Commit Graph

2716 Commits

Author SHA1 Message Date
Nick Lewycky c31ceda7d9 Make Value::isDereferenceablePointer() handle unreachable code blocks. (This
returns false in the event the computation feeding into the pointer is
unreachable, which maybe ought to be true -- but this is at least consistent
with undef->isDereferenceablePointer().) Fixes PR11825!

llvm-svn: 148671
2012-01-23 00:05:17 +00:00
Andrew Trick b9c822ab0b Handle a corner case with IV chain collection with bailout instead of assert.
Fixes PR11783: bad cast to AddRecExpr.

llvm-svn: 148572
2012-01-20 21:23:40 +00:00
Andrew Trick 16abc8a1e2 Test case comments missing from my previous checkin.
llvm-svn: 148571
2012-01-20 21:21:27 +00:00
Nick Lewycky e8415fea4b Fix CountCodeReductionForAlloca to more accurately represent what SROA can and
can't handle. Also don't produce non-zero results for things which won't be
transformed by SROA at all just because we saw the loads/stores before we saw
the use of the address.

llvm-svn: 148536
2012-01-20 08:35:20 +00:00
Andrew Trick c908b43d9f SCEVExpander fixes. Affects LSR and indvars.
LSR has gradually been improved to more aggressively reuse existing code, particularly existing phi cycles. This exposed problems with the SCEVExpander's sloppy treatment of its insertion point. I applied some rigor to the insertion point problem that will hopefully avoid an endless bug cycle in this area. Changes:

- Always used properlyDominates to check safe code hoisting.

- The insertion point provided to SCEV is now considered a lower bound. This is usually a block terminator or the use itself. Under no cirumstance may SCEVExpander insert below this point.

- LSR is reponsible for finding a "canonical" insertion point across expansion of different expressions.

- Robust logic to determine whether IV increments are in "expanded" form and/or can be safely hoisted above some insertion point.

Fixes PR11783: SCEVExpander assert.

llvm-svn: 148535
2012-01-20 07:41:13 +00:00
Dan Gohman 8ee108bf98 Set the "tail" flag on pattern-matched objc_storeStrong calls.
rdar://10531041.

llvm-svn: 148490
2012-01-19 19:14:36 +00:00
Dan Gohman 82041c2e60 Use llvm.global_ctors to locate global constructors instead
of recognizing them by name.

llvm-svn: 148416
2012-01-18 21:19:38 +00:00
Andrew Trick c193b16ea2 Test case rename
llvm-svn: 148344
2012-01-17 22:27:45 +00:00
Dan Gohman e7a243fea5 Add a new ObjC ARC optimization pass to eliminate unneeded
autorelease push+pop pairs.

llvm-svn: 148330
2012-01-17 20:52:24 +00:00
Andrew Trick 12728f04ca LSR fix: broaden the check for loop preheaders.
It's becoming clear that LoopSimplify needs to unconditionally create loop preheaders. But that is a bigger fix. For now, continuing to hack LSR.
Fixes rdar://10701050 "Cannot split an edge from an IndirectBrInst" assert.

llvm-svn: 148288
2012-01-17 06:45:52 +00:00
Andrew Trick 23ef0d6c40 Fix a corner case hit by redundant phi elimination running after LSR.
Fixes PR11761: bad IR w/ redundant Phi elim

llvm-svn: 148177
2012-01-14 03:17:23 +00:00
Dan Gohman 728db4997a Implement proper ObjC ARC objc_retainBlock "escape" analysis, so that
the optimizer doesn't eliminate objc_retainBlock calls which are needed
for their side effect of copying blocks onto the heap.
This implements rdar://10361249.

llvm-svn: 148076
2012-01-13 00:39:07 +00:00
Duncan Sands 0bf46b5363 Don't try to create a GEP when the pointee type is unsized (such GEPs
are invalid).  Fixes a crash on array1.C from the GCC testsuite when
compiled with dragonegg.

llvm-svn: 147946
2012-01-11 12:20:08 +00:00
Stepan Dyatkovskiy 8216569812 Improved compile time:
1. Size heuristics changed. Now we calculate number of unswitching
branches only once per loop.
2. Some checks was moved from UnswitchIfProfitable to
processCurrentLoop, since it is not changed during processCurrentLoop
iteration. It allows decide to skip some loops at an early stage.
Extended statistics:
- Added total number of instructions analyzed.

llvm-svn: 147935
2012-01-11 08:40:51 +00:00
Bill Wendling c79155192d If the global variable is removed by the linker, then don't constant merge it
with other symbols.

