Commit Graph

1061 Commits

Author SHA1 Message Date
Bjorn Pettersson 3961603921 [ValueTracking] Teach computeKnownBits and ComputeNumSignBits to look through ExtractElement.
Summary:
The computeKnownBits and ComputeNumSignBits functions in ValueTracking can now do a simple look-through of ExtractElement.

Reviewers: majnemer, spatel

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D24955

llvm-svn: 283434
2016-10-06 09:56:21 +00:00
Eli Friedman 74bed9d757 Make GlobalsAA ignore dead constant expressions.
Slightly improves the precision of GlobalsAA in certain situations, and
makes the behavior of optimization passes more predictable.

Differential Revision: https://reviews.llvm.org/D24104

llvm-svn: 283165
2016-10-04 00:03:55 +00:00
Sanjay Patel d27a21874b [x86, SSE/AVX] allow 128/256-bit lowering for copysign vector intrinsics (PR30433)
This should fix:
https://llvm.org/bugs/show_bug.cgi?id=30433

There are a couple of open questions about the codegen:
1. Should we let scalar ops be scalars and avoid vector constant loads/splats?
2. Should we have a pass to combine constants such as the inverted pair that we have here?

Differential Revision: https://reviews.llvm.org/D25165
 

llvm-svn: 283119
2016-10-03 16:38:27 +00:00
Simon Pilgrim 1ec20e9b0a [CostModel][X86] Added tests for current fptosi/fptoui costs
llvm-svn: 283047
2016-10-01 19:09:59 +00:00
Simon Pilgrim e0ec5c1f05 [CostModel][X86] Added fcopysign costs
llvm-svn: 283044
2016-10-01 16:41:52 +00:00
Simon Pilgrim 8b021c382d [CostModel][X86] Added fabs costs
llvm-svn: 283042
2016-10-01 16:30:13 +00:00
Jun Bum Lim 3822939ba7 Enhance calcColdCallHeuristics for InvokeInst
Summary: When identifying cold blocks, consider only the edge to the normal destination if the terminator is InvokeInst and let calcInvokeHeuristics() decide edge weights for the InvokeInst.

Reviewers: mcrosier, hfinkel, davidxl

Subscribers: mcrosier, llvm-commits

Differential Revision: https://reviews.llvm.org/D24868

llvm-svn: 282262
2016-09-23 17:26:14 +00:00
Simon Pilgrim 9178059826 [CostModel][X86] Added scalar float op costs
llvm-svn: 281864
2016-09-18 21:01:20 +00:00
Sanjay Patel e104a5f35a auto-generate checks
llvm-svn: 281756
2016-09-16 17:54:52 +00:00
David L Kreitzer 8bbabee21a Reapplying r278731 after fixing the problem that caused it to be reverted.
Enhance SCEV to compute the trip count for some loops with unknown stride.

Patch by Pankaj Chawla

Differential Revision: https://reviews.llvm.org/D22377

llvm-svn: 281732
2016-09-16 14:38:13 +00:00
Wei Mi 24662395df Create a getelementptr instead of sub expr for ValueOffsetPair if the
value is a pointer.

This patch is to fix PR30213. When expanding an expr based on ValueOffsetPair,
if the value is of pointer type, we can only create a getelementptr instead
of sub expr.

Differential Revision: https://reviews.llvm.org/D24088

llvm-svn: 281439
2016-09-14 04:39:50 +00:00
Elena Demikhovsky 3622fbfc68 [Loop Vectorizer] Fixed memory confilict checks.
Fixed a bug in run-time checks for possible memory conflicts inside loop.
The bug is in Low <-> High boundaries calculation. The High boundary should be calculated as "last memory access pointer + element size".

Differential revision: https://reviews.llvm.org/D23176

llvm-svn: 279930
2016-08-28 08:53:53 +00:00
Wei Mi 59ca96636d [UNROLL] Postpone ScalarEvolution::forgetLoop after TripCountSC is expanded
when unroll runtime iteration loop.

In llvm::UnrollRuntimeLoopRemainder, if the loop to be unrolled is the inner
loop inside a loop nest, the scalar evolution needs to be dropped for its
parent loop which is done by ScalarEvolution::forgetLoop. However, we can
postpone forgetLoop to the end of UnrollRuntimeLoopRemainder so TripCountSC
expansion can still reuse existing value.

Differential Revision: https://reviews.llvm.org/D23572

llvm-svn: 279748
2016-08-25 16:17:18 +00:00
Evgeny Stupachenko d7f9c3564a The patch improves ValueTracking on left shift with nsw flag.
Summary:
The patch fixes PR28946.

