llvm-project/llvm/test/Transforms/InstSimplify
Roman Lebedev 6102310d81
[InstSimplify][EarlyCSE] Try to CSE PHI nodes in the same basic block
Apparently, we don't do this, neither in EarlyCSE, nor in InstSimplify,
nor in (old) GVN, but do in NewGVN and SimplifyCFG of all places..

While i could teach EarlyCSE how to hash PHI nodes,
we can't really do much (anything?) even if we find two identical
PHI nodes in different basic blocks, same-BB case is the interesting one,
and if we teach InstSimplify about it (which is what i wanted originally,
https://reviews.llvm.org/D86530), we get EarlyCSE support for free.

So i would think this is pretty uncontroversial.

On vanilla llvm test-suite + RawSpeed, this has the following effects:
```
| statistic name                                     | baseline  | proposed  |      Δ |        % |    \|%\| |
|----------------------------------------------------|-----------|-----------|-------:|---------:|---------:|
| instsimplify.NumPHICSE                             | 0         | 23779     |  23779 |    0.00% |    0.00% |
| asm-printer.EmittedInsts                           | 7942328   | 7942392   |     64 |    0.00% |    0.00% |
| assembler.ObjectBytes                              | 273069192 | 273084704 |  15512 |    0.01% |    0.01% |
| correlated-value-propagation.NumPhis               | 18412     | 18539     |    127 |    0.69% |    0.69% |
| early-cse.NumCSE                                   | 2183283   | 2183227   |    -56 |    0.00% |    0.00% |
| early-cse.NumSimplify                              | 550105    | 542090    |  -8015 |   -1.46% |    1.46% |
| instcombine.NumAggregateReconstructionsSimplified  | 73        | 4506      |   4433 | 6072.60% | 6072.60% |
| instcombine.NumCombined                            | 3640264   | 3664769   |  24505 |    0.67% |    0.67% |
| instcombine.NumDeadInst                            | 1778193   | 1783183   |   4990 |    0.28% |    0.28% |
| instcount.NumCallInst                              | 1758401   | 1758799   |    398 |    0.02% |    0.02% |
| instcount.NumInvokeInst                            | 59478     | 59502     |     24 |    0.04% |    0.04% |
| instcount.NumPHIInst                               | 330557    | 330533    |    -24 |   -0.01% |    0.01% |
| instcount.TotalInsts                               | 8831952   | 8832286   |    334 |    0.00% |    0.00% |
| simplifycfg.NumInvokes                             | 4300      | 4410      |    110 |    2.56% |    2.56% |
| simplifycfg.NumSimpl                               | 1019808   | 999607    | -20201 |   -1.98% |    1.98% |
```
I.e. it fires ~24k times, causes +110 (+2.56%) more `invoke` -> `call`
transforms, and counter-intuitively results in *more* instructions total.

That being said, the PHI count doesn't decrease that much,
and looking at some examples, it seems at least some of them
were previously getting PHI CSE'd in SimplifyCFG of all places..

I'm adjusting `Instruction::isIdenticalToWhenDefined()` at the same time.
As a comment in `InstCombinerImpl::visitPHINode()` already stated,
there are no guarantees on the ordering of the operands of a PHI node,
so if we just naively compare them, we may false-negatively say that
the nodes are not equal when the only difference is operand order,
which is especially important since the fold is in InstSimplify,
so we can't rely on InstCombine sorting them beforehand.

Fixing this for the general case is costly (geomean +0.02%),
and does not appear to catch anything in test-suite, but for
the same-BB case, it's trivial, so let's fix at least that.