An object in the __cfstring section is suppoed to be filled with CFString
objects, which have a pointer to ___CFConstantStringClassReference followed by a
pointer to a __cstring. If we allow the object in the __cstring section to be
merged with another global, then it could end up in any section. Because the
linker is going to remove these symbols in the final executable, we shouldn't
bother to merge them.
<rdar://problem/10564621>

llvm-svn: 147899
2012-01-11 00:13:08 +00:00
Andrew Trick d5d2db9af9 Enable LSR IV Chains with sufficient heuristics.
These heuristics are sufficient for enabling IV chains by
default. Performance analysis has been done for i386, x86_64, and
thumbv7. The optimization is rarely important, but can significantly
speed up certain cases by eliminating spill code within the
loop. Unrolled loops are prime candidates for IV chains. In many
cases, the final code could still be improved with more target
specific optimization following LSR. The goal of this feature is for
LSR to make the best choice of induction variables.

Instruction selection may not completely take advantage of this
feature yet. As a result, there could be cases of slight code size
increase.

Code size can be worse on x86 because it doesn't support postincrement
addressing. In fact, when chains are formed, you may see redundant
address plus stride addition in the addressing mode. GenerateIVChains
tries to compensate for the common cases.

On ARM, code size increase can be mitigated by using postincrement
addressing, but downstream codegen currently misses some opportunities.

llvm-svn: 147826
2012-01-10 01:45:08 +00:00
Andrew Trick 248d410e3e Adding IV chain generation to LSR.
After collecting chains, check if any should be materialized. If so,
hide the chained IV users from the LSR solver. LSR will only solve for
the head of the chain. GenerateIVChains will then materialize the
chained IV users by computing the IV relative to its previous value in
the chain.

In theory, chained IV users could be exposed to LSR's solver. This
would be considerably complicated to implement and I'm not aware of a
case where we need it. In practice it's more important to
intelligently prune the search space of nontrivial loops before
running the solver, otherwise the solver is often forced to prune the
most optimal solutions. Hiding the chained users does this well, so
that LSR is more likely to find the best IV for the chain as a whole.

llvm-svn: 147801
2012-01-09 21:18:52 +00:00
Benjamin Kramer f9d0cc0160 InstCombine: Teach foldLogOpOfMaskedICmpsHelper that sign bit tests are bit tests.
This subsumes several other transforms while enabling us to catch more cases.

llvm-svn: 147777
2012-01-09 17:23:27 +00:00
Benjamin Kramer 6609f741b9 Tweak my last commit to be less conservative about uses.
We still save an instruction when just the "and" part is replaced.
Also change the code to match comments more closely.

llvm-svn: 147753
2012-01-08 21:12:51 +00:00
Benjamin Kramer da37e15345 InstCombine: If we have a bit test and a sign test anded/ored together, merge the sign bit into the bit test.
This is common in bit field code, e.g. checking if the first or the last bit of a bit field is set.

llvm-svn: 147749
2012-01-08 18:32:24 +00:00
Andrew Trick 732ad80dbb LSR: Don't optimize loops if an outer loop has no preheader.
LoopSimplify may not run on some outer loops, e.g. because of indirect
branches. SCEVExpander simply cannot handle outer loops with no preheaders.
Fixes rdar://10655343 SCEVExpander segfault.

llvm-svn: 147718
2012-01-07 03:16:50 +00:00
Andrew Trick 5adedf5d47 Extended replaceCongruentPhis to handle mixed phi types.
llvm-svn: 147707
2012-01-07 01:12:09 +00:00
Andrew Trick cbf2fe066a comment typo
llvm-svn: 147701
2012-01-07 00:29:20 +00:00
Dan Gohman 5ab9c0a927 Fix SpeculativelyExecuteBB to either speculate all or none of the phis
present in the bottom of the CFG triangle, as the transformation isn't
ever valuable if the branch can't be eliminated.

