By llvm-mca analysis, Haswell/Broadwell has the worst v4i64 recip-throughput cost of the AVX2 targets at 6 (vs the currently used cost of 8). Similarly SkylakeServer (our only AVX512 target model) implements PMULLQ with an average cost of 1.5 (rounded up to 2.0), and the PMULUDQ-sequence (without AVX512DQ) as a cost of 6.
This patch adds a new combine that tries to scalarize chains of
`extractelement (load %ptr), %idx` to `load (gep %ptr, %idx)`. This is
profitable when extracting only a few elements out of a large vector.
At the moment, `store (extractelement (load %ptr), %idx), %ptr`
operations on large vectors result in huge code in the backend.
This can easily be triggered by using the matrix extension, e.g.
https://clang.godbolt.org/z/qsccPdPf4
This should complement D98240.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D100273
If we simplify values we sometimes end up with type mismatches. If the
value is a constant we can often cast it though to still allow
propagation. The logic is now put into a helper and it replaces some
ad hoc things we did before.
This also introduces the AA namespace for abstract attribute related
functions and types.
We have seen various problems when the call graph was not updated or
the updated did not succeed because it involved functions outside the
SCC. This patch adds assertions and checks to avoid accidentally
changing something outside the SCC that would impact the call graph.
It also prevents us from reanalyzing functions outside the current
SCC which could cause problems on its own. Note that the transformations
we do might cause the CG to be "more precise" but the original one would
always be a super set of the most precise one. Since the call graph is
by nature an approximation, it is good enough to have a super set of all
call edges.
The constant value lattice looks like this
```
<None>
|
<undef>
/ | \
... <0> ...
\ | /
<unknown>
```
We did not account for the undef and assumed a value meant we could not
change anymore. Now we actually check if we have the same value as
before, which will signal CHANGED to the users when we go from undef to
a specific constant.
This fixes, among other things, the bug exposed by @ipccp4 in
`value-simplify.ll`.
The state of AAPotentialValues tracks if undef is contained. It should
fold undef into the first non-undef value. However we missed a case
before. There was also a shadowing definition of two variables that
caused trouble. The test exposes both problems.
This makes it possible for targets to define their own MCObjectFileInfo.
This MCObjectFileInfo is then used to determine things like section alignment.
This is a follow up to D101462 and prepares for the RISCV backend defining the
text section alignment depending on the enabled extensions.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D101921
This test case would get miscompiled by the current version of
D102982, because unrolling does not respect the PreserveCondBr
flag for partial unrolling.
The prevailing style does not add the message. The directive name is not useful
because the next line replicates the error line which includes the directive.
* Removes unnecessary loop hints.
* Use RUN line with '-scalable-vectorization=preferred' instead of 'on'
for the maximize-bandwidth behaviour. This prepares the test for enabling
scalable vectorization; With a forced instruction-cost of 1, 'on' will
always favour fixed-width VF to be chosen, whereas with 'preferred'
we can check that the maximize-bandwidth option in combination with
scalable-vectorization=preferred actually picks a scalable VF.
* Renamed to scalable-vectorization.ll, because a follow-up patch will
test more than just analysis.
This patch changes LoopUnrollAndJamPass from FunctionPass to LoopNest pass.
The next patch will utilize LoopNest to effectively handle loop nests.
Reviewed By: Whitney
Differential Revision: https://reviews.llvm.org/D99149
This test case requires unrolling against a non-latch exit in
a multiple-exit loop with exiting latch. It's not covered by
exiting heuristics or the extension in D102635.
If exiting using _Exit or ExitProcess, DLLs are still unloaded
cleanly before exiting, running destructors and other cleanup in those
DLLs. When the caller expects to exit without cleanup, running
destructors in some loaded DLLs (which can be either libLLVM.dll or
e.g. libc++.dll) can cause deadlocks occasionally.
This is an alternative to D102684.
Differential Revision: https://reviews.llvm.org/D102944
These functions were marked unavailable for MSVC targets before,
within an "T.isOSWindows() && !T.isOSCygMing()" block, but these ones
are unavailable on MinGW targets too.
This avoids generating calls to stpcpy for MinGW targets, which has
been happening since 6dbf0cfcf789365493f70ae69df8a7a59be41c75 (in
some cases).
This fixes https://github.com/mstorsjo/llvm-mingw/issues/201.
Differential Revision: https://reviews.llvm.org/D102946
Based on worst case of sandybridge (vs btver2 + bdver2) llvm-mca analysis - which is a lot less than what we were predicting (I think based off total uop count).
When removing an AttrBuilder from an AttributeSet, first check
whether there is any overlap. If nothing is being removed, we can
directly return the original set.
Don't run tasks until their corresponding thread has been added to the running
threads vector. This is an extention to fda4300da82, which doesn't seem to have
been enough to fix the synchronization issues on its own.
Keeping these bitfields from Block to Addressable allows them to be packed with
the bitfields at the end of Addressable, reducing the size of Block by eight
bytes.
Add options -[no-]offload-lto and -foffload-lto=[thin,full] for controlling
LTO for offload compilation. Allow LTO for AMDGPU target.
AMDGPU target does not support codegen of object files containing
call of external functions, therefore the LLVM module passed to
AMDGPU backend needs to contain definitions of all the callees.
An LLVM option is added to allow function importer to import
functions with noinline attribute.
HIP toolchain passes proper LLVM options to lld to make sure
function importer imports definitions of all the callees.
Reviewed by: Teresa Johnson, Artem Belevich
Differential Revision: https://reviews.llvm.org/D99683
This was reverted due to performance regressions in ARM benchmarks,
which have since been addressed by D101196 (SCEV analysis improvement)
and D101778 (CGP reverse transform).
-----
The single-use case is handled implicity by converting the icmp
into a mask check first. When comparing with zero in particular,
we don't need the one-use restriction, as we only produce a single
icmp.
https://alive2.llvm.org/ce/z/MSixcmhttps://alive2.llvm.org/ce/z/GwpG0M
If there are no matrix intrinsics in a function, we can directly bail
out, as there's nothing left to do.
Reviewed By: anemet
Differential Revision: https://reviews.llvm.org/D102931
BTVER2 has a 2 cycle throughput for v4i32 multiplies (same as SSE41 targets), which is only partially hidden by the subvector extracts/insert when splitting v8i32.
Now that getMemoryOpCost() correctly handles all the vector variants,
we should no longer hand-roll our own version of it, but use it directly.
The AVX512 variant probably needs a similar change,
but there it is less obvious.
This was initially landed in 69ed93a4355123a45c1d7216aea7cd53d07a361b,
but was reverted in 6b95fd199d96e3ba5c28a23b17b74203522bdaa8
because the patch it depends on was reverted.
