ff714f9992
Clarify a comment.
...
llvm-svn: 114487
2010-09-21 21:44:14 +00:00
a06741b356
do not rely on the implicit-dereference semantics of dyn_cast_or_null
...
llvm-svn: 114278
2010-09-18 11:55:34 +00:00
aaa22cf1b6
do not rely on the implicit-dereference semantics of dyn_cast_or_null
...
llvm-svn: 114277
2010-09-18 11:53:39 +00:00
d104806575
Use a depth-first iteratation in CorrelatedValuePropagation to avoid wasting time trying
...
to optimize unreachable blocks.
llvm-svn: 114105
2010-09-16 18:35:07 +00:00
f95f59a0c2
When substituting sunkaddrs into indirect arguments an asm, we were
...
walking the asm arguments once and stashing their Values. This is
wrong because the same memory location can be in the list twice, and
if the first one has a sunkaddr substituted, the stashed value for the
second one will be wrong (use-after-free). PR 8154.
llvm-svn: 114104
2010-09-16 18:30:55 +00:00
67e534505d
fix PR8144, a bug where constant merge would merge globals marked
...
attribute(used).
llvm-svn: 113911
2010-09-15 00:30:11 +00:00
d361aac3d0
Remove the option to disable LazyValueInfo in JumpThreading, as it is now
...
on by default and has received significant testing.
llvm-svn: 113852
2010-09-14 20:57:41 +00:00
f1144f0929
fix PR8102, a case where we'd copyValue from a value that we already
...
deleted. Fix this by doing the copyValue's before we delete stuff!
The testcase only repros the problem on my system with valgrind.
llvm-svn: 113820
2010-09-14 00:19:00 +00:00
93c9b2ea93
Revert "CMake: Get rid of LLVMLibDeps.cmake and export the libraries normally."
...
This reverts commit r113632
Conflicts:
cmake/modules/AddLLVM.cmake
llvm-svn: 113819
2010-09-13 23:59:48 +00:00
e3a89f9f9c
Remove unused variable.
...
llvm-svn: 113769
2010-09-13 18:27:59 +00:00
1094c80281
Added skeleton for inline asm multiple alternative constraint support.
...
llvm-svn: 113766
2010-09-13 18:15:37 +00:00
c237a849e3
Re-apply r113679, which was reverted in r113720, which added a paid of new instcombine transforms
...
to expose greater opportunities for store narrowing in codegen. This patch fixes a potential
infinite loop in instcombine caused by one of the introduced transforms being overly aggressive.
llvm-svn: 113763
2010-09-13 17:59:27 +00:00
26abd3e0c2
Revert 113679, it was causing an infinite loop in a testcase that I've sent
...
on to Owen.
llvm-svn: 113720
2010-09-12 06:09:23 +00:00
70f4524427
Invert and-of-or into or-of-and when doing so would allow us to clear bits of the and's mask.
...
This can result in increased opportunities for store narrowing in code generation. Update a number of
tests for this change. This fixes <rdar://problem/8285027>.
Additionally, because this inverts the order of ors and ands, some patterns for optimizing or-of-and-of-or
no longer fire in instances where they did originally. Add a simple transform which recaptures most of these
opportunities: if we have an or-of-constant-or and have failed to fold away the inner or, commute the order
of the two ors, to give the non-constant or a chance for simplification instead.
llvm-svn: 113679
2010-09-11 05:48:06 +00:00
2f5f696b66
typoes
...
llvm-svn: 113647
2010-09-10 22:25:58 +00:00
dc38d36ccb
CMake: Get rid of LLVMLibDeps.cmake and export the libraries normally.
...
llvm-svn: 113632
2010-09-10 21:14:25 +00:00
77ab138f84
This transform is also performed by InstructionSimplify, remove the duplicate.