Also, unify some heuristics between SimplifyCFG's multiple
if-converters, for consistency.

This fixes rdar://10627242.

llvm-svn: 147630
2012-01-05 23:58:56 +00:00
Eli Friedman 55fa49f32d PR11705, part 2: globalopt shouldn't put inttoptr/ptrtoint operations into global initializers if there's an implied extension or truncation.
llvm-svn: 147625
2012-01-05 23:03:32 +00:00
Dan Gohman 5267211899 Revert r56315. When the instruction to speculate is a load, this
code can incorrectly move the load across a store. This never
happens in practice today, but only because the current
heuristics accidentally preclude it.

llvm-svn: 147623
2012-01-05 22:54:35 +00:00
Benjamin Kramer aca1885695 FileCheck hygiene.
llvm-svn: 147580
2012-01-05 00:43:34 +00:00
Nick Lewycky 0c48afa0ed Teach instcombine all sorts of great stuff about shifts that have exact, nuw or
nsw bits on them.

llvm-svn: 147528
2012-01-04 09:28:29 +00:00
Andrew Trick cbcc98fb50 Fix SCEVExpander to handle loops with no preheader when LSR gives it a
"phony" insertion point.

Fixes rdar://10619599: "SelectionDAGBuilder shouldn't visit PHI nodes!" assert

llvm-svn: 147439
2012-01-02 21:25:10 +00:00
Nick Lewycky b59008c694 Make use of the exact bit when optimizing '(X >>exact 3) << 1' to eliminate the
'and' that would zero out the trailing bits, and to produce an exact shift
ourselves.

llvm-svn: 147391
2011-12-31 21:30:22 +00:00
Nick Lewycky 4c378a4453 Change CaptureTracking to pass a Use* instead of a Value* when a value is
captured. This allows the tracker to look at the specific use, which may be
especially interesting for function calls.

Use this to fix 'nocapture' deduction in FunctionAttrs. The existing one does
not iterate until a fixpoint and does not guarantee that it produces the same
result regardless of iteration order. The new implementation builds up a graph
of how arguments are passed from function to function, and uses a bottom-up walk
on the argument-SCCs to assign nocapture. This gets us nocapture more often, and
does so rather efficiently and independent of iteration order.

llvm-svn: 147327
2011-12-28 23:24:21 +00:00
Nick Lewycky a8e84fb56b Turn cos(-x) into cos(x). Patch by Alexander Malyshev!
llvm-svn: 147291
2011-12-27 18:25:50 +00:00
Nick Lewycky c554a9b58e Teach simplifycfg to recompute branch weights when merging some branches, and
to discard weights when appropriate. Still more to do (and a new TODO), but
it's a start!

llvm-svn: 147286
2011-12-27 04:31:52 +00:00
Nick Lewycky 8d302df4a4 Update the branch weight metadata when reversing the order of a branch.
llvm-svn: 147280
2011-12-26 20:54:14 +00:00
Chandler Carruth 8b7e71ffd6 Add an explicit test that we now fold cttz.i32(..., true) >> 5 -> 0.
This is a result of Benjamin's work on ValueTracking.

llvm-svn: 147259
2011-12-24 22:34:15 +00:00
Benjamin Kramer b16bd77bd2 InstCombine: Add a combine that turns (2^n)-1 ^ x back into (2^n)-1 - x iff x is smaller than 2^n and it fuses with a following add.
This was intended to undo the sub canonicalization in cases where it's not profitable, but it also
finds some cases on it's own.

