Commit Graph

370270 Commits

Author SHA1 Message Date
Fangrui Song d69ada30e2 [BranchProbabilityInfo] Make MaxSuccIdx[Src] efficient and add a comment about the subtle eraseBlock. NFC
Follow-up to D90272.
2020-10-27 16:29:23 -07:00
River Riddle c09d10437f [mlir][NFC] Fix incorrect header comments.
Resolves missed comments in D89103
2020-10-27 16:22:31 -07:00
River Riddle 935d708568 [mlir][NFC] Remove unnecessary PatternRewriter::create methods
At this point, these methods are just carbon copies of OpBuilder::create and aren't necessary given that PatternRewriter inherits from OpBuilder.

Differential Revision: https://reviews.llvm.org/D90087
2020-10-27 16:16:51 -07:00
River Riddle eacac2679d [mlir][Interfaces] Optimize the implementation of InterfaceMap to reduce generated code size.
An InterfaceMap is generated for every single operation type, and is responsible for a large amount of the code size from MLIR given that its internals highly utilize templates. This revision refactors the internal implementation to use bare malloc/free for interface instances as opposed to static variables and moves as much code out of templates as possible. This led to a decrease of over >1mb (~12% of total MLIR related code size) for a downstream MLIR library with a large amount of operations.

Differential Revision: https://reviews.llvm.org/D90086
2020-10-27 16:16:51 -07:00
River Riddle ef728eaf6e [mlir][Interfaces] Generate a struct containing function pointers instead of a class with vtables
When compiling for code size, the use of a vtable causes a destructor(and constructor in certain cases) to be generated for the class. Interface models don't need a complex constructor or a destructor, so this can lead to many megabytes of code size increase(even in opt). This revision switches to a simpler struct of function pointers approach that accomplishes the same API requirements as before. This change requires no updates to user code, or any other code aside from the generator, as the user facing API is still exactly the same.

Differential Revision: https://reviews.llvm.org/D90085
2020-10-27 16:16:51 -07:00
River Riddle d989ae9069 [mlir][SIdeEffectInterface][NFC] Move several InterfaceMethods to the extraClassDeclarations instead
All InterfaceMethods will have a corresponding entry in the interface model, and by extension have an implementation generated for every operation type. This can result in large binary size increases when a large amount of operations use an interface, such as the side effect interface.

Differential Revision: https://reviews.llvm.org/D90084
2020-10-27 16:16:51 -07:00
MaheshRavishankar 9d5239d39e [mlir][Linalg] Add fusion of IndexedGenericOp with TensorReshapeOp by expansion.
This patch adds support for fusing linalg.indexed_generic op with
linalg.tensor_reshape op by expansion, i.e.
- linalg.indexed_generic op -> linalg.tensor_reshape op when the
  latter is expanding.
- linalg.tensor_reshape op -> linalg.indexed_generic op when the
  former is folding.

Differential Revision: https://reviews.llvm.org/D90082
2020-10-27 16:15:34 -07:00
Kazu Hirata a7b662d0f4 [BranchProbabilityInfo] Fix eraseBlock
This patch ensures that BranchProbabilityInfo::eraseBlock(BB) deletes
all entries in Probs associated with with BB.

Without this patch, stale entries for BB may remain in Probs after
eraseBlock(BB), leading to a situation where a newly created basic
block has an edge probability associated with it even before the pass
responsible for creating the basic block adds any edge probability to
it.

Consider the current implementation of eraseBlock(BB):

  for (const_succ_iterator I = succ_begin(BB), E = succ_end(BB); I != E; ++I) {
    auto MapI = Probs.find(std::make_pair(BB, I.getSuccessorIndex()));
    if (MapI != Probs.end())
      Probs.erase(MapI);
  }

Notice that it uses succ_begin(BB) and succ_end(BB), which are based
on BB->getTerminator().  This means that if the terminator changes
between calls to setEdgeProbability and eraseBlock, then we may not
examine all pairs associated with BB.

