Peephole opt has a one use limitation which appears to be accidental. The function being used was incorrectly documented as returning whether the def had one *user*, but instead returned true only when there was one *use*. Add a corresponding hasOneNonDbgUser helper, and adjust peephole-opt to use the appropriate one.
All of the actual folding code handles multiple uses within a single instruction. That codepath is well exercised through instruction selection.
Differential Revision: https://reviews.llvm.org/D63656
llvm-svn: 364336
We currently have some isel patterns for treating vzmovl+load the same as vzload, but that shrinks the load which we shouldn't do if the load is volatile.
Rather than adding isel checks for volatile. This patch removes the patterns and teachs DAG combine to merge them into vzload when its legal to do so.
Differential Revision: https://reviews.llvm.org/D63665
llvm-svn: 364333
Change the generic ctpop expansion to more efficiently handle a
check for not-a-power-of-two value:
(ctpop x) != 1 --> (x == 0) || ((x & x-1) != 0)
This is the inverted predicate sibling pattern that was added with:
D63004
This should have been done before I changed IR canonicalization to
favor this form with:
rL364246
...so if this requires revert/changing, the earlier commit may also
need to modified.
llvm-svn: 364319
Simplify ZERO_EXTEND_VECTOR_INREG if the extended bits are not required.
Matches what we already do for ZERO_EXTEND.
Reapplies rL363850 but now with legality checks added at rL364290
llvm-svn: 364303
Simplify SIGN_EXTEND_VECTOR_INREG if the extended bits are not required/known zero.
Matches what we already do for SIGN_EXTEND.
Reapplies rL363802 but now with legality checks added at rL364290
llvm-svn: 364299
As part of the fix for rL364264 + rL364272 - limit the *_EXTEND conversion to !TLO.LegalOperations || isOperationLegal cases.
We'll improve X86 legality in future commits.
llvm-svn: 364290
Summary:
This addresses the regression that is being exposed by D50222 in `test/CodeGen/X86/jump_sign.ll`
The missing fold, at least partially, looks trivial:
https://rise4fun.com/Alive/Zsln
i.e. if we are comparing with zero, and comparing the `urem`-by-non-power-of-two,
and the `urem` is of something that may at most have a single bit set (or no bits set at all),
the `urem` is not needed.
Reviewers: RKSimon, craig.topper, xbolva00, spatel
Reviewed By: xbolva00, spatel
Subscribers: xbolva00, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D63390
llvm-svn: 364286
I've been trying to fix this, but hit some roadblocks. So I'm
committing the test case for now so we'll at least avoid
recreating that failure.
llvm-svn: 364272
This reverts the following patches.
"[TargetLowering] SimplifyDemandedBits SIGN_EXTEND_VECTOR_INREG -> ANY/ZERO_EXTEND_VECTOR_INREG"
"[TargetLowering] SimplifyDemandedBits ZERO_EXTEND_VECTOR_INREG -> ANY_EXTEND_VECTOR_INREG"
"[TargetLowering] SimplifyDemandedBits - add ANY_EXTEND_VECTOR_INREG support"
We can end up with an any_extend_vector_inreg with a 256 bit result type
and a 128 bit result type. This is allowed by the ISD opcode, but the
generic operation legalizer is only able to expand cases where the
total vector width is the same.
The X86 backend creates these mismatched cases for zext_vec_inreg/sext_vec_inreg.
The SimplifyDemandedBits changes are allowing those nodes to become
aext_vec_inreg. For the zext/sext cases, the X86 backend has Custom
handling and never lets them get to the generic legalizer. We need to do the same
for aext_vec_inreg.
llvm-svn: 364264
Widen vector result type for ctlz_zero_undef and cttz_zero_undef the same as
ctlz and cttz.
Differential Revision: https://reviews.llvm.org/D63463
llvm-svn: 364221
This patch does a few things to start cleaning up the isFNEG function.
-Remove the Op0/Op1 peekThroughBitcast calls that seem unnecessary. getTargetConstantBitsFromNode has its own peekThroughBitcast inside. And we have a separate peekThroughBitcast on the return value.
-Add a check of the scalar size after the first peekThroughBitcast to ensure we haven't changed the element size and just did something like f32->i32 or f64->i64.
-Remove an unnecessary check that Op1's type is floating point after the peekThroughBitcast. We're just going to look for a bit pattern from a constant. We don't care about its type.