llvm-svn: 147256
2011-12-24 17:31:53 +00:00
Benjamin Kramer 4ee5747fdd ComputeMaskedBits: Make knownzero computation more aggressive for ctlz with undef zero.
unsigned foo(unsigned x) { return 31 - __builtin_clz(x); }
now compiles into a single "bsrl" instruction on x86.

llvm-svn: 147255
2011-12-24 17:31:46 +00:00
Benjamin Kramer 010337c838 InstCombine: Canonicalize (2^n)-1 - x into (2^n)-1 ^ x iff x is known to be smaller than 2^n.
This has the obvious advantage of being commutable and is always a win on x86 because
const - x wastes a register there. On less weird architectures this may lead to
a regression because other arithmetic doesn't fuse with it anymore. I'll address that
problem in a followup.

llvm-svn: 147254
2011-12-24 17:31:38 +00:00
Nick Lewycky 854c869c36 Move this test from date-name to feature-name, and port it to FileCheck.
llvm-svn: 147223
2011-12-23 18:41:31 +00:00
Chad Rosier 388769427d Reinstate r146578; it doesn't appear to be the cause of some recent execution-
time regressions.  In general, it is beneficial to compile-time.

Original commit message:
Fix for bug #11429: Wrong behaviour for switches. Small improvement for code
size heuristics.

llvm-svn: 147175
2011-12-22 21:06:36 +00:00
Benjamin Kramer f1fd6e394d Give string constants generated by IRBuilder private linkage.
Fixes PR11640.

llvm-svn: 147144
2011-12-22 14:22:14 +00:00
Chad Rosier 1b7e2baf47 Speculatively revert r146578 to determine if it is the cause of a number of
performance regressions (both execution-time and compile-time) on our
nightly testers.

Original commit message:
Fix for bug #11429: Wrong behaviour for switches. Small improvement for code
size heuristics.

llvm-svn: 147131
2011-12-22 02:40:57 +00:00
Nick Lewycky b4039f633c Make some intrinsics safe to speculatively execute.
llvm-svn: 147036
2011-12-21 05:52:02 +00:00
Andrew Trick a34a8c45b4 Unit test for r146950: LSR postinc expansion, PR11571.
llvm-svn: 146951
2011-12-20 01:43:20 +00:00
Joerg Sonnenberger d6cb7649d8 Allow inlining of functions with returns_twice calls, if they have the
attribute themselve.

llvm-svn: 146851
2011-12-18 20:35:43 +00:00
Kevin Enderby 8b3deabd2d Revert r146822 at Pete Cooper's request as it broke clang self hosting.
Hope I did this correctly :)

llvm-svn: 146834
2011-12-17 19:48:52 +00:00
Pete Cooper eadf124d2b SimplifyCFG now predicts some conditional branches to true or false depending on previous branch on same comparison operands.
For example, 

if (a == b) {
    if (a > b) // this is false
    
Fixes some of the issues on <rdar://problem/10554090>

llvm-svn: 146822
2011-12-17 06:32:38 +00:00
Pete Cooper b33c297f14 Added InstCombine for "select cond, ~cond, x" type patterns
These can be reduced to "~cond & x" or "~cond | x"

llvm-svn: 146624
2011-12-15 00:56:45 +00:00
Eli Friedman 16ad2905a3 Make loop preheader insertion in LoopSimplify handle the case where the loop header is a landing pad correctly (by splitting the landingpad out of the loop header). Make some adjustments to the rest of LoopSimplify to make it clear that the rest of LoopSimplify isn't making bad assumptions about the presence of landing pads. PR11575.
llvm-svn: 146621
2011-12-15 00:50:34 +00:00
Dan Gohman 75d7d5e988 Move Instruction::isSafeToSpeculativelyExecute out of VMCore and
into Analysis as a standalone function, since there's no need for
it to be in VMCore. Also, update it to use isKnownNonZero and
other goodies available in Analysis, making it more precise,
enabling more aggressive optimization.

llvm-svn: 146610
2011-12-14 23:49:11 +00:00