oota-llvm.git
10 years agoRevert "Revert "YAMLIO: Encode ambiguous hex strings explicitly""
David Majnemer [Wed, 9 Apr 2014 17:04:27 +0000 (17:04 +0000)]
Revert "Revert "YAMLIO: Encode ambiguous hex strings explicitly""

Don't quote octal compatible strings if they are only two wide, they
aren't ambiguous.

This reverts commit r205857 which reverted r205857.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205914 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoobj2yaml: Don't crash if the characteristics field is zero
David Majnemer [Wed, 9 Apr 2014 16:38:15 +0000 (16:38 +0000)]
obj2yaml: Don't crash if the characteristics field is zero

obj2yaml would fail when seeing a Weak External auxiliary record with a
characteristics field holding zero instead of one of
IMAGE_WEAK_EXTERN_SEARCH_NOLIBRARY, IMAGE_WEAK_EXTERN_SEARCH_NOLIBRARY,
or IMAGE_WEAK_EXTERN_SEARCH_NOLIBRARY.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205911 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[NVPTX] Add preliminary intrinsics and codegen support for textures/surfaces
Justin Holewinski [Wed, 9 Apr 2014 15:39:15 +0000 (15:39 +0000)]
[NVPTX] Add preliminary intrinsics and codegen support for textures/surfaces

This commit adds intrinsics and codegen support for the surface read/write and texture read instructions that take an explicit sampler parameter. Codegen operates on image handles at the PTX level, but falls back to direct replacement of handles with kernel arguments if image handles are not enabled. Note that image handles are explicitly disabled for all target architectures in this change (to be enabled later).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205907 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[NVPTX] Add support for addrspacecast in global variable initializers, including...
Justin Holewinski [Wed, 9 Apr 2014 15:39:11 +0000 (15:39 +0000)]
[NVPTX] Add support for addrspacecast in global variable initializers, including emitting generic() when casting to address space 0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205906 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[NVPTX] Add query support for read-write images and managed variables
Justin Holewinski [Wed, 9 Apr 2014 15:38:52 +0000 (15:38 +0000)]
[NVPTX] Add query support for read-write images and managed variables

This also fixes a bug in the annotation cache where the cache will not be cleared between modules if multiple modules are compiled in the same process.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205905 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix some doc and comment typos
Alp Toker [Wed, 9 Apr 2014 14:47:27 +0000 (14:47 +0000)]
Fix some doc and comment typos

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205899 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Change SYS without a register to an alias to make disassembling more consistant.
Bradley Smith [Wed, 9 Apr 2014 14:44:58 +0000 (14:44 +0000)]
[ARM64] Change SYS without a register to an alias to make disassembling more consistant.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205898 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Correctly disassemble ISB operand as ISB not DBarrier.
Bradley Smith [Wed, 9 Apr 2014 14:44:54 +0000 (14:44 +0000)]
[ARM64] Correctly disassemble ISB operand as ISB not DBarrier.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205897 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Properly support both apple and standard syntax for FMOV
Bradley Smith [Wed, 9 Apr 2014 14:44:49 +0000 (14:44 +0000)]
[ARM64] Properly support both apple and standard syntax for FMOV

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205896 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Flag setting logical/add/sub immediate instructions don't use SP.
Bradley Smith [Wed, 9 Apr 2014 14:44:44 +0000 (14:44 +0000)]
[ARM64] Flag setting logical/add/sub immediate instructions don't use SP.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205895 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Conditional branches must always print their condition code, even AL.
Bradley Smith [Wed, 9 Apr 2014 14:44:39 +0000 (14:44 +0000)]
[ARM64] Conditional branches must always print their condition code, even AL.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205894 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Fix disassembly logic for extended loads/stores with 32-bit registers.
Bradley Smith [Wed, 9 Apr 2014 14:44:36 +0000 (14:44 +0000)]
[ARM64] Fix disassembly logic for extended loads/stores with 32-bit registers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205893 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] When printing a pre-indexed address with #0, the ', #0' is not optional.
Bradley Smith [Wed, 9 Apr 2014 14:44:31 +0000 (14:44 +0000)]
[ARM64] When printing a pre-indexed address with #0, the ', #0' is not optional.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205892 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Add missing shifted register MVN alias to ORN
Bradley Smith [Wed, 9 Apr 2014 14:44:26 +0000 (14:44 +0000)]
[ARM64] Add missing shifted register MVN alias to ORN