Reviewers: majnemer, sanjoy

Differential Revision: http://reviews.llvm.org/D23296

From: Li Huang
llvm-svn: 279684
2016-08-24 23:01:33 +00:00
Artur Pilipenko a1d9a67496 Remove missing file from r279433 reversal
llvm-svn: 279434
2016-08-22 13:18:19 +00:00
Simon Pilgrim 89e375a95e [CostModel][X86] Removed shift tests
There are more thorough tests found in vshift-*-cost.ll 

llvm-svn: 279406
2016-08-21 19:56:02 +00:00
Simon Pilgrim 6ad12ec629 [CostModel][X86] Added costs for vXi16 and vXi8 vectors for add/sub/mul/and/or/xor tests
llvm-svn: 279405
2016-08-21 19:44:44 +00:00
Simon Pilgrim b0a0576ffc [CostModel][X86] Replaced SSSE3 with SSE2 costs to create a better baseline
llvm-svn: 279404
2016-08-21 19:14:48 +00:00
Simon Pilgrim 07d7a21ea1 [CostModel][X86] Added fsqrt and fma costs
llvm-svn: 279403
2016-08-21 19:06:25 +00:00
Simon Pilgrim 3cd61a084f [CostModel][X86] Split off float arithmetic cost tests
llvm-svn: 279402
2016-08-21 18:34:47 +00:00
Simon Pilgrim 054e7d2ec1 [CostModel][X86] Added sub, or, and, fadd and fsub costs and missing 512-bit mul costs
llvm-svn: 279301
2016-08-19 19:07:10 +00:00
Simon Pilgrim fbfa3ee4f6 [CostModel][X86] Added some AVX512 and 512-bit vector cost tests
llvm-svn: 279291
2016-08-19 18:24:10 +00:00
Simon Pilgrim e309d2d0c3 [CostModel][X86] Add fdiv + frem cost tests
llvm-svn: 279283
2016-08-19 17:39:00 +00:00
Michael Kuperstein 41898f0396 [AliasSetTracker] Degrade AliasSetTracker when may-alias sets get too large.
Repeated inserts into AliasSetTracker have quadratic behavior - inserting a
pointer into AST is linear, since it requires walking over all "may" alias
sets and running an alias check vs. every pointer in the set.

We can avoid this by tracking the total number of pointers in "may" sets,
and when that number exceeds a threshold, declare the tracker "saturated".
This lumps all pointers into a single "may" set that aliases every other
pointer.

(This is a stop-gap solution until we migrate to MemorySSA)

This fixes PR28832.
Differential Revision: https://reviews.llvm.org/D23432

llvm-svn: 279274
2016-08-19 17:05:22 +00:00
Hans Wennborg 3879035e66 SCEV: Don't assert about non-SCEV-able value in isSCEVExprNeverPoison() (PR28932)
Differential Revision: https://reviews.llvm.org/D23594

llvm-svn: 278999
2016-08-17 22:50:18 +00:00
Reid Kleckner b99b709068 Revert "Enhance SCEV to compute the trip count for some loops with unknown stride."
This reverts commit r278731. It caused http://crbug.com/638314

llvm-svn: 278853
2016-08-16 21:02:04 +00:00
Sanjoy Das 78db2963f6 Revert "[ValueTracking] Improve ValueTracking on left shift with nsw flag"
This reverts commit r278172.  It causes PR28946.

llvm-svn: 278740
2016-08-15 21:01:31 +00:00
David L Kreitzer 7fe18251a5 Enhance SCEV to compute the trip count for some loops with unknown stride.
Patch by Pankaj Chawla

Differential Revision: https://reviews.llvm.org/D22377

llvm-svn: 278731
2016-08-15 20:21:41 +00:00
Eli Friedman a6707f56b5 [DSE] Don't remove stores made live by a call which unwinds.
Issue exposed by noalias or more aggressive alias analysis.

Fixes http://llvm.org/PR25422.

