We had an zexti32 after a sign_extend_inreg. The AND X, 0xffffffff
part of the zexti32 should never occur since SimplifyDemandedBits
from the sign_extend_inreg would have removed it.
We also had sexti32 as the root node of a pattern, but SelectionDAGISel
matches assertsext early before the tablegen based patterns are
evaluated.
This patch updates algorithms in <numeric> to use std::move
based on p0616r0. Moving values instead of copying them
creates huge speed improvements (see the paper for details).
Differential Revision: https://reviews.llvm.org/D61170
SUMMARY:
Change geNumberOfVRSaved function name to getNumberOfVRSaved of class TBVectorExt
Reviewers: hubert.reinterpretcast, Jason Liu
Differential Revision: https://reviews.llvm.org/D92225
Also sync help texts for the option between elf and coff ports.
Decisions:
- Do this even if /lldignoreenv is passed. /reproduce: does not affect
the main output, and this makes the env var more convenient to use.
(On the other hand, it's now possible to set this env var and forget
about it, and all future builds in the same shell will be much slower.
That's true for ld.lld, but posix shells have an easy way to set an
env var for a single command; in cmd.exe this is not possible without
contortions. Then again, lld-link runs in posix shells too.)
Original patch rebased across D68378 and D68381.
Differential Revision: https://reviews.llvm.org/D67707
This patch implements the definition of __ARM_FEATURE_ATOMICS and fixes the
missing definition of __ARM_FEATURE_CRC32 for Armv8.1-A.
Differential Revision: https://reviews.llvm.org/D91438
We create threads using std::thread in various places in the test suite.
However, the usual std::thread constructor may not work on all platforms,
e.g. on platforms where passing a stack size is required to create a thread.
This commit introduces a simple indirection that makes it easier to tweak
how threads are created inside the test suite on various platforms. Note
that tests that are purposefully calling std::thread's constructor directly
(e.g. because that is what they're testing) were not modified.
Followup to D92112 now that I've learnt about HVX type splitting.
This is some necessary cleanup work for min/max ops to eventually help us move the add/sub sat patterns into DAGCombine - D91876.
Differential Revision: https://reviews.llvm.org/D92169
[libomptarget][cuda] Detect missing symbols in plugin at build time
Passes -z,defs to the linker. Error on unresolved symbol references.
Otherwise, those unresolved symbols present as target code running on the host
as the plugin fails to load. This is significantly harder to debug than a link
time error. Flag matches that passed by amdgcn and ve plugins.
Reviewed By: tianshilei1992
Differential Revision: https://reviews.llvm.org/D92143
By encoding ABI-affecting properties in the name of the ABI list, it
makes it clear when an ABI list test should or should not be available,
and what results we should expect.
Note that we clearly don't encode all ABI-affecting parameters in the
name right now -- I just ported over what we supported in the code that
was there previously. As we encounter configurations that we wish to
support but produce different ABI lists, we can add those to the ABI
identifier and start supporting them.
This commit also starts checking the ABI list in the CI jobs that run
a supported configuration. Eventually, all configurations should have
a generated ABI list and the test should even run implicitly as part of
the Lit test suite.
Differential Revision: https://reviews.llvm.org/D92194
This patch enables vector type arguments on AIX. All non-aggregate Altivec vector types are 16bytes in size and are 16byte aligned.
Reviewed By: Xiangling_L
Differential Revision: https://reviews.llvm.org/D92117
A splat attribute have a single element during printing so we should
treat it as such when we decide if we elide it or not based on the flag
intended to elide large attributes.
Reviewed By: rriddle, mehdi_amini
Differential Revision: https://reviews.llvm.org/D92165
Calling any of the OpenCLOptions::is*() functions (except isKnown())
with an unknown extension string results in a seg fault. This patch
checks that the extension exists in the map before attempting to access
it.
Reviewed By: Anastasia
Differential Revision: https://reviews.llvm.org/D90928
Many pages have had their titles renamed over time,
causing broken links to spread throughout the documentation.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D92093
This patch starts emitting the `EShNum` key, when the `e_shnum = 0`
and the section header table exists.
`e_shnum` might be 0, when the the number of entries in the section header
table is larger than or equal to SHN_LORESERVE (0xff00).
In this case the real number of entries
in the section header table is held in the `sh_size`
member of the initial entry in section header table.
Currently, obj2yaml crashes when an object has `e_shoff != 0` and the `sh_size`
member of the initial entry in section header table is `0`.
This patch fixes it.
