This reverts commit r299287 plus clean-ups.
The localizer pass is a helper pass that could be run at O0 in the GISel
pipeline to work around the deficiency of the fast register allocator.
It basically shortens the live-ranges of the constants so that the
allocator does not spill all over the place.
Long term fix would be to make the greedy allocator fast.
llvm-svn: 304051
- Rewrite livein calculation to use the computeLiveIns() helper
function. This is slightly less efficient but easier to reason about
and doesn't unnecessarily add pristine and reserved registers[1]
- Zero the status register at the beginning of the loop to make sure it
has a defined value.
- Remove kill flags of values that need to stay alive throughout the loop.
[1] An upcoming commit of mine will tighten the MachineVerifier to catch
these.
llvm-svn: 304048
It's a workaround because the test was flakey passing to begin with, but
it looks like (going off commit history) it really did want to test in
the presence of debug info, so keep that behavior (by adding something
to the CU so it's not dropped) & restore the flakey pass in the process.
(added a FIXME in case someone else decides to look at it later)
llvm-svn: 304042
[AMDGPU] add intrinsic for s_getpc
Summary: The s_getpc instruction is exposed as intrinsic llvm.amdgcn.s.getpc.
Patch by Tim Corringham
llvm-svn: 304031
Re-commit r303938 and r303954 with a fix for addLiveIns(): the internal
addPristines() function must be called on an empty set or it may
accidentally reset saved registers.
- addLiveOutsNoPristines() needs to add callee saved registers that are
actually saved and restored somewhere to the set (they are not
pristine).
- Cleanup/rewrite the code for addLiveOuts()/addLiveOutsNoPristines().
This fixes the problem from D32156.
Differential Revision: https://reviews.llvm.org/D32464
llvm-svn: 304001
In the best case:
extract (binop (concat X1, X2), (concat Y1, Y2)), N --> binop XN, YN
...we kill all of the extract/concat and just have narrow binops remaining.
If only one of the binop operands is amenable, this transform is still
worthwhile because we kill some of the extract/concat.
Optional bitcasting makes the code more complicated, but there doesn't
seem to be a way to avoid that.
The TODO about extending to more than bitwise logic is there because we really
will regress several x86 tests including madd, psad, and even a plain
integer-multiply-by-2 or shift-left-by-1. I don't think there's anything
fundamentally wrong with this patch that would cause those regressions; those
folds are just missing or brittle.
If we extend to more binops, I found that this patch will fire on at least one
non-x86 regression test. There's an ARM NEON test in
test/CodeGen/ARM/coalesce-subregs.ll with a pattern like:
t5: v2f32 = vector_shuffle<0,3> t2, t4
t6: v1i64 = bitcast t5
t8: v1i64 = BUILD_VECTOR Constant:i64<0>
t9: v2i64 = concat_vectors t6, t8
t10: v4f32 = bitcast t9
t12: v4f32 = fmul t11, t10
t13: v2i64 = bitcast t12
t16: v1i64 = extract_subvector t13, Constant:i32<0>
There was no functional change in the codegen from this transform from what I
could see though.
For the x86 test changes:
1. PR32790() is the closest call. We don't reduce the AVX1 instruction count in that case,
but we improve throughput. Also, on a core like Jaguar that double-pumps 256-bit ops,
there's an unseen win because two 128-bit ops have the same cost as the wider 256-bit op.
SSE/AVX2/AXV512 are not affected which is expected because only AVX1 has the extract/concat
ops to match the pattern.
2. do_not_use_256bit_op() is the best case. Everyone wins by avoiding the concat/extract.
Related bug for IR filed as: https://bugs.llvm.org/show_bug.cgi?id=33026
3. The SSE diffs in vector-trunc-math.ll are just scheduling/RA, so nothing real AFAICT.
4. The AVX1 diffs in vector-tzcnt-256.ll are all the same pattern: we reduced the instruction
count by one in each case by eliminating two insert/extract while adding one narrower logic op.
https://bugs.llvm.org/show_bug.cgi?id=32790
Differential Revision: https://reviews.llvm.org/D33137
llvm-svn: 303997
Currently getOptimalMemOpType returns i32 for large enough sizes without
checking for alignment, leading to poor code generation when misaligned accesses
aren't permitted as we generate a word store then later split it up into byte
stores. This means we inadvertantly go over the MaxStoresPerMemcpy limit and for
memset we splat the memset value into a word then immediately split it up
again.