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205891 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] SXTW/UXTW are only valid aliases for 32-bit operations.
Bradley Smith [Wed, 9 Apr 2014 14:44:22 +0000 (14:44 +0000)]
[ARM64] SXTW/UXTW are only valid aliases for 32-bit operations.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205890 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Fix canonicalisation of MOVs. MOV is too complex to be modelled by a dumb...
Bradley Smith [Wed, 9 Apr 2014 14:44:18 +0000 (14:44 +0000)]
[ARM64] Fix canonicalisation of MOVs. MOV is too complex to be modelled by a dumb alias.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205889 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Fixup ADR/ADRP parsing such that they accept immediates and all labels types
Bradley Smith [Wed, 9 Apr 2014 14:44:12 +0000 (14:44 +0000)]
[ARM64] Fixup ADR/ADRP parsing such that they accept immediates and all labels types

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205888 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Ensure sp is decoded as SP, not XZR in LD1 instructions.
Bradley Smith [Wed, 9 Apr 2014 14:44:07 +0000 (14:44 +0000)]
[ARM64] Ensure sp is decoded as SP, not XZR in LD1 instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205887 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Tighten up the special casing in emitting arithmetic extends. UXTW should...
Bradley Smith [Wed, 9 Apr 2014 14:44:03 +0000 (14:44 +0000)]
[ARM64] Tighten up the special casing in emitting arithmetic extends. UXTW should only be translated when the instruction uses WSP, not SP. Vice versa for UXTX and 64-bit instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205886 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Rename LR to the UAL-compliant 'X30'.
Bradley Smith [Wed, 9 Apr 2014 14:43:59 +0000 (14:43 +0000)]
[ARM64] Rename LR to the UAL-compliant 'X30'.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205885 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Rename FP to the UAL-compliant 'X29'.
Bradley Smith [Wed, 9 Apr 2014 14:43:50 +0000 (14:43 +0000)]
[ARM64] Rename FP to the UAL-compliant 'X29'.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205884 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Add a PostEncoderMethod to FCMP - the Rm field should canonically be zero...
Bradley Smith [Wed, 9 Apr 2014 14:43:40 +0000 (14:43 +0000)]
[ARM64] Add a PostEncoderMethod to FCMP - the Rm field should canonically be zero but should be decoded/disassembled with any value.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205883 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] SCVTF and FCVTZS/U are undefined if scale<5> == 0.
Bradley Smith [Wed, 9 Apr 2014 14:43:35 +0000 (14:43 +0000)]
[ARM64] SCVTF and FCVTZS/U are undefined if scale<5> == 0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205882 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] EXT and EXTR instructions on v8i8 and W regs respectively must have the top...
Bradley Smith [Wed, 9 Apr 2014 14:43:31 +0000 (14:43 +0000)]
[ARM64] EXT and EXTR instructions on v8i8 and W regs respectively must have the top bit of their immediate clear.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205881 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Scaled fixed-point FCVTZSs should also have bit 29 set to zero.
Bradley Smith [Wed, 9 Apr 2014 14:43:27 +0000 (14:43 +0000)]
[ARM64] Scaled fixed-point FCVTZSs should also have bit 29 set to zero.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205880 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] UBFM/BFM is undefined on w registers when imms<5> or immr<5> is 1.
Bradley Smith [Wed, 9 Apr 2014 14:43:24 +0000 (14:43 +0000)]
[ARM64] UBFM/BFM is undefined on w registers when imms<5> or immr<5> is 1.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205879 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Floating point to fixed point scaled conversions are only available on fcvtzs...
Bradley Smith [Wed, 9 Apr 2014 14:43:20 +0000 (14:43 +0000)]
[ARM64] Floating point to fixed point scaled conversions are only available on fcvtzs and fcvtzu.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205878 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Port over the PostEncoderMethod fix for SMULH/UMULH from AArch64.
Bradley Smith [Wed, 9 Apr 2014 14:43:15 +0000 (14:43 +0000)]
[ARM64] Port over the PostEncoderMethod fix for SMULH/UMULH from AArch64.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205877 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Add missing tlbi operands and error for extra/missing register on tlbi aliases.
Bradley Smith [Wed, 9 Apr 2014 14:43:11 +0000 (14:43 +0000)]
