shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/GVN
History
Florian Hahn 78c6b1488f
[CaptureTracking] Increase limit and use it for all visited uses. 
Currently the MaxUsesToExplore limit only applies to the number of users
per value, not the total number of users to explore.

The current limit of 20 pessimizes IR with opaque pointers in some
cases. Without opaque pointers, we have deeper pointer def-use chains in
general due to extra bitcasts and geps for structs with index 0.

With opaque pointers the def-use chain is not as deep but wider, due to
bitcasts & 0-geps missing.

To improve the situation for opaque pointers, this patch does 2 things:

 1. Apply the limit to the total number of uses visited. From the
    wording in the description of the option it seems like this may be
    the original intention. With the current implementation we could
    still end up walking a lot of uses.
 2. Increase the limit to 100. This is quite arbitrary, but enables
    a good number of additional optimizations.

Those adjustments have a noticeable compile-time impact though. In part
that is likely due to additional transformations (and conversely
the current baseline misses optimizations after switching to opaque
pointers).

This recovers some regressions that showed up after enabling opaque
pointers.

Limit=100:

* NewPM-O3: +0.21%
* NewPM-ReleaseThinLTO: +0.87%
* NewPM-ReleaseLTO-g: +0.46%

https://llvm-compile-time-tracker.com/compare.php?from=2e50ecb2ef4e1da1aeab05bcf66380068e680991&to=7e6fbe519d958d09f32f01d5d44a622f551e2031&stat=instructions

Limit=60:

* NewPM-O3: +0.14%
* NewPM-ReleaseThinLTO: +0.41%
* NewPM-ReleaseLTO-g: +0.21%

https://llvm-compile-time-tracker.com/compare.php?from=aeb19817d66f1a15754163c7f48e01e9ebdd6d45&to=520563fdc146319aae90d06f88d87f2e9e1247b7&stat=instructions

Limit=40:
* NewPM-O3: +0.11%
* NewPM-ReleaseThinLTO: +0.12%
* NewPM-ReleaseLTO-g: +0.09%

https://llvm-compile-time-tracker.com/compare.php?from=aeb19817d66f1a15754163c7f48e01e9ebdd6d45&to=c9182576e9fe3f1c84a71479665aef91a416318c&stat=instructions

