Stella Laurenzo [Sat, 7 Nov 2020 17:35:49 +0000 (09:35 -0800)]
Revert "NFC: Fix some post-review nits for the Tosa dialect."
* Introduced issue in debug builds.
This reverts commit
b5fcd06105dec2a7b0e4114d6ad4524fc54498c5.
Nikita Popov [Sat, 7 Nov 2020 16:48:41 +0000 (17:48 +0100)]
[BasicAA] Use smul_ov helper (NFCI)
Instead of performing the multiplication in double the bit width
and using active bits to determine overflow, use the existing
smul_ov() APInt method to detect overflow.
The smul_ov() implementation is not particularly efficient, but
it's still better than doing this a wide, usually 128-bit, type.
Stella Laurenzo [Sat, 7 Nov 2020 17:05:51 +0000 (09:05 -0800)]
[mlir][Python] Adapt to include property change in pybind11 v2.6.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D90955
Stella Laurenzo [Sat, 7 Nov 2020 16:54:31 +0000 (08:54 -0800)]
NFC: Fix some post-review nits for the Tosa dialect.
* Moved various loose functions to either the mlir::tosa namespace or made static
* Fixed an unused variable warning in TosaMakeBroadcastable.cpp.
Suraj Sudhir [Sat, 7 Nov 2020 16:35:27 +0000 (08:35 -0800)]
TOSA MLIR Dialect
This is the TOSA MLIR Dialect described in the following MLIR RFC: https://llvm.discourse.group/t/rfc-tosa-dialect-in-mlir/1971/24
Reviewed By: stellaraccident
Differential Revision: https://reviews.llvm.org/D90411
Melanie Blower [Sat, 7 Nov 2020 15:40:24 +0000 (07:40 -0800)]
[clang] Fix length threshold for MicrosoftMangle md5 hash
Reviewers: rnk, dblaikie
Differential Revision: https://reviews.llvm.org/D90714
Melanie Blower [Sat, 7 Nov 2020 15:32:34 +0000 (07:32 -0800)]
[clang] Improve Microsoft mangling lit test with dblaikie's suggestions
Nikita Popov [Sat, 7 Nov 2020 11:25:55 +0000 (12:25 +0100)]
[CaptureTracking] Add statistics (NFC)
Add basic statistics on the number of pointers that have been
determined to maybe capture / not capture.
Nikita Popov [Sat, 7 Nov 2020 10:47:31 +0000 (11:47 +0100)]
[CaptureTracking] Early abort on too many uses (NFCI)
If there are too many uses, we should directly return -- there's
no point in inspecting the remaining uses in the worklist, as we
have to conservatively assume a capture anyway. This also means
that tooManyUses() gets called exactly once, rather than
potentially many times.
This restores the behavior prior to
e9832dfdf366ddffba68164adb6855d17c9f87c1,
where this was accidentally changed while moving the AddUses logic
into a closure, thus making the return a return from the closure
rather than the whole function.
Nikita Popov [Sat, 7 Nov 2020 10:44:02 +0000 (11:44 +0100)]
[CaptureTrackingTest] Add missing override marker (NFC)
Nikita Popov [Sat, 7 Nov 2020 10:19:27 +0000 (11:19 +0100)]
[CaptureTracking] Correctly handle multiple uses in one instruction
If the same value is used multiple times in the same instruction,
CaptureTracking may end up reporting the wrong use as being captured,
and/or report the same use as being captured multiple times.
Make sure that all checks take the use operand number into account,
rather than performing unreliable comparisons against the used value.
I'm not sure whether this can cause any problems in practice, but
at least some capture trackers (ArgUsesTracker, AACaptureUseTracker)
do care about which call argument is captured.
George Mitenkov [Sat, 7 Nov 2020 09:27:03 +0000 (12:27 +0300)]
[MLIR][SPIRVToLLVM] Added module name conversion
Since SPIR-V module has an optional name, this patch
makes a change to pass it to `ModuleOp` during conversion.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D90904
Nikita Popov [Sat, 7 Nov 2020 09:13:49 +0000 (10:13 +0100)]
[CaptureTracking] Avoid duplicate shouldExplode() check (NFCI)
We check shouldExplore() before adding uses to the worklist, so
uses that should not be explored will not reach captured() in the
first place.
Jonas Devlieghere [Sat, 7 Nov 2020 07:22:29 +0000 (23:22 -0800)]
[DWARFLinker] Use union to reduce sizeof(WorklistItem) (NFC)
Reduce the size of the WorklistItem struct by using a struct.
Kazu Hirata [Sat, 7 Nov 2020 06:47:22 +0000 (22:47 -0800)]
[BranchProbabilityInfo] Simplify getEdgeProbability (NFC)
The patch simplifies BranchProbabilityInfo::getEdgeProbability by
handling two cases separately, depending on whether we have edge
probabilities.
- If we have edge probabilities, then add up probabilities for
successors being equal to Dst.
- Otherwise, return the number of ocurrences divided by the total
number of successors.
Differential Revision: https://reviews.llvm.org/D90980
Fangrui Song [Sat, 7 Nov 2020 06:19:37 +0000 (22:19 -0800)]
[test] Fix Other/new-pass-manager.ll with has different behaviors whether or not Polly is enabled
after D89158
Fangrui Song [Sat, 7 Nov 2020 05:55:10 +0000 (21:55 -0800)]
[test] Fix Other/new-pass-manager.ll & clang/test/Misc/loop-opt-setup.c
Atmn Patel [Sat, 7 Nov 2020 05:31:35 +0000 (00:31 -0500)]
Revert "[LoopDeletion] Allows deletion of possibly infinite side-effect free loops"
This reverts commit
0b17c6e4479d62bd4ff05c48d6cdf340b198832f. This patch
causes a compile-time error in SCEV.