[ARM64] Add missing tlbi operands and error for extra/missing register on tlbi aliases.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205876 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Rework system register parsing to overcome SPSel clash in MSR variants.
Bradley Smith [Wed, 9 Apr 2014 14:43:06 +0000 (14:43 +0000)]
[ARM64] Rework system register parsing to overcome SPSel clash in MSR variants.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205875 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Port over the PostEncoderMethod from AArch64 for exclusive loads and stores...
Bradley Smith [Wed, 9 Apr 2014 14:43:01 +0000 (14:43 +0000)]
[ARM64] Port over the PostEncoderMethod from AArch64 for exclusive loads and stores, so the unused register fields are set to all-ones canonically but are recognised with any value.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205874 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Use PStateMapper to ensure that MSRcpsr operands are validated during disasse...
Bradley Smith [Wed, 9 Apr 2014 14:42:56 +0000 (14:42 +0000)]
[ARM64] Use PStateMapper to ensure that MSRcpsr operands are validated during disassembly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205873 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Remove PrefetchOp and use ARM64PRFM instead.
Bradley Smith [Wed, 9 Apr 2014 14:42:53 +0000 (14:42 +0000)]
[ARM64] Remove PrefetchOp and use ARM64PRFM instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205872 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Add WZR to isGPR32Register, since every use needs to check for this anyway.
Bradley Smith [Wed, 9 Apr 2014 14:42:49 +0000 (14:42 +0000)]
[ARM64] Add WZR to isGPR32Register, since every use needs to check for this anyway.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205871 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Remove ARM64SYS.
Bradley Smith [Wed, 9 Apr 2014 14:42:45 +0000 (14:42 +0000)]
[ARM64] Remove ARM64SYS.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205870 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Move CPSRField and DBarrier operands over to AArch64-style disassembly and...
Bradley Smith [Wed, 9 Apr 2014 14:42:42 +0000 (14:42 +0000)]
[ARM64] Move CPSRField and DBarrier operands over to AArch64-style disassembly and assembly. This removes the last users of namespace ARM64SYS.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205869 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Switch the decoder, disassembler, instprinter and asmparser over to using...
Bradley Smith [Wed, 9 Apr 2014 14:42:36 +0000 (14:42 +0000)]
[ARM64] Switch the decoder, disassembler, instprinter and asmparser over to using AArch64-style system registers, and fix up test failures discovered in the process.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205868 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Move ARM64BaseInfo.{cpp,h} into a Utils/ subdirectory, a la AArch64. These...
Bradley Smith [Wed, 9 Apr 2014 14:42:27 +0000 (14:42 +0000)]
[ARM64] Move ARM64BaseInfo.{cpp,h} into a Utils/ subdirectory, a la AArch64. These files are required in the decoder, disassembler and parser, and a layering violation was imminent.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205867 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Copy the named immediate operand mapping logic and enums from AArch64. AArch6...
Bradley Smith [Wed, 9 Apr 2014 14:42:16 +0000 (14:42 +0000)]
[ARM64] Copy the named immediate operand mapping logic and enums from AArch64. AArch64's named immediate mapping and parsing is much more advanced than ARM64's. No functionality change - they're currently living side by side while I switch uses over.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205866 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Shifted register ALU ops are reserved if sf=0 and imm6<5>=1, and also (for...
Bradley Smith [Wed, 9 Apr 2014 14:42:11 +0000 (14:42 +0000)]
[ARM64] Shifted register ALU ops are reserved if sf=0 and imm6<5>=1, and also (for add/sub only) if shift=11.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205865 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Add support for NV condition code (exists only for valid assembly/disassembly...
Bradley Smith [Wed, 9 Apr 2014 14:42:07 +0000 (14:42 +0000)]
[ARM64] Add support for NV condition code (exists only for valid assembly/disassembly, equivilant to AL)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205864 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Add missing 1Q -> 1q vector kind alias
Bradley Smith [Wed, 9 Apr 2014 14:42:01 +0000 (14:42 +0000)]
[ARM64] Add missing 1Q -> 1q vector kind alias