Fix this by leaving it up to FindOptimalMemOpLowering to figure out which type
to use, but also fix a bug there where it wasn't correctly checking if
misaligned memory accesses are allowed.
Differential Revision: https://reviews.llvm.org/D33442
llvm-svn: 303990
I forgot to forward the chain, causing some missing instruction
dependencies. The test crashes the compiler without this patch.
Inspired by the test case, D33519 also tries to remove the extra sync.
Differential Revision: https://reviews.llvm.org/D33573
llvm-svn: 303931
Rename the DEBUG_TYPE to match the names of corresponding passes where
it makes sense. Also establish the pattern of simply referencing
DEBUG_TYPE instead of repeating the passname where possible.
llvm-svn: 303921
Summary:
This is used in the Linux kernel, and effectively just means "print an
address". This brings back r193593.
Reviewed by: Renato Golin
Reviewers: t.p.northover, rengolin, richard.barton.arm, kristof.beyls
Subscribers: aemerson, javed.absar, llvm-commits, eraman
Differential Revision: https://reviews.llvm.org/D33558
llvm-svn: 303901
AVX512_VPOPCNTDQ is a new feature set that was published by Intel.
The patch represents the LLVM side of the addition of two new intrinsic based instructions (vpopcntd and vpopcntq).
Differential Revision: https://reviews.llvm.org/D33169
llvm-svn: 303858
There are some VectorShuffle Nodes in SDAG which can be selected to XXSLDWI
instruction, this patch recognizes them and does the selection to improve the
PPC performance.
llvm-svn: 303822
Use ADDframe pseudo instruction instead.
This will fix machine verifier error, and will help to fix PR32146.
Differential Revision: https://reviews.llvm.org/D33452
llvm-svn: 303758
Summary:
This is a fix for PR32538. MachineCSE first looks at MO.isDead(), but
if it is not marked dead, MachineCSE still wants to do its own check
to see if it is trivially dead. This check for the trivial case
assumed that physical registers cannot be live out of a block.
Patch by Mattias Eriksson.
Reviewers: qcolombet, jbhateja
Reviewed By: qcolombet, jbhateja
Subscribers: jbhateja, llvm-commits
Differential Revision: https://reviews.llvm.org/D33408
llvm-svn: 303731
I suspect this buildbot has slow-incdec set by default, most likely due to
the default CPU having this set. This feature bit can prevent optsize from
having an effect on this IR.
llvm-svn: 303720
This fixes 17 of the 41 -verify-machineinstrs test failures identified in PR33045
Differential Revision: https://reviews.llvm.org/D33451
llvm-svn: 303691
Summary:
Promoting Alloca to Vector and Promoting Alloca to LDS are two independent handling of Alloca and should not affect each other.
As a result, we should not give up promoting to vector if there is not enough LDS. This patch factors out the local memory usage
related checking out and replace it after the calling convention checking.
Reviewer:
arsenm
Differential Revision:
http://reviews.llvm.org/D33139
llvm-svn: 303684
Perform DAG combine:
and (srl x, c), mask => shl (bfe x, nb + c, mask >> nb), nb
Where nb is a number of trailing zeroes in mask.
It replaces two instructions with two and BFE is generally a more
expensive one. However this is only done if we are selecting a byte
or word at an aligned boundary which results in a proper SDWA
operand pattern. It is only done if SDWA is supported.
TODO: improve SDWA pass to actually convert this pattern. It is not
done now because we have an immediate in the instruction, which has
be moved into a VGPR.
Differential Revision: https://reviews.llvm.org/D33455
llvm-svn: 303681
Summary:
A temporary workaround for PR32780 - rematerialized instructions accessing the same promoted global through different constant pool entries.
The patch turns off the globals promotion optimization leaving all its code in place, so that it can be easily turned on once PR32780 is fixed.
Since this is a miscompilation issue causing generation of misbehaving code, and the problem is very subtle, the patch might be valuable enough to get into 4.0.1.
Reviewers: efriedma, jmolloy
Reviewed By: efriedma
Subscribers: aemerson, javed.absar, llvm-commits, rengolin, asl, tstellar
Differential Revision: https://reviews.llvm.org/D33446
llvm-svn: 303679
Summary:
It's rare but a small number of patterns use IntInit's at the root of the match.