Atmn Patel [Sat, 7 Nov 2020 05:29:42 +0000 (00:29 -0500)]
Revert "[CodeGen] Fixes sanitizer test"
This reverts commit
b1878b4641e06baa754ce6e3e0387b1e7d7dc143. This does
fix the test but it means that
ac73b73c1652 is not implemented
correctly. Reverting for now, and will be reverting the commit that
causes this to fail.
Fangrui Song [Sat, 7 Nov 2020 05:00:28 +0000 (21:00 -0800)]
AsmPrinter/Dwarf*: Use llvm::Register instead of unsigned
Atmn Patel [Sat, 7 Nov 2020 04:50:55 +0000 (23:50 -0500)]
[CodeGen] Fixes sanitizer test
By turning the loop into an infinite one, the loop can't be deleted
anymore so the test will continue to pass.
Fangrui Song [Sat, 7 Nov 2020 04:09:43 +0000 (20:09 -0800)]
[OpenMP] Fix -Wmisleading-indentation after D84192
Fangrui Song [Sat, 7 Nov 2020 04:02:56 +0000 (20:02 -0800)]
[AsmPrinter] Rename ByteStreamer::EmitInt8 to emitInt8
to be consistent with other emit*
Atmn Patel [Sat, 7 Nov 2020 03:36:47 +0000 (22:36 -0500)]
[LoopDeletion] Fixes failing test
The commit
0b17c6e4479d62bd4ff05c48d6cdf340b198832f occasionally causes this test to
fail, this fixes it.
Jonas Devlieghere [Sat, 7 Nov 2020 03:35:06 +0000 (19:35 -0800)]
[DWARFLinker] Add CompileUnit::getInfo helper that takes a DWARFDie (NFC)
Eliminate the need to go through the DIE index by passing the DIE to
CompileUnit::getInfo directly.
Before:
unsigned Idx = Unit->getOrigUnit().getDIEIndex(Die);
CompileUnit::DIEInfo &Info = Unit->getInfo(Idx);
After:
CompileUnit::DIEInfo &Info = Unit->getInfo(Die);
Atmn Patel [Sun, 25 Oct 2020 22:24:48 +0000 (18:24 -0400)]
[LoopDeletion] Allows deletion of possibly infinite side-effect free loops
From C11 and C++11 onwards, a forward-progress requirement has been
introduced for both languages. In the case of C, loops with non-constant
conditionals that do not have any observable side-effects (as defined by
6.8.5p6) can be assumed by the implementation to terminate, and in the
case of C++, this assumption extends to all functions. The clang
frontend will emit the `mustprogress` function attribute for C++
functions (D86233, D85393, D86841) and emit the loop metadata
`llvm.loop.mustprogress` for every loop in C11 or later that has a
non-constant conditional.
This patch modifies LoopDeletion so that only loops with
the `llvm.loop.mustprogress` metadata or loops contained in functions
that are required to make progress (`mustprogress` or `willreturn`) are
checked for observable side-effects. If these loops do not have an
observable side-effect, then we delete them.
Loops without observable side-effects that do not satisfy the above
conditions will not be deleted.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86844
cchen [Sat, 7 Nov 2020 03:03:55 +0000 (21:03 -0600)]
[OpenMP5.0] map item can be non-contiguous for target update
In order not to modify the `tgt_target_data_update` information but still be
able to pass the extra information for non-contiguous map item (offset,
count, and stride for each dimension), this patch overload `arg` when
the maptype is set as `OMP_MAP_DESCRIPTOR`. The origin `arg` is for
passing the pointer information, however, the overloaded `arg` is an
array of descriptor_dim:
struct descriptor_dim {
int64_t offset;
int64_t count;
int64_t stride
};
and the array size is the same as dimension size. In addition, since we
have count and stride information in descriptor_dim, we can replace/overload the
`arg_size` parameter by using dimension size.
For supporting `stride` in array section, we use a dummy dimension in
descriptor to store the unit size. The formula for counting the stride
in dimension D_n: `unit size * (D_0 * D_1 ... * D_n-1) * D_n.stride`.
Demonstrate how it works:
```
double arr[3][4][5];
D0: { offset = 0, count = 1, stride = 8 } // offset, count, dimension size always be 0, 1, 1 for this extra dimension, stride is the unit size
D1: { offset = 0, count = 2, stride = 8 * 1 * 2 = 16 } // stride = unit size * (product of dimension size of D0) * D1.stride = 4 * 1 * 2 = 8
D2: { offset = 2, count = 2, stride = 8 * (1 * 5) * 1 = 40 } // stride = unit size * (product of dimension size of D0, D1) * D2.stride = 4 * 5 * 1 = 20
D3: { offset = 0, count = 2, stride = 8 * (1 * 5 * 4) * 2 = 320 } // stride = unit size * (product of dimension size of D0, D1, D2) * D3.stride = 4 * 25 * 2 = 200
// X here means we need to offload this data, therefore, runtime will transfer
// data from offset 80, 96, 120, 136, 400, 416, 440, 456
// Runtime patch: https://reviews.llvm.org/D82245
// OOOOO OOOOO OOOOO
// OOOOO OOOOO OOOOO
// XOXOO OOOOO XOXOO
// XOXOO OOOOO XOXOO
```
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D84192
cchen [Sat, 7 Nov 2020 02:54:49 +0000 (20:54 -0600)]
[libomptarget] Add support for target update non-contiguous
This patch is the runtime support for https://reviews.llvm.org/D84192.
In order not to modify the tgt_target_data_update information but still be
able to pass the extra information for non-contiguous map item (offset,
count, and stride for each dimension), this patch overload arg when
the maptype is set as OMP_TGT_MAPTYPE_DESCRIPTOR. The origin arg is for
passing the pointer information, however, the overloaded arg is an
array of descriptor_dim:
```
struct descriptor_dim {
int64_t offset;
int64_t count;
int64_t stride
};
```
and the array size is the dimension size. In addition, since we
have count and stride information in descriptor_dim, we can replace/overload the
arg_size parameter by using dimension size.