This is exactly what happens in MaybeMergeBasicBlockIntoOnlyPred,
which merges basic blocks A into B if A is the sole predecessor of B,
and B is the sole successor of A.  It replaces the terminator of A
with UnreachableInst before (indirectly) calling eraseBlock(A).

The patch fixes the problem by keeping track of all edge probablities
entered with setEdgeProbability in a map from BasicBlock* to a
successor index.

Differential Revision: https://reviews.llvm.org/D90272
2020-10-27 16:14:25 -07:00
Kazu Hirata c91487769d [JumpThreading] Set edge probabilities when creating basic blocks
This patch teaches the jump threading pass to set edge probabilities
whenever the pass creates new basic blocks.

Without this patch, the compiler sometimes produces non-deterministic
results.  The non-determinism comes from the jump threading pass using
stale edge probabilities in BranchProbabilityInfo.  Specifically, when
the jump threading pass creates a new basic block, we don't initialize
its outgoing edge probability.

Edge probabilities are maintained in:

  DenseMap<Edge, BranchProbability> Probs;

in class BranchProbabilityInfo, where Edge is an ordered pair of
BasicBlock * and a successor index declared as:

  using Edge = std::pair<const BasicBlock *, unsigned>;

Probs maps edges to their corresponding probabilities.

Now, we rarely remove entries from this map, so if we happen to
allocate a new basic block at the same address as a previously deleted
basic block with an edge probability assigned, the newly created basic
block appears to have an edge probability, albeit a stale one.

This patch fixes the problem by explicitly setting edge probabilities
whenever the jump threading pass creates new basic blocks.

Differential Revision: https://reviews.llvm.org/D90106
2020-10-27 16:07:27 -07:00
Sam Clegg 84129150ce [lld][WebAssembly] Fix memory size in dylink section for -pie exectuables
This field to represents the amount of static data needed by
an dynamic library or executable it should not include things
like heap or stack areas, which in the case of `-pie` are
not determined until runtime (e.g. __stack_pointer is imported).

Differential Revision: https://reviews.llvm.org/D90261
2020-10-27 16:05:52 -07:00
Eugene Zhulenev f6c9f6eccd [mlir] JitRunner: add a config option to register symbols with ExecutionEngine at runtime
Reviewed By: mehdi_amini

Differential Revision: https://reviews.llvm.org/D90264
2020-10-27 15:57:34 -07:00
Sanjay Patel 50dfa19cc7 [CostModel] remove cost-kind predicate for FP add/mul vector reduction costs
This was originally part of:
f2c25c7079
but that was reverted because there was an underlying bug in
processing the vector type of these intrinsics. That was
fixed with:
74ffc823ed

This is similar in spirit to 01ea93d85d (memcpy) except that
here the underlying caller assumptions were created for vectorizer
use (throughput) rather than other passes.

That meant targets could have an enormous throughput cost with no
corresponding size, latency, or blended cost increase.

Paraphrasing from the previous commits:
This may not make sense for some callers, but at least now the
costs will be consistently wrong instead of mysteriously wrong.

Targets should provide better overrides if the current modeling
is not accurate.
2020-10-27 18:00:20 -04:00
Sanjay Patel 138fda5dd2 [CostModel] add tests for FP reductions; NFC 2020-10-27 18:00:20 -04:00
Martin Storsjö 70bba9ef35 [libcxx] Don't truncate intermediates to wchar_t when widening
On windows, wchar_t is 16 bit, while we might be widening chars to
char32_t.

This cast had been present since the initial commit, and removing it
doesn't seem to make any tests fail.

Differential Revision: https://reviews.llvm.org/D90228
2020-10-27 23:58:27 +02:00
Nemanja Ivanovic 5459d08795 [PowerPC] Fix single-use check and update chain users for ld-splat
When converting a BUILD_VECTOR or VECTOR_SHUFFLE to a splatting load
as of 1461fb6e78, we inaccurately check
for a single user of the load and neglect to update the users of the
output chain of the original load. As a result, we can emit a new
load when the original load is kept and the new load can be reordered
after a dependent store. This patch fixes those two issues.