On X86, one such rule is enabled by the OptForSize predicate and causes the
compiler to use the smaller:
%0 = MOV32r1
instead of the usual:
%0 = MOV32ri 1
This patch adds support for matching IntInit's at the root and uses this as a
test case for the optsize attribute that was implemented in r301750
Reviewers: qcolombet, ab, t.p.northover, rovka, kristof.beyls, aditya_nandakumar
Reviewed By: qcolombet
Subscribers: igorb, llvm-commits
Differential Revision: https://reviews.llvm.org/D32791
llvm-svn: 303678
shl (or|add x, c2), c1 => or|add (shl x, c1), (c2 << c1)
This allows to fold a constant into an address in some cases as
well as to eliminate second shift if the expression is used as
an address and second shift is a result of a GEP.
Differential Revision: https://reviews.llvm.org/D33432
llvm-svn: 303641
Summary:
This patch makes instruction fusion more aggressive by
* adding artificial edges between the successors of FirstSU and
SecondSU, similar to BaseMemOpClusterMutation::clusterNeighboringMemOps.
* updating PostGenericScheduler::tryCandidate to keep clusters together,
similar to GenericScheduler::tryCandidate.
This change increases the number of AES instruction pairs generated on
Cortex-A57 and Cortex-A72. This doesn't change code at all in
most benchmarks or general code, but we've seen improvement on kernels
using AESE/AESMC and AESD/AESIMC.
Reviewers: evandro, kristof.beyls, t.p.northover, silviu.baranga, atrick, rengolin, MatzeB
Reviewed By: evandro
Subscribers: aemerson, rengolin, MatzeB, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D33230
llvm-svn: 303618
This commit fixes a bug introduced in r301019 where optimizeLogicalImm
would replace a logical node's immediate operand that was CSE'd and
was also an operand of another node.
This commit fixes the bug by replacing the logical node instead of its
immediate operand.
rdar://problem/32295276
llvm-svn: 303607
Turn expensive 64 bit shift into 32 bit if shift does not overflow int:
shl (ext x) => zext (shl x)
Differential Revision: https://reviews.llvm.org/D33367
llvm-svn: 303569
This patch adds handling of the `micromips` and `nomicromips` attributes
passed by front-end. The patch depends on D33363.
Differential revision: https://reviews.llvm.org/D33364
llvm-svn: 303545
PPC backend eliminates compare instructions by using record-form instructions in PPCInstrInfo::optimizeCompareInstr, which is called from peephole optimization pass.
This patch improves this optimization to eliminate more compare instructions in two types of common case.
- comparison against a constant 1 or -1
The record-form instructions set CR bit based on signed comparison against 0. So, the current implementation does not exploit the record-form instruction for comparison against a non-zero constant.
This patch enables record-form optimization for constant of 1 or -1 if possible; it changes the condition "greater than -1" into "greater than or equal to 0" and "less than 1" into "less than or equal to 0".
With this patch, compare can be eliminated in the following code sequence, as an example.
uint64_t a, b;
if ((a | b) & 0x8000000000000000ull) { ... }
else { ... }
- andi for 32-bit comparison on PPC64
Since record-form instructions execute 64-bit signed comparison and so we have limitation in eliminating 32-bit comparison, i.e. with cmplwi, using the record-form. The original implementation already has such checks but andi. is not recognized as an instruction which executes implicit zero extension and hence safe to convert into record-form if used for equality check.
%1 = and i32 %a, 10
%2 = icmp ne i32 %1, 0
br i1 %2, label %foo, label %bar
In this simple example, LLVM generates andi. + cmplwi + beq on PPC64.
This patch make it possible to eliminate the cmplwi for this case.
I added andi. for optimization targets if it is safe to do so.
Differential Revision: https://reviews.llvm.org/D30081
llvm-svn: 303500
Summary:
While this makes some case better and some case worse - so it's unclear if it is a worthy combine just by itself - this is a useful canonicalisation.
As per discussion in D32756 .
Reviewers: jyknight, nemanjai, mkuper, spatel, RKSimon, zvi, bkramer
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32916
llvm-svn: 303441
This patch defines the i1 type as illegal in the X86 backend for AVX512.
