Jeroen Ketema [Mon, 9 Oct 2017 19:43:04 +0000 (19:43 +0000)]
Implement mem_fence on ptx
PTX does not differentiate between read and write fences. Hence, these a
lowered to a mem_fence call. The mem_fence function compiles to the
“member.cta” instruction, which commits all outstanding reads and writes
of a thread such that these become visible to all other threads in the same
CTA (i.e., work-group). The instruction does not differentiate between
global and local memory. Hence, the flags parameter is ignored, except
for deciding whether a “member.cta” instruction should be issued at all.
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 315235
Shoaib Meenai [Mon, 9 Oct 2017 19:25:17 +0000 (19:25 +0000)]
[libc++] Support Microsoft ABI without vcruntime headers
The vcruntime headers are hairy and clash with both libc++ headers
themselves and other libraries. libc++ normally deals with the clashes
by deferring to the vcruntime headers and silencing its own definitions,
but for clients which don't want to depend on vcruntime headers, it's
desirable to support the opposite, i.e. have libc++ provide its own
definitions.
Certain operator new/delete replacement scenarios are not currently
supported in this mode, which requires some tests to be marked XFAIL.
The added documentation has more details.
Differential Revision: https://reviews.llvm.org/D38522
llvm-svn: 315234
Alexey Bataev [Mon, 9 Oct 2017 19:08:15 +0000 (19:08 +0000)]
[SLP] Add test for reversed load, NFC.
llvm-svn: 315232
Hamza Sood [Mon, 9 Oct 2017 19:07:09 +0000 (19:07 +0000)]
Testing commit access.
llvm-svn: 315231
Zachary Turner [Mon, 9 Oct 2017 18:50:29 +0000 (18:50 +0000)]
Fix some C++ value / reference semantics issues.
Some functions were taking Twine's not by const&, these are all
fixed to take by const&. We also had a case where some functions
were overloaded to accept by const& and &&. Now there is only
one version which accepts by value and move's the value.
llvm-svn: 315229
Jeroen Ketema [Mon, 9 Oct 2017 18:36:48 +0000 (18:36 +0000)]
Make ptx barrier work irrespective of the cl_mem_fence_flags
This generates a "bar.sync 0” instruction, which not only causes the
threads to wait, but does acts as a memory fence, as required by
OpenCL. The fence does not differentiate between local and global
memory. Unfortunately, there is no similar instruction which does
not include a memory fence. Hence, we cannot optimize the case
where neither CLK_LOCAL_MEM_FENCE nor CLK_GLOBAL_MEM_FENCE is
passed.
llvm-svn: 315228
Petr Hosek [Mon, 9 Oct 2017 18:29:52 +0000 (18:29 +0000)]
[sanitizer] Don't intercept signal and sigaction on Fuchsia
Fuchsia doesn't support signals, so don't use interceptors for signal or
sigaction.
Differential Revision: https://reviews.llvm.org/D38669
llvm-svn: 315227
Daniel Sanders [Mon, 9 Oct 2017 18:14:53 +0000 (18:14 +0000)]
[globalisel] Add support for ValueType operands in patterns.
It's rare but there are a small number of patterns like this:
(set i64:$dst, (add i64:$src1, i64:$src2))
These should be equivalent to register classes except they shouldn't check for
a specific register bank.
This doesn't occur in AArch64/ARM/X86 but does occasionally come up in other
in-tree targets such as BPF.
llvm-svn: 315226
Lang Hames [Mon, 9 Oct 2017 18:11:04 +0000 (18:11 +0000)]
[MC] Use a unique_ptr<MCAssembler> for MCObjectStreamer's Assembler member.
Removes manual new/delete.
llvm-svn: 315225
Sanjay Patel [Mon, 9 Oct 2017 17:54:46 +0000 (17:54 +0000)]
[InstCombine] fix formatting; NFC
llvm-svn: 315223
Adrian McCarthy [Mon, 9 Oct 2017 17:50:01 +0000 (17:50 +0000)]
Fix after r315079
Microsoft's debug implementation of std::copy checks if the destination is an
array and then does some bounds checking. This was causing an assertion
failure in fs::rename_internal which copies to a buffer of the appropriate
size but that's type-punned to an array of length 1 for API compatibility
reasons.