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205863 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Add parsing for vector lists such as {v0.8b-v3.8b}
Bradley Smith [Wed, 9 Apr 2014 14:41:58 +0000 (14:41 +0000)]
[ARM64] Add parsing for vector lists such as {v0.8b-v3.8b}

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205862 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Correctly alias LSL to UXTW for 32bit instruction variants, rather than UXTX
Bradley Smith [Wed, 9 Apr 2014 14:41:53 +0000 (14:41 +0000)]
[ARM64] Correctly alias LSL to UXTW for 32bit instruction variants, rather than UXTX

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205861 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] STRHro and STRBro were not being decoded at all.
Bradley Smith [Wed, 9 Apr 2014 14:41:49 +0000 (14:41 +0000)]
[ARM64] STRHro and STRBro were not being decoded at all.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205860 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] MOVK with sf=0 and hw<1>=1 is unallocated. Shift amount for ADD/SUB instructi...
Bradley Smith [Wed, 9 Apr 2014 14:41:45 +0000 (14:41 +0000)]
[ARM64] MOVK with sf=0 and hw<1>=1 is unallocated. Shift amount for ADD/SUB instructions is unallocated if shift > 4.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205859 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Register-offset loads and stores with the 'option' field equal to 00x or...
Bradley Smith [Wed, 9 Apr 2014 14:41:38 +0000 (14:41 +0000)]
[ARM64] Register-offset loads and stores with the 'option' field equal to 00x or 10x are undefined.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205858 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRevert "YAMLIO: Encode ambiguous hex strings explicitly"
Filipe Cabecinhas [Wed, 9 Apr 2014 14:35:17 +0000 (14:35 +0000)]
Revert "YAMLIO: Encode ambiguous hex strings explicitly"

This reverts commit r205839.

It broke several tests in lld.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205857 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSLPVectorizer: Only vectorize intrinsics whose operands are widened equally
Arnold Schwaighofer [Wed, 9 Apr 2014 14:20:47 +0000 (14:20 +0000)]
SLPVectorizer: Only vectorize intrinsics whose operands are widened equally

The vectorizer only knows how to vectorize intrinics by widening all operands by
the same factor.

Patch by Tyler Nowicki!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205855 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAVX-512: insert element to mask vector; store i1 data
Elena Demikhovsky [Wed, 9 Apr 2014 12:37:50 +0000 (12:37 +0000)]
AVX-512: insert element to mask vector; store i1 data

Implemented INSERT_VECTOR_ELT operation for v16i1 and v8i1 vectors;
Implemented "store" for i1 type

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205850 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd support for building LLVM on FreeBSD 9.2
Viktor Kutuzov [Wed, 9 Apr 2014 11:43:34 +0000 (11:43 +0000)]
Add support for building LLVM on FreeBSD 9.2

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205847 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRe-commit: [mips] abs.[ds], and neg.[ds] should be allowed regardless of -enable...
Daniel Sanders [Wed, 9 Apr 2014 09:56:43 +0000 (09:56 +0000)]
Re-commit: [mips] abs.[ds], and neg.[ds] should be allowed regardless of -enable-no-nans-fp-math

Summary:
They behave in accordance with the Has2008 and ABS2008 configuration bits of the processor which are used to select between the 1985 and 2008 versions of IEEE 754. In 1985 mode, these instructions are arithmetic (i.e. they raise invalid operation exceptions when given NaN), in 2008 mode they are non-arithmetic (i.e. they are copies).

nmadd.[ds], and nmsub.[ds] are still subject to -enable-no-nans-fp-math because the ISA spec does not explicitly state that they obey Has2008 and ABS2008.

Fixed the issue with the previous version of this patch (r205628). A pre-existing 'let Predicate =' statement was removing some predicates that were necessary for FP64 to behave correctly.

Reviewers: matheusalmeida

Reviewed By: matheusalmeida

Differential Revision: http://llvm-reviews.chandlerc.com/D3274

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205844 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoYAMLIO: Encode ambiguous hex strings explicitly
David Majnemer [Wed, 9 Apr 2014 07:56:27 +0000 (07:56 +0000)]
YAMLIO: Encode ambiguous hex strings explicitly

YAMLIO would turn a BinaryRef into the string 0000000004000000.
However, the leading zero causes parsers to interpret it as being an
octal number instead of a hexadecimal one.

