is a
transitive dependency of libedit. If a system had libedit on it that
was implemented in terms of libbsd.so, then the arc4random test,
previously implemented as a linker test, would succeed with -ledit.
However, on Ubuntu this would also require a #include <bsd/stdlib.h>.
This caused a build breakage on configure-based Ubuntu 12.04 with
libedit installed.
FIX:
This fix changes configure to test for arc4random by searching for it
in the standard header files. On Ubuntu 12.04, this test now properly
fails to find arc4random as it is not defined in the default header
locations. It also tweaks the #define names to match the output of the
header check command, which is slightly different than the linker
function check #defines.
I tested the following scenarios:
(1) Ubuntu 12.04 without the libedit package [did not find arc4random,
as expected]
(2) Ubuntu 12.04 with libedit package [properly did not find
arc4random, as expected]
(3) Ubuntu 12.04 with most recent libedit, custom built, and not
dependent on libbsd.so [properly did not find arc4random, as
expected].
(4) FreeBSD 10.0B1 [properly found arc4random, as expected]
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200819
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Wed, 5 Feb 2014 04:36:47 +0000 (04:36 +0000)]
Add a test for printing absolute symbols in ELF.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200818
91177308-0d34-0410-b5e6-
96231b3b80d8
Jordan Rose [Wed, 5 Feb 2014 00:02:42 +0000 (00:02 +0000)]
[CMake] Deliberately get all LLVM library dependencies for standalone builds.
CMake won't expand the dependency graph for us if the dependencies are in
another project, which leads to link errors in the standalone build.
This is a refinement of r200765.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200812
91177308-0d34-0410-b5e6-
96231b3b80d8
Jordan Rose [Wed, 5 Feb 2014 00:02:37 +0000 (00:02 +0000)]
[CMake] Move -stdlib=libc++ handling into its own file.
r200744 moved this into cmake/config-ix.cmake, so that it would happen very
early in the build process. However, standalone builds of Clang and other
external projects never include this file (which is correct).
Now, -stdlib=libc++ and the LLVM_COMPILER_IS_GCC_COMPATIBLE option are
both set in a new include file, HandleLLVMStdlib, which is included by
both config-ix.cmake and HandleLLVMOptions.cmake. This preserves existing
behavior for projects relying on HandleLLVMOptions and still does the
right thing for builds of LLVM itself.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200811
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Tue, 4 Feb 2014 23:53:15 +0000 (23:53 +0000)]
Small fix for llvm-nm handling of weak symbols on ELF (print 'v').
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200808
91177308-0d34-0410-b5e6-
96231b3b80d8
Manman Ren [Tue, 4 Feb 2014 23:53:12 +0000 (23:53 +0000)]
Update testing case for r200806.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200807
91177308-0d34-0410-b5e6-
96231b3b80d8
Manman Ren [Tue, 4 Feb 2014 23:49:02 +0000 (23:49 +0000)]
Fix wording of warning message about invalid debug info.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200806
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Tue, 4 Feb 2014 23:18:52 +0000 (23:18 +0000)]
Add a test for common symbols in coff.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200803
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Tue, 4 Feb 2014 22:53:45 +0000 (22:53 +0000)]
Silence a warning:
In file included from ../unittests/Support/ProcessTest.cpp:11:
../utils/unittest/googletest/include/gtest/gtest.h:1448:28: warning: comparison of integers of different signs: 'const unsigned int' and 'const int' [-Wsign-compare]
GTEST_IMPL_CMP_HELPER_(NE, !=);
~~~~~~~~~~~~~~~~~~~~~~~~~~^~~
../utils/unittest/googletest/include/gtest/gtest.h:1433:12: note: expanded from macro 'GTEST_IMPL_CMP_HELPER_'
if (val1 op val2) {\
^
../unittests/Support/ProcessTest.cpp:46:3: note: in instantiation of function template specialization 'testing::internal::CmpHelperNE<unsigned int, int>' requested here
EXPECT_NE((r1 | r2), 0);
^
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200801
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Tue, 4 Feb 2014 22:50:47 +0000 (22:50 +0000)]
Remove unused SF_ThreadLocal.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200800
91177308-0d34-0410-b5e6-
96231b3b80d8
Justin Bogner [Tue, 4 Feb 2014 21:03:17 +0000 (21:03 +0000)]
llvm-cov: Fix include order in GCOV.cpp
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200796
91177308-0d34-0410-b5e6-
96231b3b80d8
Benjamin Kramer [Tue, 4 Feb 2014 20:27:23 +0000 (20:27 +0000)]
SimplifyLibCalls: Push TLI through the exp2->ldexp transform.
For the odd case of platforms with exp2 available but not ldexp.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200795
91177308-0d34-0410-b5e6-
96231b3b80d8
Peter Collingbourne [Tue, 4 Feb 2014 20:04:46 +0000 (20:04 +0000)]
Avoid using EL_GETFP.
This should fix the build against old versions of libedit.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200794
91177308-0d34-0410-b5e6-
96231b3b80d8
Lang Hames [Tue, 4 Feb 2014 19:42:47 +0000 (19:42 +0000)]
[X86] Only 213 FMA3 variants should be marked commutable.
Commuting the 231 and 132 variants would swap addends and
multiplicands/multipliers, which isn't valid.
I'm still trying to reduce a decent test case for this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200792
91177308-0d34-0410-b5e6-
96231b3b80d8