Fix is to make make the destination a pointer rather than an array.
llvm-svn: 315222
Stephane Sezer [Mon, 9 Oct 2017 17:49:32 +0000 (17:49 +0000)]
Update ABISysV_arm64::RegisterIsVolatile to accept registers prefixed with r
Summary:
While the specification says that the 64bit registers are prefixed with
`x`, it seems that many people still use `r`. Until recently, we had been using
the `r` prefix instead of the `x` prefix in ds2. This caused lldb to fail during
unwinding. I think it's reasonable to check for a register prefixed with `r`,
since some people still choose to use `r`.
Reviewers: sas, fjricci, clayborg
Reviewed By: sas, clayborg
Subscribers: aemerson, javed.absar, kristof.beyls
Differential Revision: https://reviews.llvm.org/D38376
Change by Alex Langford <apl@fb.com>
llvm-svn: 315221
Evgeniy Stepanov [Mon, 9 Oct 2017 17:45:03 +0000 (17:45 +0000)]
[ubsan] Disable one test on Android.
llvm-svn: 315220
Matt Arsenault [Mon, 9 Oct 2017 17:44:18 +0000 (17:44 +0000)]
AMDGPU: Fix missing declaration for __builtin_amdgcn_dispatch_ptr
llvm-svn: 315219
Francis Ricci [Mon, 9 Oct 2017 17:27:47 +0000 (17:27 +0000)]
[dsymutil] Emit valid debug locations when no symbol flags are set
Summary:
swiftc emits symbols without flags set, which led dsymutil to ignore
them when searching for global symbols, causing dwarf location data
to be omitted. Xcode's dsymutil handles this case correctly, and emits
valid location data. Add this functionality to llvm-dsymutil by
allowing parsing of symbols with no flags set.
Reviewers: aprantl, friss, JDevlieghere
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38587
llvm-svn: 315218
Alexey Bataev [Mon, 9 Oct 2017 17:14:03 +0000 (17:14 +0000)]
[SLP] Test for wrongly vectorized set of extractelements, NFC.
llvm-svn: 315217
Vlad Tsyrklevich [Mon, 9 Oct 2017 17:11:44 +0000 (17:11 +0000)]
Limit CFI blacklist entries to specific CFI mode
Summary:
Since D37924 and D37925 were merged, it's now possible to specify
individual sanitizers or CFI modes in sanitizer blacklists. Update the
CFI blacklist entries to only apply to cfi-unrelated-cast checks.
Reviewers: eugenis, pcc
Reviewed By: eugenis
Subscribers: kcc
Differential Revision: https://reviews.llvm.org/D38385
llvm-svn: 315216
Peter Collingbourne [Mon, 9 Oct 2017 17:07:47 +0000 (17:07 +0000)]
Make the cfi target available on more platforms.
On non-Linux targets it just installs the blacklist.
Differential Revision: https://reviews.llvm.org/D38661
llvm-svn: 315215
Ilya Biryukov [Mon, 9 Oct 2017 16:58:16 +0000 (16:58 +0000)]
[clangd] Added a command-line arg to mirror clangd input into a file.
Summary: The arg is useful for debugging and creating test cases.
Reviewers: bkramer, krasimir
Reviewed By: bkramer
Subscribers: klimek, cfe-commits
Differential Revision: https://reviews.llvm.org/D37970
llvm-svn: 315214
Ilya Biryukov [Mon, 9 Oct 2017 16:53:00 +0000 (16:53 +0000)]
[clangd] Added a test for r315212.
llvm-svn: 315213
Ilya Biryukov [Mon, 9 Oct 2017 16:52:12 +0000 (16:52 +0000)]
Set PreprocessorOpts.GeneratePreamble=true in PrecompiledPreamble.
Summary:
It was previsouly set only in ASTUnit, but it should be set for all client of
PrecompiledPreamble.
Reviewers: erikjv, bkramer, klimek
Reviewed By: bkramer
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D38617
llvm-svn: 315212
Ilya Biryukov [Mon, 9 Oct 2017 16:26:26 +0000 (16:26 +0000)]
[clangd] Added move-only function helpers.
Summary:
They are now used in ClangdScheduler instead of deferred std::async
computations.
The results of `std::async` are much less effective and do not provide
a good abstraction for similar purposes, i.e. for storing additional callbacks
to clangd async tasks. The actual callback API will follow a bit later.
