Bob Haarman [Wed, 5 Apr 2017 00:43:54 +0000 (00:43 +0000)]
[COFF] support /ERRORLIMIT option
Summary:
This adds support for reporting multiple errors in a single invocation of lld-link. The limit defaults to 20 and can be changed with the /ERRORLIMIT command line parameter, or set to unlimited by passing a value of 0.
This is a new attempt after r295507, which was reverted because opening files raced with exiting early, causing the test to be flaky. This version avoids the race by exiting before calling enqueuePath.
Reviewers: pcc, ruiu
Reviewed By: ruiu
Subscribers: llvm-commits, dblaikie
Differential Revision: https://reviews.llvm.org/D31688
llvm-svn: 299496
Rui Ueyama [Wed, 5 Apr 2017 00:43:45 +0000 (00:43 +0000)]
Remove `=` from a lambda that doesn't capture any variable.
llvm-svn: 299495
Rui Ueyama [Wed, 5 Apr 2017 00:43:25 +0000 (00:43 +0000)]
Inline small functions that are used only once as lambdas.
llvm-svn: 299494
Rui Ueyama [Wed, 5 Apr 2017 00:43:05 +0000 (00:43 +0000)]
Add newlines.
llvm-svn: 299493
Rui Ueyama [Wed, 5 Apr 2017 00:42:45 +0000 (00:42 +0000)]
Make dummy variable's scope smaller.
llvm-svn: 299492
Bob Haarman [Wed, 5 Apr 2017 00:42:07 +0000 (00:42 +0000)]
ThinLTOBitcodeWriter: handle aliases first in filterModule
Summary: This change fixes a "local linkage requires default visibility" assert when attempting to build LLVM with ThinLTO on Windows.
Reviewers: pcc, tejohnson, mehdi_amini
Reviewed By: pcc
Subscribers: llvm-commits, Prazek
Differential Revision: https://reviews.llvm.org/D31632
llvm-svn: 299491
Ahmed Bougacha [Wed, 5 Apr 2017 00:14:39 +0000 (00:14 +0000)]
[X86] Relax assert in broadcast-of-subvector lowering.
Before r294774, there was a problem when lowering broadcasts to use
128-bit subvectors.
When we looked through a bitcast to find the broadcast input, we'd keep
using the original type, so you'd end up with things like:
(v8f32 (broadcast
(v4f32 (extract_subvector
(v8i32 V),
...))
))
r294774 fixed it to always emit subvectors with the scalar type of the
original source.
It also introduced some asserts, to check that we use scalars with
the same size, and vectors with the same number of elements.
The scalar size equality is checked earlier when looking through bitcasts,
and is a useful assert.
However, the number of elements don't have to be identical: we're always
going to extract a 128-bit subvector, and we can have different size
inputs if we looked through a concat_vector to find a 256-bit source.
Relax the overzealous assert.
Replace it with a check of the original source vector being 256 or 512
bits. If it's 128 bits, we can't extract_subvector from it.
Fixes PR32371.
llvm-svn: 299490
Jim Ingham [Wed, 5 Apr 2017 00:08:21 +0000 (00:08 +0000)]
Reverting r299374 & r299402 due to testsuite failure.
This caused a failure in the test case:
functionalities/breakpoint/objc/TestObjCBreakpoints.py
When we are parsing up names we stick interesting parts of the names
in various buckets, one of which is the ObjC selector bucket. The new
C++ name parser must be interfering with this process somehow.
<rdar://problem/
31439305>
llvm-svn: 299489
Adam Nemet [Tue, 4 Apr 2017 23:46:34 +0000 (23:46 +0000)]
Another attempt to fix the sphinx warning from r299470
llvm-svn: 299488
Matt Arsenault [Tue, 4 Apr 2017 23:44:46 +0000 (23:44 +0000)]
Allow targets to opt-in to codegen in SCC order
Decouple this setting from EnableIRPA.
To support function calls on AMDGPU, it is necessary to
report the global register usage throughout the kernel's
call graph, so callees need to be handled first.
llvm-svn: 299487
Daniel Berlin [Tue, 4 Apr 2017 23:43:10 +0000 (23:43 +0000)]
Re-apply MemorySSA: Add support for caching clobbering access in
stores with some fixes.
Summary:
This enables us to cache the clobbering access for stores, despite the
fact that we can't rewrite the use-def chains themselves.
Early testing shows that, after this change, for larger testcases, it
will be a significant net positive (memory and time) to remove the
walker caching.
Reviewers: george.burgess.iv, davide
Subscribers: Prazek, llvm-commits
Differential Revision: https://reviews.llvm.org/D31567
llvm-svn: 299486
Daniel Berlin [Tue, 4 Apr 2017 23:43:04 +0000 (23:43 +0000)]
Revert "MemorySSA: Add support for caching clobbering access in stores"
This reverts revision r299322.
llvm-svn: 299485
Petr Hosek [Tue, 4 Apr 2017 23:32:45 +0000 (23:32 +0000)]
[MC] Set defaults based on section names and support name suffixes
Set correct default flags and section type based on its name for .text,
.data, .bss, .init_array, .fini_array, .preinit_array, .tdata, and .tbss
and support section name suffixes for .data.*, .rodata.*, .text.*,
.bss.*, .tdata.* and .tbss.* which matches the behavior of GAS.