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D126236
2022-06-02 21:43:58 +01:00
..
PRE
[GVN] Enable enable-split-backedge-in-load-pre option by default
2022-05-30 09:55:58 +02:00
2007-07-25-DominatedLoop.ll
…
2007-07-25-InfiniteLoop.ll
…
2007-07-25-Loop.ll
…
2007-07-25-NestedLoop.ll
…
2007-07-25-SinglePredecessor.ll
…
2007-07-26-InterlockingLoops.ll
…
2007-07-26-NonRedundant.ll
…
2007-07-26-PhiErasure.ll
…
2007-07-30-PredIDom.ll
…
2007-07-31-NoDomInherit.ll
…
2007-07-31-RedundantPhi.ll
…
2008-02-12-UndefLoad.ll
…
2008-02-13-NewPHI.ll
…
2008-07-02-Unreachable.ll
…
2008-12-09-SelfRemove.ll
…
2008-12-12-RLE-Crash.ll
…
2008-12-14-rle-reanalyze.ll
…
2008-12-15-CacheVisited.ll
…
2009-01-21-SortInvalidation.ll
…
2009-01-22-SortInvalidation.ll
…
2009-03-10-PREOnVoid.ll
OpaquePtr: Bulk update tests to use typed sret
2020-11-20 17:58:26 -05:00
2009-07-13-MemDepSortFail.ll
…
2009-11-12-MemDepMallocBitCast.ll
…
2010-03-31-RedundantPHIs.ll
…
2010-05-08-OneBit.ll
…
2010-11-13-Simplify.ll
…
2011-04-27-phioperands.ll
…
2011-07-07-MatchIntrinsicExtract.ll
…
2011-09-07-TypeIdFor.ll
…
2012-05-22-PreCrash.ll
…
2016-08-30-MaskedScatterGather-inseltpoison.ll
Precommit transform tests that have poison as insertelement's placeholder
2020-12-24 11:46:17 +09:00
2016-08-30-MaskedScatterGather.ll
…
addrspacecast.ll
…
assume-equal.ll
[GVN] Set phi entries of unreachable predecessors to poison instead of undef
2021-12-30 14:47:24 +00:00
assume.ll
[GVN] Use that assume(!X) implies X==false (PR47496)
2020-09-17 21:34:44 +02:00
basic-undef-test.ll
…
basic.ll
…
big-endian.ll
[test, GVN] Fix use of var defined in CHECK-NOT
2021-04-06 17:00:21 +01:00
bitcast-of-call.ll
…
br-identical.ll
…
callbr-loadpre-critedge.ll
[llvm][test] rewrite callbr to use i rather than X constraint NFC
2022-01-11 11:31:08 -08:00
callbr-scalarpre-critedge.ll
[llvm][test] rewrite callbr to use i rather than X constraint NFC
2022-01-11 11:31:08 -08:00
calloc-load-removal.ll
Extract utility function for checking initial value of allocation [NFC, try 2]
2022-01-07 08:44:08 -08:00
calls-nonlocal.ll
[GVN] Set phi entries of unreachable predecessors to poison instead of undef
2021-12-30 14:47:24 +00:00
calls-readonly.ll
…
capture-tracking-limit.ll
[CaptureTracking] Increase limit and use it for all visited uses.
2022-06-02 21:43:58 +01:00
commute.ll
[IR][GVN] add/allow commutative intrinsics with >2 args
2020-09-03 10:14:53 -04:00
cond_br2.ll
Revert "[NFC] remove explicit default value for strboolattr attribute in tests"
2021-05-24 19:43:40 +02:00
cond_br.ll
…
condprop-memdep-invalidation.ll
[GVN] Enable enable-split-backedge-in-load-pre option by default
2022-05-30 09:55:58 +02:00
condprop.ll
[GVN] Set phi entries of unreachable predecessors to poison instead of undef
2021-12-30 14:47:24 +00:00
constexpr-vector-constainsundef-crash-inseltpoison.ll
Precommit transform tests that have poison as insertelement's placeholder
2020-12-24 11:46:17 +09:00
constexpr-vector-constainsundef-crash.ll
[ValueTracking] Check for ConstantExpr before using recursive helpers.
2020-07-24 17:37:09 +01:00
crash-no-aa.ll
…
crash.ll
…
critical-edge-split-failure.ll
[llvm][test] rewrite callbr to use i rather than X constraint NFC
2022-01-11 11:31:08 -08:00
critical-edge-split-indbr-pred-in-loop.ll
…
dbg-redundant-load.ll
…
debugloc.ll
…
duplicate-phis.ll
[Test] [GVN] Add test showing equivalent PHIs generation by GVN
2021-12-09 16:40:23 +07:00
edge.ll
…
equality-assume.ll
[GVN] Set phi entries of unreachable predecessors to poison instead of undef
2021-12-30 14:47:24 +00:00
fence.ll
…
flags.ll
…
fold-const-expr.ll
…
fpmath.ll
…
freeze.ll
…
funclet.ll
…
gc_relocate.ll
[Statepoint] Update gc.statepoint calls in tests with elementtype (NFC)
2022-02-04 14:15:41 +01:00
gvn-loop-load-pre-order.ll