Reviewers: klimek, bkramer, sammccall, krasimir
Reviewed By: sammccall
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D38627
llvm-svn: 315210
Javed Absar [Mon, 9 Oct 2017 16:21:25 +0000 (16:21 +0000)]
[TableGen] remove make_range where not necessary
llvm-svn: 315209
Zachary Turner [Mon, 9 Oct 2017 15:46:13 +0000 (15:46 +0000)]
[llvm-rc] Have the tokenizer discard single & block comments.
This allows rc files to have comments. Eventually we should
just use clang's c preprocessor, but that's a bit larger
effort for minimal gain, and this is straightforward.
Differential Revision: https://reviews.llvm.org/D38651
llvm-svn: 315207
Sanjay Patel [Mon, 9 Oct 2017 15:22:20 +0000 (15:22 +0000)]
[DAG] combine assertsexts around a trunc
This was a suggested follow-up to:
D37017 / https://reviews.llvm.org/rL313577
llvm-svn: 315206
Amara Emerson [Mon, 9 Oct 2017 15:15:09 +0000 (15:15 +0000)]
[AArch64] Improve codegen for inverted overflow checking intrinsics
E.g. if we have a (xor(overflow-bit), 1) where overflow-bit comes from an
intrinsic like llvm.sadd.with.overflow then we can kill the xor and use the
inverted condition code for the CSEL.
rdar://
28495949
Reviewed By: kristof.beyls
Differential Revision: https://reviews.llvm.org/D38160
llvm-svn: 315205
Sanjay Patel [Mon, 9 Oct 2017 15:01:58 +0000 (15:01 +0000)]
[x86] regenerate test checks; NFC
llvm-svn: 315204
Sanjay Patel [Mon, 9 Oct 2017 01:29:54 +0000 (01:29 +0000)]
[AArch64] fix typos in test assertions
llvm-svn: 315203
Craig Topper [Mon, 9 Oct 2017 01:05:16 +0000 (01:05 +0000)]
[X86] Remove a setLoadExtAction from the AVX512 section that uses an AVX512BW type and is alraedy present in the AVX512BW section.
llvm-svn: 315202
Craig Topper [Mon, 9 Oct 2017 01:05:15 +0000 (01:05 +0000)]
[X86] Enable extended comparison predicate support for SETUEQ/SETONE when targeting AVX instructions.
We believe that despite AMD's documentation, that they really do support all 32 comparision predicates under AVX.
Differential Revision: https://reviews.llvm.org/D38609
llvm-svn: 315201
Davide Italiano [Mon, 9 Oct 2017 00:18:45 +0000 (00:18 +0000)]
[DWARFDIE] Rewrite `operator !=` using `operator ==`. NFCI.
llvm-svn: 315200
Davide Italiano [Mon, 9 Oct 2017 00:11:49 +0000 (00:11 +0000)]
[SymbolFile/DWARF] Simplify two functions. NFCI.
llvm-svn: 315199
Benjamin Kramer [Sun, 8 Oct 2017 21:28:47 +0000 (21:28 +0000)]
Certain versions of clang require an explicit initialization for literal const members.
include/clang/Lex/PreprocessorLexer.h:79:3: error: constructor for
'clang::PreprocessorLexer' must explicitly initialize the const member
'FID'
llvm-svn: 315197
Benjamin Kramer [Sun, 8 Oct 2017 21:23:02 +0000 (21:23 +0000)]
Remove unused variables. No functionality change.
llvm-svn: 315196
Simon Pilgrim [Sun, 8 Oct 2017 20:58:14 +0000 (20:58 +0000)]
[X86][SSE] Don't call combineTo inside combineX86ShufflesRecursively. NFCI.
Return the combined shuffle from combineX86ShufflesRecursively and perform the combineTo in the caller.
Makes it easier for future patches to use this in functions that aren't actually shuffles themselves.
llvm-svn: 315195
Benjamin Kramer [Sun, 8 Oct 2017 20:53:36 +0000 (20:53 +0000)]
Make SourceLocation, QualType and friends have constexpr constructors.