For DAG operations on <N x i1> types (build vector, extract vector element, ...) i8 is used, and should be truncated/extended.
This should produce better scalar code for i1 types since GPRs will be used instead of mask registers.
Differential Revision: https://reviews.llvm.org/D32273
llvm-svn: 303421
pruneSubRegValues() needs to remove subregister ranges starting at
instructions that later get removed by eraseInstrs(). It missed to check
one case in which eraseInstrs() would remove an instruction.
Fixes http://llvm.org/PR32688
llvm-svn: 303396
This also reverts follow-ups r303292 and r303298.
It broke some Chromium tests under MSan, and apparently also internal
tests at Google.
llvm-svn: 303369
This provides a new way to access the TargetMachine through
TargetPassConfig, as a dependency.
The patterns replaced here are:
* Passes handling a null TargetMachine call
`getAnalysisIfAvailable<TargetPassConfig>`.
* Passes not handling a null TargetMachine
`addRequired<TargetPassConfig>` and call
`getAnalysis<TargetPassConfig>`.
* MachineFunctionPasses now use MF.getTarget().
* Remove all the TargetMachine constructors.
* Remove INITIALIZE_TM_PASS.
This fixes a crash when running `llc -start-before prologepilog`.
PEI needs StackProtector, which gets constructed without a TargetMachine
by the pass manager. The StackProtector pass doesn't handle the case
where there is no TargetMachine, so it segfaults.
Related to PR30324.
Differential Revision: https://reviews.llvm.org/D33222
llvm-svn: 303360
Summary:
There should be no intesection between SDWA operands and potential MIs. E.g.:
```
v_and_b32 v0, 0xff, v1 -> src:v1 sel:BYTE_0
v_and_b32 v2, 0xff, v0 -> src:v0 sel:BYTE_0
v_add_u32 v3, v4, v2
```
In that example it is possible that we would fold 2nd instruction into 3rd (v_add_u32_sdwa) and then try to fold 1st instruction into 2nd (that was already destroyed). So if SDWAOperand is also a potential MI then do not apply it.
Reviewers: vpykhtin, arsenm
Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye
Differential Revision: https://reviews.llvm.org/D32804
llvm-svn: 303347
According to Intel's Optimization Reference Manual for SNB+:
" For LEA instructions with three source operands and some specific situations, instruction latency has increased to 3 cycles, and must
dispatch via port 1:
- LEA that has all three source operands: base, index, and offset
- LEA that uses base and index registers where the base is EBP, RBP,or R13
- LEA that uses RIP relative addressing mode
- LEA that uses 16-bit addressing mode "
This patch currently handles the first 2 cases only.
Differential Revision: https://reviews.llvm.org/D32277
llvm-svn: 303333
Partially implement callee-side for arguments and return values.
byval doesn't work properly, and most likely sret or other on-stack
return values most as well.
llvm-svn: 303308
When legalizing vector operations on vNi128, they will be split to v1i128
because that is a legal type on ppc64, but then the compiler will crash in
selection dag because it fails to select for these operations. This patch fixes
shift operations. Logical shift right and left shift can be performed in the
vector unit, but algebraic shift right requires being split.
Differential Revision: https://reviews.llvm.org/D32774
llvm-svn: 303307
- '-verify-mahcineinstrs' starts to complain allocatable live-in physical
registers on non-entry or non-landing-pad basic blocks.
- Refactor the XBEGIN translation to define EAX on a dedicated fallback code
path due to XABORT. Add a pseudo instruction to define EAX explicitly to
avoid add physical register live-in.
Differential Revision: https://reviews.llvm.org/D33168
llvm-svn: 303306
Summary: Moving LiveRangeShrink to x86 as this pass is mostly useful for archtectures with great register pressure.
Reviewers: MatzeB, qcolombet
Reviewed By: qcolombet
Subscribers: jholewinski, jyknight, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D33294
llvm-svn: 303292
Avoids instructions to pack a vector when the source is really
a scalar being broadcast.
Also be smarter and look for per-component fneg.
Doesn't yet handle scalar from upper half of register
or other swizzles.
llvm-svn: 303291
The variables MinGPR/MinG8R were not updated properly when resetting the
offsets, which in the included testcase lead to saving the CR register
in the same location as R30.