Reviewed By: grokos
Differential Revision: https://reviews.llvm.org/D82245
Mehdi Amini [Sat, 7 Nov 2020 01:34:43 +0000 (01:34 +0000)]
Don't link any LLVM/MLIR library to the C API unit-test
The tests are intended to exercise the public C API and will link to a
specific shared library exposing only the C API, this library itself may
link to libMLIR.so.
If we link some LLVM library statically in the test themselves, we end
up with duplicated cl::opt registrations in LLVM. A possible setup if
these libraries were needed could be to link libMLIR.so directly when
available and link statically when it isn't available (in which case the
libary exposing the C API would be statically link and isolated from the
cl::opt registry, hopefully).
Differential Revision: https://reviews.llvm.org/D90993
Atmn Patel [Fri, 6 Nov 2020 05:10:23 +0000 (00:10 -0500)]
[Inliner] Handle `mustprogress` functions
When inlining `mustprogress` functions, if the caller or the callee has
the attribute, we drop the function attribute. The loops that have the
`llvm.loop.mustprogress` metadata keep their metadata. We do not need to
add new loop metadata to inlined functions because the patch in D86841
already adds the relevant loop metadata in all of the necessary places.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D87262
Fangrui Song [Sat, 7 Nov 2020 00:49:51 +0000 (16:49 -0800)]
[test] -mtriple=x86_64-* -> -mtriple=x86_64
Stella Laurenzo [Sat, 7 Nov 2020 00:47:26 +0000 (16:47 -0800)]
[mlir][Python] Fix cmake typo keeping the extension from installing.
Siva Chandra Reddy [Fri, 6 Nov 2020 23:43:45 +0000 (15:43 -0800)]
[libc] Extend WrapperGen to emit aliases.
Reviewed By: michaelrj
Differential Revision: https://reviews.llvm.org/D90985
Zequan Wu [Fri, 6 Nov 2020 22:53:57 +0000 (14:53 -0800)]
[llvm-cov] Fix missing slash in -path-equivalence
Elvina Yakubova [Fri, 6 Nov 2020 22:50:43 +0000 (01:50 +0300)]
[AArch64] Add driver tests for HiSilicon's TSV110
Sean Silva [Thu, 5 Nov 2020 02:29:08 +0000 (18:29 -0800)]
[mlir][Linalg] Canonicalize duplicate args.
I ran into this pattern when converting elementwise ops like
`addf %arg0, %arg : tensor<?xf32>` to linalg. Redundant arguments can
also easily arise from linalg-fusion-for-tensor-ops.
Also, fix some small bugs in the logic in
LinalgStructuredOpsInterface.td.
Differential Revision: https://reviews.llvm.org/D90812
Louis Dionne [Fri, 6 Nov 2020 22:26:42 +0000 (17:26 -0500)]
Revert "Allow running back-deployment testing against libc++abi"
This reverts commit
4d79ef814aedb63, which broke a few build bots.
I'm reverting until I have time to investigate.
Elvina Yakubova [Fri, 6 Nov 2020 22:22:35 +0000 (01:22 +0300)]
[AArch64] Add pipeline model for HiSilicon's TSV110
This patch adds the scheduling and cost model for TSV110.
Reviewed by: SjoerdMeijer, bryanpkc
Differential Revision: https://reviews.llvm.org/D89972
Kazu Hirata [Fri, 6 Nov 2020 22:18:00 +0000 (14:18 -0800)]
[TableGen] Use llvm::is_contained (NFC)
Eric Astor [Fri, 6 Nov 2020 20:18:15 +0000 (15:18 -0500)]
[ms] [llvm-ml] Allow arbitrary strings as integer constants
MASM interprets strings in expression contexts as integers expressed in big-endian base-256, treating each character as its ASCII representation.
This completely eliminates the need to special-case single-character strings.
Reviewed By: thakis
Differential Revision: https://reviews.llvm.org/D90788
Atmn Patel [Sun, 25 Oct 2020 22:30:09 +0000 (18:30 -0400)]
[LoopDeletion] Remove dead loops with no exit blocks
Currently, LoopDeletion refuses to remove dead loops with no exit blocks
because it cannot statically determine the control flow after it removes
the block. This leads to miscompiles if the loop is an infinite loop and
should've been removed.
Differential Revision: https://reviews.llvm.org/D90115
Alexander Shaposhnikov [Fri, 6 Nov 2020 21:10:41 +0000 (13:10 -0800)]
[llvm-objcopy][MachO] Skip sections with zero offset
Some binaries can contain regular sections with zero offset and zero size.
This diff makes llvm-objcopy's handling of such sections consistent with
cctools's strip (which doesn't modify them),
previously the tool would allocate file space for them.
Test plan: make check-all
Differential revision: https://reviews.llvm.org/D90796
Jonas Devlieghere [Fri, 6 Nov 2020 21:04:55 +0000 (13:04 -0800)]
[debuginfo-tests] Skip optnone-loops.cpp on Darwin
This is failing on GreenDragon:
http://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake/24745/
Stephen Kelly [Sat, 31 Oct 2020 00:41:16 +0000 (00:41 +0000)]
Add new matchers for dependent names in templates
Differential Revision: https://reviews.llvm.org/D90767
Mehdi Amini [Fri, 6 Nov 2020 20:46:09 +0000 (20:46 +0000)]
Fix gcc warning by removing extra `;` after a macro (NFC)
Rahman Lavaee [Fri, 6 Nov 2020 20:44:24 +0000 (12:44 -0800)]
[obj2yaml] [yaml2obj] Add yaml support for SHT_LLVM_BB_ADDR_MAP section.
YAML support allows us to better test the feature in the subsequent patches. The implementation is quite similar to the .stack_sizes section.
Reviewed By: jhenderson, grimar
Differential Revision: https://reviews.llvm.org/D88717
Nathan James [Fri, 6 Nov 2020 19:58:19 +0000 (19:58 +0000)]
[clangd] Set the User option for clang-tidy to mimick its behaviour
Probably not essential as afaik only one check uses this field. but still good to have consistent behaviour.