No functionality change intended.
llvm-svn: 315194
Jan Vesely [Sun, 8 Oct 2017 20:07:58 +0000 (20:07 +0000)]
travis: Make sure we report failure even if only earlier checked files fail
for loop would only report status of the last command
v2: return '1'
call test instead of '['
Reviewer: Jeroen Ketema
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 315193
Jan Vesely [Sun, 8 Oct 2017 20:07:56 +0000 (20:07 +0000)]
check_external_calls.sh: Print number of calls in tested file.
Reviewer: Jeroen Ketema
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 315192
Jan Vesely [Sun, 8 Oct 2017 19:34:00 +0000 (19:34 +0000)]
ptx: Use __clc_nextafter to implement nextafter
using clang builtin results in external library call
Reviewer: Jeroen Ketema
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 315191
Jan Vesely [Sun, 8 Oct 2017 19:33:58 +0000 (19:33 +0000)]
Do not include clc_nextafter header globally
Drop unused clc/math/clc_nextafter.h header
Reviewer: Jeroen Ketema
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 315190
Jan Vesely [Sun, 8 Oct 2017 19:33:55 +0000 (19:33 +0000)]
math/nextafter: Use custom declaration inc file
Reviewer: Jeroen Ketema
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 315189
Jan Vesely [Sun, 8 Oct 2017 19:33:53 +0000 (19:33 +0000)]
math/binary_decl.inc: Do not declare mixed float/double functions
fmin/fmax only need vector/scalar mix
Reviewer: Jeroen Ketema
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 315188
Simon Pilgrim [Sun, 8 Oct 2017 19:24:30 +0000 (19:24 +0000)]
Tidyup with clang-format. NFCI.
llvm-svn: 315187
Simon Pilgrim [Sun, 8 Oct 2017 19:18:10 +0000 (19:18 +0000)]
[X86][SSE] Add test case for PR27708
llvm-svn: 315186
Benjamin Kramer [Sun, 8 Oct 2017 19:11:02 +0000 (19:11 +0000)]
Remove unused variables. No functionality change.
llvm-svn: 315185
Craig Topper [Sun, 8 Oct 2017 17:54:50 +0000 (17:54 +0000)]
[X86] Regenerate fast-isel-select-pseudo-cmov.ll to prepare for D38609.
llvm-svn: 315184
Javed Absar [Sun, 8 Oct 2017 17:23:30 +0000 (17:23 +0000)]
[TableGen] Simplify, add range_loop in CodeGenSchedule
llvm-svn: 315183
Simon Pilgrim [Sun, 8 Oct 2017 17:21:18 +0000 (17:21 +0000)]
[X86] getTargetConstantBitsFromNode - add support for decoding scalar constants
llvm-svn: 315182
Craig Topper [Sun, 8 Oct 2017 16:57:23 +0000 (16:57 +0000)]
[X86] Prefer MOVSS/SD over BLENDI during legalization. Remove BLENDI versions of scalar arithmetic patterns
Summary:
We currently disable some converting of shuffles to MOVSS/MOVSD during legalization if SSE41 is enabled. But later during shuffle combining we go back to prefering MOVSS/MOVSD.
Additionally we have patterns that look for BLENDIs to detect scalar arithmetic operations. I believe due to the combining using MOVSS/MOVSD these are unnecessary.
Interestingly, we still codegen blend instructions even though lowering/isel emit movss/movsd instructions. Turns out machine CSE commutes them to blend, and then commuting those blends back into blends that are equivalent to the original movss/movsd.
This patch fixes the inconsistency in legalization to prefer MOVSS/MOVSD. The one test change was caused by this change. The problem is that we have integer types and are mostly selecting integer instructions except for the shufps. This shufps forced the execution domain, but the vpblendw couldn't have its domain changed with a naive instruction swap. We could fix this by special casing VPBLENDW based on the immediate to widen the element type.
The rest of the patch is removing all the excess scalar patterns.
Long term we should probably add isel patterns to make MOVSS/MOVSD emit blends directly instead of relying on the double commute. We may also want to consider emitting movss/movsd for optsize. I also wonder if we should still use the VEX encoded blendi instructions even with AVX512. Blends have better throughput, and that may outweigh the register constraint.
Reviewers: RKSimon, zvi
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38023
llvm-svn: 315181
Benjamin Kramer [Sun, 8 Oct 2017 15:59:35 +0000 (15:59 +0000)]
Make more constructors constexpr or use =default.
