Block merging in MLIR will incorrectly merge blocks with operations whose values are used outside of that block. This change forbids this behavior and provides a test where it is illegal to perform such a merge.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D91745
Currently, `node` only includes the semicolon for (some) statements. However,
declarations have the same issue of (potentially) trailing semicolons, so `node`
should behave the same for them.
Differential Revision: https://reviews.llvm.org/D91872
The default version only works if the returned node has a single
result. The X86 and PowerPC versions support multiple results
and allow a single result to be returned from a node with
multiple outputs. And allow a single result that is not result 0
of the node.
Also replace the Mips version since the new version should work
for it. The original version handled multiple results, but only
if the new node and original node had the same number of results.
Differential Revision: https://reviews.llvm.org/D91846
Add canoncalization patterns to remove zero-iteration 'for' loops, replace
single-iteration 'for' loops with their bodies; remove known-false conditionals
with no 'else' branch and replace conditionals with known value by the
respective region. Although similar transformations are performed at the CFG
level, not all flows reach that level, e.g., the GPU flow may want to remove
single-iteration loops before deciding on loop mapping to thread dimensions.
Reviewed By: herhut
Differential Revision: https://reviews.llvm.org/D91865
Update the help string for `target.source-map` to remove the use of the word
"duple" and to add examples. Additionally I rewrote parts with the goal of
making the description more concrete.
rdar://68736012
Differential Revision: https://reviews.llvm.org/D91742
Use the OR(CMP,ADD) / AND(CMP,SUB) patterns like we do on pre-SSE4 targets.
We're still using X86ISD::BLENDV on some AVX targets as we don't do custom lowering for >= 256-bit vectors.
Really this (and combineVSelectWithAllOnesOrZeros) needs moving to DAGCombiner, but pre-SSE42 we see the vXi64 comparison type as a 2 x 32-bits result so we can't just rely on ComputeNumSignBits to give us the 'all bits' result we need.
This is re-applying a combination of f7eac51b9b and 8ec7ea3ddc as one patch
to avoid regressions now that we have better testing in place.
Those were reverted with 32dd5870ee because of crashing in experimental intrinsics.
That bug should be fixed with 7ae346434.
Paraphrased original commit messages:
This is the last step in removing cost-kind as a consideration in the
basic class model for intrinsics.
See D89461 for the start of that.
Subsequent commits dealt with each of the special-case intrinsics that
had customization here in the basic class. This should remove a barrier
to retrying D87188 (canonicalization to the abs intrinsic).
The ARM and x86 cost diffs seen here may be wrong because the
target-specific overrides have their own bugs, but we hope this is
less wrong - if something has a significant throughput cost, then it
should have a significant size / blended cost too by default.
The only behavioral diff in current regression tests is shown in the
x86 scatter-gather test (which is misplaced or broken because it runs
the entire -O3 pipeline) - we unrolled less, and we assume that is
a improvement.
Exception: in general, we want the *size* cost for a scalar call to be
cheap even if the other costs are expensive - we expect it to just be
a branch with some optional stack manipulation.
It is likely that we will want to carve out some
exceptions/overrides to this rule as follow-up patches for
calls that have some general and/or target-specific difference
to the expected lowering.
This was noticed as a regression in unrolling, so we have a test
for that now along with a couple of direct cost model tests.
If the assumed scalarization costs for the oversized vector
calls are not realistic, that would be another follow-up
refinement of the cost models.
Differential Revision: https://reviews.llvm.org/D90554
This will ensure that passes that add new global variables will create them
in address space 1 once the passes have been updated to no longer default
to the implicit address space zero.
This also changes AutoUpgrade.cpp to add -G1 to the DataLayout if it wasn't
already to present to ensure bitcode backwards compatibility.
Reviewed by: arsenm
Differential Revision: https://reviews.llvm.org/D84345
This is similar to the existing alloca and program address spaces (D37052)
and should be used when creating/accessing global variables.
We need this in our CHERI fork of LLVM to place all globals in address space 200.
This ensures that values are accessed using CHERI load/store instructions
instead of the normal MIPS/RISC-V ones.
The problem this is trying to fix is that most of the time the type of
globals is created using a simple PointerType::getUnqual() (or ::get() with
the default address-space value of 0). This does not work for us and we get
assertion/compilation/instruction selection failures whenever a new call
is added that uses the default value of zero.
In our fork we have removed the default parameter value of zero for most
address space arguments and use DL.getProgramAddressSpace() or
DL.getGlobalsAddressSpace() whenever possible. If this change is accepted,
I will upstream follow-up patches to use DL.getGlobalsAddressSpace() instead
of relying on the default value of 0 for PointerType::get(), etc.
This patch and the follow-up changes will not have any functional changes
for existing backends with the default globals address space of zero.
A follow-up commit will change the default globals address space for
AMDGPU to 1.
Reviewed By: dylanmckay
Differential Revision: https://reviews.llvm.org/D70947
Also added diffing of a few more math functions. Combining the diff check
for all of these functions helps us meet the OSS fuzz bar of a minimum of
100 program edges.