Fixes PR31888.
Differential Revision: https://reviews.llvm.org/D30229
llvm-svn: 299484
Ahmed Bougacha [Tue, 4 Apr 2017 22:55:53 +0000 (22:55 +0000)]
[AArch64] Avoid partial register deps on insertelt of load into lane 0.
This improves upon r246462: that prevented FMOVs from being emitted
for the cross-class INSERT_SUBREGs by disabling the formation of
INSERT_SUBREGs of LOAD. But the ld1.s that we started selecting
caused us to introduce partial dependencies on the vector register.
Avoid that by using SCALAR_TO_VECTOR: it's a first-class citizen that
is folded away by many patterns, including the scalar LDRS that we
want in this case.
Credit goes to Adam for finding the issue!
llvm-svn: 299482
Adam Nemet [Tue, 4 Apr 2017 22:45:20 +0000 (22:45 +0000)]
Fix sphinx warning from r299470
llvm-svn: 299481
Evgeniy Stepanov [Tue, 4 Apr 2017 22:35:16 +0000 (22:35 +0000)]
Change section flag character for SHF_LINK_ORDER to "o".
See matching MC change in https://reviews.llvm.org/D31554.
llvm-svn: 299480
Evgeniy Stepanov [Tue, 4 Apr 2017 22:35:08 +0000 (22:35 +0000)]
Change section flag character for SHF_LINK_ORDER to "o".
GAS uses "m" as a compatibility alias for "M" (SHF_MERGE).
"o" is free, except on ia64, where it already means SHF_LINK_ORDER.
llvm-svn: 299479
Rafael Espindola [Tue, 4 Apr 2017 22:33:02 +0000 (22:33 +0000)]
Avoid calling basename to compute xdynamiclib_namespec.
This also exposes a xdynamiclib_filename that can be used to simplify
a few tests.
llvm-svn: 299478
Craig Topper [Tue, 4 Apr 2017 22:22:30 +0000 (22:22 +0000)]
[InstCombine] Add test cases for various add/subtracts of constants(scalar, splat, and vector) with phis and selects. Improvements coming in a future commit.
llvm-svn: 299476
Rafael Espindola [Tue, 4 Apr 2017 22:20:18 +0000 (22:20 +0000)]
[lit] Add a minimum export implementation.
llvm-svn: 299475
Sanjay Patel [Tue, 4 Apr 2017 22:06:03 +0000 (22:06 +0000)]
[InstCombine] rename variable for easier reading; NFC
We usually give constants a 'C' somewhere in the name...
llvm-svn: 299474
Meador Inge [Tue, 4 Apr 2017 21:46:50 +0000 (21:46 +0000)]
[Driver] Add option to print the resource directory
This patch adds the option -print-resource-dir. It simply
prints the resource directory. This information will eventually
be used in compiler-rt to setup COMPILER_RT_LIBRARY_INSTALL_DIR.
Patch by Catherine Moore!
Differential Revision: https://reviews.llvm.org/D31447
llvm-svn: 299473
Craig Topper [Tue, 4 Apr 2017 21:44:56 +0000 (21:44 +0000)]
[InstCombine] Turn subtract of vectors of i1 into xor like we do for scalar i1. Matches what we already do for add.
llvm-svn: 299472
Rafael Espindola [Tue, 4 Apr 2017 21:42:59 +0000 (21:42 +0000)]
Don't remove the cwd.
This works with a regular shell since the kernel can keep track of a
deleted cwd. Since we just keep a path string, the following
subprocess invocations fail.
I think this would also fail on windows.
llvm-svn: 299471
Adam Nemet [Tue, 4 Apr 2017 21:18:36 +0000 (21:18 +0000)]
Add #pragma clang fp
This adds the new pragma and the first variant, contract(on/off/fast).
The pragma has the same block scope rules as STDC FP_CONTRACT, i.e. it can be
placed at the beginning of a compound statement or at file scope.
Similarly to STDC FP_CONTRACT there is no need to use attributes. First an
annotate token is inserted with the parsed details of the pragma. Then the
annotate token is parsed in the proper contexts and the Sema is updated with
the corresponding FPOptions using the shared ActOn function with STDC
FP_CONTRACT.
After this the FPOptions from the Sema is propagated into the AST expression
nodes. There is no change here.
I was going to add a 'default' option besides 'on/off/fast' similar to STDC
FP_CONTRACT but then decided against it. I think that we'd have to make option
uppercase then to avoid using 'default' the keyword. Also because of the
scoped activation of pragma I am not sure there is really a need a for this.
Differential Revision: https://reviews.llvm.org/D31276
llvm-svn: 299470
Adam Nemet [Tue, 4 Apr 2017 21:18:30 +0000 (21:18 +0000)]
Set FMF for -ffp-contract=fast
With this, FMF(contract) becomes an alternative way to express the request to
contract.
These are currently only propagated for FMul, FAdd and FSub. The rest will be
added as more FMFs are hooked up for this.
This is toward fixing PR25721.