Fixes https://bugs.llvm.org/show_bug.cgi?id=47891
2020-10-27 16:49:38 -05:00
Joseph Huber d981c7b758 [OpenMP] Add Support for Mapping Names in Libomptarget RTL
Summary:
This patch adds basic support for priting the source location and names for the
mapped variables. This patch does not support names for custom mappers. This is
based on D89802. The names information currently will be printed out only in
debug mode or using env LIBOMPTARGET_INFO during execution. But the information
is added when availible to the Device and Private data structures. To get the
information out the code must be built with debug symbols on using -g or
-Rpass=openmp-opt

Reviewers: jdoerfert

Differential Revision: https://reviews.llvm.org/D90172
2020-10-27 16:53:05 -04:00
Louis Dionne 2089e762d0 [libc++] Try to fix cross-loading of lit.local.cfg on Windows
On windows, the previous path replacement using forward slashes wouldn't
work, and so we'd end up including the same file again. We would do that
until we'd hit the recursion limit of the Python interpreter.

Instead, use `os.path` to properly replace without assuming a specific
path separator.
2020-10-27 16:28:29 -04:00
Tony 801aeb5d51 [AMDGPU] Order AMDGPU ELF machine numbers in ascending order
Differential Revision: https://reviews.llvm.org/D90259
2020-10-27 20:16:30 +00:00
Joseph Huber a87d7b3d44 [OpenMP] Add Passing in Original Declaration Names To Mapper API
Summary:
This patch adds support for passing in the original delcaration name in the
source file to the libomptarget runtime. This will allow the runtime to provide
more intelligent debugging messages. This patch takes the original expression
parsed from the OpenMP map / update clause and provides a textual
representation if it was explicitly mapped, otherwise it takes the name of the
variable declaration as a fallback. The information in passed to the runtime in
a global array of strings that matches the existing ident_t source location
strings using ";name;filename;column;row;;". See
clang/test/OpenMP/target_map_names.cpp for an example of the generated output
for a given map clause.

Reviewers: jdoervert

Differential Revision: https://reviews.llvm.org/D89802
2020-10-27 16:09:19 -04:00
Louis Dionne c56bbb3961 [libc++] Make sure we include a header when checking compiler macros
Otherwise, it's possible for some __config_site macros not to be
picked up.
2020-10-27 15:58:43 -04:00
Fangrui Song 6a972ffabd [test] Delete Feature/load_module.ll which is covered by load_extension.ll 2020-10-27 12:50:53 -07:00
LLVM GN Syncbot fa562cb224 [gn build] Port e025d09b21 2020-10-27 19:40:30 +00:00
LLVM GN Syncbot 6495896ede [gn build] Port ce6900c6cb 2020-10-27 19:40:29 +00:00
Fangrui Song 62536fae61 [test] Make ThinLTO/X86/crash_debuginfo.ll work with -enable-new-pm=1
LegacyInlinerBase::doFinalization runs removeDeadFunctions() to remove bar but
the new PM inliner doesn't. Improve the test to use llvm-nm -U.
2020-10-27 12:37:19 -07:00
Aaron En Ye Shi 3700556ecb [HIP][NFC] Use correct max in cuda_complex_builtins
Update the clang complex builtins for OpenMP to use the
correct max function from either __nv_* or __ocml_*.
2020-10-27 19:35:09 +00:00
Nicolai Hähnle e025d09b21 Revert multiple patches based on "Introduce CfgTraits abstraction"
These logically belong together since it's a base commit plus
followup fixes to less common build configurations.

The patches are:

Revert "CfgInterface: rename interface() to getInterface()"

This reverts commit a74fc48158.

Revert "Wrap CfgTraitsFor in namespace llvm to please GCC 5"

This reverts commit f2a06875b6.

