Matthias Braun [Mon, 1 Oct 2018 17:50:52 +0000 (17:50 +0000)]
MIRParser: Check that instructions only reference DILocation metadata
llvm-svn: 343505
Fangrui Song [Mon, 1 Oct 2018 17:21:07 +0000 (17:21 +0000)]
Remove redundant null pointer check in operator delete
Summary:
C89 4.10.3.2 The free function
C99 7.20.3.2 The free function
C11 7.22.3.3 The free function
If ptr is a null pointer, no action shall occur.
_aligned_free on MSDN:
If memblock is a NULL pointer, this function simply performs no actions.
Reviewers: EricWF, mclow.lists, khng300, hotpxl
Reviewed By: mclow.lists, khng300, hotpxl
Subscribers: lichray, llvm-commits, hotpxl, khng300, christof, ldionne, cfe-commits, libcxx-commits
Differential Revision: https://reviews.llvm.org/D52401
llvm-svn: 343503
Jonas Devlieghere [Mon, 1 Oct 2018 17:20:57 +0000 (17:20 +0000)]
Fix tests affected by printing change.
I forgot to update some tests that were affected by the escaping of
backticks in the format string, landed in r343471.
llvm-svn: 343502
Wouter van Oortmerssen [Mon, 1 Oct 2018 17:20:31 +0000 (17:20 +0000)]
[WebAssembly] Fixed AsmParser not allowing instructions with /
Summary:
The AsmParser Lexer regards these as a seperate token.
Here we expand the instruction name with them if they are
adjacent (no whitespace).
Tested: the basic-assembly.s test case has one case with a / in it.
The currently are also instructions with : in them, which we intend
to rename rather than fix them here.
Reviewers: tlively, dschuff
Subscribers: sbc100, jgravelle-google, aheejin, sunfish, llvm-commits
Differential Revision: https://reviews.llvm.org/D52442
llvm-svn: 343501
Tatyana Krasnukha [Mon, 1 Oct 2018 17:14:12 +0000 (17:14 +0000)]
Fix build with GCC < 5.0 (PR39131)
llvm-svn: 343500
Craig Topper [Mon, 1 Oct 2018 17:10:50 +0000 (17:10 +0000)]
[X86] Enable load folding in the test shrinking code
This patch adds load folding support to the test shrinking code. This was noticed missing in the review for D52669
Differential Revision: https://reviews.llvm.org/D52699
llvm-svn: 343499
Craig Topper [Mon, 1 Oct 2018 17:10:45 +0000 (17:10 +0000)]
[X86] Improve test instruction shrinking when the sign flag is used and the output of the and is truncated
Currently we skip looking through truncates if the sign flag is used. But that's overly restrictive.
It's safe to look through the truncate as long as we ensure one of the 3 things when we shrink. Either the MSB of the mask at the shrunken size isn't set. If the mask bit is set then either the shrunk size needs to be equal to the compare size or the sign flag needs to be unused.
There are still missed opportunities to shrink a load and fold it in here. This will be fixed in a future patch.
Differential Revision: https://reviews.llvm.org/D52669
llvm-svn: 343498
Aaron Smith [Mon, 1 Oct 2018 17:08:51 +0000 (17:08 +0000)]
[lldb] Start a new line for the next output if there are no symbols in the current symtab
Summary:
If there is no newline the "lldb" prompt could be on the wrong line. To reproduce the missing newline you can do 'image dump smytab' on any binary.
Previously
Symtab, file = D:\upstream\build\Debug\bin\clang-diff.exe, num_symbols = 0(lldb)
Now
Symtab, file = D:\upstream\build\Debug\bin\clang-diff.exe, num_symbols = 0
(lldb)
Reviewers: zturner, aleksandr.urakov, lldb-commits
Subscribers: abidh
Differential Revision: https://reviews.llvm.org/D52627
llvm-svn: 343497
Dan Liew [Mon, 1 Oct 2018 16:51:01 +0000 (16:51 +0000)]
[NFC] Fix `-Wcast-qual` warnings in Darwin `internal_sysctl(...)` implementation.
Summary: The warnings were introduced in r341187 and r341192.
Reviewers: kubamracek, george.karpenkov, krytarowski
Subscribers: #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D52717
llvm-svn: 343496
Rui Ueyama [Mon, 1 Oct 2018 16:41:58 +0000 (16:41 +0000)]
Add an explanation about cross linking.