Differential Revision: https://reviews.llvm.org/D31168
llvm-svn: 299469
Balaram Makam [Tue, 4 Apr 2017 21:15:53 +0000 (21:15 +0000)]
[AArch64] Add missing schedinfo, check completeness for Falkor.
llvm-svn: 299468
Keno Fischer [Tue, 4 Apr 2017 20:30:47 +0000 (20:30 +0000)]
[ExecutionDepsFix] Don't revisit true dependencies
If an instruction has a true dependency, it makes sense for to use that
register for any undef read operands in the same instruction (we'll have
to wait for that register to become available anyway). This logic
was already implemented. However, the code would then still try to
revisit that instruction and break the dependency (and always fail,
since by definition a true dependency has to be live before the
instruction). Avoid revisiting such instructions as a performance
optimization. No functional change.
Differential Revision: https://reviews.llvm.org/D30173
llvm-svn: 299467
Craig Topper [Tue, 4 Apr 2017 20:26:25 +0000 (20:26 +0000)]
[InstCombine] Support folding and/or/xor with a constant vector RHS into selects and phis
Currently we only fold with ConstantInt RHS. This generalizes to any Constant RHS.
Differential Revision: https://reviews.llvm.org/D31610
llvm-svn: 299466
Matthias Gehre [Tue, 4 Apr 2017 20:11:13 +0000 (20:11 +0000)]
[clang-format] fix crash in NamespaceEndCommentsFixer (PR32438)
Summary:
The new test case was crashing before. Now it passes
as expected.
Reviewers: djasper
Subscribers: klimek, cfe-commits
Differential Revision: https://reviews.llvm.org/D31441
llvm-svn: 299465
Rafael Espindola [Tue, 4 Apr 2017 20:03:34 +0000 (20:03 +0000)]
Don't resolve hidden undef to a DSO.
The ELF spec says:
all of the non-default visibility attributes, when applied to a symbol
reference, imply that a definition to satisfy that reference must be
provided within the current executable or shared object.
But we were trying to resolve those undef references to shared
symbols. That causes odd results like creating a got entry with
a relocation pointing to 0.
llvm-svn: 299464
Dominic Chen [Tue, 4 Apr 2017 19:52:25 +0000 (19:52 +0000)]
[analyzer] Add new Z3 constraint manager backend
Summary: Implement new Z3 constraint manager backend.
Reviewers: zaks.anna, dcoughlin, NoQ, xazax.hun
Subscribers: mgorny, cfe-commits
Differential Revision: https://reviews.llvm.org/D28952
llvm-svn: 299463
Petr Hosek [Tue, 4 Apr 2017 19:51:53 +0000 (19:51 +0000)]
[AArch64][Fuchsia] Allow -mcmodel=kernel for --target=aarch64-fuchsia
This mode is just like -mcmodel=small except that it moves the
thread pointer from TPIDR_EL0 to TPIDR_EL1.
Patch by Roland McGrath.
Differential Revision: https://reviews.llvm.org/D31624
llvm-svn: 299462
Jonas Devlieghere [Tue, 4 Apr 2017 19:42:29 +0000 (19:42 +0000)]
[clangd] Link against clangSema
Fixes linking issue introduced by rL299421 when building LLVM with
shared libraries.
llvm-svn: 299461
Craig Topper [Tue, 4 Apr 2017 19:31:21 +0000 (19:31 +0000)]
[InstCombine] Add test cases for missing combines of phis with and/or/xor with constant argument. NFC
llvm-svn: 299460
Coby Tayree [Tue, 4 Apr 2017 19:20:21 +0000 (19:20 +0000)]
[fixup][X86][inline-asm] Add support for MS 'EVEN' directive
refining tested targets resolution, to amend failures caused by rL299454
llvm-svn: 299459
Yi Kong [Tue, 4 Apr 2017 19:06:04 +0000 (19:06 +0000)]
Implement host CPU detection for AArch64
This shares detection logic with ARM(32), since AArch64 capable CPUs may
also run in 32-bit system mode.
We observe weird /proc/cpuinfo output for MSM8992 and MSM8994, where
they report all CPU cores as one single model, depending on which CPU
core the kernel is running on. As a workaround, we hardcode the known
CPU part name for these SoCs.
For big.LITTLE systems, this patch would only return the part name of
the first core (usually the little core). Proper support will be added
in a follow-up change.
Differential Revision: D31675
llvm-svn: 299458
Matt Arsenault [Tue, 4 Apr 2017 18:43:11 +0000 (18:43 +0000)]
Verifier: Check some amdgpu calling convention restrictions
llvm-svn: 299457
Balaram Makam [Tue, 4 Apr 2017 18:42:14 +0000 (18:42 +0000)]
[AArch64] Refine Falkor Machine Model - Part 2
llvm-svn: 299456
Coby Tayree [Tue, 4 Apr 2017 17:58:28 +0000 (17:58 +0000)]
[X86][inline-asm] Add support for MS 'EVEN' directive
MS assembly syntax provide us with the 'EVEN' directive as a synonymous to at&t '.even'.
This patch include the (small, simple) changes need to allow it.
llvm-side:
https://reviews.llvm.org/D27417
Differential Revision: https://reviews.llvm.org/D27418
llvm-svn: 299454
Coby Tayree [Tue, 4 Apr 2017 17:57:23 +0000 (17:57 +0000)]
[X86][inline-asm] Add support for MS 'EVEN' directive
MS assembly syntax provide us with the 'EVEN' directive as a synonymous to at&t '.even'.
This patch include the (small, simple) changes need to allow it.