As per http://llvm-compile-time-tracker.com/compare.php?from=04879086b44348cad600a0a1ccbe1f7776cc3cf9&to=82bdedb888b945df1e9f130dd3ac4dd3c96e2925&stat=instructions
this appears to cause geomean +0.03% compile time increase (regression),
but geomean -0.01%..-0.04% code size decrease (improvement).
2020-08-27 18:47:04 +03:00
..
ConstProp [ConstProp] Remove ConstantPropagation 2020-08-26 15:51:30 -07:00
2010-12-20-Boolean.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
2011-01-14-Thread.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
2011-02-01-Vector.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
2011-09-05-InsertExtractValue.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
2011-10-27-BinOpCrash.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
2011-11-23-MaskedBitsCrash.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
2013-04-19-ConstantFoldingCrash.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
AndOrXor.ll [InstSimplify] add tests for logic-of-icmp with min/max constant; NFC 2020-04-19 08:24:38 -04:00
abs_intrinsic.ll [ValueTracking] Add abs intrinsics support to computeConstantRange() 2020-08-12 22:28:46 +02:00
add-mask.ll [ValueTracking] Add computeKnownBits DemandedElts support to masked add instructions (PR36319) 2020-03-18 21:50:56 +00:00
add.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
addsub.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
and-icmps-same-ops.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
and-or-icmp-min-max.ll [InstSimplify] fold and/or of compares with equality to min/max constant 2020-04-23 09:16:10 -04:00
and-or-icmp-nullptr.ll [InstSimplify] fold and/or of compares with equality to min/max constant 2020-04-23 09:16:10 -04:00
and-or-icmp-zero.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
assume-non-zero.ll [ValueTracking] isKnownNonZero() should take non-null-ness assumptions into consideration (PR43267) 2019-12-20 01:47:57 +03:00
assume_icmp.ll [InstSimplify] Fold icmp with dominating assume 2020-07-03 18:53:58 +02:00
bitcast-vector-fold.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
bitreverse-fold.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
bitreverse.ll [ValueTracking] Add computeKnownBits DemandedElts support to EXTRACTELEMENT/OR/BSWAP/BITREVERSE instructions (PR36319) 2020-03-18 18:49:58 +00:00
bswap.ll [ValueTracking] Add computeKnownBits DemandedElts support to EXTRACTELEMENT/OR/BSWAP/BITREVERSE instructions (PR36319) 2020-03-18 18:49:58 +00:00
call.ll [InstSimplify] Extract abs intrinsic tests into separate file (NFC) 2020-08-12 22:28:46 +02:00
cast-unsigned-icmp-cmp-0.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
cast.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
cmp_ext.ll [InstSimplify] Simplify comparison between zext(x) and sext(x) 2020-07-04 11:03:00 +02:00
cmp_of_min_max.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
compare.ll [InstSimplify] Peephole optimization for icmp (urem X, Y), X 2020-08-04 20:48:37 +02:00
constantfold-add-nuw-allones-to-allones.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
constantfold-shl-nuw-C-to-C.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
constfold-constrained.ll [FEnv] Constfold some unary constrained operations 2020-03-28 12:28:33 +07:00
dead-code-removal.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
distribute.ll [InstSimplify] add tests for expandCommutativeBinOp; NFC 2020-07-29 23:21:39 +09:00
div-by-0-guard-before-smul_ov-not.ll [InstSimplify] Drop leftover "division-by-zero guard" around `@llvm.umul.with.overflow` inverted overflow bit 2019-08-29 12:48:04 +00:00
div-by-0-guard-before-smul_ov.ll [InstSimplify] Drop leftover "division-by-zero guard" around `@llvm.umul.with.overflow` overflow bit 2019-08-29 12:47:50 +00:00
div-by-0-guard-before-umul_ov-not.ll [InstSimplify] Drop leftover "division-by-zero guard" around `@llvm.umul.with.overflow` inverted overflow bit 2019-08-29 12:48:04 +00:00
div-by-0-guard-before-umul_ov.ll [InstSimplify] Drop leftover "division-by-zero guard" around `@llvm.umul.with.overflow` overflow bit 2019-08-29 12:47:50 +00:00
div.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
exact-nsw-nuw.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
extract-element.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
fast-math.ll [ValueTracking] fix CannotBeNegativeZero() to disregard 'nsz' FMF 2020-05-05 16:04:59 -04:00
fcmp-select.ll [InstSimplify] use FMF to improve fcmp+select fold 2019-11-04 08:29:56 -05:00
fdiv.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
floating-point-arithmetic.ll [ValueTracking] fix miscompile in maxnum case of cannotBeOrderedLessThanZeroImpl (PR46627) 2020-07-14 08:08:09 -04:00
floating-point-compare.ll [InstSimplify] fold fcmp using isKnownNeverInfinity + isKnownNeverNaN 2020-07-26 09:04:37 -04:00
fold-intrinsics.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
fp-nan.ll [InstSimplify] simplify FP ops harder with FMF 2020-03-12 09:13:28 -04:00
fp-undef.ll [InstSimplify] simplify FP ops harder with FMF (part 2) 2020-03-12 09:53:20 -04:00
freeze-noundef.ll [ValueTracking] Let getGuaranteedNonPoisonOp find multiple non-poison operands 2020-08-26 04:40:21 +09:00
freeze.ll [ValueTracking] Instruction::isBinaryOp should be used for constexprs 2020-07-26 21:48:51 +09:00
gep.ll [SVE] Make ConstantFoldGetElementPtr work for scalable vectors of indices 2020-06-25 07:28:19 +01:00