Differential Revision: https://reviews.llvm.org/D52567
llvm-svn: 343495
Simon Pilgrim [Mon, 1 Oct 2018 16:31:30 +0000 (16:31 +0000)]
[X86][Btver2] Fix BT(C|R|S)mr & BT(C|R|S)mi schedule latency + uop counts
Match AMD Fam16h SOG + llvm-exegesis tests
llvm-svn: 343494
Matthias Braun [Mon, 1 Oct 2018 16:25:50 +0000 (16:25 +0000)]
DAGCombiner: StoreMerging: Fix bad index calculating when adjusting mismatching vector types
This fixes a case of bad index calculation when merging mismatching
vector types. This changes the existing code to just use the existing
extract_{subvector|element} and a bitcast (instead of bitcast first and
then newly created extract_xxx) so we don't need to adjust any indices
in the first place.
rdar://
44584718
Differential Revision: https://reviews.llvm.org/D52681
llvm-svn: 343493
Alexey Bataev [Mon, 1 Oct 2018 16:20:57 +0000 (16:20 +0000)]
[OPENMP][NVPTX] Handle `requires datasharing` flag correctly with
lightweight runtime.
The datasharing flag must be set to `1` when executing SPMD-mode compatible directive with reduction|lastprivate clauses.
llvm-svn: 343492
Sanjay Patel [Mon, 1 Oct 2018 16:17:18 +0000 (16:17 +0000)]
[x86] add tests for 256- and 512-bit vector types for scalar-to-vector transform; NFC
llvm-svn: 343491
Simon Pilgrim [Mon, 1 Oct 2018 16:12:44 +0000 (16:12 +0000)]
[X86] Create schedule classes for BT(C|R|S)mi and BT(C|R|S)mr instructions
llvm-svn: 343490
Evandro Menezes [Mon, 1 Oct 2018 16:11:19 +0000 (16:11 +0000)]
[AArch64] Refactor cheap cost model
Refactor the order in `TII::isAsCheapAsAMove()` to ease future development
and maintenance. Practically NFC.
llvm-svn: 343489
Sam McCall [Mon, 1 Oct 2018 16:07:03 +0000 (16:07 +0000)]
[Basic] Update clang tests (really testing sys::fs) that broke with r343460
llvm-svn: 343488
Simon Pilgrim [Mon, 1 Oct 2018 15:01:00 +0000 (15:01 +0000)]
[X86] Remove unnecessary BTmi/BTmr scheduler overrides
llvm-svn: 343487
Jesper Antonsson [Mon, 1 Oct 2018 14:59:25 +0000 (14:59 +0000)]
[InstCombine] Handle vector compares in foldGEPIcmp(), take 2
Summary:
This is a continuation of the fix for PR34627 "InstCombine assertion at vector gep/icmp folding". (I just realized bugpoint had fuzzed the original test for me, so I had fixed another trigger of the same assert in adjacent code in InstCombine.)
This patch avoids optimizing an icmp (to look only at the base pointers) when the resulting icmp would have a different type.
The patch adds a testcase and also cleans up and shrinks the pre-existing test for the adjacent assert trigger.
Reviewers: lebedev.ri, majnemer, spatel
Reviewed By: lebedev.ri
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D52494
llvm-svn: 343486
Simon Atanasyan [Mon, 1 Oct 2018 14:43:07 +0000 (14:43 +0000)]
[mips] Generate tests expectations using update_llc_test_checks. NFC
Generate tests expectations using update_llc_test_checks and reduce
number of "check prefixes" used in the tests.
llvm-svn: 343485
Simon Pilgrim [Mon, 1 Oct 2018 14:42:16 +0000 (14:42 +0000)]
[X86][Btver2] Fix BTmr schedule uop counts
Match AMD Fam16h SOG + llvm-exegesis tests
llvm-svn: 343484
Alexey Bataev [Mon, 1 Oct 2018 14:40:06 +0000 (14:40 +0000)]
[OPENMP] Simplify code, NFC.
llvm-svn: 343483
Sanjay Patel [Mon, 1 Oct 2018 14:40:00 +0000 (14:40 +0000)]
[InstCombine] try to convert vector insert+extract to trunc; 2nd try
This was originally committed at rL343407, but reverted at
rL343458 because it crashed trying to handle a case where
the destination type is FP. This version of the patch adds
a check for that possibility. Tests added at rL343480.
Original commit message:
This transform is requested for the backend in:
https://bugs.llvm.org/show_bug.cgi?id=39016
...but I figured it was worth doing in IR too, and it's probably
easier to implement here, so that's this patch.
In the simplest case, we are just truncating a scalar value. If the
extract index doesn't correspond to the LSBs of the scalar, then we
have to shift-right before the truncate. Endian-ness makes this tricky,
but hopefully the ASCII-art helps visualize the transform.
Differential Revision: https://reviews.llvm.org/D52439
llvm-svn: 343482
Haojian Wu [Mon, 1 Oct 2018 14:38:43 +0000 (14:38 +0000)]
[Preprocessor] Fix a crash when handling non-alpha include header.
Summary: the crash is casued by an assertion in StringRef.
(llvm::StringRef::front() const: Assertion `!empty()' failed.)