Reviewed By: lntue
Differential Revision: https://reviews.llvm.org/D91817
Summary:
Expand existing loopsink testing to also test loopsinking using new pass
manager. Enable memoryssa for loopsink with new pass manager. This
combination exposed a bug that was previously fixed for loopsink
without memoryssa. When sinking an instruction into a loop, the source
block may not be part of the loop but still needs to be checked for
pointer invalidation. This is the fix for bugzilla #39695 (PR 54659)
expanded to also work with memoryssa.
Respond to review comments. Enable Memory SSA in legacy Loop Sink pass
under EnableMSSALoopDependency option control. Update tests accordingly.
Respond to review comments. Add options controlling whether memoryssa is
used for loop sink, defaulting to off. Expand testing based on these
options.
Respond to review comments. Properly indicated preserved analyses.
This relanding addresses a compile-time performance problem by moving
test for profile data earlier to avoid unnecessary computations.
Author: Jamie Schmeiser <schmeise@ca.ibm.com>
Reviewed By: asbirlea (Alina Sbirlea)
Differential Revision: https://reviews.llvm.org/D90249
The constrained intrinsics have metadata arguments, so the
tests here were crashing as noted in D90554 (and that was
reverted even though this bug exists independently of that
change).
std::string_view("") produces a string_view instance that compares
equal to std::string_view(), but requires more complex initialization
(storing the address of the string literal, rather than zeroing).
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D91009
Summary:
[NFC intended] Refactor the code for printChanged for reuse and to facilitate
subsequent reporters of changes to the IR in the new pass manager.
Create abstract template base classes for common functionality and give
classes more appropriate names. The base classes handle all of the
determination of when a function or pass is "interesting" and should be
reported or filtered out. They have pure virtual functions which are called
when a change by a pass has been recognized so the derived class need only
provide the overrides to present the information about the changing IR.
There are at least 2 more change reporters to come (which were presented
in my tutorial at the 2020 llvm developer's meeting) that derive from
these classes.
Respond to review comments: move function out of line, remove inline keyword,
remove unneeded qualifiers, simplify comparison.
Author: Jamie Schmeiser <schmeise@ca.ibm.com>
Reviewed By: aeubanks (Arthur Eubanks), madhur13490 (Madhur Amilkanthwar)
Differential Revision: https://reviews.llvm.org/D87000
wchar_t can be signed (thus hasSignedIntegerRepresentation() returns
true), but it doesn't have an unsigned type, which would lead to a crash
when trying to get it.
With this fix, we special-case WideChar types in the pointer assignment
code.
Differential Revision: https://reviews.llvm.org/D91625
Just something I forgot when I added the R82. Need to have a look
at crypto and fusing, but will do that as a follow up.
Differential Revision: https://reviews.llvm.org/D91848
This checks to see if the loop will likely become a tail predicated loop
and disables wls loop generation if so, as the likelihood for reverting
is currently too high. These should be fairly rare situations anyway due
to the way iterations and element counts are used during lowering. Just
not trying can alter how SCEV's are materialized however, leading to
different codegen.
It also adds a option to disable all while low overhead loops, for
debugging.
Differential Revision: https://reviews.llvm.org/D91663
This patch implements out of line atomics for LSE deployment
mechanism. Details how it works can be found in llvm/docs/Atomics.rst
Options -moutline-atomics and -mno-outline-atomics to enable and disable it
were added to clang driver. This is clang and llvm part of out-of-line atomics
interface, library part is already supported by libgcc. Compiler-rt
support is provided in separate patch.
Differential Revision: https://reviews.llvm.org/D91157
This is a partial un-revert of 32dd5870ee (originally df09f82599 ).
I'm adding back the baseline tests first, so we don't have
to back-track as much in case there are still problems.
See discussion in D90554.
This is a partial un-revert of 32dd5870ee. I'm adding
back the baseline tests first, so we don't have to
back-track as much in case there are still problems.
During the initial Solaris sanitizer port, I missed to enable the
`sanitizer_common` and `ubsan_minimal` testsuites. This patch fixes this,
correcting a few unportabilities:
- `Posix/getpass.cpp` failed to link since Solaris lacks `libutil`.
Omitting the library lets the test `PASS`, but I thought adding `%libutil`
along the lines of `%librt` to be overkill.
- One subtest of `Posix/getpw_getgr.cpp` is disabled because Solaris
`getpwent_r` has a different signature than expected.
- `/dev/null` is a symlink on Solaris.
- XPG7 specifies that `uname` returns a non-negative value on success.
Tested on `amd64-pc-solaris2.11` and `sparcv9-sun-solaris2.11`.
Differential Revision: https://reviews.llvm.org/D91606
This canonicalization is useful to resolve loads into scalar values when
doing partial bufferization.
Differential Revision: https://reviews.llvm.org/D91855
Commit f3aa9e36d9 fixed the embedded OS
build by removing all passed args for `GetName`/`GetDemangledName`. The motivation
for this was that these arguments were apparently removed in
commit 22b044877d. However, only `GetName`'s language
argument was removed but the mangling preference argument was *not* removed
(and unfortunately had a default argument). So when that commit removed all
the args it didn't just fix the build but it also changed all the mangling
preferences to 'demangled' for all `GetName` calls.
