LLVM and SPIRV-LLVM-Translator pulldown (WW30) #2138
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
We were previously expanding vselect and matching on the expansion to generate bitselects, but in some cases the expansion would be further combined and a bitselect would not get generated. This patch improves codegen in those cases by legalizing vselect and lowering it to v128.bitselect. The old pattern that matches the expansion is still useful for lowering IR that already uses the expansion rather than a select operation. Differential Revision: https://reviews.llvm.org/D83734
Updating the simd-select.ll tests manually with consistent named regexps for the register numbers was taking more time than it was worth, so this patch updates that test file to have autogenerated output. This is not a significant readability regression because the tests in that file are all very small. Depends on D83734. Differential Revision: https://reviews.llvm.org/D83736
Many tests use opt's -analyze feature, which does not translate well to
NPM and has better alternatives. The alternative here is to explicitly
add a pass that calls ScalarEvolution::print().
The legacy pass manager RUNs aren't changing, but they are now pinned to
the legacy pass manager. For each legacy pass manager RUN, I added a
corresponding NPM RUN using the 'print<scalar-evolution>' pass. For
compatibility with update_analyze_test_checks.py and existing test
CHECKs, 'print<scalar-evolution>' now prints what -analyze prints per
function.
This was generated by the following Python script and failures were
manually fixed up:
import sys
for i in sys.argv:
with open(i, 'r') as f:
s = f.read()
with open(i, 'w') as f:
for l in s.splitlines():
if "RUN:" in l and ' -analyze ' in l and '\\' not in l:
f.write(l.replace(' -analyze ', ' -analyze -enable-new-pm=0 '))
f.write('\n')
f.write(l.replace(' -analyze ', ' -disable-output ').replace(' -scalar-evolution ', ' "-passes=print<scalar-evolution>" ').replace(" | ", " 2>&1 | "))
f.write('\n')
else:
f.write(l)
There are a couple failures still in ScalarEvolution under NPM, but
those are due to other unrelated naming conflicts.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D83798
There is currently no COMPILER_PATH test. A subsequent --ld-path patch will improve the coverage here.
This is in preparation for fixing multiple problems with the way AGPR copies are handled, but this change is NFC itself. First, it's relying on recursively calling copyPhysReg, which is losing information necessary to get correct super register handling. Second, it's constructing a new RegScavenger and doing a O(N^2) walk on every single sub-spill for every AGPR tuple copy. Third, it's using the forward form of the scavenger, and not using the preferred backwards scan.
The size of VTList that is pushed into this container is usually 1, but often 6 or 7. Change the vector to SmallVector to eliminate frequent mallocs. This happens hundreds of thousands of times in each tablegen execution during the LLVM/clang build. https://reviews.llvm.org/D83849
Although the SIMD spec proposal does not specifically include a select instruction, the select instruction in MVP WebAssembly is polymorphic over the selected types, so it is able to work on v128 values when they are enabled. This patch introduces a new variant of the select instruction for each legal vector type. Additional ISel patterns are adapted from the SELECT_I32 and SELECT_I64 patterns. Depends on D83736. Differential Revision: https://reviews.llvm.org/D83737
Replace std::vector with SmallVector to reduce the number of mallocs. This method is frequently executed, and the number of elements in the vector is typically small. https://reviews.llvm.org/D83920
Summary: 1. gcc uses `-march` and `-mtune` flag to chose arch and pipeline model, but clang does not have `-mtune` flag, we uses `-mcpu` to chose both infos. 2. Add SiFive e31 and u54 cpu which have default march and pipeline model. 3. Specific `-mcpu` with rocket-rv[32|64] would select pipeline model only, and use the driver's arch choosing logic to get default arch. Reviewers: lenary, asb, evandro, HsiangKai Reviewed By: lenary, asb, evandro Tags: #llvm, #clang Differential Revision: https://reviews.llvm.org/D71124