This fixes another issue reported in PR26519.
Differential Revision: https://reviews.llvm.org/D33017
llvm-svn: 303257
Summary:
This fixes pr32392.
The lowering pipeline is:
llvm.ppc.cfence in IR -> PPC::CFENCE8 in isel -> Actual instructions in
expandPostRAPseudo.
The reason why expandPostRAPseudo is chosen is because previous passes
are likely eliminating instructions like cmpw 3, 3 (early CSE) and bne-
7, .+4 (some branch pass(s)).
Differential Revision: https://reviews.llvm.org/D32763
llvm-svn: 303205
Summary:
In SelectionDAG, when a store is immediately chained to another store
to the same address, elide the first store as it has no observable
effects. This is causes small improvements dealing with intrinsics
lowered to stores.
Test notes:
* Many testcases overwrite store addresses multiple times and needed
minor changes, mainly making stores volatile to prevent the
optimization from optimizing the test away.
* Many X86 test cases optimized out instructions associated with
associated with va_start.
* Note that test_splat in CodeGen/AArch64/misched-stp.ll no longer has
dependencies to check and can probably be removed and potentially
replaced with another test.
Reviewers: rnk, john.brawn
Subscribers: aemerson, rengolin, qcolombet, jyknight, nemanjai, nhaehnle, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D33206
llvm-svn: 303198
According to Intel's Optimization Reference Manual for SNB+:
" For LEA instructions with three source operands and some specific situations, instruction latency has increased to 3 cycles, and must
dispatch via port 1:
- LEA that has all three source operands: base, index, and offset
- LEA that uses base and index registers where the base is EBP, RBP,or R13
- LEA that uses RIP relative addressing mode
- LEA that uses 16-bit addressing mode "
This patch currently handles the first 2 cases only.
Differential Revision: https://reviews.llvm.org/D32277
llvm-svn: 303183
Shrink-wrapping uses post-dominators to find a restore point that
post-dominates all the uses of CSR / stack.
The way dominator trees are modeled in LLVM today is that unreachable
blocks are not present in a generic dominator tree, so, an unreachable node is
dominated by anything: include/llvm/Support/GenericDomTree.h:467.
Since for post-dominators, a no-return block is considered
"unreachable", calling findNearestCommonDominator on an unreachable node
A and a non-unreachable node B, will return B, which can be false. If we
find such node, we bail out since there is no good restore point
available.
rdar://problem/30186931
llvm-svn: 303130
We don't use section-relative relocations on AArch64, so all symbols must be at
least visible to the linker (i.e. properly global or l_whatever, but not
L_whatever).
llvm-svn: 303118
This caused PR33053.
Original commit message:
> The new experimental reduction intrinsics can now be used, so I'm enabling this
> for AArch64. We will need this for SVE anyway, so it makes sense to do this for
> NEON reductions as well.
>
> The existing code to match shufflevector patterns are replaced with a direct
> lowering of the reductions to AArch64-specific nodes. Tests updated with the
> new, simpler, representation.
>
> Differential Revision: https://reviews.llvm.org/D32247
llvm-svn: 303115
At O3 we are more willing to increase size if we believe it will improve
performance. The current threshold for tail-duplication of 2 instructions is
conservative, and can be relaxed at O3.
Benchmark results:
llvm test-suite:
6% improvement in aha, due to duplication of loop latch
3% improvement in hexxagon
2% slowdown in lpbench. Seems related, but couldn't completely diagnose.
Internal google benchmark:
Produces 4% improvement on internal google protocol buffer serialization
benchmarks.
Differential-Revision: https://reviews.llvm.org/D32324
llvm-svn: 303084
Follow up to D33147
NVPTXTargetLowering::LowerCall was trusting the default argument values.
Fixes another 17 of the NVPTX '-verify-machineinstrs with EXPENSIVE_CHECKS' errors in PR32146.
Differential Revision: https://reviews.llvm.org/D33189
llvm-svn: 303082
This patch enables fusing dependent AESE/AESMC and AESD/AESIMC
instruction pairs on Cortex-A72, as recommended in the Software
Optimization Guide, section 4.10.
llvm-svn: 303073
Doing this means that if an LEApcrel is used in two places we will rematerialize
instead of generating two MOVs. This is particularly useful for printfs using
the same format string, where we want to generate an address into a register
that's going to get corrupted by the call.