Test is provided at the following (clang-side) review:
https://reviews.llvm.org/D27418
Differential Revision: https://reviews.llvm.org/D27417
llvm-svn: 299453
Rafael Espindola [Tue, 4 Apr 2017 17:49:45 +0000 (17:49 +0000)]
Avoid sub shell.
Another step in getting these tests to run with the integrated one.
llvm-svn: 299452
Jim Ingham [Tue, 4 Apr 2017 17:48:21 +0000 (17:48 +0000)]
Tone down the "lldb types" log a bit.
Change the get shared class info function to only
dump its results to the inferior stdout when the
log is verbose. This matches the lldb side of the
same process, which only logs what it found if the
log is on verbose.
llvm-svn: 299451
Craig Topper [Tue, 4 Apr 2017 17:48:08 +0000 (17:48 +0000)]
[InstCombine] Add more test cases for missing combines of selects with and/or/xor with constant argument. NFC
llvm-svn: 299450
Lang Hames [Tue, 4 Apr 2017 17:03:49 +0000 (17:03 +0000)]
[RuntimeDyld] Make RuntimeDyld honor the ProcessAllSections flag.
When the ProcessAllSections flag (introduced in r204398) is set RuntimeDyld is
supposed to make a call to the client's memory manager for every section in each
object that is loaded. Due to some missing checks, this was not happening in all
cases. This patch adds the missing cases, and fixes the Orc unit test that
verifies correct behavior for ProcessAllSections (The unit test had been
silently bailing out due to an ordering issue: a change in the test order meant
that this unit-test was running before the native target was registered. This
issue has also been fixed in this patch).
This fixes <rdar://problem/
22789965>
llvm-svn: 299449
Sanjay Patel [Tue, 4 Apr 2017 16:54:58 +0000 (16:54 +0000)]
[x86] remove dead select-of-constants transform; NFCI
https://reviews.llvm.org/D30537 / https://reviews.llvm.org/rL296977 added these transforms
and other related transforms to the generic DAGCombiner (with a hook that x86 sets to true),
so these patterns should not exist by the time we reach the target-specific combiner hook.
llvm-svn: 299448
Anastasia Stulova [Tue, 4 Apr 2017 16:50:46 +0000 (16:50 +0000)]
[Bug 25404] Fix crash on typedef in OpenCL 2.0
Fixed the assertion due to absence of source location for
implicitly defined types (using addImplicitTypedef()).
During Sema checks the source location is being expected
and therefore an assertion is triggered.
The change is not specific to OpenCL. But it is particularly
common for OpenCL types to be declared implicitly in Clang
to support the mode without the standard header.
Differential Revision: https://reviews.llvm.org/D31397
llvm-svn: 299447
Rong Xu [Tue, 4 Apr 2017 16:42:20 +0000 (16:42 +0000)]
[PGO] Memory intrinsic calls optimization based on profiled size
This patch optimizes two memory intrinsic operations: memset and memcpy based
on the profiled size of the operation. The high level transformation is like:
mem_op(..., size)
==>
switch (size) {
case s1:
mem_op(..., s1);
goto merge_bb;
case s2:
mem_op(..., s2);
goto merge_bb;
...
default:
mem_op(..., size);
goto merge_bb;
}
merge_bb:
Differential Revision: http://reviews.llvm.org/D28966
llvm-svn: 299446
Jin-Gu Kang [Tue, 4 Apr 2017 16:40:25 +0000 (16:40 +0000)]
Preserve vec3 type.
Summary: Preserve vec3 type with CodeGen option.
Reviewers: Anastasia, bruno
Reviewed By: Anastasia
Subscribers: bruno, ahatanak, cfe-commits
Differential Revision: https://reviews.llvm.org/D30810
llvm-svn: 299445
Matt Arsenault [Tue, 4 Apr 2017 16:34:39 +0000 (16:34 +0000)]
AMDGPU: Remove legacy export intrinsic
llvm-svn: 299444
Matt Arsenault [Tue, 4 Apr 2017 16:34:35 +0000 (16:34 +0000)]
AMDGPU: Remove legacy image intrinsics
llvm-svn: 299443
Michael Zuckerman [Tue, 4 Apr 2017 15:44:06 +0000 (15:44 +0000)]
Fix problem with test.
llvm-svn: 299442
Krasimir Georgiev [Tue, 4 Apr 2017 15:08:42 +0000 (15:08 +0000)]
[clangd] Fix completion test to not depend on the standard library
llvm-svn: 299440
Coby Tayree [Tue, 4 Apr 2017 14:43:23 +0000 (14:43 +0000)]
[X86][MS-compatability]Allow named synonymous for MS-assembly operators
This patch enhances X86AsmParser's immediate expression parsing abilities, to include a named synonymous for selected binary/unary bitwise operators: {and,shl,shr,or,xor,not}, ultimately achieving better MS-compatability
MASM reference:
https://msdn.microsoft.com/en-us/library/94b6khh4.aspx
Differential Revision: D31277
llvm-svn: 299439
Simon Pilgrim [Tue, 4 Apr 2017 14:40:53 +0000 (14:40 +0000)]
Strip trailing whitespace
llvm-svn: 299438
Olga Malysheva [Tue, 4 Apr 2017 14:39:52 +0000 (14:39 +0000)]
Test cancellation_for_sections.c expectedly fails on GCC
llvm-svn: 299437
Daniel Sanders [Tue, 4 Apr 2017 14:27:06 +0000 (14:27 +0000)]
[globalisel][tablegen] Fix non-determinism introduced in r299430.