…on (NFCI) Summary: This refactors some common support related to shadow memory setup from asan and hwasan into sanitizer_common. This should not only reduce code duplication but also make these facilities available for new compiler-rt uses (e.g. heap profiling). In most cases the separate copies of the code were either identical, or at least functionally identical. A few notes: In ProtectGap, the asan version checked the address against an upper bound (kZeroBaseMaxShadowStart, which is (2^18). I have created a copy of kZeroBaseMaxShadowStart in hwasan_mapping.h, with the same value, as it isn't clear why that code should not do the same check. If it shouldn't, I can remove this and guard this check so that it only happens for asan. In asan's InitializeShadowMemory, in the dynamic shadow case it was setting __asan_shadow_memory_dynamic_address to 0 (which then sets both macro SHADOW_OFFSET as well as macro kLowShadowBeg to 0) before calling FindDynamicShadowStart(). AFAICT this is only needed because FindDynamicShadowStart utilizes kHighShadowEnd to get the shadow size, and kHighShadowEnd is a macro invoking MEM_TO_SHADOW(kHighMemEnd) which in turn invokes: (((kHighMemEnd) >> SHADOW_SCALE) + (SHADOW_OFFSET)) I.e. it computes the shadow space needed by kHighMemEnd (the shift), and adds the offset. Since we only want the shadow space here, the earlier setting of SHADOW_OFFSET to 0 via __asan_shadow_memory_dynamic_address accomplishes this. In the hwasan version, it simply gets the shadow space via "MemToShadowSize(kHighMemEnd)", where MemToShadowSize just does the shift. I've simplified the asan handling to do the same thing, and therefore was able to remove the setting of the SHADOW_OFFSET via __asan_shadow_memory_dynamic_address to 0. Reviewers: vitalybuka, kcc, eugenis Subscribers: dberris, #sanitizers, llvm-commits, davidxl Tags: #sanitizers Differential Revision: https://reviews.llvm.org/D83247
This reverts commit b16dfbe. Accidental push, reverting and creating a new revision.
The patch that introduces handling iterators implemented as pointers may cause crash in some projects because pointer difference is mistakenly handled as pointer decrement. (Similair case for iterators implemented as class instances are already handled correctly.) This patch fixes this issue. The second case that causes crash is comparison of an iterator implemented as pointer and a null-pointer. This patch contains a fix for this issue as well. The third case which causes crash is that the checker mistakenly considers all integers as nonloc::ConcreteInt when handling an increment or decrement of an iterator implemented as pointers. This patch adds a fix for this too. The last case where crashes were detected is when checking for success of an std::advance() operation. Since the modeling of iterators implemented as pointers is still incomplete this may result in an assertion. This patch replaces the assertion with an early exit and adds a FIXME there. Differential Revision: https://reviews.llvm.org/D83295
…gnment assumptions" due to the performance bugs filed in https://bugs.llvm.org/show_bug.cgi?id=46753. An SROA change soon may obviate some of these problems. This reverts commit 8d09f20.
Differential Revision: https://reviews.llvm.org/D83713
Differential Revision: https://reviews.llvm.org/D83438
Python is now handled in CMake with different variables, thus the intel plugin needs a corresponding update. Test Plan: Differential Revision: https://phabricator.intern.facebook.com/D22555992
Summary: Following guidance in https://llvm.org/docs/TestingGuide.html#testing-analysis Reviewers: mehdi_amini Subscribers: mgorny, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D83918
Some time ago, I introduced shortcut features like dylib-has-no-shared_mutex to encode whether the deployment target supported shared_mutex (say). This made the test suite annotations cleaner. However, the problem with building Lit features on top of other Lit features is that it's easier for them to become stale, especially when they are generated programmatically. Furthermore, it makes the bar for defining configurations from scratch higher, since more features have to be defined. Instead, I think it's better to put the XFAILs in the tests directly, which allows cleaning them up with a simple grep.
This test has been failing on some SDKs for a long time because we lack a proper way of identifying the SDK version in Lit. Until that is possible, mark the test as unsupported on Apple to restore the CI.
… its the default 32-bit cpu in clang Alternative to D83897. I believe the big change here is that I removed slow unaligned memory 16 Down side that it may adversely effect tuning if someone explicitly targets -march=pentium4 and expects pentium4 tuned code. Of course pentium4 is so old our default behavior with the previous settings may not have been the best either. Reviewed By: echristo, RKSimon Differential Revision: https://reviews.llvm.org/D83913
Differential Revision: https://reviews.llvm.org/D83857
Statepoint has no static operands, so it cannot be verified against MCInstrDescr. Revert NumDefs change introduced by ef658eb.