-Add VT checks on several places that consume the return value of isFNEG. Due to the peekThroughBitcasts inside, the type of the return value isn't guaranteed. So its not safe to use it to build other nodes without ensuring the type matches the type being used to build the node. We might be able to replace these checks with bitcasts instead, but I don't have a test case so a bail out check seemed better for now.
Differential Revision: https://reviews.llvm.org/D63683
llvm-svn: 364206
This can occur under certain circumstances when undefs are created later on in the constant multipliers (e.g. in this case due to SimplifyDemandedVectorElts). Its better to let the shift by zero to occur and perform any cleanup afterward.
Fixes OSS Fuzz #15429
llvm-svn: 364179
Ideally we'd be able to represent this truncate as a any_extend to
v16i32 and a truncate, but SelectionDAG doens't know how to not
fold those together.
We have isel patterns to use a vpmovzxwd+vpdmovdb for the truncate,
but we aren't able to simultaneously fold the load and the store
from the isel pattern. By pulling the truncate into the store we
can successfully hide it from the DAG combiner. Then we can isel
pattern match the truncstore and load+any_extend separately.
llvm-svn: 364163
128/256 bit scalar_to_vectors are canonicalized to (insert_subvector undef, (scalar_to_vector), 0). We have isel patterns that try to match this pattern being used by a vzmovl to use a 128-bit instruction and a subreg_to_reg.
This patch detects the insert_subvector undef portion of this and pulls it through the vzmovl, creating a narrower vzmovl and an insert_subvector allzeroes. We can then match the insertsubvector into a subreg_to_reg operation by itself. Then we can fall back on existing (vzmovl (scalar_to_vector)) patterns.
Note, while the scalar_to_vector case is the motivating case I didn't restrict to just that case. I'm also wondering about shrinking any 256/512 vzmovl to an extract_subvector+vzmovl+insert_subvector(allzeros) but I fear that would have bad implications to shuffle combining.
I also think there is more canonicalization we can do with vzmovl with loads or scalar_to_vector with loads to create vzload.
Differential Revision: https://reviews.llvm.org/D63512
llvm-svn: 364095
We already use vmovq for v2i64/v2f64 vzmovl. But we were using a
blendpd+xorpd for v4i64/v4f64/v8i64/v8f64 under opt speed. Or
movsd+xorpd under optsize.
I think the blend with 0 or movss/d is only needed for
vXi32 where we don't have an instruction that can move 32
bits from one xmm to another while zeroing upper bits.
movq is no worse than blendpd on any known CPUs.
llvm-svn: 364079
We sometimes get poor code size because constants of types < 32b are legalized
as 32 bit G_CONSTANTs with a truncate to fit. This works but means that the
localizer can no longer sink them (although it's possible to extend it to do so).
On AArch64 however s8 and s16 constants can be selected in the same way as s32
constants, with a mov pseudo into a W register. If we make s8 and s16 constants
legal then we can avoid unnecessary truncates, they can be CSE'd, and the
localizer can sink them as normal.
There is a caveat: if the user of a smaller constant has to widen the sources,
we end up with an anyext of the smaller typed G_CONSTANT. This can cause
regressions because of the additional extend and missed pattern matching. To
remedy this, there's a new artifact combiner to generate the wider G_CONSTANT
if it's legal for the target.
Differential Revision: https://reviews.llvm.org/D63587
llvm-svn: 364075
The sat add/sub tests still have unnecessary extract_subvector((vandnps ymm, ymm), 0) uses that should be split to (vandnps (extract_subvector(ymm, 0), extract_subvector(ymm, 0)), but its getting better.
llvm-svn: 364038
Summary:
BLSI sets the C flag is the input is not zero. So if its followed
by a TEST of the input where only the Z flag is consumed, we can
replace it with the opposite check of the C flag.
We should be able to do the same for BLSMSK and BLSR, but the
naive test case for those is being optimized to a subo by
CodeGenPrepare.
Reviewers: spatel, RKSimon
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D63589
llvm-svn: 363957
This is an exception to the rule that we should prefer xmm ops to ymm ops.
As shown in PR42305:
https://bugs.llvm.org/show_bug.cgi?id=42305
...the store folding opportunity with vextractf128 may result in better
perf by reducing the instruction count.