This lets the compiler reason about the type more easily. No
functionality change intended.
llvm-svn: 315180
Amara Emerson [Sun, 8 Oct 2017 15:29:35 +0000 (15:29 +0000)]
[AArch64][GlobalISel] Add a test case for G_PHI of p0 instruction selection.
llvm-svn: 315179
Amara Emerson [Sun, 8 Oct 2017 15:29:31 +0000 (15:29 +0000)]
[AArch64][GlobalISel] Add a test case for G_PHI of p0 regbank selection.
llvm-svn: 315178
Amara Emerson [Sun, 8 Oct 2017 15:29:11 +0000 (15:29 +0000)]
[AArch64][GlobalISel] Make G_PHI of p0 types legal.
Differential Revision: https://reviews.llvm.org/D38621
llvm-svn: 315177
Simon Pilgrim [Sun, 8 Oct 2017 12:58:15 +0000 (12:58 +0000)]
[X86][XOP] Add XOP oddshuffles tests
XOP codegen is often different to generic AVX - thank you vpperm!
llvm-svn: 315176
Gadi Haber [Sun, 8 Oct 2017 12:52:54 +0000 (12:52 +0000)]
[X86][SKX] Adding the scheduling information for the SKX target.
Adding the scheduling information for the SkylakeServer (SKX) target.
This patch adds the instruction scheduling information for the SkylakeServer (SKX) architecture target by adding the file X86SchedSkylakeServer.td located under the X86 Target.
We used the scheduling information retrieved from the Skylake architects in order to create the file.
The scheduling information includes latency, number of micro-Ops and used ports by each SKL instruction.
The patch continues the scheduling replacement and insertion effort started with the SNB target in r310792, the HSW target in r311879 and the SkylakeClient (SKL) target in rL313613.
Please expect some performance fluctuations due to code alignment effects.
Reviewers: zvi, RKSimon, craig.topper, chandlerc, aymanmu
Differential Revision: https://reviews.llvm.org/D38443
Change-Id: I5c228fcc09e9e5a99b6116e62b356c4f9b971185
llvm-svn: 315175
Ayman Musa [Sun, 8 Oct 2017 09:46:50 +0000 (09:46 +0000)]
[X86] Add missing entries in 'MemoryFoldTable2Addr' to get complete form of the table.
Get the folding table 'MemoryFoldTable2Addr' to a complete state as part of the process explained in https://reviews.llvm.org/D38028
Differential Revision: https://reviews.llvm.org/D38500
llvm-svn: 315174
Ayman Musa [Sun, 8 Oct 2017 09:20:32 +0000 (09:20 +0000)]
[X86][TableGen] Recommitting the X86 memory folding tables TableGen backend while disabling it by default.
After the original commit ([[ https://reviews.llvm.org/rL304088 | rL304088 ]]) was reverted, a discussion in llvm-dev was opened on 'how to accomplish this task'.
In the discussion we concluded that the best way to achieve our goal (which is to automate the folding tables and remove the manually maintained tables) is:
# Commit the tablegen backend disabled by default.
# Proceed with an incremental updating of the manual tables - while checking the validity of each added entry.
# Repeat previous step until we reach a state where the generated and the manual tables are identical. Then we can safely remove the manual tables and include the generated tables instead.
# Schedule periodical (1 week/2 weeks/1 month) runs of the pass:
- if changes appear (new entries):
- make sure the entries are legal
- If they are not, mark them as illegal to folding
- Commit the changes (if there are any).
CMake flag added for this purpose is "X86_GEN_FOLD_TABLES". Building with this flags will run the pass and emit the X86GenFoldTables.inc file under build/lib/Target/X86/ directory which is a good reference for any developer who wants to take part in the effort of completing the current folding tables.
Differential Revision: https://reviews.llvm.org/D38028
llvm-svn: 315173
Craig Topper [Sun, 8 Oct 2017 08:50:59 +0000 (08:50 +0000)]
[X86] Stop LowerSIGN_EXTEND_AVX512 from creating v8i16/v16i16/v16i8 vselects with a v8i1/v16i1 condition when BWI is not available.
Some of the tests in vector-shuffle-v1.ll would get into an infinite loop without this.
llvm-svn: 315172
Ayman Musa [Sun, 8 Oct 2017 08:32:56 +0000 (08:32 +0000)]
[X86] Add new attribute to X86 instructions to enable marking them as "not memory foldable"
This attribute will be used in a tablegen backend that generated the X86 memory folding tables which will be added in a future pass.