This should fix the last issue on llvm-clang-x86_64-expensive-checks-win.
llvm-svn: 299436
Asiri Rathnayake [Tue, 4 Apr 2017 14:03:54 +0000 (14:03 +0000)]
Fix exception address alignment test for EHABI
This test fails on ARM bare-metal targets because it assumes the Itanium ABI,
whereas EHABI requires the exception address to be 8-byte aligned.
I was a bit puzzled at first because this should've failed on the public
arm-linux builder too. I think the reason it passes there is because we don't
include libunwind headers in the include path when running the libcxxabi tests,
so the system unwind.h gets picked up.
Reviewers: rengolin, EricWF
Differential revision: https://reviews.llvm.org/D31178
llvm-svn: 299435
Olga Malysheva [Tue, 4 Apr 2017 13:56:50 +0000 (13:56 +0000)]
Reset cancellation status for 'parallel', 'sections' and 'for' constracts.
Without this fix cancellation status for parallel, sections and for persists
across construct boundaries.
Differential Revision: https://reviews.llvm.org/D31419
llvm-svn: 299434
Daniel Sanders [Tue, 4 Apr 2017 13:52:00 +0000 (13:52 +0000)]
[globalisel][tablegen] Try to make MSVC happy with r299430
Fix other cases of 'const StringRef' creeping back in at the same time.
This should fix the llvm-clang-x86_64-expensive-checks-win buildbot.
llvm-svn: 299433
Michael Zuckerman [Tue, 4 Apr 2017 13:32:14 +0000 (13:32 +0000)]
[X86][LLVM] Converting __mm{|256|512}_movm_epi{8|16|32|64} LLVMIR call into generic intrinsics.
This patch is a part one of two reviews, one for the clang and the other for LLVM.
The patch deletes the back-end intrinsics and adds support for them in the auto upgrade.
Differential Revision: https://reviews.llvm.org/D31393
llvm-svn: 299432
Michael Zuckerman [Tue, 4 Apr 2017 13:29:53 +0000 (13:29 +0000)]
[X86][Clang] Converting __mm{|256|512}_movm_epi{8|16|32|64} LLVMIR call into generic intrinsics.
This patch is a part two of two reviews, one for the clang and the other for LLVM.
In this patch, I covered the clang side, by introducing the intrinsic to the front end.
This is done by creating a generic replacement.
Differential Revision: https://reviews.llvm.org/D31394a
llvm-svn: 299431
Daniel Sanders [Tue, 4 Apr 2017 13:25:23 +0000 (13:25 +0000)]
[tablegen][globalisel] Add support for nested instruction matching.
Summary:
Lift the restrictions that prevented the tree walking introduced in the
previous change and add support for patterns like:
(G_ADD (G_MUL (G_SEXT $src1), (G_SEXT $src2)), $src3) -> SMADDWrrr $dst, $src1, $src2, $src3
Also adds support for G_SEXT and G_ZEXT to support these cases.
One particular aspect of this that I should draw attention to is that I've
tried to be overly conservative in determining the safety of matches that
involve non-adjacent instructions and multiple basic blocks. This is intended
to be used as a cheap initial check and we may add a more expensive check in
the future. The current rules are:
* Reject if any instruction may load/store (we'd need to check for intervening
memory operations.
* Reject if any instruction has implicit operands.
* Reject if any instruction has unmodelled side-effects.
See isObviouslySafeToFold().
Reviewers: t.p.northover, javed.absar, qcolombet, aditya_nandakumar, ab, rovka
Reviewed By: ab
Subscribers: igorb, dberris, llvm-commits, kristof.beyls
Differential Revision: https://reviews.llvm.org/D30539
llvm-svn: 299430
Siddharth Bhat [Tue, 4 Apr 2017 13:08:23 +0000 (13:08 +0000)]
[Polly] [DependenceInfo] change WAR, WAW generation to correct semantics
= Change of WAR, WAW generation: =
- `buildFlow(Sink, MustSource, MaySource, Sink)` treates any flow of the form
`sink <- may source <- must source` as a *may* dependence.
- we used to call:
```lang=cpp, name=old-flow-call.cpp
Flow = buildFlow(MustWrite, MustWrite, Read, Schedule);
WAW = isl_union_flow_get_must_dependence(Flow);
WAR = isl_union_flow_get_may_dependence(Flow);
```
- This caused some WAW dependences to be treated as WAR dependences.
- Incorrect semantics.
- Now, we call WAR and WAW correctly.
== Correct WAW: ==
```lang=cpp, name=new-waw-call.cpp
Flow = buildFlow(Write, MustWrite, MayWrite, Schedule);
WAW = isl_union_flow_get_may_dependence(Flow);
isl_union_flow_free(Flow);
```
== Correct WAR: ==
```lang=cpp, name=new-war-call.cpp
Flow = buildFlow(Write, Read, MustaWrite, Schedule);
WAR = isl_union_flow_get_must_dependence(Flow);
isl_union_flow_free(Flow);
```
- We want the "shortest" WAR possible (exact dependences).
- We mark all the *must-writes* as may-source, reads as must-souce.