Differential revision: https://reviews.llvm.org/D92098
The following line asserts when `sh_addralign > MAX_UINT32 && (uint32_t)sh_addralign == 0`:
```
ExpectedOffset = alignTo(ExpectedOffset,
SecHdr.sh_addralign ? SecHdr.sh_addralign : 1);
```
it happens because `sh_addralign` is truncated to 32-bit value, but `alignTo`
doesn't accept `Align == 0`. We should change `1` to `1uLL`.
Differential revision: https://reviews.llvm.org/D92163
If usubsat() is legal, this is likely to result in smaller codegen expansion than the default cmp+select codegen expansion.
Allows us to move the x86-specific lowering to the generic expansion code.
Differential Revision: https://reviews.llvm.org/D92183
The test case isn't using the AST matchers for all checks as there doesn't seem to be support for
matching NonTypeTemplateParmDecl default arguments. Otherwise this is simply importing the
default arguments.
Reviewed By: martong
Differential Revision: https://reviews.llvm.org/D92106
This patch adds tests for things that happened to be fixed by previous
patches, but that should continue working if we do decide to treat
sizeless types as incomplete types.
Differential Revision: https://reviews.llvm.org/D79584
Folding a select of vector constants that include undef elements only
applies to fixed vectors, but there's no earlier check the type is not
scalable so it crashes for scalable vectors. This adds a check so this
optimization is only attempted for fixed vectors.
Reviewed By: sdesmalen
Differential Revision: https://reviews.llvm.org/D92046
This was orginally committed in 2245fb8aaa.
but was immediately reverted in f3abd54958
because of a PHI handling issue.
Original commit message:
1. It doesn't make sense to enforce that the bonus instruction
is only used once in it's basic block. What matters is
whether those user instructions fit within our budget, sure,
but that is another question.
2. It doesn't make sense to enforce that said bonus instructions
are only used within their basic block. Perhaps the branch
condition isn't using the value computed by said bonus instruction,
and said bonus instruction is simply being calculated
to be used in successors?
So iff we can clone bonus instructions, to lift these restrictions,
we just need to carefully update their external uses
to use the new cloned instructions.
Notably, this transform (even without this change) appears to be
poison-unsafe as per alive2, but is otherwise (including the patch) legal.
We don't introduce any new PHI nodes, but only "move" the instructions
around, i'm not really seeing much potential for extra cost modelling
for the transform, especially since now we allow at most one such
bonus instruction by default.
This causes the fold to fire +11.4% more (13216 -> 14725)
as of vanilla llvm test-suite + RawSpeed.
The motivational pattern is IEEE-754-2008 Binary16->Binary32
extension code:
ca57d77fb2/src/librawspeed/common/FloatingPoint.h (L115-L120)
^ that should be a switch, but it is not now: https://godbolt.org/z/bvja5v
That being said, even thought this seemed like this would fix it: https://godbolt.org/z/xGq3TM
apparently that fold is happening somewhere else afterall,
so something else also has a similar 'artificial' restriction.
The ops are very similar to the std variants, but support async GPU execution.
gpu.alloc does not currently support an alignment attribute, and the new ops do not have
canonicalizers/folders like their std siblings do.
Reviewed By: herhut
Differential Revision: https://reviews.llvm.org/D91698
This is related to MIPS. Currently we might report an error and exit,
though there is no problem to report a warning and try to continue dumping
an object. The code uses `MipsGOTParser<ELFT> Parser`, which is isolated
in this method.
Differential revision: https://reviews.llvm.org/D92090
These patterns are using zexti32 which matches either assertzexti32
or (and X, 0xffffffff). But if we match (and X, 0xffffffff) it will
remove the AND and the inputs may no longer have the zero bits
needed to guarantee the result has enough zeros.
This commit changes the patterns to only match assertzexti32.
I'm not sure how to test the broken case since the DIVUW/REMUW nodes
are created during type legalization, but type legalization won't
create an (and X, 0xfffffffff) directly on the inputs.
I've also changed the zexti32 on the root of the pattern to just
checking for AND. We were previously also matching assertzexti32,
but I doubt that pattern would ever occur.
When widening an IndVar that has LCSSA Phi users outside
the loop, we can safely widen it as usual and then truncate
the result outside the loop without hurting the performance.
Differential Revision: https://reviews.llvm.org/D91593
Reviewed By: skatkov
For now, we will hardcode the result as 0.0 if the input is denormal or 0. That will
have the impact the precision. As the fsqrt added belong to the cold path of the
cmp+branch, it won't impact the performance for normal inputs for PowerPC, but improve
the precision if the input is denormal.
Reviewed By: Spatel
Differential Revision: https://reviews.llvm.org/D80974
Craig Topper