…ion remarks Summary: This change added a new inline advisor that takes optimization remarks from previous inlining as input, and provides the decision as advice so current inlining can replay inline decisions of a different compilation. Dwarf inline stack with line and discriminator is used as anchor for call sites including call context. The change can be useful for Inliner tuning as it provides a channel to allow external input for tweaking inline decisions. Existing alternatives like alwaysinline attribute is per-function, not per-callsite. Per-callsite inline intrinsic can be another solution (not yet existing), but it's intrusive to implement and also does not differentiate call context. A switch -sample-profile-inline-replay=<inline_remarks_file> is added to hook up the new inline advisor with SampleProfileLoader's inline decision for replay. Since SampleProfileLoader does top-down inlining, inline decision can be specialized for each call context, hence we should be able to replay inlining accurately. However with a bottom-up inliner like CGSCC inlining, the replay can be limited due to lack of specialization for different call context. Apart from that limitation, the new inline advisor can still be used by regular CGSCC inliner later if needed for tuning purpose. Subscribers: mgorny, aprantl, hiraditya, llvm-commits Tags: #llvm Resubmit for https://reviews.llvm.org/D84086
…ptimization remarks" This reverts commit 2d6ecfa.
Hopefully this will make the bot a little less noisy. Rationale for each:
AlignTrailingComments: We don't want to force-align the various expected-error
and friends.
CommentPragmas: Tell clang-format to leave the "// CHECK:" and the
"// expected-" alone.
AlwaysBreakTemplateDeclarations: Templates in tests often have no break between
the template-head and the declaration.
Differential Revision: https://reviews.llvm.org/D83901
…eturn value to actually say "consteval". This warning was modified in 796ed03 to use the term "consteval" for consteval functions. However the warning has never worked as intended since the diagnostic's arguments are used in the wrong order. This was unfortunately missed by 796ed03 since no test did exercise this specific warning. Additionally send the NamedDecl* into the diagnostic instead of just the IdentifierInfo* to correctly work with special names and template arguments.
…Operator This patch - adds `canCreateUndefOrPoison` - refactors `canCreatePoison` so it can deal with constantexprs `canCreateUndefOrPoison` will be used at D83926. Reviewed By: nikic, jdoerfert Differential Revision: https://reviews.llvm.org/D84007
The function extractArgumentsFromModule() was passing a one-based index to, but replaceFunctionCalls() was expecting a zero-based argument index. This resulted in assertion errors when reducing function call arguments with different types. Additionally, the Reviewed By: lebedev.ri Differential Revision: https://reviews.llvm.org/D84099
The getAllOnesValue can only handle things that are bitcast from a ConstantInt, while here we bitcast through a pointer, so we may see more complex objects (like Array or Struct). Differential Revision: https://reviews.llvm.org/D83870
…OR relocs. When processing a MachO SUBTRACTOR/UNSIGNED pair, if the UNSIGNED target is non-extern then check the r_symbolnum field of the relocation to find the targeted section and use the section's address to find 'ToSymbol'. Previously 'ToSymbol' was found by loading the initial value stored at the fixup location and treating this as an address to search for. This is incorrect, however: the initial value includes the addend and will point to the wrong block if the addend is less than zero or greater than the block size. rdar://65756694
The map initialization unnecessarily ends up in almost every translation unit due to transitive inclusion.
Its use was removed in dff07df ("[SPIRV] Fix assertion due to
duplicate operands in opencl version metadata. This happens when two
SPIR modules are linked together. Remove check for OCL version in
OCLTypeToSPIRV. Only check if language is C or C++.", 2015-12-16).
a7b763b introduces a bug, when it could be possible to map a single pointer type in LLVM IR to two different pointer types in SPIR-V (when SPV_INTEL_usm_storage_classes extension is not allowed). This patch fixes this bug. Signed-off-by: Dmitry Sidorov <[email protected]>
Extension is published at intel#1612
vladimirlaz
requested review from
AGindinson,
AlexeySachkov,
AlexeySotkin,
Fznamznon,
bader,
elizabethandrews,
mdtoguchi and
premanandrao
as code owners
Contributor
|
/summary:run |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
92 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
LLVM: llvm/llvm-project@f7a57153
SPIRV-LLVM-Translator: KhronosGroup/SPIRV-LLVM-Translator@5202e9c