Revert "Try to make GCC5 happy about the CfgTraits thing"

This reverts commit 03a5f7ce12.

Revert "Introduce CfgTraits abstraction"

This reverts commit c0cdd22c72.
2020-10-27 20:33:30 +01:00
Nicolai Hähnle ce6900c6cb Revert "DomTree: Extract (mostly) read-only logic into type-erased base classes"
This reverts commit 848a68a032.
2020-10-27 20:33:29 +01:00
Florian Hahn 7d6b58302a [llvm-reduce] Add test with some aliases. 2020-10-27 19:29:48 +00:00
Stella Laurenzo 013b9322de [mlir][Python] Custom python op view wrappers for building and traversing.
* Still rough edges that need more sugar but the bones are there. Notes left in the test case for things that can be improved.
* Does not actually yield custom OpViews yet for traversing. Will rework that in a followup.

Differential Revision: https://reviews.llvm.org/D89932
2020-10-27 12:23:34 -07:00
Stanislav Mekhanoshin 78ae1f6c90 [AMDGPU] Change predicate for fma/fmac legacy
I do not exactly like the use of a negative predicate to
enable instructions' support. Change HasNoMadMacF32Insts
with HasFmaLegacy32.

Differential Revision: https://reviews.llvm.org/D90250
2020-10-27 12:03:52 -07:00
Victor Huang 2e1a737f46 [PowerPC][PCRelative] Turn on TLS support for PCRel by default
Turn on TLS support for PCRel by default and update the test cases.

Differential Revision: https://reviews.llvm.org/D88738
Reviewed by: stefanp, kamaub
2020-10-27 13:58:44 -05:00
Louis Dionne 88ffc72717 [libc++] Add a libc++ configuration that does not support localization
When porting libc++ to embedded systems, it can be useful to drop support
for localization, which these systems don't implement or care about.

Differential Revision: https://reviews.llvm.org/D90072
2020-10-27 14:56:30 -04:00
Duncan P. N. Exon Smith 917acac960 FileManager: Shrink FileEntryRef to the size of a pointer
Shrink `FileEntryRef` to the size of a pointer, by having it directly
reference the `StringMapEntry` the same way that `DirectoryEntryRef`
does. This makes `FileEntryRef::FileEntryRef` private as a side effect
(`FileManager` is a friend!).

There are two helper types added within `FileEntryRef`:

- `FileEntryRef::MapValue` is the type stored in
  `FileManager::SeenFileEntries`. It's a replacement for
  `SeenFileEntryOrRedirect`, where the second pointer type has been
  changed from `StringRef*` to `MapEntry*` (see next bullet).
- `FileEntryRef::MapEntry` is the instantiation of `StringMapEntry<>`
  where `MapValue` is stored. This is what `FileEntryRef` has a pointer
  to, in order to grab the name in addition to the value.

Differential Revision: https://reviews.llvm.org/D89488
2020-10-27 14:55:41 -04:00
Sean Silva 83154c5418 [mlir] Add bufferization for std.select op.
Differential Revision: https://reviews.llvm.org/D90204
2020-10-27 11:46:33 -07:00
Jay Foad 4b1ea84a1d [AMDGPU] Fix check prefix for VOP3 VI disassembler tests
Also, following D81841, don't try to encode f16 literals in i16/u16
instructions.

Differential Revision: https://reviews.llvm.org/D90242
2020-10-27 18:43:25 +00:00
Christian Sigg 8c176b6029 [mlir] Catch async.yield operands not matching the number of async.execute results.
Reviewed By: ezhulenev

Differential Revision: https://reviews.llvm.org/D90211
2020-10-27 19:39:34 +01:00
Martin Storsjö 0be27302d4 [libcxx] Fix typo in spelling of 'sentinel'. NFC.
Differential Revision: https://reviews.llvm.org/D90220
2020-10-27 20:38:32 +02:00
Simon Pilgrim 48555cd494 [llvm-readobj] Remove duplicate inner if() condition. NFCI.
This should have been removed when rG445c3fdd2ae8 simplified the conditions.