Reviewers: sammccall
Subscribers: jsji, cfe-commits
Differential Revision: https://reviews.llvm.org/D52721
llvm-svn: 343481
Sanjay Patel [Mon, 1 Oct 2018 14:29:09 +0000 (14:29 +0000)]
[InstCombine] add more insert-extract tests for D52439; NFC
The first attempt at this transform:
rL343407
...was reverted:
rL343458
...because it did not handle the case where we bitcast to FP.
The patch was already limited to avoid the case where we
bitcast from FP, but we might want to transform that too.
llvm-svn: 343480
Alexey Bataev [Mon, 1 Oct 2018 14:26:31 +0000 (14:26 +0000)]
[OPENMP] Fix enum identifier, NFC.
llvm-svn: 343479
Simon Pilgrim [Mon, 1 Oct 2018 14:23:37 +0000 (14:23 +0000)]
[X86] Create schedule classes for BTmi and BTmr instructions
llvm-svn: 343478
Jonas Hahnfeld [Mon, 1 Oct 2018 14:16:55 +0000 (14:16 +0000)]
[libomptarget-nvptx] Enable asserts in bclib
If the user requested LIBOMPTARGET_NVPTX_DEBUG, include asserts in
the bitcode library. Everything else will have very unpleasent
effects because asserts will appear when falling back to the static
library libomptarget-nvptx.a.
Differential Revision: https://reviews.llvm.org/D52701
llvm-svn: 343477
Jonas Hahnfeld [Mon, 1 Oct 2018 14:14:26 +0000 (14:14 +0000)]
[libomptarget-nvptx] reduction: Determine if runtime uninitialized
Pass in the correct value of isRuntimeUninitialized() which solves
parallel reductions as reported on the mailing list.
For reference: r333285 did the same for loop scheduling.
Differential Revision: https://reviews.llvm.org/D52725
llvm-svn: 343476
Andrey Churbanov [Mon, 1 Oct 2018 14:08:50 +0000 (14:08 +0000)]
Fixed workaround made in https://reviews.llvm.org/D51694.
Patch suggested by Kelvin Li: removed optional "kind=" part of kind-selector
for variables with long names and kind names.
Differential Revision: https://reviews.llvm.org/D52712
llvm-svn: 343475
Haojian Wu [Mon, 1 Oct 2018 14:02:02 +0000 (14:02 +0000)]
[clangd] Add "check-clangd" target
Summary:
So we don't have to run "check-clang-tools" (which builds and tests all
clang tools) to verify our clangd-related change. It'd save waiting time for
clangd developers.
check-clangd (build ~1000 files, run ~340 tests) vs check-clang-tools (build
~3000 files, run ~1000 tests).
In the future, we probably want to add similar target for other
clang-tools (e.g. clang-tidy).
Reviewers: sammccall
Subscribers: mgorny, ilya-biryukov, ioeric, MaskRay, jkorous, arphaman, kadircet, cfe-commits
Differential Revision: https://reviews.llvm.org/D52710
llvm-svn: 343474
Haojian Wu [Mon, 1 Oct 2018 14:00:51 +0000 (14:00 +0000)]
Move llvm util dependencies from clang-tools-extra to add_lit_target.
Summary:
Address fixme in r301762. And would simplify the cmake file in
clang-tools-extra.
Reviewers: sammccall
Subscribers: mgorny, llvm-commits, cfe-commits
Differential Revision: https://reviews.llvm.org/D52713
llvm-svn: 343473
Patrick Lyster [Mon, 1 Oct 2018 13:47:43 +0000 (13:47 +0000)]
Add support for unified_shared_memory clause on requires directive
llvm-svn: 343472
Jonas Devlieghere [Mon, 1 Oct 2018 13:22:24 +0000 (13:22 +0000)]
[Interpreter] Escape backticks when dumping format entities.
Currently we reject our own default disassembly-format string because it
contains two backticks which causes everything in between to be
interpreter as an expression by the command interpreter. This patch
fixes that by escaping backticks when dumping format strings.
llvm-svn: 343471
Jonas Devlieghere [Mon, 1 Oct 2018 13:20:15 +0000 (13:20 +0000)]
Escape newlines in default disassembly format.
We can safely escape newlines in format strings because they will be
ignored by the format entity parser.
llvm-svn: 343470
Robert Widmann [Mon, 1 Oct 2018 13:15:09 +0000 (13:15 +0000)]
[LLVM-C] Add an accessor for the kind of a Metadata Node
Summary: Allows for retrieving the type of a metadata node. Has the added benefit of ensuring that the C and C++ kind APIs stay in sync as a failure to add a corresponding LLVMMetadataKind will result in the switch in the accessor being semantically malformed.
Reviewers: whitequark, deadalnix
Reviewed By: whitequark
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D52693
llvm-svn: 343469
Simon Pilgrim [Mon, 1 Oct 2018 13:12:05 +0000 (13:12 +0000)]
[X86][Btver2] Fix masked load schedule
JFPU01 resource usage should match JFPX
Match AMD Fam16h SOG + llvm-exegesis tests
llvm-svn: 343468
Guillaume Chatelet [Mon, 1 Oct 2018 12:19:10 +0000 (12:19 +0000)]
[llvm-exegesis][NFC] Move random functions from CodeTemplate to SnippetGenerator.