Instead, escape such strings as needed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205839 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDelinearize: Extend informationin -analyze output
Tobias Grosser [Wed, 9 Apr 2014 07:53:49 +0000 (07:53 +0000)]
Delinearize: Extend informationin -analyze output

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205838 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Match not instruction.
Matt Arsenault [Wed, 9 Apr 2014 07:16:16 +0000 (07:16 +0000)]
R600/SI: Match not instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205837 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM64: scalarize v1i64 mul operation
Tim Northover [Wed, 9 Apr 2014 07:07:02 +0000 (07:07 +0000)]
ARM64: scalarize v1i64 mul operation

This is the second part of fixing PR19367.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205836 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM64: add pattern for <1 x i64> custom not node.
Tim Northover [Wed, 9 Apr 2014 06:55:39 +0000 (06:55 +0000)]
ARM64: add pattern for <1 x i64> custom not node.

This should fix PR19367.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205835 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoObject: add type names for ARM/COFF relocations
Saleem Abdulrasool [Wed, 9 Apr 2014 06:18:28 +0000 (06:18 +0000)]
Object: add type names for ARM/COFF relocations

Add type name mappings for the ARM COFF relocations.  This allows for objdump to
provide a more useful description of relocations in disassembly inline form.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205834 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM MC: 80 column
Saleem Abdulrasool [Wed, 9 Apr 2014 06:18:26 +0000 (06:18 +0000)]
ARM MC: 80 column

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205833 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM MC: sort source files in CMakeLists
Saleem Abdulrasool [Wed, 9 Apr 2014 06:18:23 +0000 (06:18 +0000)]
ARM MC: sort source files in CMakeLists

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205832 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[C++11] More 'nullptr' conversion or in some cases just using a boolean check instead...
Craig Topper [Wed, 9 Apr 2014 06:08:46 +0000 (06:08 +0000)]
[C++11] More 'nullptr' conversion or in some cases just using a boolean check instead of comparing to nullptr.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205831 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[C++11] Make use of 'nullptr' in TableGen library.
Craig Topper [Wed, 9 Apr 2014 04:50:04 +0000 (04:50 +0000)]
[C++11] Make use of 'nullptr' in TableGen library.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205830 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[C++11] Replace some comparisons with 'nullptr' with simple boolean checks to reduce...
Craig Topper [Wed, 9 Apr 2014 04:20:00 +0000 (04:20 +0000)]
[C++11] Replace some comparisons with 'nullptr' with simple boolean checks to reduce verbosity.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205829 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoWinCOFF: Emit common symbols as specified in the COFF spec
David Majnemer [Tue, 8 Apr 2014 22:33:40 +0000 (22:33 +0000)]
WinCOFF: Emit common symbols as specified in the COFF spec

Summary:
Local common symbols were properly inserted into the .bss section.
However, putting external common symbols in the .bss section would give
them a strong definition.

Instead, encode them as undefined, external symbols who's symbol value
is equivalent to their size.

Reviewers: Bigcheese, rafael, rnk

CC: llvm-commits
Differential Revision: http://reviews.llvm.org/D3324

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205811 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoBug 19348: Check for legal ExtLoad operation before folding
Matt Arsenault [Tue, 8 Apr 2014 21:40:37 +0000 (21:40 +0000)]
Bug 19348: Check for legal ExtLoad operation before folding

(aext (zextload x)) -> (aext (truncate (*extload x)))

Patch by Stanislav Mekhanoshin!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205805 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agodivide by the result of the gcd
Sebastian Pop [Tue, 8 Apr 2014 21:21:13 +0000 (21:21 +0000)]
divide by the result of the gcd

used to fail with 'Step should divide Start with no remainder.'

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205802 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agohandle special cases when findGCD returns 1
Sebastian Pop [Tue, 8 Apr 2014 21:21:10 +0000 (21:21 +0000)]
handle special cases when findGCD returns 1

used to fail with 'Step should divide Start with no remainder.'