Reviewed By: sammccall
Differential Revision: https://reviews.llvm.org/D90552
Valentin Churavy [Fri, 6 Nov 2020 19:26:04 +0000 (14:26 -0500)]
Fix unwind info relocation with large code model on AArch64
Makes sure that the unwind info uses 64bits pcrel relocation if a large code model is specified and handle the corresponding relocation in the ExecutionEngine. This can happen with certain kernel configuration (the same as the one in https://reviews.llvm.org/D27609, found at least on the ArchLinux stock kernel and the one used on https://www.packet.net/) using the builtin JIT memory manager.
Co-authored-by: Yichao Yu <yyc1992@gmail.com>
Differential Revision: https://reviews.llvm.org/D27629
Siva Chandra Reddy [Fri, 6 Nov 2020 01:07:17 +0000 (17:07 -0800)]
[libc][WrapperGen] Replace the C _Noreturn annotation with C++ [[noreturn]].
Reviewed By: michaelrj
Differential Revision: https://reviews.llvm.org/D90900
Valentin Churavy [Fri, 6 Nov 2020 19:08:30 +0000 (14:08 -0500)]
[RTDYLD] support absolute relocations where needed
These appear in some sections, such as DWARF tables, since
RuntimeDyldELF explicitly maps to this as a sentinel value:
https://github.com/llvm/llvm-project/blob/
29d1fba7b5335d969e3e5daa84b7a25cd1fa75ef/llvm/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp#L1199
That could then be a source of problems if it tried to examine these
sections (for example, with either setProcessAllSections(true) or ORCv2 on i686).
Replaces https://reviews.llvm.org/D89241
Reviewed By: lhames, vchuravy
Differential Revision: https://reviews.llvm.org/D90722
Kazu Hirata [Fri, 6 Nov 2020 19:05:35 +0000 (11:05 -0800)]
[BranchProbabilityInfo] Use succ_size (NFC)
Craig Topper [Fri, 6 Nov 2020 18:29:22 +0000 (10:29 -0800)]
[RISCV] Add test case to show incorrect matching to sroiw when the or mask does not have 1s in the upper 32 bits.
The matching code for sroiw is truncating the mask to 32 bits before
checking its value. We need to check all 64 bits.
Quentin Colombet [Fri, 6 Nov 2020 18:25:16 +0000 (10:25 -0800)]
Prevent LICM and machineLICM from hoisting convergent operations
Results of convergent operations are implicitly affected by the
enclosing control flows and should not be hoisted out of arbitrary
loops.
Patch by Xiaoqing Wu <xiaoqing_wu@apple.com>
Differential Revision: https://reviews.llvm.org/D90361
Richard Howell [Fri, 6 Nov 2020 18:04:28 +0000 (10:04 -0800)]
[lldb] add a missing dependency on intrinsics_gen
Sometimes builds will fail with errors like:
```
In file included from /build/external/llvm-project/lldb/source/Symbol/SwiftASTContext.cpp:52:
In file included from /build/external/swift/include/swift/IRGen/Linking.h:22:
In file included from /build/external/swift/include/swift/SIL/SILFunction.h:24:
In file included from /build/external/swift/include/swift/SIL/SILBasicBlock.h:23:
In file included from /build/external/swift/include/swift/SIL/SILInstruction.h:21:
In file included from /build/external/swift/include/swift/AST/Builtins.h:24:
**/build/external/llvm-project/llvm/include/llvm/IR/Attributes.h:74:14: fatal error: 'llvm/IR/Attributes.inc' file not found**
**^~~~~~~~~~~~~~~~~~~~~~~~**
```
This change ensures the `Attributes.inc` file is generated before building `SwiftASTContext.cpp`.
Differential Revision: https://reviews.llvm.org/D90857
Alex Zinenko [Fri, 6 Nov 2020 10:00:13 +0000 (11:00 +0100)]
[mlir] Use PyValue instead of PyOpResult in Python operand container
The PyOpOperands container was erroneously constructing objects for
individual operands as PyOpResult. Operands in fact are just values,
which may or may not be results of another operation. The code would
eventually crash if the operand was a block argument. Add a test that
exercises the behavior that previously led to crashes.
Reviewed By: stellaraccident
Differential Revision: https://reviews.llvm.org/D90917
Alex Zinenko [Fri, 6 Nov 2020 17:59:58 +0000 (18:59 +0100)]
Revert "[mlir][CAPI] Proposal: Always building a libMLIRPublicAPI.so."
This reverts commit
80fe2f61fac9f72825f907637a9c63aca32d8f14.
Broke linkage with GNU ld. See original review thread for more details.
Nawrin Sultana [Fri, 6 Nov 2020 16:47:33 +0000 (10:47 -0600)]
[OpenMP] Fix potential division by 0
This patch fixes potential division by 0 in case hwloc does not
recognize cores (or architecture has no cores).
Patch by Andrey Churbanov
Differential Revision: https://reviews.llvm.org/D90954
Kostya Kortchinsky [Tue, 3 Nov 2020 19:21:15 +0000 (11:21 -0800)]
[scudo][standalone] Simplify populateFreelist
`populateFreelist` was more complicated that it needed to be. We used
to call to `populateBatches` that would do some internal shuffling and
add pointers one by one to the batches, but ultimately this was not
needed. We can get rid of `populateBatches`, and do processing in
bulk. This doesn't necessarily make things faster as this is not on the
hot path, but it makes the function cleaner.
Additionally clean up a couple of items, like `UNLIKELY`s and setting
`Exhausted` to `false` which can't happen.
Differential Revision: https://reviews.llvm.org/D90700
Simon Pilgrim [Fri, 6 Nov 2020 17:26:56 +0000 (17:26 +0000)]
[InstCombine] computeKnownBitsMul - use KnownBits::isNonZero() helper.