Differential Revision: https://reviews.llvm.org/D32858
llvm-svn: 303054
Doing this lets us hoist it out of loops, and I've also marked it as
rematerializable the same as the thumb1 and thumb2 counterparts.
It looks like it being marked as such was just a mistake, as the commit that
made that change only mentions LEApcrelJT and in thumb1 and thumb2 only the
LEApcrelJT instructions were marked as having side-effects, so it looks like
the intent was to only mark LEApcrelJT as having side-effects but LEApcrel was
accidentally marked as such also.
Differential Revision: https://reviews.llvm.org/D32857
llvm-svn: 303053
Currently, when masked load, store, gather or scatter intrinsics are used, we check in CodeGenPrepare pass if the subtarget support these intrinsics, if not we replace them with scalar code - this is a functional transformation not an optimization (not optional).
CodeGenPrepare pass does not run when the optimization level is set to CodeGenOpt::None (-O0).
Functional transformation should run with all optimization levels, so here I created a new pass which runs on all optimization levels and does no more than this transformation.
Differential Revision: https://reviews.llvm.org/D32487
llvm-svn: 303050
I noticed the 512-bit lzcnts don't use the X86 specific lookup table code and instead use the EXPAND case in LegalizeDAG. I was toying around with fixing this and noticed it would require compare instructions that generate i1 masks and then converting from mask to vector. Then I noticed that we don't test which compares are used with avx512vl and no avx512cd.
llvm-svn: 303020
Remove an unneeded prefix from the 32-bit command line. Make all the 64-bit triples match. Replace ALL with X64 and remove it from the 32-bit test.
llvm-svn: 303019
Summary:
We should not change volatile loads/stores in promoting alloca to vector.
Reviewers:
arsenm
Differential Revision:
http://reviews.llvm.org/D33107
llvm-svn: 302943
This fixes 47 of the 75 NVPTX '-verify-machineinstrs with EXPENSIVE_CHECKS' errors in PR32146.
Differential Revision: https://reviews.llvm.org/D33147
llvm-svn: 302942
Summary: LiveRangeShrink pass moves instruction right after the definition with the same BB if the instruction and its operands all have more than one use. This pass is inexpensive and guarantees optimal live-range within BB.
Reviewers: davidxl, wmi, hfinkel, MatzeB, andreadb
Reviewed By: MatzeB, andreadb
Subscribers: hiraditya, jyknight, sanjoy, skatkov, gberry, jholewinski, qcolombet, javed.absar, krytarowski, atrick, spatel, RKSimon, andreadb, MatzeB, mehdi_amini, mgorny, efriedma, davide, dberlin, llvm-commits
Differential Revision: https://reviews.llvm.org/D32563
llvm-svn: 302938
Llvm-stress discovered that a COPY may end up in ExpandPostRA::LowerCopy()
with an undef source operand. It is not possible for the target to handle
this, as this flag is not passed to TII->copyPhysReg().
This patch solves this by treating such a COPY as an identity COPY.
Review: Matthias Braun
https://reviews.llvm.org/D32892
llvm-svn: 302877
Summary:
Instead of using RemoveExtraEdges (which uses analyzeBranch, which cannot
always be trusted) at the end to fixup the CFG we keep the CFG updated as
we go along and remove or add branches and merge blocks.
This way we won't have any problems if the involved MBBs contain
unanalyzable instructions.
This fixes PR32721.
In that case we had a triangle
EBB
| \
| |
| TBB
| /
FBB
where FBB didn't have any successors at all since it ended with an
unconditional return. Then TBB and FBB were be merged into EBB, but EBB
would still keep its successors, and the use of analyzeBranch and
CorrectExtraCFGEdges wouldn't help to remove them since the return
instruction is not analyzable (at least not on ARM).
Reviewers: kparzysz, iteratee, MatzeB
Reviewed By: iteratee
Subscribers: aemerson, rengolin, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D33037
llvm-svn: 302876
According to Power ISA V3.0 document, the first source operand of mtvsrdd is constant 0 if r0 is specified. So the corresponding register constraint should be g8rc_nox0.
This bug caused wrong output generated by 401.bzip2 when -mcpu=power9 and fdo are specified.