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205801 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoin findGCD of multiply expr return the gcd
Sebastian Pop [Tue, 8 Apr 2014 21:21:05 +0000 (21:21 +0000)]
in findGCD of multiply expr return the gcd

we used to return 1 instead of the gcd

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205800 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[docs] VCS contains a record of authorship
Sean Silva [Tue, 8 Apr 2014 21:12:56 +0000 (21:12 +0000)]
[docs] VCS contains a record of authorship

No need to explicitly mention the author in the document.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205793 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[docs] Fix up some links to the preferred style.
Sean Silva [Tue, 8 Apr 2014 21:06:22 +0000 (21:06 +0000)]
[docs] Fix up some links to the preferred style.

:doc:`...` and :ref:`...` links help Sphinx keep track the dependencies
between documents and ensure that they are not pointing to nowhere.

Raw HTML links work just fine and are easier for people less familiar
with reST/Sphinx. They are easy to change over to the :doc:/:ref: style
after the fact so this is not a problem.

This commit doesn't fix all of them.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205792 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Constant Hoisting][ARM64] Enable constant hoisting for ARM64.
Juergen Ributzka [Tue, 8 Apr 2014 20:39:59 +0000 (20:39 +0000)]
[Constant Hoisting][ARM64] Enable constant hoisting for ARM64.

This implements the target-hooks for ARM64 to enable constant hoisting.

This fixes <rdar://problem/14774662> and <rdar://problem/16381500>.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205791 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRegAlloc: Account for a variable entry block frequency
Duncan P. N. Exon Smith [Tue, 8 Apr 2014 19:18:56 +0000 (19:18 +0000)]
RegAlloc: Account for a variable entry block frequency

Until r197284, the entry frequency was constant -- i.e., set to 2^14.
Although current ToT still has a constant entry frequency, since r197284
that has been an implementation detail (which is soon going to change).

  - r204690 made the wrong assumption for the CSRCost metric.  Adjust
    callee-saved register cost based on entry frequency.

  - r185393 made the wrong assumption (although it was valid at the
    time).  Update SpillPlacement.cpp::Threshold to be relative to the
    entry frequency.

Since ToT still has 2^14 entry frequency, this should have no observable
functionality change.

<rdar://problem/14292693>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205789 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PowerPC] Don't return false from PPC::isVSLDOIShuffleMask
Hal Finkel [Tue, 8 Apr 2014 19:00:27 +0000 (19:00 +0000)]
[PowerPC] Don't return false from PPC::isVSLDOIShuffleMask

PPC::isVSLDOIShuffleMask should return -1, not false, when the shuffle
predicate should be false.

Noticed by inspection; no test case (yet).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205787 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix the ARM VLD3 (single 3-element structure to all lanes)
Kevin Enderby [Tue, 8 Apr 2014 18:00:52 +0000 (18:00 +0000)]
Fix the ARM VLD3 (single 3-element structure to all lanes)
size 16 double-spaced registers instruction printing.

This:
vld3.16 {d0[], d2[], d4[]}, [r4]!

was being printed as:

vld3.16 {d0[], d1[], d2[]}, [r4]!

rdar://16531387

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205779 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoVerifier: Give the right message for bad atomic loads
Duncan P. N. Exon Smith [Tue, 8 Apr 2014 17:07:44 +0000 (17:07 +0000)]
Verifier: Give the right message for bad atomic loads

Talk about load (not store) on an invalid atomic load.

<rdar://problem/16287567>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205777 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd -pass-remarks flag to 'opt'.
Diego Novillo [Tue, 8 Apr 2014 16:42:38 +0000 (16:42 +0000)]
Add -pass-remarks flag to 'opt'.

Summary:
This adds support in 'opt' to filter pass remarks emitted by
optimization passes. A new flag -pass-remarks specifies which
passes should emit a diagnostic when LLVMContext::emitOptimizationRemark
is invoked.

This will allow the front end to simply pass along the regular
expression from its own -Rpass flag when launching the backend.

Depends on D3227.

Reviewers: qcolombet

CC: llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D3291

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205775 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd support for optimization reports.
Diego Novillo [Tue, 8 Apr 2014 16:42:34 +0000 (16:42 +0000)]
Add support for optimization reports.