- Then, we ask for *must* dependence.
- This removes all the reads that flow through a *must-write*
before reaching a sink.
- Note that we only block ealier writes with *must-writes*. This is
intuitively correct, as we do not want may-writes to block
must-writes.
- Leaves us with direct (R -> W).
- This affects reduction generation since RED is built using WAW and WAR.
= New StrictWAW for Reductions: =
- We used to call:
```lang=cpp,name=old-waw-war-call.cpp
Flow = buildFlow(MustWrite, MustWrite, Read, Schedule);
WAW = isl_union_flow_get_must_dependence(Flow);
WAR = isl_union_flow_get_may_dependence(Flow);
```
- This *is* the right model of WAW we need for reductions, just not in general.
- Reductions need to track only *strict* WAW, without any interfering reductions.
= Explanation: Why the new WAR dependences in tests are correct: =
- We no longer set WAR = WAR - WAW
- Hence, we will have WAR dependences that were originally removed.
- These may look incorrect, but in fact make sense.
== Code: ==
```lang=llvm, name=new-war-dependence.ll
; void manyreductions(long *A) {
; for (long i = 0; i < 1024; i++)
; for (long j = 0; j < 1024; j++)
; S0: *A += 42;
;
; for (long i = 0; i < 1024; i++)
; for (long j = 0; j < 1024; j++)
; S1: *A += 42;
;
```
=== WAR dependence: ===
{ S0[1023, 1023] -> S1[0, 0] }
- Between `S0[1023, 1023]` and `S1[0, 0]`, we will have the dependences:
```lang=cpp, name=dependence-incorrect, counterexample
S0[1023, 1023]:
*-- tmp = *A (load0)--*
WAR 2 add = tmp + 42 |
*-> *A = add (store0) |
WAR 1
S1[0, 0]: |
tmp = *A (load1) |
add = tmp + 42 |
A = add (store1)<-*
```
- One may assume that WAR2 *hides* WAR1 (since store0 happens before
store1). However, within a statement, Polly has no idea about the
ordering of loads and stores.
- Hence, according to Polly, the code may have looked like this:
```lang=cpp, name=dependence-correct
S0[1023, 1023]:
A = add (store0)
tmp = A (load0) ---*
add = A + 42 |
WAR 1
S1[0, 0]: |
tmp = A (load1) |
add = A + 42 |
A = add (store1) <-*
```
- So, Polly generates (correct) WAR dependences. It does not make sense
to remove these dependences, since they are correct with respect to
Polly's model.
Reviewers: grosser, Meinersbur
tags: #polly
Differential revision: https://reviews.llvm.org/D31386
llvm-svn: 299429
Olga Malysheva [Tue, 4 Apr 2017 12:56:55 +0000 (12:56 +0000)]
Test check-in, comment changed
llvm-svn: 299428
Simon Dardis [Tue, 4 Apr 2017 11:28:53 +0000 (11:28 +0000)]
[mips] Deal with empty blocks in the mips hazard scheduler
This patch teaches the hazard scheduler how to handle empty blocks
when search for the next real instruction when dealing with forbidden
slots.
Reviewers: slthakur
Differential Revision: https://reviews.llvm.org/D31293
llvm-svn: 299427
Krasimir Georgiev [Tue, 4 Apr 2017 10:42:22 +0000 (10:42 +0000)]
[clangd] Remove private vector fields from completion test.
llvm-svn: 299426
Oren Ben Simhon [Tue, 4 Apr 2017 10:23:18 +0000 (10:23 +0000)]
[X86] Add 64 bit pattern matching for PSADBW
PSADBW pattern currently supports the 32 bit IR pattern and only GLT (greather than) comparison.
The patch extends the pattern to catch also 64 bit IR pattern and includes all other comparison types (not only GLT).
Differential Revision: https://reviews.llvm.org/D31577
llvm-svn: 299425
Philip Pfaffe [Tue, 4 Apr 2017 10:22:17 +0000 (10:22 +0000)]
Fix formatting in LoopGenerators
llvm-svn: 299424
Philip Pfaffe [Tue, 4 Apr 2017 10:01:53 +0000 (10:01 +0000)]
[Polly][NewPM] Pull references to the legacy PM interface from utilities and helpers
Summary:
A couple of the utilities used to analyze or build IR make explicit use of the legacy PM on their interface, to access analysis results. This patch removes the legacy PM from the interface, and just passes the required results directly.
This shouldn't introduce any function changes, although the API technically allowed to obtain two different analysis results before, one passed by reference and one through the PM. I don't believe that was ever intended, however.
Reviewers: grosser, Meinersbur
Reviewed By: grosser
Subscribers: nemanjai, pollydev, llvm-commits
Tags: #polly
Differential Revision: https://reviews.llvm.org/D31653
llvm-svn: 299423
Haojian Wu [Tue, 4 Apr 2017 09:53:55 +0000 (09:53 +0000)]
Fix windows buildbot error.
llvm-svn: 299422
Krasimir Georgiev [Tue, 4 Apr 2017 09:46:39 +0000 (09:46 +0000)]
[clangd] Add code completion support
Summary: Adds code completion support to clangd.