Avoid an expensive isKnownNonZero() call - this is a small cleanup before moving the extra NSW functionality from computeKnownBitsMul into KnownBits::computeForMul.
Simon Pilgrim [Fri, 6 Nov 2020 17:22:28 +0000 (17:22 +0000)]
[SLP][AMDGPU] Regenerate packed-math tests and remove unused check prefix
Simon Pilgrim [Fri, 6 Nov 2020 17:14:10 +0000 (17:14 +0000)]
[VectorCombine][X86] Removed unused check prefixes
Stella Laurenzo [Fri, 6 Nov 2020 06:37:47 +0000 (22:37 -0800)]
[mlir][CAPI] Proposal: Always building a libMLIRPublicAPI.so.
We were discussing on discord regarding the need for extension-based systems like Python to dynamically link against MLIR (or else you can only have one extension that depends on it). Currently, when I set that up, I piggy-backed off of the flag that enables build libLLVM.so and libMLIR.so and depended on libMLIR.so from the python extension if shared library building was enabled. However, this is less than ideal.
In the current setup, libMLIR.so exports both all symbols from the C++ API and the C-API. The former is a kitchen sink and the latter is curated. We should be splitting them and for things that are properly factored to depend on the C-API, they should have the option to *only* depend on the C-API, and we should build that shared library no matter what. Its presence isn't just an optimization: it is a key part of the system.
To do this right, I needed to:
* Introduce visibility macros into mlir-c/Support.h. These should work on both *nix and windows as-is.
* Create a new libMLIRPublicAPI.so with just the mlir-c object files.
* Compile the C-API with -fvisibility=hidden.
* Conditionally depend on the libMLIR.so from libMLIRPublicAPI.so if building libMLIR.so (otherwise, also links against the static libs and will produce a mondo libMLIRPublicAPI.so).
* Disable re-exporting of static library symbols that come in as transitive deps.
This gives us a dynamic linked C-API layer that is minimal and should work as-is on all platforms. Since we don't support libMLIR.so building on Windows yet (and it is not very DLL friendly), this will fall back to a mondo build of libMLIRPublicAPI.so, which has its uses (it is also the most size conscious way to go if you happen to know exactly what you need).
Sizes (release/stripped, Ubuntu 20.04):
Shared library build:
libMLIRPublicAPI.so: 121Kb
_mlir.cpython-38-x86_64-linux-gnu.so: 1.4Mb
mlir-capi-ir-test: 135Kb
libMLIR.so: 21Mb
Static build:
libMLIRPublicAPI.so: 5.5Mb (since this is a "static" build, this includes the MLIR implementation as non-exported code).
_mlir.cpython-38-x86_64-linux-gnu.so: 1.4Mb
mlir-capi-ir-test: 44Kb
Things like npcomp and circt which bring their own dialects/transforms/etc would still need the shared library build and code that links against libMLIR.so (since it is all C++ interop stuff), but hopefully things that only depend on the public C-API can just have the one narrow dep.
I spot checked everything with nm, and it looks good in terms of what is exporting/importing from each layer.
I'm not in a hurry to land this, but if it is controversial, I'll probably split off the Support.h and API visibility macro changes, since we should set that pattern regardless.
Reviewed By: mehdi_amini, benvanik
Differential Revision: https://reviews.llvm.org/D90824
Kevin P. Neal [Fri, 6 Nov 2020 15:44:13 +0000 (10:44 -0500)]
[FPEnv] Use strictfp metadata in casting nodes
The strictfp metadata was added to the casting AST nodes in D85960, but
we aren't using that metadata yet. This patch adds that support.
In order to avoid lots of ad-hoc passing around of the strictfp bits I
updated the IRBuilder when moving from a function that has the Expr* to a
function that lacks it. I believe we should switch to this pattern to keep
the strictfp support from being overly invasive.
For the purpose of testing that we're picking up the right metadata, I
also made my tests use a pragma to make the AST's strictfp metadata not
match the global strictfp metadata. This exposes issues that we need to
deal with in subsequent patches, and I believe this is the right method
for most all of our clang strictfp tests.
Differential Revision: https://reviews.llvm.org/D88913
Jay Foad [Fri, 6 Nov 2020 14:19:59 +0000 (14:19 +0000)]
[TableGen] Indentation and whitespace fixes in generated code. NFC.
Some of these were found by running clang-format over the generated
code, although that complains about far more issues than I have fixed
here.
Differential Revision: https://reviews.llvm.org/D90937
Jay Foad [Fri, 6 Nov 2020 15:25:43 +0000 (15:25 +0000)]
[AMDGPU] Simplify exp target parsing
Treat any identifier as a potential exp target and diagnose them all the
same way as "invalid exp target"s.
Differential Revision: https://reviews.llvm.org/D90947
Balazs Benics [Fri, 6 Nov 2020 15:51:34 +0000 (16:51 +0100)]
[analyzer][NFCi] Mark CallEvent::getOriginExpr virtual, some cleanup
Previously we just shadowed the original implementation with a virtual
declaration - which is really bugprone in a long run.
This patch marks `CallEvent::getOriginExpr` virtual to let subclasses
override it's behavior.
At the same time, I checked all virtual functions of this class hierarchy
to make sure we don't suffer from this elsewhere.
Removes redundant declarations of `virtual` if `override` is already present.
In theory, this patch is a functional change, but no tests were broken.
I suspect that there were no meaningful changes in behavior in the
subclasses compared to the shadowed `CallEvent::getOriginExpr`.
That being said, I had a hard time coming up with unit-tests covering this.
Motivation: https://reviews.llvm.org/D74735#2370909
Reviewed By: NoQ
Differential Revision: https://reviews.llvm.org/D90754
Stephen Kelly [Mon, 2 Nov 2020 23:56:32 +0000 (23:56 +0000)]
Traverse-ignore explicit template instantiations
Continue to dump and match on explicit template specializations, but
omit explicit instantiation declarations and definitions.