Summary: Just moving methods around.
Reviewers: courbet
Subscribers: tschuett, llvm-commits
Differential Revision: https://reviews.llvm.org/D52720
llvm-svn: 343461
Sam McCall [Mon, 1 Oct 2018 12:17:05 +0000 (12:17 +0000)]
[Support] Listing a directory containing dangling symlinks is not an error.
Summary:
Reporting this as an error required stat()ing every file, as well as seeming
semantically questionable.
Reviewers: vsk, bkramer
Subscribers: mgrang, kristina, llvm-commits, liaoyuke
Differential Revision: https://reviews.llvm.org/D52648
llvm-svn: 343460
Haojian Wu [Mon, 1 Oct 2018 12:16:38 +0000 (12:16 +0000)]
Build clang-headers when building clang tools.
Summary:
clang tools require clang headers to work on real project, e.g. when we
build clangd via `ninja clangd`, we expect the binary can run on
real-world project (without running another command `ninja clang-headers`).
Reviewers: sammccall
Reviewed By: sammccall
Subscribers: mgorny, ilya-biryukov, ioeric, kadircet, cfe-commits
Differential Revision: https://reviews.llvm.org/D52714
llvm-svn: 343459
Hans Wennborg [Mon, 1 Oct 2018 12:07:45 +0000 (12:07 +0000)]
Revert r343407 "[InstCombine] try to convert vector insert+extract to trunc"
This caused Chromium builds to fail with "Illegal Trunc" assertion.
See https://crbug.com/890723 for repro.
> This transform is requested for the backend in:
> https://bugs.llvm.org/show_bug.cgi?id=39016
> ...but I figured it was worth doing in IR too, and it's probably
> easier to implement here, so that's this patch.
>
> In the simplest case, we are just truncating a scalar value. If the
> extract index doesn't correspond to the LSBs of the scalar, then we
> have to shift-right before the truncate. Endian-ness makes this tricky,
> but hopefully the ASCII-art helps visualize the transform.
>
> Differential Revision: https://reviews.llvm.org/D52439
llvm-svn: 343458
Sam McCall [Mon, 1 Oct 2018 11:56:42 +0000 (11:56 +0000)]
[CodeComplete] #include completion treats -I as non-system (require header-like extension).
llvm-svn: 343457
Guillaume Chatelet [Mon, 1 Oct 2018 11:46:06 +0000 (11:46 +0000)]
[llvm-exegesis][NFC] Make randomizeUnsetVariables a free function.
Summary: This is prelimineary to moving random functions to SnippetGenerator.
Reviewers: courbet
Subscribers: tschuett, llvm-commits
Differential Revision: https://reviews.llvm.org/D52718
llvm-svn: 343456
Alexander Timofeev [Mon, 1 Oct 2018 11:06:35 +0000 (11:06 +0000)]
[AMDGPU] Divergence driven instruction selection. Shift operations.
Summary: This change enables VOP3 shifts to be explicitly selected
dependent on the divergence.
Differential Revision: https://reviews.llvm.org/D52559
Reviewers: rampitec
llvm-svn: 343455
Puyan Lotfi [Mon, 1 Oct 2018 10:50:23 +0000 (10:50 +0000)]
[NFC] Adding "REQUIRES: zlib" to a llvm-objcopy test for bots without zlib.
M test/tools/llvm-objcopy/compress-and-decompress-debug-sections-error.test
llvm-svn: 343454
Sam McCall [Mon, 1 Oct 2018 10:42:51 +0000 (10:42 +0000)]
[clangd] Query dex index using query-style trigrams, not identifier-style trigrams
llvm-svn: 343453
Andrea Di Biagio [Mon, 1 Oct 2018 10:35:13 +0000 (10:35 +0000)]
[X86][BtVer2] Teach how to identify zero-idiom VPERM2F128rr instructions.
This patch adds another variant class to identify zero-idiom VPERM2F128rr
instructions.
On Jaguar, a VPERM wih bit 3 and 7 of the mask set, is a zero-idiom.
Differential Revision: https://reviews.llvm.org/D52663
llvm-svn: 343452
Puyan Lotfi [Mon, 1 Oct 2018 10:29:41 +0000 (10:29 +0000)]
[llvm-objcopy] Adding support for decompressing zlib compressed dwarf sections.