Reviewers: bkramer, malaperle-ericsson
Reviewed By: bkramer, malaperle-ericsson
Subscribers: stanionascu, malaperle-ericsson, cfe-commits
Differential Revision: https://reviews.llvm.org/D31328
llvm-svn: 299421
James Henderson [Tue, 4 Apr 2017 09:42:24 +0000 (09:42 +0000)]
[ELF] Fail the link early if the map file path is invalid
As with the changes made in r297645, we do not want a potentially long link to
be run, if it will ultimately fail because the map file is not writable. This
change reuses the same functionality as the output file path check. See
https://reviews.llvm.org/D30449 for further justification and explanations.
Reviewers: ruiu
Differential Revision: https://reviews.llvm.org/D31603
llvm-svn: 299420
Haojian Wu [Tue, 4 Apr 2017 09:30:06 +0000 (09:30 +0000)]
[clang-rename] Support renaming qualified symbol
Summary:
The patch adds a new feature for renaming qualified symbol references.
Unlike orginal clang-rename behavior, when renaming a qualified symbol to a new
qualified symbol (e.g "A::Foo" => "B::Bar"), this new rename behavior will
consider the prefix qualifiers of the symbol, and calculate the new prefix
qualifiers. It aims to add as few additional qualifiers as possible.
As this is an early version (only supports renaming classes), I don't change
current clang-rename interfaces at the moment, and would like to keep its
(command-line tool) behavior. So I added new interfaces for the prototype.
In the long run, these interfaces should be unified.
No functionality changes in original clang-rename command-line tool.
This patch also contains a few bug fixes of clang-rename which are discovered by
the new unittest:
* fix a potential nullptr accessment when class declaration doesn't have definition.
* add USRs of nested declartaions in "getNamedDeclFor".
Reviewers: ioeric
Reviewed By: ioeric
Subscribers: alexfh, cfe-commits, mgorny
Differential Revision: https://reviews.llvm.org/D31176
llvm-svn: 299419
Peter Smith [Tue, 4 Apr 2017 09:29:36 +0000 (09:29 +0000)]
[ELF] Rename ARM Thunks in anticipation of Range Thunks
The existing names for the ARM and Thumb Thunks highlight their current
use as interworking Thunks. These Thunks can also be used for range
extension Thunks where there is no state change. This change makes the name
more generic so it is suitable for range extension.
Differential Revision: https://reviews.llvm.org/D31605
llvm-svn: 299418
Ilia K [Tue, 4 Apr 2017 08:00:28 +0000 (08:00 +0000)]
Enable lldm-mi commands -stack-list-locals -stack-list-variables and -var-create to work only with variables in scope
Patch by ayuckhulk
Reviewers: abidh, lldb-commits, ki.stfu
Reviewed By: ki.stfu
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D31073
llvm-svn: 299417
Craig Topper [Tue, 4 Apr 2017 06:38:44 +0000 (06:38 +0000)]
[X86] Remove some code that tries to disable HLE feature. This feature flag was removed from the backend.
llvm-svn: 299416
Jonas Hahnfeld [Tue, 4 Apr 2017 06:02:32 +0000 (06:02 +0000)]
Align all scalar numbers to LLVM_YAML_IS_FLOW_SEQUENCE_VECTOR
Otherwise, yamlize in YAMLTraits.h might be wrongly defined.
This makes some AMDGPU tests fail when LLVM_LINK_LLVM_DYLIB is set.
Differential Revision: https://reviews.llvm.org/D30508
llvm-svn: 299415
Mehdi Amini [Tue, 4 Apr 2017 05:38:38 +0000 (05:38 +0000)]
Uses quote to include cxxabi.h to make sure the local one is included
llvm-svn: 299414
Craig Topper [Tue, 4 Apr 2017 05:03:02 +0000 (05:03 +0000)]
[InstCombine] Use setAllBits in place of getAllOnesValue since we know the bitwidths are the same. NFCI
llvm-svn: 299413
Zvi Rackover [Tue, 4 Apr 2017 04:47:57 +0000 (04:47 +0000)]
InstCombine: Use the InstSimplify hook for shufflevector
Summary: Start using the recently added InstSimplify hook for shuffles in the respective InstCombine visitor.
Reviewers: spatel, RKSimon, craig.topper, majnemer
Reviewed By: majnemer
Subscribers: majnemer, llvm-commits
Differential Revision: https://reviews.llvm.org/D31526
llvm-svn: 299412
Eric Fiselier [Tue, 4 Apr 2017 02:54:27 +0000 (02:54 +0000)]
Fix more -Wshadow warnings introduced by recent Clang change
llvm-svn: 299411
Galina Kistanova [Tue, 4 Apr 2017 02:50:40 +0000 (02:50 +0000)]
Modules/builtins.m requires shell.
llvm-svn: 299410
Jason Molenda [Tue, 4 Apr 2017 01:09:20 +0000 (01:09 +0000)]
Skip three test cases that are asserting on macosx as of r299199. A quick
look showed that the target's arch has no core / byte order and so when
AuxVector::AuxVector calls into a dataextractor and sets the byte size to 0,
it asserts. e.g.
m_arch = {
m_triple = (Data = "x86_64--linux", Arch = x86_64, SubArch = NoSubArch, Vendor = UnknownVendor, OS = Linux, Environment = UnknownEnvironment, ObjectFormat = ELF)
m_core = kCore_invalid
m_byte_order = eByteOrderInvalid
m_flags = 0x00000000
m_distribution_id = <no value available>
}
<rdar://problem/
31380097>
llvm-svn: 299408
Eric Fiselier [Tue, 4 Apr 2017 01:05:59 +0000 (01:05 +0000)]
Work around recent -Wshadow changes in Clang
llvm-svn: 299407
Reid Kleckner [Tue, 4 Apr 2017 00:56:34 +0000 (00:56 +0000)]
[PDB] Save one type record copy
Summary:
The TypeTableBuilder provides stable storage for type records. We don't
need to copy all of the bytes into a flat vector before adding it to the
TpiStreamBuilder.