Differential Revision: https://reviews.llvm.org/D90763
David Spickett [Fri, 6 Nov 2020 10:20:20 +0000 (10:20 +0000)]
[Arm][MC] Remove unused prefixes in .arch_extension fp tests
idiv: There is no difference between Armv7m and Thumbv7M
behaviour so the specific CHECKs are not needed.
The errors for Armv7-a and Thumbv7-a will always
include "ARM" or "THUMB" respectively so they need their
own CHECK prefix, making CHECK-V7 redundant.
mp: Behaviour is dependent on whether the triple is v6/v7/v7M
regardless of being Arm or Thumb. So we don't need the more
specific CHECK-ARMv7M etc.
simd: Errors are either v7 only, or v7 and v8 so CHECK-V8
is not needed.
fp: Same as simd
Reviewed By: ostannard
Differential Revision: https://reviews.llvm.org/D90918
Louis Dionne [Fri, 6 Nov 2020 15:06:44 +0000 (10:06 -0500)]
[libc++] Try fixing the oss-fuzz build
See https://github.com/google/oss-fuzz/issues/4586.
Jay Foad [Fri, 6 Nov 2020 14:57:51 +0000 (14:57 +0000)]
[AMDGPU] Run exp tests on GFX9 and GFX10 too. NFC.
Roman Lebedev [Fri, 6 Nov 2020 14:35:17 +0000 (17:35 +0300)]
[NFC][InstCombine] Update few comment updates i missed in
0ac56e8eaaeb
As pointed out in post-commit review in that commit
Arnold Schwaighofer [Mon, 2 Nov 2020 17:37:15 +0000 (09:37 -0800)]
llvm.coro.id.async lowering: Parameterize how-to restore the current's continutation context and restart the pipeline after splitting
The `llvm.coro.suspend.async` intrinsic takes a function pointer as its
argument that describes how-to restore the current continuation's
context from the context argument of the continuation function. Before
we assumed that the current context can be restored by loading from the
context arguments first pointer field (`first_arg->caller_context`).
This allows for defining suspension points that reuse the current
context for example.
Also:
llvm.coro.id.async lowering: Add llvm.coro.preprare.async intrinsic
Blocks inlining until after the async coroutine was split.
Also, change the async function pointer's context size position
struct async_function_pointer {
uint32_t relative_function_pointer_to_async_impl;
uint32_t context_size;
}
And make the position of the `async context` argument configurable. The
position is specified by the `llvm.coro.id.async` intrinsic.
rdar://
70097093
Differential Revision: https://reviews.llvm.org/D90783
Paul C. Anagnostopoulos [Thu, 5 Nov 2020 14:53:19 +0000 (09:53 -0500)]
[NVPTX] [TableGen] Use new features of TableGen to simplify and clarify.
Differential Revision: https://reviews.llvm.org/D90861
Simon Moll [Fri, 6 Nov 2020 14:09:32 +0000 (15:09 +0100)]
[VE] Add v(m)regs to preserve_all reg mask
V(m)regs where defined before CSR_preserve_all was, add them now.
Reviewed By: kaz7
Differential Revision: https://reviews.llvm.org/D90912
David Spickett [Fri, 6 Nov 2020 10:59:16 +0000 (10:59 +0000)]
[Clang][AArch64] Remove unused prefix in constrained rounding test
This test was added in
7f38812d5b7e0fd5145e44d5ae831a67d782dca6
and all the other tests make use of the COMMONIR check. So I think
this was left in by mistake for this particular test.
Reviewed By: kpn
Differential Revision: https://reviews.llvm.org/D90921
Than McIntosh [Thu, 5 Nov 2020 14:09:44 +0000 (09:09 -0500)]
[NFC] Fix typo in comment.
Differential Revision: https://reviews.llvm.org/D90846
Paul C. Anagnostopoulos [Thu, 5 Nov 2020 21:44:15 +0000 (16:44 -0500)]
[TableGen] Clarify text and fix errors in the Programmer's Reference
Differential Revision: https://reviews.llvm.org/D90881
Michał Górny [Thu, 5 Nov 2020 20:24:56 +0000 (21:24 +0100)]
[lldb] [test] Avoid double negation in llgs/debugserver logic
Use positive logic (i.e. llgs_platform/debugserver_platform) for
indicating which platforms use the particular server variant.
Deduplicate the lists — it is rather expected that none of the platforms
using LLGS would use debugserver.
Differential Revision: https://reviews.llvm.org/D90875
Simon Moll [Fri, 6 Nov 2020 13:25:13 +0000 (14:25 +0100)]
[VE][NFC] Refactor to support more than one calling conv
Prepare for supporting different calling conventions by factoring out
things into CC-dependent selection functions (getParamCC, getReturnCC).
Reviewed By: kaz7
Differential Revision: https://reviews.llvm.org/D90911
Louis Dionne [Fri, 6 Nov 2020 13:17:02 +0000 (08:17 -0500)]
[debuginfo-tests] NFC: Move test that was committed to the wrong location
Differential Revision: https://reviews.llvm.org/D90931
Louis Dionne [Wed, 8 Jul 2020 20:38:54 +0000 (16:38 -0400)]
Allow running back-deployment testing against libc++abi
Summary:
Before this patch, we could only link against the back-deployment libc++abi
dylib. This patch allows linking against the just-built libc++abi, but
running against the back-deployment one -- just like we do for libc++.
Also, add XFAIL markup to flag expected errors.
Florian Hahn [Fri, 6 Nov 2020 09:39:55 +0000 (09:39 +0000)]
[SLP] Also try to vectorize incoming values of PHIs .
Currently we do not consider incoming values of PHIs as roots for SLP
vectorization. This means we miss scenarios like the one in the test
case and PR47670.
It appears quite straight-forward to consider incoming values of PHIs as
roots for vectorization, but I might be missing something that makes
this problematic.