Instructions with this attribute unset will be excluded from the full set of X86 instructions available for the pass.
Differential Revision: https://reviews.llvm.org/D38027
llvm-svn: 315171
Jan Vesely [Sun, 8 Oct 2017 06:56:14 +0000 (06:56 +0000)]
ldexp: Fix double precision function return type
Fixes ~1200 external calls from nvtpx library.
Reviewer: Jeroen Ketema
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 315170
Rui Ueyama [Sun, 8 Oct 2017 03:52:15 +0000 (03:52 +0000)]
Rename ignoreInterpSection -> needsInterpSection.
This should improve consistency. Also added comment.
llvm-svn: 315169
Rui Ueyama [Sun, 8 Oct 2017 03:45:49 +0000 (03:45 +0000)]
Use llvm::Optional instead of UINT_MAX to represent a null value.
llvm-svn: 315168
Rui Ueyama [Sun, 8 Oct 2017 02:54:42 +0000 (02:54 +0000)]
Make a helper function a non-member function. NFC.
llvm-svn: 315167
Rui Ueyama [Sun, 8 Oct 2017 02:44:08 +0000 (02:44 +0000)]
Use llvm::Optional instead of a magic number -1 to represent "no result".
llvm-svn: 315166
Rui Ueyama [Sun, 8 Oct 2017 02:27:02 +0000 (02:27 +0000)]
Make ScriptParser::checkSection a non-member function.
This patch also make its return type to `void` because
it always returns a given value as-is.
llvm-svn: 315165
Rui Ueyama [Sun, 8 Oct 2017 02:06:03 +0000 (02:06 +0000)]
Remove a trivial function.
llvm-svn: 315164
Rui Ueyama [Sun, 8 Oct 2017 01:55:29 +0000 (01:55 +0000)]
Add comment.
llvm-svn: 315163
Craig Topper [Sun, 8 Oct 2017 01:33:42 +0000 (01:33 +0000)]
[X86] Simplify some code in getInsertVINSERTImmediate and getExtractVEXTRACTImmediate. NFC
Replace one of the divides with a multiply.
llvm-svn: 315162
Craig Topper [Sun, 8 Oct 2017 01:33:41 +0000 (01:33 +0000)]
[X86] If we see an insert of a bitcast into zero vector, canonicalize it to move the bitcast to the other side of the insert.
This improves detection of zeroing of upper bits during isel.
llvm-svn: 315161
Craig Topper [Sun, 8 Oct 2017 01:33:40 +0000 (01:33 +0000)]
[X86] Remove ISD::INSERT_SUBVECTOR handling from combineBitcastForMaskedOp. Add isel patterns to make up for it.
This will allow for some flexibility in canonicalizing bitcasts around insert_subvector.
llvm-svn: 315160
Craig Topper [Sun, 8 Oct 2017 01:33:38 +0000 (01:33 +0000)]
[X86] Use getConstantOperandVal to simplify some code. NFC
llvm-svn: 315159
Rui Ueyama [Sun, 8 Oct 2017 00:45:34 +0000 (00:45 +0000)]
Remove redundant initialization code.
llvm-svn: 315158
Vedant Kumar [Sat, 7 Oct 2017 20:20:42 +0000 (20:20 +0000)]
cmake: Fix one more usage of append()
append() isn't available with some cmake versions, so I need to use a
different construct. I missed this case in r315144.
http://lab.llvm.org:8080/green/job/clang-stage1-configure-RA/39355
llvm-svn: 315157
Simon Pilgrim [Sat, 7 Oct 2017 17:57:22 +0000 (17:57 +0000)]
[X86][SSE] Match bitcasted BUILD_VECTOR of constants for v2i64 shifts on 64-bit targets (PR34855)
Extension to rL315155, generate constant shifts on 64-bits as well as 32-bits.
llvm-svn: 315156
Simon Pilgrim [Sat, 7 Oct 2017 17:42:17 +0000 (17:42 +0000)]
[X86][SSE] Match bitcasted v4i32 BUILD_VECTORS for v2i64 shifts on 64-bit targets (PR34855)
We were already doing this for 32-bit targets, but we can generate these on 64-bits as well.
llvm-svn: 315155
Craig Topper [Sat, 7 Oct 2017 17:07:48 +0000 (17:07 +0000)]
[SelectionDAG} Use KnownBits::isUnknown and hasConflict. NFC
llvm-svn: 315154
Craig Topper [Sat, 7 Oct 2017 16:51:19 +0000 (16:51 +0000)]
[X86] Add X86ISD::CMOV to computeKnownBitsForTargetNode and ComputeNumSignBitsForTargetNode.