Differential Revision: https://reviews.llvm.org/D32880
llvm-svn: 302834
Updates the MSP430 target to generate EABI-compatible libcall names.
As a byproduct, adjusts the hardware multiplier options available in
the MSP430 target, adds support for promotion of the ISD::MUL operation
for 8-bit integers, and correctly marks R11 as used by call instructions.
Patch by Andrew Wygle.
Differential Revision: https://reviews.llvm.org/D32676
llvm-svn: 302820
We don't use it and it was removed in gfx9, and the encoding
bit repurposed.
Additionally actually using it requires changing the output register
class, which wasn't done anyway.
llvm-svn: 302814
This patch is the first in a series of patches to provide code gen for
doing compares in GPRs when the compare result is required in a GPR.
It adds the infrastructure to select GPR sequences for i1->i32 and i1->i64
extensions. This first patch handles equality comparison on i32 operands with
the result sign or zero extended.
Differential Revision: https://reviews.llvm.org/D31847
llvm-svn: 302810
manages to form a VSELECT with a non-i1 element type condition. Those
are technically allowed in SDAG (at least, the generic type legalization
logic will form them and I wouldn't want to try to audit everything te
preclude forming them) so we need to be able to lower them.
This isn't too hard to implement. We mark VSELECT as custom so we get
a chance in C++, add a fast path for i1 conditions to get directly
handled by the patterns, and a fallback when we need to manually force
the condition to be an i1 that uses the vptestm instruction to turn
a non-mask into a mask.
This, unsurprisingly, generates awful code. But it at least doesn't
crash. This was actually impacting open source packages built with LLVM
for AVX-512 in the wild, so quickly landing a patch that at least stops
the immediate bleeding.
I think I've found where to fix the codegen quality issue, but less
confident of that change so separating it out from the thing that
doesn't change the result of any existing test case but causes mine to
not crash.
llvm-svn: 302785
This is the same as r292827 for AArch64: we widen 8- and 16-bit ADD, SUB
and MUL to 32 bits since we only have TableGen patterns for 32 bits.
See the commit message for r292827 for more details.
At this point we could just remove some of the tests for regbankselect
and instruction-select, since we're not going to see any narrow
operations at those levels anymore. Instead I decided to update them
with G_ANYEXT/G_TRUNC operations, so we can validate the full sequences
generated by the legalizer.
llvm-svn: 302782
G_ANYEXT can be introduced by the legalizer when widening scalars. Add
support for it in the register bank info (same mapping as everything
else) and in the instruction selector.
When selecting it, we treat it as a COPY, just like G_TRUNC. On this
occasion we get rid of some assertions in selectCopy so we can reuse it.
This shouldn't be a problem at the moment since we're not supporting any
complicated cases (e.g. FPR, different register banks). We might want to
separate the paths when we do.
llvm-svn: 302778
This reverts r302712.
The change fails with ASAN enabled:
ERROR: AddressSanitizer: use-after-poison on address ... at ...
READ of size 2 at ... thread T0
#0 ... in llvm::SDNode::getNumValues() const <snip>/include/llvm/CodeGen/SelectionDAGNodes.h:855:42
#1 ... in llvm::SDNode::hasAnyUseOfValue(unsigned int) const <snip>/lib/CodeGen/SelectionDAG/SelectionDAG.cpp:7270:3
#2 ... in llvm::SDValue::use_empty() const <snip> include/llvm/CodeGen/SelectionDAGNodes.h:1042:17
#3 ... in (anonymous namespace)::DAGCombiner::MergeConsecutiveStores(llvm::StoreSDNode*) <snip>/lib/CodeGen/SelectionDAG/DAGCombiner.cpp:12944:7
Reviewers: niravd
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D33081
llvm-svn: 302746
Summary:
Allow consecutive stores whose values come from consecutive loads to
merged in the presense of other uses of the loads. Previously this was
disallowed as in general the merged load cannot be shared with the
other uses. Merging N stores into 1 may cause as many as N redundant
loads. However in the context of caching this should have neglible
affect on memory pressure and reduce instruction count making it
almost always a win.
Fixes PR32086.