Summary:
This patch adds backend support for -Rpass=, which indicates the name
of the optimization pass that should emit remarks stating when it
made a transformation to the code.

Pass names are taken from their DEBUG_NAME definitions.

When emitting an optimization report diagnostic, the lack of debug
information causes the diagnostic to use "<unknown>:0:0" as the
location string.

This is the back end counterpart for

http://llvm-reviews.chandlerc.com/D3226

Reviewers: qcolombet

CC: llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D3227

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205774 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoX86MCAsmInfoGNUCOFF: Set PointerSize as 8 for targeting x64. It caused DW_LNE_set_add...
NAKAMURA Takumi [Tue, 8 Apr 2014 15:28:50 +0000 (15:28 +0000)]
X86MCAsmInfoGNUCOFF: Set PointerSize as 8 for targeting x64. It caused DW_LNE_set_address was misemitted on x64.

FIXME: I haven't investigate whether CalleeSaveStackSlotSize should be 8.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205772 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM64: fix fmsub patterns which assumed accum operand was first
Tim Northover [Tue, 8 Apr 2014 12:23:51 +0000 (12:23 +0000)]
ARM64: fix fmsub patterns which assumed accum operand was first

Confusingly, the NEON fmla instructions put the accumulator first but the
scalar versions put it at the end (like the fma lib function & LLVM's
intrinsic).

This should fix PR19345, assuming there's only one issue.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205758 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoThe LLVM C API shouldn't be including a file from the C++ API. Especially not a
Richard Smith [Tue, 8 Apr 2014 10:47:04 +0000 (10:47 +0000)]
The LLVM C API shouldn't be including a file from the C++ API. Especially not a
file that it doesn't use.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205755 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAVX-512: Added fp_to_uint and uint_to_fp patterns.
Elena Demikhovsky [Tue, 8 Apr 2014 07:24:02 +0000 (07:24 +0000)]
AVX-512: Added fp_to_uint and uint_to_fp patterns.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205754 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix a (legacy) PassManager crash that occurs when a ModulePass
Andrew Trick [Tue, 8 Apr 2014 03:40:34 +0000 (03:40 +0000)]
Fix a (legacy) PassManager crash that occurs when a ModulePass
indirectly requires a function analysis.

This bug was reported by Jason Kim. He included a test case here:
http://reviews.llvm.org/D3312

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205753 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoX86: Split the relocation selection up
David Majnemer [Tue, 8 Apr 2014 02:15:13 +0000 (02:15 +0000)]
X86: Split the relocation selection up

Before, we would have conditional operators where one side of the
operator would be of type RelocationTypeAMD64 and the other is of type
RelocationTypeI386.  GCC would noisly warn with -Wenum-compare
diagnostic.

Instead, refactor the code so it is more like the X86 ELF object writer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205752 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTidy up comments a bit.
Jim Grosbach [Mon, 7 Apr 2014 23:47:23 +0000 (23:47 +0000)]
Tidy up comments a bit.

Punctuation, grammar, formatting, etc..

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205749 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM64: Range based for loop in ARM64PromoteConstant pass
Jim Grosbach [Mon, 7 Apr 2014 23:47:21 +0000 (23:47 +0000)]
ARM64: Range based for loop in ARM64PromoteConstant pass

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205748 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM64: Clean up file header comment a bit.
Jim Grosbach [Mon, 7 Apr 2014 23:14:38 +0000 (23:14 +0000)]
ARM64: Clean up file header comment a bit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205747 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoobj2yaml: Use the correct relocation type for different machine types
David Majnemer [Mon, 7 Apr 2014 23:12:20 +0000 (23:12 +0000)]
obj2yaml: Use the correct relocation type for different machine types

The IO normalizer would essentially lump I386 and AMD64 relocations
together.  Relocation types with the same numeric value would then get
mapped in appropriately.