Summary: Implementations based on ISD::SELECT.
Reviewers: RKSimon, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38663
llvm-svn: 315153
Sanjay Patel [Sat, 7 Oct 2017 15:39:06 +0000 (15:39 +0000)]
[InstSimplify] add tests to show we can do better at folding poison; NFC
llvm-svn: 315152
Simon Pilgrim [Sat, 7 Oct 2017 14:34:24 +0000 (14:34 +0000)]
[TableGen] Avoid repeated find calls in CodeGenDAGPatterns getters. NFCI.
The assertion tests were using count() instead of testing the find result, resulting in double the number of searches in debug/assert builds.
Instead, call find once (like the release builds do) and assert the result against end().
llvm-svn: 315151
Simon Pilgrim [Sat, 7 Oct 2017 12:42:23 +0000 (12:42 +0000)]
[X86][SSE] Improve shuffling combining with horizontal operations
Recognise cases when we can merge the shuffles with their horizontal (HADD/HSUB/PACK) instruction inputs.
Replaces an older implementation which performed some of this during lowering, expanding an existing target shuffle combine stage instead.
Differential Revision: https://reviews.llvm.org/D38506
llvm-svn: 315150
Simon Pilgrim [Sat, 7 Oct 2017 12:24:10 +0000 (12:24 +0000)]
Fix signed/unsigned warning
llvm-svn: 315149
Simon Pilgrim [Sat, 7 Oct 2017 12:08:43 +0000 (12:08 +0000)]
[TableGen] Avoid unnecessary std::string creations
Avoid unnecessary std::string creations in the TreePredicateFn getters and in CodeGenDAGPatterns::getSDNodeNamed
Differential Revision: https://reviews.llvm.org/D38624
llvm-svn: 315148
Jeroen Ketema [Sat, 7 Oct 2017 09:10:44 +0000 (09:10 +0000)]
Delete empy directory
llvm-svn: 315147
Martin Storsjo [Sat, 7 Oct 2017 06:00:32 +0000 (06:00 +0000)]
[X86] Update an outdated comment about SjLj
The SjLj intrinsics in the X86 backend are intended for use with
SjLj exception handling as well, since SVN r271244.
Differential Revision: https://reviews.llvm.org/D38532
llvm-svn: 315146
Craig Topper [Sat, 7 Oct 2017 05:11:59 +0000 (05:11 +0000)]
[X86] Correct result type for the flag result of RDSEED and RDRAND nodes. Correct the CC type for the CMOV used with RDSEED/RDRAND.
The flag result was MVT::Glue, but should be MVT::i32. The CC type was MVT::i8, but should be MVT::i32.
llvm-svn: 315145
Vedant Kumar [Sat, 7 Oct 2017 03:44:46 +0000 (03:44 +0000)]
Use list(APPEND) instead of append()
append() seems to be available with the version of cmake I'm using, but
not on the bots:
http://green.lab.llvm.org/green//job/clang-stage1-configure-RA/39354
llvm-svn: 315144
Vedant Kumar [Sat, 7 Oct 2017 01:46:36 +0000 (01:46 +0000)]
[ubsan] Add a static runtime on Darwin
As a follow-up to r315142, this makes it possible to use ubsan with a
static runtime on Darwin. I've also added a new StandaloneStatic testing
configuration so the new setup can be tested.
llvm-svn: 315143
Vedant Kumar [Sat, 7 Oct 2017 01:42:09 +0000 (01:42 +0000)]
[Driver] Fix -static-libsan / -shared-libsan on Darwin
Don't ignore these flags on Darwin. The old behavior of using the
dynamic runtime when neither flag is passed is preserved.
llvm-svn: 315142
Rafael Espindola [Sat, 7 Oct 2017 01:08:28 +0000 (01:08 +0000)]
Inline small function. NFC.
llvm-svn: 315141
Rui Ueyama [Sat, 7 Oct 2017 00:58:34 +0000 (00:58 +0000)]
Remove OutputSection::updateAlignment.