In terms of vectorized instructions, this applies to quite a few
benchmarks across MultiSource/SPEC2000/SPEC2006 on X86 with -O3 -flto
Same hash: 185 (filtered out)
Remaining: 52
Metric: SLP.NumVectorInstructions
Program base patch diff
test-suite...ProxyApps-C++/HPCCG/HPCCG.test 9.00 27.00 200.0%
test-suite...C/CFP2000/179.art/179.art.test 8.00 22.00 175.0%
test-suite...T2006/458.sjeng/458.sjeng.test 14.00 30.00 114.3%
test-suite...ce/Benchmarks/PAQ8p/paq8p.test 11.00 18.00 63.6%
test-suite...s/FreeBench/neural/neural.test 12.00 18.00 50.0%
test-suite...rimaran/enc-3des/enc-3des.test 65.00 95.00 46.2%
test-suite...006/450.soplex/450.soplex.test 63.00 89.00 41.3%
test-suite...ProxyApps-C++/CLAMR/CLAMR.test 177.00 250.00 41.2%
test-suite...nchmarks/McCat/18-imp/imp.test 13.00 18.00 38.5%
test-suite.../Applications/sgefa/sgefa.test 26.00 35.00 34.6%
test-suite...pplications/oggenc/oggenc.test 100.00 133.00 33.0%
test-suite...6/482.sphinx3/482.sphinx3.test 103.00 134.00 30.1%
test-suite...oxyApps-C++/miniFE/miniFE.test 169.00 213.00 26.0%
test-suite.../Benchmarks/Olden/tsp/tsp.test 59.00 73.00 23.7%
test-suite...TimberWolfMC/timberwolfmc.test 503.00 622.00 23.7%
test-suite...T2006/456.hmmer/456.hmmer.test 65.00 79.00 21.5%
test-suite...libquantum/462.libquantum.test 58.00 68.00 17.2%
test-suite...ternal/HMMER/hmmcalibrate.test 84.00 98.00 16.7%
test-suite...ications/JM/ldecod/ldecod.test 351.00 401.00 14.2%
test-suite...arks/VersaBench/dbms/dbms.test 52.00 57.00 9.6%
test-suite...ce/Benchmarks/Olden/bh/bh.test 118.00 128.00 8.5%
test-suite.../Benchmarks/Bullet/bullet.test 6355.00 6880.00 8.3%
test-suite...nsumer-lame/consumer-lame.test 480.00 519.00 8.1%
test-suite...000/183.equake/183.equake.test 226.00 244.00 8.0%
test-suite...chmarks/Olden/power/power.test 105.00 113.00 7.6%
test-suite...6/471.omnetpp/471.omnetpp.test 92.00 99.00 7.6%
test-suite...ications/JM/lencod/lencod.test 1173.00 1261.00 7.5%
test-suite...0/253.perlbmk/253.perlbmk.test 55.00 59.00 7.3%
test-suite...oxyApps-C/miniAMR/miniAMR.test 92.00 98.00 6.5%
test-suite...chmarks/MallocBench/gs/gs.test 446.00 473.00 6.1%
test-suite.../CINT2006/403.gcc/403.gcc.test 464.00 491.00 5.8%
test-suite...6/464.h264ref/464.h264ref.test 998.00 1055.00 5.7%
test-suite...006/453.povray/453.povray.test 5711.00 6007.00 5.2%
test-suite...FreeBench/distray/distray.test 102.00 107.00 4.9%
test-suite...:: External/Povray/povray.test 4184.00 4378.00 4.6%
test-suite...DOE-ProxyApps-C/CoMD/CoMD.test 112.00 117.00 4.5%
test-suite...T2006/445.gobmk/445.gobmk.test 104.00 108.00 3.8%
test-suite...CI_Purple/SMG2000/smg2000.test 789.00 819.00 3.8%
test-suite...yApps-C++/PENNANT/PENNANT.test 233.00 241.00 3.4%
test-suite...marks/7zip/7zip-benchmark.test 417.00 428.00 2.6%
test-suite...arks/mafft/pairlocalalign.test 627.00 643.00 2.6%
test-suite.../Benchmarks/nbench/nbench.test 259.00 265.00 2.3%
test-suite...006/447.dealII/447.dealII.test 4641.00 4732.00 2.0%
test-suite...lications/ClamAV/clamscan.test 106.00 108.00 1.9%
test-suite...CFP2000/177.mesa/177.mesa.test 1639.00 1664.00 1.5%
test-suite...oxyApps-C/RSBench/rsbench.test 66.00 65.00 -1.5%
test-suite.../CINT2000/252.eon/252.eon.test 3416.00 3444.00 0.8%
test-suite...CFP2000/188.ammp/188.ammp.test 1846.00 1861.00 0.8%
test-suite.../CINT2000/176.gcc/176.gcc.test 152.00 153.00 0.7%
test-suite...CFP2006/444.namd/444.namd.test 3528.00 3544.00 0.5%
test-suite...T2006/473.astar/473.astar.test 98.00 98.00 0.0%
test-suite...frame_layout/frame_layout.test NaN 39.00 nan%
On ARM64, there appears to be a slight regression on SPEC2006, which
might be interesting to investigate:
test-suite...T2006/473.astar/473.astar.test 0.9%
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D88735
David Zarzycki [Fri, 6 Nov 2020 12:13:55 +0000 (07:13 -0500)]
[lld testing] Unbreak read-only source builds
Tests must not modify the source tree.
Simon Pilgrim [Fri, 6 Nov 2020 11:35:36 +0000 (11:35 +0000)]
[InstCombine] Regenerate narrow-math.ll tests
David Spickett [Fri, 6 Nov 2020 09:54:43 +0000 (09:54 +0000)]
[AArch64][MC] Remove unused CHECK-ERROR in SVE test file
This file is only ever looking for errors so we can
just use the default CHECK.