Also some `GetName` calls were outside the TARGET_OS_EMBEDDED ifdef, so
this change ended up breaking the following tests on macOS:
lldb-api :: lang/objc/objc-static-method-stripped/TestObjCStaticMethodStripped.py
lldb-api :: lang/objc/objc-super/TestObjCSuper.py
From what I can see f3aa9e36d9 removed 12 ePreferMangled args and this patch
re-adds 12 args with roughly the same line numbers, so this *should* restore the
old behaviour and also keep the embedded build working. On the other hand,
ObjectFileMachO::ParseSymtab is a very successful attempt at writing
the longest possible function within LLVM, so this fix is partly based
on the engineering principle known as "hoping for the best".
Implement getMinimumJumpTableEntries() to specify threshold for jump
table genaration. We use 8 for the case of PIC mode to relieve the
impact of PIC calculation required to implement PIC mode jump table.
Update jump table regression test also.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D91785
This reverts commit f8284d21a8.
Revert "[mlir][Linalg] NFC: Expose some utility functions used for promotion."
This reverts commit 0c59f51592.
Revert "Remove unused isZero function"
This reverts commit 0f9f0a4046.
Change f8284d21 led to multiple failures in IREE compilation.
As reported in PR 48202, two allocator tests `FAIL` on Solaris/sparcv9,
presumably because Solaris uses the full 64-bit address space and the
allocator cannot deal with that:
SanitizerCommon-Unit :: ./Sanitizer-sparcv9-Test/SanitizerCommon.CombinedAllocator32Compact
SanitizerCommon-Unit :: ./Sanitizer-sparcv9-Test/SanitizerCommon.SizeClassAllocator32Iteration
This patch disables the tests.
Tested on `sparcv9-sun-solaris2.11`.
Differential Revision: https://reviews.llvm.org/D91622
Many of the `FastUnwindTest.*` tests `FAIL` on SPARC, both Solaris and
Linux. The issue is that the fake stacks used in those tests don't match
the requirements of the SPARC unwinder in `sanitizer_stacktrace_sparc.cpp`
which has to look at the register window save area.
I'm disabling the failing tests.
Tested on `sparcv9-sun-solaris2.11`.
Differential Revision: https://reviews.llvm.org/D91618
This extends the "memory region" command to
show tagged regions on AArch64 Linux when the MTE
extension is enabled.
(lldb) memory region the_page
[0x0000fffff7ff8000-0x0000fffff7ff9000) rw-
memory tagging: enabled
This is done by adding an optional "flags" field to
the qMemoryRegion packet. The only supported flag is
"mt" but this can be extended.
This "mt" flag is read from /proc/{pid}/smaps on Linux,
other platforms will leave out the "flags" field.
Where this "mt" flag is received "memory region" will
show that it is enabled. If it is not or the target
doesn't support memory tagging, the line is not shown.
(since majority of the time tagging will not be enabled)
Testing is added for the existing /proc/{pid}/maps
parsing and the new smaps parsing.
Minidump parsing has been updated where needed,
though it only uses maps not smaps.
Target specific tests can be run with QEMU and I have
added MTE flags to the existing helper scripts.
Reviewed By: labath
Differential Revision: https://reviews.llvm.org/D87442
Depends On D89963
**Automatic reference counting algorithm outline:**
1. `ReturnLike` operations forward the reference counted values without
modifying the reference count.
2. Use liveness analysis to find blocks in the CFG where the lifetime of
reference counted values ends, and insert `drop_ref` operations after
the last use of the value.
3. Insert `add_ref` before the `async.execute` operation capturing the
value, and pairing `drop_ref` before the async body region terminator,
to release the captured reference counted value when execution
completes.
4. If the reference counted value is passed only to some of the block
successors, insert `drop_ref` operations in the beginning of the blocks
that do not have reference coutned value uses.
Reviewed By: silvas
Differential Revision: https://reviews.llvm.org/D90716
Extract the scratch offset from the scratch buffer descriptor that is
stored in the global table.
Differential Revision: https://reviews.llvm.org/D91701
Our code that dumps groups has 3 noticeable issues:
1) It uses `unwrapOrError` in many places.
2) It doesn't allow reporting unique warnings, because the `getGroups` helper is not
a member of `DumpStyle<ELFT>`.
3) It might just crash. See the comment for `StrTableOrErr->data() + Sym.st_name` line.
In this patch I am starting addressing these points.
For start I've converted one of `unwrapOrError` calls to a unique warning.
Differential revision: https://reviews.llvm.org/D91798
Our `printStackSize` implementation currently uses
API like `RelocationRef`, `object::symbol_iterator`.
It is not ideal as it doesn't allow
to handle possible error conditions properly.
Some time ago I started rewriting it and this NFC patch is
a one more step toward to it. Here I am introducing the
`forEachRelocationDo` helper. With it it is possible to iterate
over all kinds of relocations, what is helpful for improving
the code in `printStackSize` and around.
Differential revision: https://reviews.llvm.org/D91530
William S. Moses