Differential revision: https://reviews.llvm.org/D21007

llvm-svn: 278451
2016-08-12 01:09:53 +00:00
Andrew Kaylor b10f6876cd [ValueTracking] An improvement to IR ValueTracking on Non-negative Integers
Patch by Li Huang

Differential Revision: https://reviews.llvm.org/D18777

llvm-svn: 278267
2016-08-10 18:47:19 +00:00
Andrew Kaylor 3c05edfd5e [ValueTracking] Improve ValueTracking on left shift with nsw flag
Patch by Li Huang

Differential Revison: https://reviews.llvm.org/D23296

llvm-svn: 278172
2016-08-09 22:41:35 +00:00
Wei Mi 575435012c Fix the runtime error caused by "Use ValueOffsetPair to enhance value reuse during SCEV expansion".
The patch is to fix the bug in PR28705. It was caused by setting wrong return
value for SCEVExpander::findExistingExpansion. The return values of findExistingExpansion
have different meanings when the function is used in different ways so it is easy to make
mistake. The fix creates two new interfaces to replace SCEVExpander::findExistingExpansion,
and specifies where each interface is expected to be used.

Differential Revision: https://reviews.llvm.org/D22942

llvm-svn: 278161
2016-08-09 20:40:03 +00:00
Wei Mi 785858cf6c Recommit "Use ValueOffsetPair to enhance value reuse during SCEV expansion".
The fix for PR28705 will be committed consecutively.

In D12090, the ExprValueMap was added to reuse existing value during SCEV expansion.
However, const folding and sext/zext distribution can make the reuse still difficult.

A simplified case is: suppose we know S1 expands to V1 in ExprValueMap, and
  S1 = S2 + C_a
  S3 = S2 + C_b
where C_a and C_b are different SCEVConstants. Then we'd like to expand S3 as
V1 - C_a + C_b instead of expanding S2 literally. It is helpful when S2 is a
complex SCEV expr and S2 has no entry in ExprValueMap, which is usually caused
by the fact that S3 is generated from S1 after const folding.

In order to do that, we represent ExprValueMap as a mapping from SCEV to
ValueOffsetPair. We will save both S1->{V1, 0} and S2->{V1, C_a} into the
ExprValueMap when we create SCEV for V1. When S3 is expanded, it will first
expand S2 to V1 - C_a because of S2->{V1, C_a} in the map, then expand S3 to
V1 - C_a + C_b.

Differential Revision: https://reviews.llvm.org/D21313

llvm-svn: 278160
2016-08-09 20:37:50 +00:00
Sanjoy Das d4c85af7fd [SCEV] Un-grep'ify tests; NFC
llvm-svn: 277861
2016-08-05 20:33:49 +00:00
Sanjoy Das b0b4e86215 [SCEV] Don't infinitely recurse on unreachable code
llvm-svn: 277848
2016-08-05 18:34:14 +00:00
Michael Kuperstein 3ceac2bbd5 [LV, X86] Be more optimistic about vectorizing shifts.
Shifts with a uniform but non-constant count were considered very expensive to
vectorize, because the splat of the uniform count and the shift would tend to
appear in different blocks. That made the splat invisible to ISel, and we'd
scalarize the shift at codegen time.

Since r201655, CodeGenPrepare sinks those splats to be next to their use, and we
are able to select the appropriate vector shifts. This updates the cost model to
to take this into account by making shifts by a uniform cheap again.

Differential Revision: https://reviews.llvm.org/D23049

llvm-svn: 277782
2016-08-04 22:48:03 +00:00
Simon Pilgrim c8fe132756 [X86] Dropped XOP ctbits checks - they match the AVX checks
llvm-svn: 277718
2016-08-04 11:04:13 +00:00
Simon Pilgrim 5d5ca9c0cb [X86][SSE] Add initial costs for vector CTTZ/CTLZ
llvm-svn: 277716
2016-08-04 10:51:41 +00:00
George Burgess IV 5f0e76dca6 [CFLAA] Remove modref queries from CFLAA.
As it turns out, modref queries are broken with CFLAA. Specifically,
the data source we were using for determining modref behaviors
explicitly ignores operations on non-pointer values. So, it wouldn't
note e.g. storing an i32 to an i32* (or loading an i64 from an i64*).
It also ignores external function calls, rather than acting
conservatively for them.

(N.B. These operations, where necessary, *are* tracked by CFLAA; we just
use a different mechanism to do so. Said mechanism is relatively
imprecise, so it's unlikely that we can provide reasonably good modref
answers with it as implemented.)

Patch by Jia Chen.

Differential Revision: https://reviews.llvm.org/D22978

llvm-svn: 277366
2016-08-01 18:47:28 +00:00
Adam Nemet aa3506c5f0 [BPI] Add new LazyBPI analysis
Summary:
The motivation is the same as in D22141: In order to add the hotness
attribute to optimization remarks we need BFI to be available in all
passes that emit optimization remarks.  BFI depends on BPI so unless we
make this lazy as well we would still compute BPI unconditionally.