icmp-abs-nabs.ll [ValueTracking] When calculating known bits for integer abs, make sure we're looking at a negate and not just any instruction with the nsw flag set. 2019-08-07 18:28:16 +00:00
icmp-bool-constant.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
icmp-constant.ll [InstSimplify] avoid crashing by trying to rem-by-zero 2020-08-06 16:06:31 -04:00
icmp-ranges.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
icmp.ll [NFCI][InstSimplify] Add CHECK-LABEL to new icmp.ll test 2020-06-25 01:10:35 +03:00
implies.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
insertelement.ll [InstSimplify] allow insertelement-with-undef fold if poison-safe 2020-05-01 10:34:29 -04:00
known-never-nan.ll [FPEnv] Intrinsic llvm.roundeven 2020-05-26 19:24:58 +07:00
known-non-zero.ll [ValueTracking] Allow context-sensitive nullness check for non-pointers 2019-12-09 15:15:52 -06:00
load-relative-32.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
load-relative.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
load.ll [AssumeBundles] add cannonicalisation to the assume builder 2020-06-19 10:32:26 +02:00
log-exp-intrinsic.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
log2-pow2-intrinsic.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
log10-pow10-intrinsic.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
logic-of-fcmps.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
maxmin.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
maxmin_intrinsics.ll [InstSimplify] Fold min/max intrinsic based on icmp of operands 2020-08-26 22:02:57 +02:00
mul.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
negate.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
noalias-ptr.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
null-ptr-is-valid-attribute.ll ValueTracking: Fix isKnownNonZero for non-0 null pointers for byval 2020-07-16 13:50:49 -04:00
null-ptr-is-valid.ll ValueTracking: Fix isKnownNonZero for non-0 null pointers for byval 2020-07-16 13:50:49 -04:00
or-icmps-same-ops.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
or.ll [InstSimplify] fix bug in matching or-with-not op (PR46083) 2020-06-03 13:44:29 -04:00
past-the-end.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
phi-cse.ll [InstSimplify][EarlyCSE] Try to CSE PHI nodes in the same basic block 2020-08-27 18:47:04 +03:00
phi.ll [InstSimplify] add a phi test with 1 incoming value; NFC 2019-06-19 17:23:29 +00:00
pr28725.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
pr33957.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
ptr_diff.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
reassociate.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
redundant-null-check-in-uadd_with_overflow-of-nonnull-ptr.ll [InstSimplify] Pass SimplifyQuery into simplifyUnsignedRangeCheck() and use it for isKnownNonZero() 2019-09-11 15:32:46 +00:00
rem.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
remove-dead-call.ll [InstSimplify] Report "Changed" also when only deleting dead instructions 2019-08-09 07:08:25 +00:00
require-dominator.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
result-of-add-of-negative-is-non-zero-and-no-underflow.ll [NFC][InstSimplify] Add some more tests for D67498/D67502 2019-09-13 17:58:24 +00:00
result-of-usub-by-nonzero-is-non-zero-and-no-overflow.ll [InstSimplify] simplifyUnsignedRangeCheck(): handle more cases (PR43251) 2019-09-12 09:26:17 +00:00
result-of-usub-is-non-zero-and-no-overflow.ll [InstSimplify] Handle more 'A </>/>=/<= B &&/|| (A - B) !=/== 0' patterns (PR43251) 2019-09-25 22:59:41 +00:00
returned.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
round-intrinsics.ll [FPEnv] Intrinsic llvm.roundeven 2020-05-26 19:24:58 +07:00
saturating-add-sub.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
sdiv.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
select-and-cmp.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
select-implied.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
select-or-cmp.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
select.ll Recommit "[InstSimplify] Remove select ?, undef, X -> X and select ?, X, undef -> X transforms" and its follow up patches 2020-08-12 10:45:27 -07:00
shift-128-kb.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
shift-knownbits.ll Reapply [InstSimplify] Remove known bits constant folding 2020-05-08 10:24:53 +02:00
shift.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
shr-nop.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
shr-scalar-vector-consistency.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
shufflevector.ll [InstSimplify] fold splat of inserted constant to vector constant 2019-12-15 09:32:03 -05:00
signed-div-rem.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
simplify-nested-bitcast.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
srem.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
sub.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
undef.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
vec-cmp.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
vector_gep.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
vector_ptr_bitcast.ll Revert "Temporarily Revert "Add basic loop fusion pass."" 2019-04-17 04:52:47 +00:00
vscale.ll [InstSimplify] Simplify to vector constants when possible 2020-08-26 11:40:36 -07:00