Differential Revision: https://reviews.llvm.org/D63517
llvm-svn: 363853
We already do this for ZERO_EXTEND/ZERO_EXTEND_VECTOR_INREG - this just extends the pattern matcher to recognize cases where we don't need the zeros in the extension.
llvm-svn: 363841
Use matchBinaryPredicate instead of isConstOrConstSplat to let us handle non-uniform shift cases.
This requires us to tweak matchBinaryPredicate to allow it to (optionally) handle constants with different type widths.
llvm-svn: 363792
This allows targets to make more decisions about reserved registers
after isel. For example, now it should be certain there are calls or
stack objects in the frame or not, which could have been introduced by
legalization.
Patch by Matthias Braun
llvm-svn: 363757
Part of fixing the X86 regression noted in D63281 - I've split this into X86 and generic parts - the generic commit will be coming shortly and will fix the vector-reduce-mul-widen.ll regression introduced here.
llvm-svn: 363693
This adds vector splitting for vaarg instructions during type legalization
Committed on behalf of @luke (Luke Lau)
Differential Revision: https://reviews.llvm.org/D60762
llvm-svn: 363671
The isel patterns for these use a bitcast and load/store, but
DAG combine should have canonicalized those away.
For the purposes of the memory folding table these opcodes can be
replaced by the MOVSSrm_alt/MOVSDrm_alt and MOVSSmr/MOVSDmr opcodes.
llvm-svn: 363644
Rename the old versions that use FR32/FR64 to MOVSSrm_alt/MOVSDrm_alt.
Use the new versions in patterns that previously used a COPY_TO_REGCLASS
to VR128. These patterns expect the upper bits to be zero. The
current set up appears to work, but I'm not sure we should be
enforcing upper bits being zero through a COPY_TO_REGCLASS.
I wanted to flip the arrangement and use a COPY_TO_REGCLASS to
FR32/FR64 for the patterns that need an f32/f64 result, but that
complicated fastisel and globalisel.
I've been doing some experiments with reducing some isel patterns
and ended up in a situation where I had a
(SUBREG_TO_REG (COPY_TO_RECLASS (VMOVSSrm), VR128)) and our
post-isel peephole was unable to avoid using an instruction for
the SUBREG_TO_REG due to the COPY_TO_REGCLASS. Having a VR128
instruction removes the COPY_TO_REGCLASS that was breaking this.
llvm-svn: 363643
If a XMM non-temporal store has less than natural alignment, scalarize the vector - with SSE4A we can stay on the vector and use MOVNTSD(f64), else we must move to GPRs and use MOVNTI(i32/i64).
llvm-svn: 363592
If a YMM/ZMM non-temporal store has less than natural alignment, split the vector - either they will be satisfactorily aligned or will continue to be split until they are XMMs - at which point the legalizer will scalarize it.
llvm-svn: 363582
For loads, pre-SSE41 we can't perform NT loads at all, and after that we can only perform vector aligned loads, so if the alignment is less than for a xmm we'll just end up using the regular unaligned vector loads anyway.
First step towards fixing PR42026 - the next step for stores will be to use SSE4A movntsd where possible and to avoid the stack spill on SSE2 targets.
Differential Revision: https://reviews.llvm.org/D63246
llvm-svn: 363564
This is already done in DAGCombiner::visitINSERT_SUBVECTOR, but this helps a number of shuffles across different vector widths recognise when they come from the same source.
llvm-svn: 363542
This patch changes MIR stack-id from an integer to an enum,
and adds printing/parsing support for this in MIR files. The default
stack-id '0' is now renamed to 'default'.
This should make MIR tests that have stack objects with different stack-ids
more descriptive. It also clarifies code operating on StackID.
Reviewers: arsenm, thegameg, qcolombet
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D60137
llvm-svn: 363533
This is based on the example/discussion in PR37428:
https://bugs.llvm.org/show_bug.cgi?id=37428
Proper vector shift instructions don't appear until AVX2, so we may generate several
extra instructions within a loop trying to compensate for that. It's difficult to
recover from that shift expansion later than this, so use the existing TLI hook and
splat analysis to enable better codegen.
This extends CGP functionality introduced with:
rL201655
Differential Revision: https://reviews.llvm.org/D63233
llvm-svn: 363511
This is similar logic/motivation to the select splitting in D62969.
In D63233, the pattern changes so that we no longer have an extract_subvector of vselect,
but the operands of the select are still being concatenated.