This makes addTypeRecord take an ArrayRef<uint8_t> and a hash code to go
with it, which seems like a simplification.
Reviewers: ruiu, zturner, inglorion
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31634
llvm-svn: 299406
Reid Kleckner [Mon, 3 Apr 2017 23:58:15 +0000 (23:58 +0000)]
[codeview] Cope with unsorted streams in type merging
Summary:
MASM can produce type streams that are not topologically sorted. It can
even produce type streams with circular references, but those are not
common in practice.
Reviewers: inglorion, ruiu
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31629
llvm-svn: 299403
Sean Callanan [Mon, 3 Apr 2017 23:56:41 +0000 (23:56 +0000)]
Add CPlusPlusNameParser to the xcodeproj
llvm-svn: 299402
Eric Fiselier [Mon, 3 Apr 2017 23:23:44 +0000 (23:23 +0000)]
Fix C++17 dylib build
llvm-svn: 299401
Reid Kleckner [Mon, 3 Apr 2017 23:00:25 +0000 (23:00 +0000)]
[Fuzzer] Flush std::cout before aborting in CxxStringEqTest
On Windows, abort() does not appear to flush std::cout. Should fix red
sanitizer-windows bot.
llvm-svn: 299398
Craig Topper [Mon, 3 Apr 2017 22:59:30 +0000 (22:59 +0000)]
Recommit r299321 '[X86] Add __extension__ to f16c macro intrinsics to suppress warnings about compound literals when compiled for with earlier language standards enabled.'
The bot didn't recover after the revert. So it looks like this wasn't the issue.
llvm-svn: 299397
Sanjay Patel [Mon, 3 Apr 2017 22:45:46 +0000 (22:45 +0000)]
add/move codegen tests for and/or of setcc; NFC
llvm-svn: 299396
Tim Northover [Mon, 3 Apr 2017 22:24:32 +0000 (22:24 +0000)]
Update stale doxygen links in ProgrammersManual.rst
Patch by Wei-Ren Chen.
llvm-svn: 299395
Jason Molenda [Mon, 3 Apr 2017 22:23:01 +0000 (22:23 +0000)]
The LIBLLDB_LOG_TEMPORARY channel got lost at some point where
Logging.cpp was being changed in the past. Re-add it.
llvm-svn: 299394
Zvi Rackover [Mon, 3 Apr 2017 22:05:30 +0000 (22:05 +0000)]
InstSimplify: Add a hook for shufflevector
Summary:
Add a hook for simplification of shufflevector's with the following rules:
- Constant folding - NFC, as it was already being done by the default handler.
- If only one of the operands is constant, constant fold the shuffle if the
mask does not select elements from the variable operand - to show the hook is firing and affecting the test-cases.
Reviewers: RKSimon, craig.topper, spatel, sanjoy, nlopes, majnemer
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D31525
llvm-svn: 299393
Weiming Zhao [Mon, 3 Apr 2017 21:50:04 +0000 (21:50 +0000)]
Reland r298901 with modifications (reverted in r298932)
Dont emit Mapping symbols for sections that contain only data.
Summary:
Dont emit mapping symbols for sections that contain only data.
Reviewers: rengolin, weimingz, kparzysz, t.p.northover, peter.smith
Reviewed By: t.p.northover
Patched by Shankar Easwaran <shankare@codeaurora.org>
Subscribers: alekseyshl, t.p.northover, llvm-commits
Differential Revision: https://reviews.llvm.org/D30724
llvm-svn: 299392
Matt Arsenault [Mon, 3 Apr 2017 21:45:13 +0000 (21:45 +0000)]
AMDGPU: Remove llvm.SI.vs.load.input
llvm-svn: 299391
Rui Ueyama [Mon, 3 Apr 2017 21:36:31 +0000 (21:36 +0000)]
Change the error message format for an incompatible relocation.
Previous error message style:
error: /home/alice/src/bar.c:12: relocation R_X86_64_PLT32 cannot refer to absolute symbol 'answer' defined in /home/alice/src/foo.o
New error message style:
error: relocation R_X86_64_PLT32 cannot refer to absolute symbol: foo
>>> defined in /home/alice/src/foo.o
>>> referenced by bar.c:12 (/home/alice/src/bar.c:12)
>>> /home/alice/src/bar.o:(.text+0x1)
llvm-svn: 299390
Matt Arsenault [Mon, 3 Apr 2017 21:28:13 +0000 (21:28 +0000)]
DAG: Fix missing legalization for any_extend_vector_inreg operands
llvm-svn: 299389
Reid Kleckner [Mon, 3 Apr 2017 21:25:20 +0000 (21:25 +0000)]
[codeview] Add support for label type records
MASM can produce these type records.
llvm-svn: 299388