VSX introduced some permute instructions that are direct
replacements for Altivec ones except they can target all
the VSX registers. We have added code generation for most
of these but somehow missed the low/hi word merges (XXMRG[LH]W).
This caused some additional spills on some large
computationally intensive code.
This patch simply adds the missed patterns.
This simple heuristic uses the estimated live range length combined
with the number of registers in the class to switch which heuristic to
use. This was taking the raw number of registers in the class, even
though not all of them may be available. AMDGPU heavily relies on
dynamically reserved numbers of registers based on user attributes to
satisfy occupancy constraints, so the raw number is highly misleading.
There are still a few problems here. In the original testcase that
made me notice this, the live range size is incorrect after the
scheduler rearranges instructions, since the instructions don't have
the original InstrDist offsets. Additionally, I think it would be more
appropriate to use the number of disjointly allocatable registers in
the class. For the AMDGPU register tuples, there are a large number of
registers in each tuple class, but only a small fraction can actually
be allocated at the same time since they all overlap with each
other. It seems we do not have a query that corresponds to the number
of independently allocatable registers. Relatedly, I'm still debugging
some allocation failures where overlapping tuples seem to not be
handled correctly.
The test changes are mostly noise. There are a handful of x86 tests
that look like regressions with an additional spill, and a handful
that now avoid a spill. The worst looking regression is likely
test/Thumb2/mve-vld4.ll which introduces a few additional
spills. test/CodeGen/AMDGPU/soft-clause-exceeds-register-budget.ll
shows a massive improvement by completely eliminating a large number
of spills inside a loop.
This patch includes the following updates to the load/store refactoring effort introduced in D93370:
- Update various VSX patterns that use to "force" an XForm, to instead just XForm.
This allows the ability for the patterns to compute the most optimal addressing
mode (and to produce a DForm instruction when possible)
- Update pattern and test case for the LXVD2X/STXVD2X intrinsics
- Update LIT test cases that use to use the XForm instruction to use the DForm instruction
Differential Revision: https://reviews.llvm.org/D95115
Extend shuffle canonicalization and conversion of shuffles fed by vectorized
scalars to big endian subtargets. For big endian subtargets, loads and direct
moves of scalars into vector registers put the data in the correct element for
SCALAR_TO_VECTOR if the data type is 8 bytes wide. However, if the data type is
narrower, the value still ends up in the wrong place - althouth a different
wrong place than on little endian targets.
This patch extends the combine that keeps values where they are if they feed a
shuffle to big endian targets.
Differential revision: https://reviews.llvm.org/D100478
Summary: Some constants can be handled with less instructions than our current results. And it seems our original approach is not very easy to extend. Therefore this patch proposes to materialize all 64-bit constants by enumerated patterns.
I traversed almost all constants to verified the functionality of these pattens. A traversed comparison of the number of instructions used by the original method and the new method has also been completed, where no degradation was caused by this patch. This patch also passed Bootstrap test and SPEC test.
Improvements of this patch are shown in llvm/test/CodeGen/PowerPC/constants-i64.ll
Reviewed By: steven.zhang, stefanp
Differential Revision: https://reviews.llvm.org/D92089
tryLatency compares two sched candidates. For the top zone it prefers
the one with lesser depth, but only if that depth is greater than the
total latency of the instructions we've already scheduled -- otherwise
its latency would be hidden and there would be no stall.
Unfortunately it only tests the depth of one of the candidates. This can
lead to situations where the TopDepthReduce heuristic does not kick in,
but a lower priority heuristic chooses the other candidate, whose depth
*is* greater than the already scheduled latency, which causes a stall.
The fix is to apply the heuristic if the depth of *either* candidate is
greater than the already scheduled latency.
All this also applies to the BotHeightReduce heuristic in the bottom
zone.
Differential Revision: https://reviews.llvm.org/D72392
We currently miss a number of opportunities to emit single-instruction
VMRG[LH][BHW] instructions for shuffles on little endian subtargets. Although
this in itself is not a huge performance opportunity since loading the permute
vector for a VPERM can always be pulled out of loops, producing such merge
instructions is useful to downstream optimizations.
Since VPERM is essentially opaque to all subsequent optimizations, we want to
avoid it as much as possible. Other permute instructions have semantics that can
be reasoned about much more easily in later optimizations.
This patch does the following:
- Canonicalize shuffles so that the first element comes from the first vector
(since that's what most of the mask matching functions want)
- Switch the elements that come from splat vectors so that they match the
corresponding elements from the other vector (to allow for merges)
- Adds debugging messages for when a shuffle is matched to a VPERM so that
anyone interested in improving this further can get the info for their code
Differential revision: https://reviews.llvm.org/D77448
This patch introduces a TargetLowering query, isMulhCheaperThanMulShift.
Currently in DAG Combine, it will transform mulhs/mulhu into a
wider multiply and a shift if the wide multiply is legal.
This TLI function is implemented on 64-bit PowerPC, as it is more desirable to
have multiply-high over multiply + shift for words and doublewords. Having
multiply-high can also aid in further transformations that can be done.
Differential Revision: https://reviews.llvm.org/D78271
Summary:
When doing the conversion: MachineInst -> MCInst, we should ignore the
implicit operands, it will expose more opportunity for InstiAlias.
Reviewed By: steven.zhang
Differential Revision: https://reviews.llvm.org/D77118
We can only report the knownbits for a SCALAR_TO_VECTOR node if we only demand the 0'th element - the upper elements are undefined and shouldn't be trusted.
This is causing a number of regressions that need addressing but we need to get the bugfix in first.
Added some tests testing urem and srem operations with a constant divisor.
Patch by TG908 (Tim Gymnich)
Differential Revision: https://reviews.llvm.org/D68421
llvm-svn: 373830