Summary: I had added support for compressing dwarf sections in a prior commit,
this one adds support for decompressing. Usage is:
llvm-objcopy --decompress-debug-sections input.o output.o
Reviewers: jakehehrlich, jhenderson, alexshap
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D51841
llvm-svn: 343451
Florian Hahn [Mon, 1 Oct 2018 09:59:48 +0000 (09:59 +0000)]
Recommit r343308: [LoopInterchange] Turn into a loop pass.
llvm-svn: 343450
James Henderson [Mon, 1 Oct 2018 09:57:20 +0000 (09:57 +0000)]
Fix typo in LLD man page
llvm-svn: 343449
Eric Liu [Mon, 1 Oct 2018 08:50:49 +0000 (08:50 +0000)]
[clangd] Fix header mapping for std::string. NFC
Some implementation has std::string declared in <iosfwd>.
llvm-svn: 343448
Clement Courbet [Mon, 1 Oct 2018 08:37:48 +0000 (08:37 +0000)]
[X86][Sched] Update scheduling information for VZEROALL on HWS, BDW, SKX, SNB.
Summary:
While looking at PR35606, I found out that the scheduling info is incorrect.
One can check that it's really a P5+P6 and not a 2*P56 with:
echo -e 'vzeroall\nvandps %xmm1, %xmm2, %xmm3' | ./bin/llvm-exegesis -mode=uops -snippets-file=-
(vandps executes on P5 only)
Reviewers: craig.topper, RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D52541
llvm-svn: 343447
Clement Courbet [Mon, 1 Oct 2018 08:37:37 +0000 (08:37 +0000)]
[X86][Sched] Add pfm uop counter definitions for SNB,BDW,SKX.
llvm-svn: 343446
Carlos Alberto Enciso [Mon, 1 Oct 2018 08:14:44 +0000 (08:14 +0000)]
[DebugInfo][Dexter] Incorrect DBG_VALUE after MCP dead copy instruction removal.
When MachineCopyPropagation eliminates a dead 'copy', its associated debug information becomes invalid. as the recorded register has been removed. It causes the debugger to display wrong variable value.
Differential Revision: https://reviews.llvm.org/D52614
llvm-svn: 343445
Clement Courbet [Mon, 1 Oct 2018 07:16:22 +0000 (07:16 +0000)]
[CodeGen][NFC] Add tests for heterogeneous types in MergeConsecutiveStores
Reviewers: efriedma
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D52643
llvm-svn: 343444
Craig Topper [Mon, 1 Oct 2018 07:08:41 +0000 (07:08 +0000)]
[X86] Stop X86DomainReassignment from creating copies between GR8/GR16 physical registers and k-registers.
We can only copy between a k-register and a GR32/GR64 register.
This patch detects that the copy will be illegal and prevents the domain reassignment from happening for that closure.
This probably isn't the best fix, and we should probably figure out how to handle this correctly.
Fixes PR38803.
llvm-svn: 343443
Lang Hames [Mon, 1 Oct 2018 04:59:10 +0000 (04:59 +0000)]
[ORC] Pass Symbols to ExecutionSession::lookup by value, potentially saving a
copy.
llvm-svn: 343442
Eric Fiselier [Mon, 1 Oct 2018 04:08:06 +0000 (04:08 +0000)]
Attempt to fix aligned allocation configuration under clang-cl
When we're using clang-cl and Microsoft's runtime implementation,
we don't provide align_val_t or aligned new/delete ourselves.
This patch updates the _LIBCPP_HAS_NO_ALIGNED_ALLOCATION macro
to reflect this.
llvm-svn: 343441
Eric Fiselier [Mon, 1 Oct 2018 03:59:05 +0000 (03:59 +0000)]
Remove one more warning from clang-cl build
llvm-svn: 343440
Eric Fiselier [Mon, 1 Oct 2018 02:54:08 +0000 (02:54 +0000)]
Fix Shadowing warning on Windows
llvm-svn: 343439
Eric Fiselier [Mon, 1 Oct 2018 01:59:37 +0000 (01:59 +0000)]
Fix even more Clang warnings.
This patch disables shift-sign-overflow warnings for now. It also
fixes most -Wfloat-equal warnings and -Wextra-semi warnings.
llvm-svn: 343438
Eric Fiselier [Mon, 1 Oct 2018 01:51:46 +0000 (01:51 +0000)]
Bump default dialect to C++14 for clang-cl
llvm-svn: 343437
Eric Fiselier [Mon, 1 Oct 2018 01:47:23 +0000 (01:47 +0000)]
Remove even more clang-cl warnings
llvm-svn: 343436
Eric Fiselier [Mon, 1 Oct 2018 01:43:42 +0000 (01:43 +0000)]
Mork more tests as FLAKY
llvm-svn: 343435
Eric Fiselier [Mon, 1 Oct 2018 01:31:23 +0000 (01:31 +0000)]
Remove even more warnings from clang-cl build
llvm-svn: 343434
Eric Fiselier [Mon, 1 Oct 2018 01:15:50 +0000 (01:15 +0000)]
Turn off warnings under clang-cl
llvm-svn: 343433
Eric Fiselier [Mon, 1 Oct 2018 01:05:51 +0000 (01:05 +0000)]
Fix threaded test under no-threading configuration
llvm-svn: 343432
Eric Fiselier [Mon, 1 Oct 2018 01:00:11 +0000 (01:00 +0000)]
Attempt to unbreak Windows configuration.