The solution is to use the new LazyBPI pass in LazyBFI and only compute
BPI when computation of BFI is requested by the client.

I extended the laziness test using a LoopDistribute test to also cover
BPI.

Reviewers: hfinkel, davidxl

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D22835

llvm-svn: 277083
2016-07-28 23:31:12 +00:00
George Burgess IV dbd35c44d4 [CFLAA] Add getModRefBehavior to CFLAnders.
This patch lets CFLAnders respond to mod-ref queries. It also includes
a small bugfix to CFLSteens.

Patch by Jia Chen.

Differential Revision: https://reviews.llvm.org/D22823

llvm-svn: 276939
2016-07-27 23:07:07 +00:00
Hans Wennborg 685e8ff953 Revert r276136 "Use ValueOffsetPair to enhance value reuse during SCEV expansion."
It causes Clang tests to fail after Windows self-host (PR28705).

(Also reverts follow-up r276139.)

llvm-svn: 276822
2016-07-26 23:25:13 +00:00
Wei Mi 97de034e18 Remove useless pass from the pipeline in test/Analysis/Dominators/2007-01-14-BreakCritEdges.ll.
llvm-svn: 276644
2016-07-25 16:27:34 +00:00
Sanjoy Das a7d9ec8751 [SCEV] Make isImpliedCondOperandsViaRanges smarter
This change lets us prove things like

  "{X,+,10} s< 5000" implies "{X+7,+,10} does not sign overflow"

It does this by replacing replacing getConstantDifference by
computeConstantDifference (which is smarter) in
isImpliedCondOperandsViaRanges.

llvm-svn: 276505
2016-07-23 00:54:36 +00:00
Wei Mi e04d0eff29 [PM] Port BreakCriticalEdges to the new PM.
Differential Revision: https://reviews.llvm.org/D22688

llvm-svn: 276449
2016-07-22 18:04:25 +00:00
Wei Mi 481232e991 Fix test/Analysis/ScalarEvolution/scev-expander-existing-value-offset.ll for rL276136.
The content in this testcase was accidentally duplicated. Fix the error.

llvm-svn: 276139
2016-07-20 16:54:58 +00:00
Wei Mi db80c0c77f Use ValueOffsetPair to enhance value reuse during SCEV expansion.
In D12090, the ExprValueMap was added to reuse existing value during SCEV expansion.
However, const folding and sext/zext distribution can make the reuse still difficult.

A simplified case is: suppose we know S1 expands to V1 in ExprValueMap, and
  S1 = S2 + C_a
  S3 = S2 + C_b
where C_a and C_b are different SCEVConstants. Then we'd like to expand S3 as
V1 - C_a + C_b instead of expanding S2 literally. It is helpful when S2 is a
complex SCEV expr and S2 has no entry in ExprValueMap, which is usually caused
by the fact that S3 is generated from S1 after const folding.

In order to do that, we represent ExprValueMap as a mapping from SCEV to
ValueOffsetPair. We will save both S1->{V1, 0} and S2->{V1, C_a} into the
ExprValueMap when we create SCEV for V1. When S3 is expanded, it will first
expand S2 to V1 - C_a because of S2->{V1, C_a} in the map, then expand S3 to
V1 - C_a + C_b.

Differential Revision: https://reviews.llvm.org/D21313

llvm-svn: 276136
2016-07-20 16:40:33 +00:00
Simon Pilgrim 1b4f511aaa [X86][SSE] Add cost model values for CTPOP of vectors
This patch adds costs for the vectorized implementations of CTPOP, the default values were seriously underestimating the cost of these and was encouraging vectorization on targets where serialized use of POPCNT would be much better.

Differential Revision: https://reviews.llvm.org/D22456

llvm-svn: 276104
2016-07-20 10:41:28 +00:00
George Burgess IV 8b85321bae [CFLAA] Make a test tell the truth. NFC.
Dishonesty noted by Jia Chen.

llvm-svn: 276028
2016-07-19 20:56:41 +00:00
George Burgess IV 3b059841ff [CFLAA] Add some interproc. analysis to CFLAnders.
This patch adds function summary support to CFLAnders. It also comes
with a lot of tests! Woohoo!

Patch by Jia Chen.

Differential Revision: https://reviews.llvm.org/D22450

llvm-svn: 276026
2016-07-19 20:47:15 +00:00