I feel it is easier to understand without this function.
llvm-svn: 315140
Rui Ueyama [Sat, 7 Oct 2017 00:58:19 +0000 (00:58 +0000)]
Clarify code by adding a comment.
llvm-svn: 315139
Rui Ueyama [Sat, 7 Oct 2017 00:43:31 +0000 (00:43 +0000)]
Merge addInputSec with OutputSection::addSection.
Previously, when we added an input section to an output section, we
called `OutputSectionFactory::addInputSec`. This isn't a good design
because, a factory class is intended to create a new object and
return it, but in this use case, it will never create a new object.
This patch fixes the design flaw.
llvm-svn: 315138
Rui Ueyama [Sat, 7 Oct 2017 00:22:59 +0000 (00:22 +0000)]
Return early. NFC.
llvm-svn: 315137
Jessica Paquette [Sat, 7 Oct 2017 00:16:34 +0000 (00:16 +0000)]
[MachineOutliner] Disable outlining from LinkOnceODRs by default
Say you have two identical linkonceodr functions, one in M1 and one in M2.
Say that the outliner outlines A,B,C from one function, and D,E,F from another
function (where letters are instructions). Now those functions are not
identical, and cannot be deduped. Locally to M1 and M2, these outlining
choices would be good-- to the whole program, however, this might not be true!
To mitigate this, this commit makes it so that the outliner sees linkonceodr
functions as unsafe to outline from. It also adds a flag,
-enable-linkonceodr-outlining, which allows the user to specify that they
want to outline from such functions when they know what they're doing.
Changing this handles most code size regressions in the test suite caused by
competing with linker dedupe. It also doesn't have a huge impact on the code
size improvements from the outliner. There are 6 tests that regress > 5% from
outlining WITH linkonceodrs to outlining WITHOUT linkonceodrs. Overall, most
tests either improve or are not impacted.
Not outlined vs outlined without linkonceodrs:
https://hastebin.com/raw/qeguxavuda
Not outlined vs outlined with linkonceodrs:
https://hastebin.com/raw/edepoqoqic
Outlined with linkonceodrs vs outlined without linkonceodrs:
https://hastebin.com/raw/awiqifiheb
Numbers generated using compare.py with -m size.__text. Tests run for AArch64
with -Oz -mllvm -enable-machine-outliner -mno-red-zone.
llvm-svn: 315136
Jason Molenda [Sat, 7 Oct 2017 00:13:10 +0000 (00:13 +0000)]
Remove the input file specification from the
"Prepare Swig Bindings" shell script phase; it
wasn't the actual input file and could lead
to incorrect dependency analysis by the build
system.
<rdar://problem/
34751196>
llvm-svn: 315135
Rui Ueyama [Sat, 7 Oct 2017 00:08:30 +0000 (00:08 +0000)]
Remove unused parameters.
llvm-svn: 315133
Evgeniy Stepanov [Sat, 7 Oct 2017 00:04:24 +0000 (00:04 +0000)]
[asan] Disable wcslen test on 32-bit Android.
llvm-svn: 315132
Rui Ueyama [Fri, 6 Oct 2017 23:43:54 +0000 (23:43 +0000)]
Use error() instead of warn() to report undefined symbols.
I believe the reason why we used warn() instead of error() to report
undefined symbols is because the older implementation of error() exitted
immediately. Here, we want to find as many undefined symbols as we can,
so I chose to use warn() instead of error().
Now error() does not exit immediately, so it doesn't make sense to keep
them as warnings.
Differential Revision: https://reviews.llvm.org/D38652
llvm-svn: 315131
Sanjay Patel [Fri, 6 Oct 2017 23:43:06 +0000 (23:43 +0000)]
[InstCombine] use correct type when propagating constant condition in simplifyDivRemOfSelectWithZeroOp (PR34856)
llvm-svn: 315130
Rui Ueyama [Fri, 6 Oct 2017 23:34:43 +0000 (23:34 +0000)]
Do not mutate Script->Opt.Commands from a leaf helper function.
Factory::addInputSec added an output section to Script->Opt.Commands,
but that is too subtle. This patch makes it explicit so that it is easy
to see when a new element is added to Script->Opt.Commands.
llvm-svn: 315129