Although libc++ doesn't yet support Windows we still have Windows
builders to track our progress.
Currently the clang-cl configuration seems broken because it doesn't
support -std=c++11 and instead requires /std:c++11. This patch attempts
to fix this.
llvm-svn: 343431
Lang Hames [Mon, 1 Oct 2018 00:59:28 +0000 (00:59 +0000)]
[ORC] Add convenience methods for creating DynamicLibraryFallbackGenerators for
libraries on disk, and for the current process.
Avoids more boilerplate during JIT construction.
llvm-svn: 343430
Lang Hames [Mon, 1 Oct 2018 00:59:26 +0000 (00:59 +0000)]
[ORC] Add a method to JITTargetMachineBuilder to get the default data layout
for the target machine.
This simplifies usage during setup of concurrent JIT stacks where the client
needs a DataLayout, but not a TargetMachine (TargetMachines are created on
the fly by the compile threads later).
llvm-svn: 343429
Craig Topper [Sun, 30 Sep 2018 23:43:30 +0000 (23:43 +0000)]
[X86] Change an llvm_unreachable to a report_fatal_error so the optimizer will stop making us reach the other report_fatal_error in this function.
There's a conditional report_fatal_error just above this llvm_unreachable. The optimizer when seeing the unreachable removes the conditional and just makes any other error trigger the existing report_fatal_error.
llvm-svn: 343428
Lang Hames [Sun, 30 Sep 2018 23:18:24 +0000 (23:18 +0000)]
[ORC] Add an 'intern' method to ExecutionEngine for interning symbol names.
This cuts down on boilerplate by reducing 'ES.getSymbolStringPool().intern(...)'
to 'ES.intern(...)'.
llvm-svn: 343427
Fangrui Song [Sun, 30 Sep 2018 22:31:29 +0000 (22:31 +0000)]
Use the container form llvm::sort(C, ...)
There are a few leftovers in rL343163 which span two lines. This commit
changes these llvm::sort(C.begin(), C.end, ...) to llvm::sort(C, ...)
llvm-svn: 343426
Fangrui Song [Sun, 30 Sep 2018 21:41:11 +0000 (21:41 +0000)]
Use the container form llvm::sort(C, ...)
There are a few leftovers of rC343147 that are not (\w+)\.begin but in
the form of ([-[:alnum:]>.]+)\.begin or spanning two lines. Change them
to use the container form in this commit. The 12 occurrences have been
inspected manually for safety.
llvm-svn: 343425
Simon Pilgrim [Sun, 30 Sep 2018 20:19:16 +0000 (20:19 +0000)]
[X86] Fix scheduler class for BTmi instructions
This wasn't treated as a folded load instruction
llvm-svn: 343424
Lang Hames [Sun, 30 Sep 2018 19:12:23 +0000 (19:12 +0000)]
[ORC] Extract and tidy up JITTargetMachineBuilder, add unit test.
(1) Adds comments for the API.
(2) Removes the setArch method: This is redundant: the setArchStr method on the
triple should be used instead.
(3) Turns EmulatedTLS on by default. This matches EngineBuilder's behavior.
llvm-svn: 343423
Martin Storsjo [Sun, 30 Sep 2018 18:31:03 +0000 (18:31 +0000)]
[COFF] In MinGW mode, ignore relocations against a discarded section
When GCC produces a jump table as part of a comdat function, the
jump table itself is produced as plain non-comdat rdata section. When
linked with ld.bfd, all of those rdata sections are kept, with
relocations unchanged in the sections that refer to discarded comdat
sections.
This has been observed with at least GCC 5.x and 7.x.
Differential Revision: https://reviews.llvm.org/D52600
llvm-svn: 343422
Simon Pilgrim [Sun, 30 Sep 2018 18:19:00 +0000 (18:19 +0000)]
[LLVM-MCA][X86] Add missing VCMPESTR/VCMPESTR tests
llvm-svn: 343421
Eric Fiselier [Sun, 30 Sep 2018 18:05:39 +0000 (18:05 +0000)]
Fix linkage error on ProgramPoint's dump method.
Currently, ProgramPoint::dump calls the out-of-line function ProgramPoint::print. This causes
libraries which include ProgramPoint.h to become dependent on libclangAnalysis, which in turn
causes missing symbol link error when building with -DBUILD_SHARED_LIBS=ON -DLLVM_ENABLE_MODULES=ON.
The breakage was introduced in r343160.