The closest case is represented in either the first or last test diffs here - we have an
extra instruction, but we converted 3-4 ymm instructions into 4-5 xmm instructions.
I think that's the right trade-off for most AVX1 targets.
In the example based on PR37428:
https://bugs.llvm.org/show_bug.cgi?id=37428
...this makes the loop about 30% faster (tested on Haswell by compiling with -mavx).
Differential Revision: https://reviews.llvm.org/D63364
llvm-svn: 363508
This reverts rL363474. -debug-only=isel was added to some tests that
don't specify `REQUIRES: asserts`. This causes failures on
-DLLVM_ENABLE_ASSERTIONS=off builds.
I chose to revert instead of fixing the tests because I'm not sure
whether we should add `REQUIRES: asserts` to more tests.
llvm-svn: 363482
Current findBestLoopTop can find and move one kind of block to top, a latch block has one successor. Another common case is:
* a latch block
* it has two successors, one is loop header, another is exit
* it has more than one predecessors
If it is below one of its predecessors P, only P can fall through to it, all other predecessors need a jump to it, and another conditional jump to loop header. If it is moved before loop header, all its predecessors jump to it, then fall through to loop header. So all its predecessors except P can reduce one taken branch.
Differential Revision: https://reviews.llvm.org/D43256
llvm-svn: 363471
I recently discovered a bug on the x86 platform: The fp80 type was not handled well by x86 for constrained floating point nodes, as their regular counterparts are replaced by extending loads and truncating stores during the preprocess phase. Normally, platforms don't have this issue, as they don't typically attempt to perform such legalizations during instruction selection preprocessing. Before this change, strict_fp nodes survived until they were mutated to normal nodes, which happened shortly after preprocessing on other platforms. This modification lowers these nodes at the same phase while properly utilizing the chain.5
Submitted by: Drew Wock <drew.wock@sas.com>
Reviewed by: Craig Topper, Kevin P. Neal
Approved by: Craig Topper
Differential Revision: https://reviews.llvm.org/D63271
llvm-svn: 363417
As suggested in the post-commit thread for rL363392 - it's
wasteful to have so many runs for larger tests. AVX1/AVX2
is what shows the diff and probably what matters most going
forward.
llvm-svn: 363411
Summary:
I found the following case having tail blocks with no successors merging opportunities after block placement.
Before block placement:
bb0:
...
bne a0, 0, bb2:
bb1:
mv a0, 1
ret
bb2:
...
bb3:
mv a0, 1
ret
bb4:
mv a0, -1
ret
The conditional branch bne in bb0 is opposite to beq.
After block placement:
bb0:
...
beq a0, 0, bb1
bb2:
...
bb4:
mv a0, -1
ret
bb1:
mv a0, 1
ret
bb3:
mv a0, 1
ret
After block placement, that appears new tail merging opportunity, bb1 and bb3 can be merged as one block. So the conditional constraint for merging tail blocks with no successors should be removed. In my experiment for RISC-V, it decreases code size.
Author of original patch: Jim Lin
Reviewers: haicheng, aheejin, craig.topper, rnk, RKSimon, Jim, dmgreen
Reviewed By: Jim, dmgreen
Subscribers: xbolva00, dschuff, javed.absar, sbc100, jgravelle-google, aheejin, kito-cheng, dmgreen, PkmX, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D54411
llvm-svn: 363284
Summary:
Relate bug: https://bugs.llvm.org/show_bug.cgi?id=37472
The shrink wrapping pass prematurally restores the stack, at a point where the stack might still be accessed.
Taking an exception can cause the stack to be corrupted.
As a first approach, this patch is overly conservative, assuming that any instruction that may load or store could access
the stack.
Reviewers: dmgreen, qcolombet
Reviewed By: qcolombet
Subscribers: simpal01, efriedma, eli.friedman, javed.absar, llvm-commits, eugenis, chill, carwil, thegameg
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D63152
llvm-svn: 363265
This patch uses the mechanism from D62995 to strengthen the
definitions of the reduction intrinsics by letting the scalar
result/accumulator type be overloaded from the vector element type.
For example:
; The LLVM LangRef specifies that the scalar result must equal the
; vector element type, but this is not checked/enforced by LLVM.
declare i32 @llvm.experimental.vector.reduce.or.i32.v4i32(<4 x i32> %a)
This patch changes that into:
declare i32 @llvm.experimental.vector.reduce.or.v4i32(<4 x i32> %a)
Which has the type-constraint more explicit and causes LLVM to check
the result type with the vector element type.