Reviewed By: sdesmalen
Differential Revision: https://reviews.llvm.org/D90915
David Spickett [Fri, 6 Nov 2020 09:45:16 +0000 (09:45 +0000)]
[AArch64][MC] Remove unused prefix in v8.4-a trace test
It was unused when added and the CHECK-ERROR lines
cover the possible outputs.
Reviewed By: ostannard
Differential Revision: https://reviews.llvm.org/D90913
Kazushi (Jam) Marukawa [Sun, 1 Nov 2020 01:59:28 +0000 (10:59 +0900)]
[VE] Optimize address calculation
Optimize address calculations using LEA/LEASL instructions.
Update comments in VEISelLowering.cpp also. Update an
existing regression test optimized by this modification.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D90878
Simon Moll [Fri, 6 Nov 2020 10:11:11 +0000 (11:11 +0100)]
[VE][TTI] don't advertise vregs/vops
Claim to not have any vector support to dissuade SLP, LV and friends
from generating SIMD IR for the VE target. We will take this back once
vector isel is stable.
Reviewed By: kaz7, fhahn
Differential Revision: https://reviews.llvm.org/D90462
Simon Pilgrim [Thu, 5 Nov 2020 18:33:14 +0000 (18:33 +0000)]
[X86] Regenerate zext-load tests and add 32-bit test coverage.
Sander de Smalen [Fri, 6 Nov 2020 09:03:23 +0000 (09:03 +0000)]
[VPlan] NFC: Change VFRange to take ElementCount
This patch changes the type of Start, End in VFRange to be an ElementCount
instead of `unsigned`. This is done as preparation to make VPlans for
scalable vectors, but is otherwise NFC.
Reviewed By: dmgreen, fhahn, vkmr
Differential Revision: https://reviews.llvm.org/D90715
Alex Zinenko [Fri, 6 Nov 2020 09:20:08 +0000 (10:20 +0100)]
[mlir] Add folding of memref_cast inside another memref_cast
There exists a generic folding facility that folds the operand of a memref_cast
into users of memref_cast that support this. However, it was not used for the
memref_cast itself. Fix it to enable elimination of memref_cast chains such as
%1 = memref_cast %0 : A to B
%2 = memref_cast %1 : B to A
that is achieved by combining the folding with the existing "A to A" cast
elimination.
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D90910
Krasimir Georgiev [Fri, 6 Nov 2020 09:33:23 +0000 (10:33 +0100)]
[clang-format] do not break before { in JS comments
In JavaScript some @tags can be followed by `{`, and machinery that parses
these comments will fail to understand the comment if followed by a line break.
clang-format already handles this case by not breaking before `{` in comments.
However this was not working in cases when the column limit falls within `@tag`
or between `@tag` and `{`. This adapts clang-format for this case.
Reviewed By: mprobst
Differential Revision: https://reviews.llvm.org/D90908
Sander de Smalen [Fri, 6 Nov 2020 07:56:55 +0000 (07:56 +0000)]
[TypeSize] Extend UnivariateLinearPolyBase with getWithIncrement/Decrement methods
This patch adds getWithIncrement/getWithDecrement methods to
ElementCount and TypeSize to allow:
TypeSize::getFixed(8).getWithIncrement(8) <=> TypeSize::getFixed(16)
TypeSize::getFixed(16).getWithDecrement(8) <=> TypeSize::getFixed(8)
TypeSize::getScalable(8).getWithIncrement(8) <=> TypeSize::getScalable(16)
TypeSize::getScalable(16).getWithDecrement(8) <=> TypeSize::getScalable(8)
This patch implements parts of the POC in D90342.
Reviewed By: ctetreau, dmgreen
Differential Revision: https://reviews.llvm.org/D90713
Roman Lebedev [Fri, 6 Nov 2020 08:14:03 +0000 (11:14 +0300)]
[IR] CmpInst: Add getFlippedSignednessPredicate()
And refactor a few places to use it
Roman Lebedev [Fri, 6 Nov 2020 07:40:25 +0000 (10:40 +0300)]
[IR] CmpInst: add isRelational()
Since there's CmpInst::isEquality(), it only makes sense
to have it's inverse for consistency.
Roman Lebedev [Fri, 6 Nov 2020 07:36:30 +0000 (10:36 +0300)]
[IR] CmpInst: add isEquality(Pred)
Currently there is only a member version of isEquality(),
which requires an actual [IF]CmpInst to be avaliable,
which isn't always possible, and is inconsistent with
the general pattern here.
I wanted to use it in a new patch, but it wasn't there..
Roman Lebedev [Fri, 6 Nov 2020 06:52:54 +0000 (09:52 +0300)]
[IR] CmpInst: add getUnsignedPredicate()
There's already getSignedPredicate(), it is not symmetrical to not have
it's opposite. I wanted to use it in new code, but it wasn't there..
Freddy Ye [Thu, 5 Nov 2020 07:24:21 +0000 (15:24 +0800)]
[X86] use macros to split GFNI intrinsics into different kinds
Tremont microarchitecture only has GFNI(SSE) version, not AVX and
AVX512 version. This patch is to avoid compiling fail on Windows when
using -march=tremont to invoke one of GFNI(SSE) intrinsic.
Differential Revision: https://reviews.llvm.org/D90822
Max Kazantsev [Fri, 6 Nov 2020 07:55:18 +0000 (14:55 +0700)]
[Test] One more test on IndVars with negative step
Yevgeny Rouban [Fri, 6 Nov 2020 07:46:34 +0000 (14:46 +0700)]
[BranchProbabilityInfo] Introduce method copyEdgeProbabilities(). NFC
A new method is introduced to allow bulk copy of outgoing edge
probabilities from one block to another. This can be useful when
a block is cloned from another one and we do not know if there
are edge probabilities set for the original block or not.
Copying outside of the BranchProbabilityInfo class makes the user
unconditionally set the cloned block's edge probabilities even if
they are unset for the original block.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D90839