This patch fixes the issues by moving ProgramPoint::dump's declaration out of line.
llvm-svn: 343420
Craig Topper [Sun, 30 Sep 2018 17:47:18 +0000 (17:47 +0000)]
[X86] Copy memrefs when folding a load for division instruction selection.
llvm-svn: 343419
Aaron Ballman [Sun, 30 Sep 2018 17:39:39 +0000 (17:39 +0000)]
Reverting r343415 as it breaks at least one of the bots.
http://lab.llvm.org:8011/builders/llvm-clang-lld-x86_64-scei-ps4-ubuntu-fast/builds/37336
llvm-svn: 343418
Bjorn Pettersson [Sun, 30 Sep 2018 17:26:58 +0000 (17:26 +0000)]
[PHIElimination] Lower a PHI node with only undef uses as IMPLICIT_DEF
Summary:
The lowering of PHI nodes used to detect if all inputs originated
from IMPLICIT_DEF's. If so the PHI node was replaced by an
IMPLICIT_DEF. Now we also consider undef uses when checking the
inputs. So if all inputs are implicitly defined or undef we
lower the PHI to an IMPLICIT_DEF. This makes
PHIElimination::LowerPHINode more consistent as it checks
both implicit and undef properties at later stages.
Reviewers: MatzeB, tstellar
Reviewed By: MatzeB
Subscribers: jvesely, nhaehnle, llvm-commits
Differential Revision: https://reviews.llvm.org/D52558
llvm-svn: 343417
Bjorn Pettersson [Sun, 30 Sep 2018 17:23:21 +0000 (17:23 +0000)]
[PHIElimination] Update the regression test for PR16508
Summary:
When PR16508 was solved (in rL185363) a regression test was
added as test/CodeGen/PowerPC/2013-07-01-PHIElimBug.ll.
I discovered that the test case no longer reproduced the
scenario from PR16508. This problem could have been amended
by adding an extra RUN line with "-O1" (or possibly "-O0"),
but instead I added a mir-reproducer
test/CodeGen/PowerPC/2013-07-01-PHIElimBug.mir
to get a reproducer that is less sensitive to changes in
earlier passes (including O-level).
While being at it I also corrected a code comment in
PHIElimination::EliminatePHINodes that has been incorrect
since the related bugfix from rL185363.
Reviewers: MatzeB, hfinkel
Reviewed By: MatzeB
Subscribers: nemanjai, jsji, llvm-commits
Differential Revision: https://reviews.llvm.org/D52553
llvm-svn: 343416
Aaron Ballman [Sun, 30 Sep 2018 17:22:58 +0000 (17:22 +0000)]
Allow clang-tidy to be built without a dependency on the clang static analyzer.
Patch by Stephen Kelly.
llvm-svn: 343415
Simon Pilgrim [Sun, 30 Sep 2018 17:01:59 +0000 (17:01 +0000)]
[LLVM-MCA][X86] Add some AVX512 tests
These are going to be necessary to check I don't mess up when I start cleaning up all the remaining vector integer overrides
llvm-svn: 343414
Simon Pilgrim [Sun, 30 Sep 2018 16:38:38 +0000 (16:38 +0000)]
[X86][Btver2] Fix PCmpIStrI/PCmpIStrM schedules
Missing JFPU0 pipe and double JFPU1 pipe (to match JVALU1) resources
Match AMD Fam16h SOG + llvm-exegesis tests
llvm-svn: 343413
Zachary Turner [Sun, 30 Sep 2018 16:19:18 +0000 (16:19 +0000)]
[PDB] Add native support for dumping array types.
llvm-svn: 343412
Pavel Labath [Sun, 30 Sep 2018 16:12:09 +0000 (16:12 +0000)]
Fix NetBSD build for r343409
Forgot to remove the method declaration from the header.
llvm-svn: 343411
Simon Pilgrim [Sun, 30 Sep 2018 15:58:56 +0000 (15:58 +0000)]
[X86][BtVer2] Add the ability to add additional uops for folded instructions
Some instructions take an extra load uop - but not consistently.....
llvm-svn: 343410
Pavel Labath [Sun, 30 Sep 2018 15:58:52 +0000 (15:58 +0000)]
Pull GetSoftwareBreakpointPCOffset into base class
Summary:
This function encodes the knowledge of whether the PC points to the
breakpoint instruction of the one following it after the breakpoint is
"hit". This behavior mainly(*) depends on the architecture and not on the
OS, so it makes sense for it to be implemented in the base class, where
it can be shared between different implementations (Linux and NetBSD
atm).
(*) It is possible for an OS to expose a different API, perhaps by doing
some fixups in the kernel. In this case, the implementation can override
this function to implement custom behavior.
Reviewers: krytarowski, zturner
Subscribers: lldb-commits
Differential Revision: https://reviews.llvm.org/D52532
llvm-svn: 343409
Ed Maste [Sun, 30 Sep 2018 15:08:18 +0000 (15:08 +0000)]
Update ifunc attribute support documentation
Previously we documented GNU binutils and glibc versions required for
ifunc support, but our own lld linker and FreeBSD's rtld also support
ifuncs.