Reviewers: RKSimon, arsenm, rnk, greened, aemerson
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D62996
llvm-svn: 363240
$noreg was being used way more than it should have. We also had
xmm registers in addressing modes.
Mostly found by hacking the machine verifier to do some stricter
checking that happened to work for this test, but not sure if
generally applicable for other tests or other targets.
llvm-svn: 363231
This was using its own, outdated list of possible captures. This was
at minimum not catching cmpxchg and addrspacecast captures.
One change is now any volatile access is treated as capturing. The
test coverage for this pass is quite inadequate, but this required
removing volatile in the lifetime capture test.
Also fixes some infrastructure issues to allow running just the IR
pass.
Fixes bug 42238.
llvm-svn: 363169
This patch changes how LLVM handles the accumulator/start value
in the reduction, by never ignoring it regardless of the presence of
fast-math flags on callsites. This change introduces the following
new intrinsics to replace the existing ones:
llvm.experimental.vector.reduce.fadd -> llvm.experimental.vector.reduce.v2.fadd
llvm.experimental.vector.reduce.fmul -> llvm.experimental.vector.reduce.v2.fmul
and adds functionality to auto-upgrade existing LLVM IR and bitcode.
Reviewers: RKSimon, greened, dmgreen, nikic, simoll, aemerson
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D60261
llvm-svn: 363035
This behavior was added in r130928 for both FastISel and SD, and then
disabled in r131156 for FastISel.
This re-enables it for FastISel with the corresponding fix.
This is triggered only when FastISel can't lower the arguments and falls
back to SelectionDAG for it.
FastISel contains a map of "register fixups" where at the end of the
selection phase it replaces all uses of a register with another
register that FastISel sometimes pre-assigned. Code at the end of
SelectionDAGISel::runOnMachineFunction is doing the replacement at the
very end of the function, while other pieces that come in before that
look through the MachineFunction and assume everything is done. In this
case, the real issue is that the code emitting COPY instructions for the
liveins (physreg to vreg) (EmitLiveInCopies) is checking if the vreg
assigned to the physreg is used, and if it's not, it will skip the COPY.
If a register wasn't replaced with its assigned fixup yet, the copy will
be skipped and we'll end up with uses of undefined registers.
This fix moves the replacement of registers before the emission of
copies for the live-ins.
The initial motivation for this fix is to enable tail calls for
swiftself functions, which were blocked because we couldn't prove that
the swiftself argument (which is callee-save) comes from a function
argument (live-in), because there was an extra copy (vreg to vreg).
A few tests are affected by this:
* llvm/test/CodeGen/AArch64/swifterror.ll: we used to spill x21
(callee-save) but never reload it because it's attached to the return.
We now don't even spill it anymore.
* llvm/test/CodeGen/*/swiftself.ll: we tail-call now.
* llvm/test/CodeGen/AMDGPU/mubuf-legalize-operands.ll: I believe this
test was not really testing the right thing, but it worked because the
same registers were re-used.
* llvm/test/CodeGen/ARM/cmpxchg-O0.ll: regalloc changes
* llvm/test/CodeGen/ARM/swifterror.ll: get rid of a copy
* llvm/test/CodeGen/Mips/*: get rid of spills and copies
* llvm/test/CodeGen/SystemZ/swift-return.ll: smaller stack
* llvm/test/CodeGen/X86/atomic-unordered.ll: smaller stack
* llvm/test/CodeGen/X86/swifterror.ll: same as AArch64
* llvm/test/DebugInfo/X86/dbg-declare-arg.ll: stack size changed
Differential Revision: https://reviews.llvm.org/D62361
llvm-svn: 362963
Summary:
Our default behavior is to use sign_extend for signed comparisons and zero_extend for everything else. But for equality we have the freedom to use either extension. If we can prove the input has been truncated from something with enough sign bits, we can use sign_extend instead and let DAG combine optimize it out. A similar rule is used by type legalization in LegalizeIntegerTypes.
This gets rid of the movzx in PR42189. The immediate will still take 4 bytes instead of the 2 bytes plus 0x66 prefix a cmp di, 32767 would get, but it avoids a length changing prefix.