For example:
IMAGE_REL_AMD64_ADDR64 and IMAGE_REL_I386_DIR16 both have a numeric
value of one.  We would see IMAGE_REL_I386_DIR16 in obj2yaml conversions
of object files with a machine type of IMAGE_FILE_MACHINE_AMD64.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205746 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[docs] Fix some links
Sean Silva [Mon, 7 Apr 2014 22:46:40 +0000 (22:46 +0000)]
[docs] Fix some links

The TableGen docs have changed structure

Patch by Tay Ray Chuan!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205744 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[docs] Update link title
Sean Silva [Mon, 7 Apr 2014 22:42:53 +0000 (22:42 +0000)]
[docs] Update link title

docs/TableGen/ is not really just "fundamentals" anymore, but rather
more of a portal for all things TableGen.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205743 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[docs] Fix some Sphinx warnings that have crept in.
Sean Silva [Mon, 7 Apr 2014 22:29:53 +0000 (22:29 +0000)]
[docs] Fix some Sphinx warnings that have crept in.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205742 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoReverting commit r205628 due to mips64 issues.
Reed Kotler [Mon, 7 Apr 2014 22:11:40 +0000 (22:11 +0000)]
Reverting commit r205628 due to mips64 issues.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205741 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoPut a limit on ScheduleDAGSDNodes::ClusterNeighboringLoads to avoid blowing up compil...
Andrew Trick [Mon, 7 Apr 2014 21:29:22 +0000 (21:29 +0000)]
Put a limit on ScheduleDAGSDNodes::ClusterNeighboringLoads to avoid blowing up compile time.

Fixes PR16365 - Extremely slow compilation in -O1 and -O2.

The SD scheduler has a quadratic implementation of load clustering
which absolutely blows up compile time for large blocks with constant
pool loads. The MI scheduler has a better implementation of load
clustering. However, we have not done the work yet to completely
eliminate the SD scheduler. Some benchmarks still seem to benefit from
early load clustering, although maybe by chance.

As an intermediate term fix, I just put a nice limit on the number of
DAG users to search before finding a match. With this limit there are no
binary differences in the LLVM test suite, and the PR16365 test case
does not suffer any compile time impact from this routine.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205738 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Handle INSERT_SUBREG in SIFixSGPRCopies
Tom Stellard [Mon, 7 Apr 2014 19:45:45 +0000 (19:45 +0000)]
R600/SI: Handle INSERT_SUBREG in SIFixSGPRCopies

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205732 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Match 24-bit arithmetic patterns in a Target DAGCombine
Tom Stellard [Mon, 7 Apr 2014 19:45:41 +0000 (19:45 +0000)]
R600: Match 24-bit arithmetic patterns in a Target DAGCombine

Moving these patterns from TableGen files to PerformDAGCombine()
should allow us to generate better code by eliminating unnecessary
shifts and extensions earlier.

This also fixes a bug where the MAD pattern was calling
SimplifyDemandedBits with a 24-bit mask on the first operand
even when the full pattern wasn't being matched.  This occasionally
resulted in some instructions being incorrectly deleted from the
program.

v2:
  - Fix bug with 64-bit mul

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205731 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Replace dyn_cast + assert with cast
Tom Stellard [Mon, 7 Apr 2014 19:31:13 +0000 (19:31 +0000)]
R600: Replace dyn_cast + assert with cast

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205730 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove an unused file.
Richard Smith [Mon, 7 Apr 2014 17:17:00 +0000 (17:17 +0000)]
Remove an unused file.

Using this file would result in an odr violation: it defines an llvm::Interval
class that conflicts with the one in Analysis/Interval.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205726 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoWhen a CHECK-NEXT fails because there was no match on the next line, include
Richard Smith [Mon, 7 Apr 2014 17:09:53 +0000 (17:09 +0000)]
When a CHECK-NEXT fails because there was no match on the next line, include
the non-matching next line in the diagnostic to make the problem more obvious.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205725 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse std::swap
Matt Arsenault [Mon, 7 Apr 2014 16:44:26 +0000 (16:44 +0000)]
Use std::swap

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205723 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse .data() instead of &x[0]
Matt Arsenault [Mon, 7 Apr 2014 16:44:24 +0000 (16:44 +0000)]
Use .data() instead of &x[0]

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205722 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoInvert the option to enable debug info verification. No functional
Eric Christopher [Mon, 7 Apr 2014 13:55:21 +0000 (13:55 +0000)]
Invert the option to enable debug info verification. No functional
change outside of the command line to enable it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205713 91177308-0d34-0410-b5e6-96231b3b80d8