Reviewers: spatel, jyknight, andreadb, hfinkel, efriedma
Reviewed By: efriedma
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D30471
llvm-svn: 302712
For stores, check if the stored value is defined by a floating point
instruction and if yes, we return a default mapping with FPR instead
of GPR.
llvm-svn: 302679
The new experimental reduction intrinsics can now be used, so I'm enabling this
for AArch64. We will need this for SVE anyway, so it makes sense to do this for
NEON reductions as well.
The existing code to match shufflevector patterns are replaced with a direct
lowering of the reductions to AArch64-specific nodes. Tests updated with the
new, simpler, representation.
Differential Revision: https://reviews.llvm.org/D32247
llvm-svn: 302678
Summary:
When trying to figure out if MBB could fallthrough to ToMBB (possibly by
falling through a bunch of other MBBs) we didn't actually check if there
was fallthrough between the last two blocks in the chain.
Reviewers: kparzysz, iteratee, MatzeB
Reviewed By: kparzysz, iteratee
Subscribers: javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D32996
llvm-svn: 302650
This method must return a valid register class, or the list-ilp isel
scheduler will crash. For MVT::Untyped nullptr was previously returned, but
now ADDR128BitRegClass is returned instead. This is needed just as long as
list-ilp (and probably also list-hybrid) is still there.
Review: Ulrich Weigand, A Trick
https://reviews.llvm.org/D32802
llvm-svn: 302649
This pass uses a new target hook to decide whether or not to expand a particular
intrinsic to the shuffevector sequence.
Differential Revision: https://reviews.llvm.org/D32245
llvm-svn: 302631
This is a follow-up to r302611, which moved an -O0 computation of DT
from SDAGISel to TwoAddress.
Don't use it here either, and avoid computing it completely. The only
use was forwarding the analysis as an optional argument to utility
functions.
Differential Revision: https://reviews.llvm.org/D32766
llvm-svn: 302612
Before r247167, the pass manager builder controlled which AA
implementations were used, exporting them all in the AliasAnalysis
analysis group.
Now, AAResultsWrapperPass always uses BasicAA, but still uses other AA
implementations if made available in the pass pipeline.
But regardless, SDAGISel is required at O0, and really doesn't need to
be doing fancy optimizations based on useful AA results.
Don't require AA at CodeGenOpt::None, and only use it otherwise.
This does have a functional impact (and one testcase is pessimized
because we can't reuse a load). But I think that's desirable no matter
what.
Note that this alone doesn't result in less DT computations: TwoAddress
was previously able to reuse the DT we computed for SDAG. That will be
fixed separately.
Differential Revision: https://reviews.llvm.org/D32766
llvm-svn: 302611
We currently require SCEV, which requires DT/LI. Those are expensive to
compute, but the pass only runs for functions that have the safestack
attribute.
Compute DT/LI to build SCEV lazily, only when the pass is actually going
to transform the function.
Differential Revision: https://reviews.llvm.org/D31302
llvm-svn: 302610
This should hopefully makes changes to the O0 pipeline obvious; it's
easy to require expensive passes, and this helps make informed
decisions.
Case in point: in the few weeks separating the time when I initially
wrote this patch to the time when I committed, the test regressed as
r302103 added another use of DT!
llvm-svn: 302608
Summary: computeKnownBitsForTargetNode was not defined for Lanai which resulted in additional AND's with 0x1 for the output of SETCC instructions.
Reviewers: eliben, majnemer
Reviewed By: majnemer
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D29605
llvm-svn: 302568
This patch adds more patterns that a reasonable person might write that can be compiled to BZHI.
This adds support for
(~0U >> (32 - b)) & a;
and
a << (32 - b) >> (32 - b);
This was inspired by the code in APInt::clearUnusedBits.
This can pass an index of 32 to the bzhi instruction which a quick test of Haswell hardware shows will not mask any bits. Though the description text in the Intel manual says the "index is saturated to OperandSize-1". The pseudocode in the same manual indicates no bits will be zeroed for this case.
I think this is still missing cases where the subtract portion is an 8-bit operation.
Differential Revision: https://reviews.llvm.org/D32616
llvm-svn: 302549
for scalar masked instructions only the lower bit of the mask is relevant. so for constant masks we should either do an unmasked operation or no operation, depending on the value of the lower bit.
This patch handles cases where the lower bit is '1'.
Differential Revision: https://reviews.llvm.org/D32805
llvm-svn: 302546
Quentin Colombet