Reviewers: RKSimon, spatel, xbolva00
Reviewed By: xbolva00
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D63032
llvm-svn: 362920
Previously we did the equivalent operation in isel patterns with
COPY_TO_REGCLASS operations to transition. By inserting
scalar_to_vetors and extract_vector_elts before isel we can
allow each piece to be selected individually and accomplish the
same final result.
I ideally we'd use vector operations earlier in lowering/combine,
but that looks to be more difficult.
The scalar-fp-to-i64.ll changes are because we have a pattern for
using movlpd for store+extract_vector_elt. While an f64 store
uses movsd. The encoding sizes are the same.
llvm-svn: 362914
This is the second part of the commit fixing PR38917 (hoisting
partitially redundant machine instruction). Most of PRE (partitial
redundancy elimination) and CSE work is done on LLVM IR, but some of
redundancy arises during DAG legalization. Machine CSE is not enough
to deal with it. This simple PRE implementation works a little bit
intricately: it passes before CSE, looking for partitial redundancy
and transforming it to fully redundancy, anticipating that the next
CSE step will eliminate this created redundancy. If CSE doesn't
eliminate this, than created instruction will remain dead and eliminated
later by Remove Dead Machine Instructions pass.
The third part of the commit is supposed to refactor MachineCSE,
to make it more clear and to merge MachinePRE with MachineCSE,
so one need no rely on further Remove Dead pass to clear instrs
not eliminated by CSE.
First step: https://reviews.llvm.org/D54839
Fixes llvm.org/PR38917
This is fixed recommit of r361356 after PowerPC64 multistage build failure.
llvm-svn: 362901
This is a potentially large perf win for AVX1 targets because of the way we
auto-vectorize to 256-bit but then expect the backend to legalize/optimize
for the half-implemented AVX1 ISA.
On the motivating example from PR37428 (even though this patch doesn't solve
the vector shift issue):
https://bugs.llvm.org/show_bug.cgi?id=37428
...there's a 16% speedup when compiling with "-mavx" (perf tested on Haswell)
because we eliminate the remaining 256-bit vblendv ops.
I added comments on a couple of tests that require further work. If we have
256-bit logic ops separating the vselect and extract, we should probably narrow
everything to 128-bit, but that requires a larger pattern match.
Differential Revision: https://reviews.llvm.org/D62969
llvm-svn: 362797
This primarily affects add/fadd/mul/fmul/and/or/xor/pmuludq/pmuldq/max/min/fmaxc/fminc/pmaddwd/pavg.
We already commuted the unmasked and zero masked versions.
I've added 512-bit stack folding tests for most of the instructions
affected. I've tested needing commuting and not commuting across
unmasked, merged masked, and zero masked. The 128/256 bit instructions
should behave similarly.
llvm-svn: 362746
This patch is the first step towards ensuring MergeConsecutiveStores correctly handles non-temporal loads\stores:
1 - When merging load\stores we must ensure that they all have the same non-temporal flag. This is unlikely to occur, but can in strange cases where we're storing at the end of one page and the beginning of another.
2 - The merged load\store node must retain the non-temporal flag.
Differential Revision: https://reviews.llvm.org/D62910
llvm-svn: 362723
This is intended to enable the use of an immediate blend or
more optimal instruction. But if the passthru is zero we don't
need any additional instructions.
llvm-svn: 362675
avx/avx2 masked loads only support all zeros for passthru in hardware.
So we have to emit a blend for all other values. We have an optimization
that tries to optimize this blend if the mask is constant. But we
don't need to perform this optimization if the passthru value is zero
which doesn't need the blend at all.
llvm-svn: 362674
As far as I know these should be freely reassociatable just like
the floating point MAXC/MINC instructions.
The *reduce* test changes are largely regressions and caused by
the "generic" CPU we default to not having a scheduler model.
The machine-combiner-int-vec.ll test shows the positive benefits
of this change.
Differential Revision: https://reviews.llvm.org/D62787
llvm-svn: 362629
As suggested in D62498 - collectConcatOps() matches both
concat_vectors and insert_subvector patterns, and we see
more test improvements by using the more general match.
llvm-svn: 362620
We already handle the case where we combine shuffle(extractsubvector(x),extractsubvector(x)), this relaxes the requirement to permit different sources as long as they have the same value type.
This causes a couple of cases where the VPERMV3 binary shuffles occur at a wider width than before, which I intend to improve in future commits - but as only the subvector's mask indices are defined, these will broadcast so we don't see any increase in constant size.
llvm-svn: 362599
Philip Reames