Reported as "Snippet 5" in https://www.viva64.com/en/b/0771/
2020-10-27 18:36:04 +00:00
Amy Huang 504615353f Revert "[CodeView] Emit static data members as S_CONSTANTs."
Seems like there's an assert in here that we shouldn't be running into.

This reverts commit 515973222e.
2020-10-27 11:29:58 -07:00
LLVM GN Syncbot 16ca0037c8 [gn build] Port 46c3d5cb05 2020-10-27 18:08:19 +00:00
Michael Liao 46c3d5cb05 [amdgpu] Add the late codegen preparation pass.
Summary:
- Teach that pass to widen naturally aligned but not DWORD aligned
  sub-DWORD loads.

Reviewers: rampitec, arsenm

Subscribers:

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D80364
2020-10-27 14:07:59 -04:00
Simon Pilgrim 64d3ed304f [X86] Regenerate scalar fptosi/fptoui tests. NFCI.
Merge prefixes where possible, use 'X86' instead of 'X32' (which we try to only use for gnux32 triple tests).
2020-10-27 17:44:30 +00:00
Louis Dionne cfa1376a01 [libc++] Get rid of iostreams in the to_string tests 2020-10-27 13:36:13 -04:00
Kazushi (Jam) Marukawa a65883a78a [VE] Add vector reduction instructions
Add VSUMS/VSUMX/VFSUM/VMAXS/VMAXX/VFMAX/VRAND/VROR/VRXOR isntructions.
Add regression tests too.

Reviewed By: simoll

Differential Revision: https://reviews.llvm.org/D90227
2020-10-28 02:33:21 +09:00
Vedant Kumar 5a3ef55a52 [Utils] Skip RemoveRedundantDbgInstrs in MergeBlockIntoPredecessor (PR47746)
This patch changes MergeBlockIntoPredecessor to skip the call to
RemoveRedundantDbgInstrs, in effect partially reverting D71480 due to
some compile-time issues spotted in LoopUnroll and SimplifyCFG.

The call to RemoveRedundantDbgInstrs appears to have changed the
worst-case behavior of the merging utility. Loosely speaking, it seems
to have gone from O(#phis) to O(#insts).

It might not be possible to mitigate this by scanning a block to
determine whether there are any debug intrinsics to remove, since such a
scan costs O(#insts).

So: skip the call to RemoveRedundantDbgInstrs. There's surprisingly
little fallout from this, and most of it can be addressed by doing
RemoveRedundantDbgInstrs later. The exception is (the block-local
version of) SimplifyCFG, where it might just be too expensive to call
RemoveRedundantDbgInstrs.

Differential Revision: https://reviews.llvm.org/D88928
2020-10-27 10:12:59 -07:00
Sam McCall 1a1aad9156 [ADT] Fix accidental pointer comparison in test 2020-10-27 18:11:45 +01:00
Simon Pilgrim 86bec79b15 [X86] Regenerate xor tests. NFCI.
Merge prefixes where possible, use 'X86' instead of 'X32' (which we try to only use for gnux32 triple tests).
2020-10-27 16:45:47 +00:00
Simon Pilgrim 4036551ae4 [X86] Regenerate tbm intrinsics tests. NFCI.
Merge prefixes where possible, use 'X86' instead of 'X32' (which we try to only use for gnux32 triple tests).
2020-10-27 16:45:47 +00:00
Simon Pilgrim e0c06e310c [X86] Regenerate popcnt tests. NFCI.
Merge prefixes where possible, use 'X86' instead of 'X32' (which we try to only use for gnux32 triple tests).
2020-10-27 16:45:46 +00:00
Simon Pilgrim fae1ffceae [X86] Regenerate xop tests with common prefixes. 2020-10-27 16:45:46 +00:00