[GVN] Execute performLoopLoadPRE ahead of PerformLoadPRE
2021-08-24 09:50:27 +01:00
int_sideeffect.ll
…
invariant.group.ll
…
invariant.start.ll
…
lifetime-simple.ll
…
load-constant-mem.ll
…
load-dead-block.ll
[GVN] Don't explicitly materialize undefs from dead blocks
2021-03-06 23:46:24 +01:00
load-from-unreachable-predecessor.ll
…
load-of-pointer-select-available.ll
[GVN] Add tests with redundant load of pointer select.
2022-01-28 20:15:32 +00:00
loadpre-context.ll
[tests] Stablize tests for possible change in deref semantics
2021-07-14 13:05:43 -07:00
loadpre-missed-opportunity.ll
[GVN] Remove unused check-prefixes
2020-11-08 13:36:42 +00:00
malloc-load-removal.ll
[MemoryBuiltins] Remove isNoAliasFn() in favor of isNoAliasCall()
2022-01-10 09:18:15 +01:00
masked-load-store-vn-crash.ll
[GVN] Account for masked loads/stores depending on load/store instructions
2020-09-10 10:57:33 -05:00
masked-load-store.ll
Handle masked loads and stores in MemoryLocation/Dependence
2020-09-08 19:08:44 -05:00
MemdepMiscompile.ll
…
mssa-update-dead-def.ll
Fix MemorySSAUpdater::insertDef for dead code
2022-03-31 16:32:35 -07:00
no_speculative_loads_with_asan.ll
[GVN][NewGVN] Regenerate no_speculative_loads_with_asan.ll tests
2022-04-27 10:45:39 +01:00
no-mem-dep-info.ll
…
noalias.ll
[MemCpyOpt] Correctly merge alias scopes during call slot optimization
2020-12-03 09:23:37 -08:00
non-integral-pointers-inseltpoison.ll
Reapply [ConstantFold] Handle vectors in ConstantFoldLoadThroughBitcast()
2021-03-21 17:48:21 +01:00
non-integral-pointers.ll
Reapply [ConstantFold] Handle vectors in ConstantFoldLoadThroughBitcast()
2021-03-21 17:48:21 +01:00
non-local-offset.ll
…
nonescaping-malloc.ll
Revert "[BasicAA] Remove unneeded special case for malloc/calloc"
2022-05-18 07:35:27 -07:00
nonescaping.ll
[tests] Add test coverage for issue causing revert f7988d0
2022-05-18 08:17:57 -07:00
null-aliases-nothing.ll
…
opaque-ptr.ll
[GVN] Encode GEPs in offset representation
2022-04-28 09:32:05 +02:00
opt-remarks-multiple-users.ll
[GVN] Improve analysis for missed optimization remark
2021-05-17 21:51:15 -07:00
opt-remarks-non-dominating.ll
[GVN] Improve analysis for missed optimization remark
2021-05-17 21:51:15 -07:00
opt-remarks.ll
…
phi-translate-partial-alias.ll
…
phi.ll
[gvn] Precisely propagate equalities to phi operands
2021-03-08 08:59:00 -08:00
pr10820.ll
…
pr12979.ll
…
pr14166.ll
…
pr17732.ll
…
pr17852.ll
OpaquePtr: Bulk update tests to use typed byval
2020-11-20 14:00:46 -05:00
pr24397.ll
…
pr24426.ll
…
pr25440.ll
…
pr28562.ll
…
pr28879.ll
…
pr32314.ll
…
pr33549.ll
Revert "[NFC] remove explicit default value for strboolattr attribute in tests"
2021-05-24 19:43:40 +02:00
pr34908.ll
…
pr36063.ll
…
pr42605.ll
Revert "[NFC] remove explicit default value for strboolattr attribute in tests"
2021-05-24 19:43:40 +02:00
pr46054-md-nonlocaldefcache-cleanup.ll
…
pr49193.ll
[GVN] Add missing ICF update
2021-04-06 10:13:42 -07:00
pre-compare.ll
…
pre-new-inst.ll
…
pre-skip-convergent.ll
[gvn] PRE needs to skip convergent intrinsics/calls.
2020-10-30 11:24:40 -04:00
preserve-analysis.ll
[NFC] Rename GVN -> GVNPass and SROA -> SROAPass
2021-11-09 10:35:58 -08:00
preserve-memoryssa.ll
[GVN] Set phi entries of unreachable predecessors to poison instead of undef
2021-12-30 14:47:24 +00:00
propagate-ir-flags.ll
…
range.ll
…
readattrs.ll
…
rle-must-alias.ll
…
rle-no-phi-translate.ll
…
rle-nonlocal.ll
…
simplify-icf-cache-invalidation.ll
[GVN] Properly invalidate ICF cache when we simplify a value
2021-04-08 14:01:57 -07:00
stale-loop-info.ll
…
storeinvgroup.ll
[MemoryDependence] Fix invariant group store
2021-03-09 19:03:39 -05:00
tbaa.ll
…
unreachable_block_infinite_loop.ll
…
unreachable-predecessor.ll
…
volatile-nonvolatile.ll
…
vscale.ll
…