Differential Revision: https://reviews.llvm.org/D52696
llvm-svn: 343408
Sanjay Patel [Sun, 30 Sep 2018 14:34:01 +0000 (14:34 +0000)]
[InstCombine] try to convert vector insert+extract to trunc
This transform is requested for the backend in:
https://bugs.llvm.org/show_bug.cgi?id=39016
...but I figured it was worth doing in IR too, and it's probably
easier to implement here, so that's this patch.
In the simplest case, we are just truncating a scalar value. If the
extract index doesn't correspond to the LSBs of the scalar, then we
have to shift-right before the truncate. Endian-ness makes this tricky,
but hopefully the ASCII-art helps visualize the transform.
Differential Revision: https://reviews.llvm.org/D52439
llvm-svn: 343407
Sanjay Patel [Sun, 30 Sep 2018 13:50:42 +0000 (13:50 +0000)]
[InstCombine] allow lengthening of insertelement to eliminate shuffles
As noted in post-commit comments for D52548, the limitation on
increasing vector length can be applied by opcode.
As a first step, this patch only allows insertelement to be
widened because that has no logical downsides for IR and has
little risk of pessimizing codegen.
This may cause PR39132 to go into hiding during a full compile,
but that bug is not fixed.
llvm-svn: 343406
Simon Pilgrim [Sun, 30 Sep 2018 12:46:42 +0000 (12:46 +0000)]
[DAG] Don't perform SINT_TO_FP<->UINT_TO_FP custom conversion after legalization
The SINT_TO_FP<->UINT_TO_FP combines for non-negative integers should only occur for legal ops once LegalOperations = true
No test case to hand, noticed when investigating PR38226 + PR38970
llvm-svn: 343405
Roman Lebedev [Sun, 30 Sep 2018 12:42:08 +0000 (12:42 +0000)]
[NFC][CodeGen][X86][AArch64] Add 64-bit constant bit field extract pattern tests
llvm-svn: 343404
Simon Pilgrim [Sun, 30 Sep 2018 09:42:04 +0000 (09:42 +0000)]
[X86] Regenerate MMX coalescing test
Exposes another extractelement(bitcast(scalartovector())) pattern
llvm-svn: 343403
Jonas Hahnfeld [Sun, 30 Sep 2018 09:23:21 +0000 (09:23 +0000)]
[libomptarget-nvptx] Align data sharing stack
NVPTX requires addresses of pointer locations to be 8-byte aligned
or there will be an exception during runtime.
This could happen without this patch as shown in the added test:
getId() requires 4 byte of stack and putValueInParallel() uses 16
bytes to store the addresses of the captured variables.
Differential Revision: https://reviews.llvm.org/D52655
llvm-svn: 343402
Jonas Hahnfeld [Sun, 30 Sep 2018 09:23:14 +0000 (09:23 +0000)]
[libomptarget-nvptx] Fix ancestor_thread_num and team_size (non-SPMD)
According to OpenMP 4.5, p250:12-14:
If the requested nest level is outside the range of 0 and the
nest level of the current thread, as returned by the omp_get_level
routine, the routine returns -1.
The SPMD code path will need a similar fix.
Differential Revision: https://reviews.llvm.org/D51787
llvm-svn: 343401
Zachary Turner [Sun, 30 Sep 2018 03:57:49 +0000 (03:57 +0000)]
[PDB] Fix this test for real.
I was able to test this fix on an actual Windows machine
so this should get the bot green again.
llvm-svn: 343400
Craig Topper [Sun, 30 Sep 2018 03:01:46 +0000 (03:01 +0000)]
[X86] Disable BMI BEXTR in X86DAGToDAGISel::matchBEXTRFromAnd unless we're on compiling for a CPU with single uop BEXTR
Summary:
This function turns (X >> C1) & C2 into a BMI BEXTR or TBM BEXTRI instruction. For BMI BEXTR we have to materialize an immediate into a register to feed to the BEXTR instruction.
The BMI BEXTR instruction is 2 uops on Intel CPUs. It looks like on SKL its one port 0/6 uop and one port 1/5 uop. Despite what Agner's tables say. I know one of the uops is a regular shift uop so it would have to go through the port 0/6 shifter unit. So that's the same or worse execution wise than the shift+and which is one 0/6 uop and one 0/1/5/6 uop. The move immediate into register is an additional 0/1/5/6 uop.
For now I've limited this transform to AMD CPUs which have a single uop BEXTR. If may also might make sense if we can fold a load or if the and immediate is larger than 32-bits and can't be encoded as a sign extended 32-bit value or if LICM or CSE can hoist the move immediate and share it. But we'd need to look more carefully at that. In the regression I looked at it doesn't look load folding or large immediates were occurring so the regression isn't caused by the loss of those. So we could try to be smarter here if we find a compelling case.
Reviewers: RKSimon, spatel, lebedev.ri, andreadb
Reviewed By: RKSimon
Subscribers: llvm-commits, andreadb, RKSimon
Differential Revision: https://reviews.llvm.org/D52570
llvm-svn: 343399