...
llvm-svn: 113608
2010-09-10 19:52:35 +00:00
d85c9ccdba
Lower the unrolling theshold to 150. Empirical tests indicate that this is a sweet spot in the performance per
...
code size increase curve.
llvm-svn: 113595
2010-09-10 17:57:00 +00:00
04cf3fd761
What the loop unroller cares about, rather than just not unrolling loops with calls, is
...
not unrolling loops that contain calls that would be better off getting inlined. This mostly
comes up when an interleaved devirtualization pass has devirtualized a call which the inliner
will inline on a future pass. Thus, rather than blocking all loops containing calls, add
a metric for "inline candidate calls" and block loops containing those instead.
llvm-svn: 113535
2010-09-09 20:32:23 +00:00
6270515918
Revert r113439, which relaxed the requirement that loops containing calls cannot be unrolled. After some discussion,
...
there seems to be a better way to achieve the same effect.
llvm-svn: 113528
2010-09-09 20:02:23 +00:00
11ab204fdc
r113526 introduced an unintended change to the loop unrolling threshold. Revert it.
...
llvm-svn: 113527
2010-09-09 19:11:57 +00:00
b61b1647e2
Fix typo in code to cap the loop code size reduction calculation.
...
llvm-svn: 113526
2010-09-09 19:08:59 +00:00
62ea1b718c
Use code-size reduction metrics to estimate the amount of savings we'll get when we unroll a loop.
...
Next step is to recalculate the threshold values given this new heuristic.
llvm-svn: 113525
2010-09-09 19:07:31 +00:00
8084dbaf8e
Relax the "don't unroll loops containing calls" rule. Instead, when a loop contains a call, lower the
...
unrolling threshold to the optimize-for-size threshold. Basically, for loops containing calls, unrolling
can still be profitable as long as the loop is REALLY small.
llvm-svn: 113439
2010-09-08 23:10:07 +00:00
3fe002dfb5
Generalize instcombine's support for combining multiple bit checks into a single test. Patch by Dirk Steinke!
...
llvm-svn: 113423
2010-09-08 22:16:17 +00:00
a4d9c78aa1
Add a separate unrolling threshold when the current function is being optimized for size.
...
The threshold value of 50 is arbitrary, and I chose it simply by analogy to the inlining thresholds, where
the baseline unrolling threshold is slightly smaller than the baseline inlining threshold. This could
undoubtedly use some tuning.
llvm-svn: 113306
2010-09-07 23:15:30 +00:00
6e27b3e004
Fix a serious performance regression introduced by r108687 on linux:
...
turning (fptrunc (sqrt (fpext x))) -> (sqrtf x) is great, but we have
to delete the original sqrt as well. Not doing so causes us to do
two sqrt's when building with -fmath-errno (the default on linux).
llvm-svn: 113260
2010-09-07 20:01:38 +00:00
71972d45dc
Fix major bug in thunk detection. Also verify the calling convention.
...
Switch from isWeakForLinker to mayBeOverridden which is more accurate.
Add more statistics and debugging info. Add comments. Move static function
outside anonymous namespace.
llvm-svn: 113190
2010-09-07 01:42:10 +00:00
be9019090e
fix PR8067, an over-aggressive assertion in LICM.
...
llvm-svn: 113146
2010-09-06 05:11:24 +00:00
b01c24a945
Teach loop rotate to hoist trivially invariant instructions
...
in the duplicated block instead of duplicating them.
Duplicating them into the end of the loop and the preheader
means that we got a phi node in the header of the loop,
which prevented LICM from hoisting them. GVN would
usually come around later and merge the duplicated
instructions so we'd get reasonable output... except that
anything dependent on the shoulda-been-hoisted value can't
be hoisted. In PR5319 (which this fixes), a memory value
didn't get promoted.
llvm-svn: 113134
2010-09-06 01:10:22 +00:00
da24b9a49a
pull a simple method out of LICM into a new
...
Loop::hasLoopInvariantOperands method. Remove
a useless and confusing Loop::isLoopInvariant(Instruction)
method, which didn't do what you thought it did.
No functionality change.
llvm-svn: 113133
2010-09-06 01:05:37 +00:00
1edf7434cf
more cleanups
...
llvm-svn: 113115
2010-09-05 20:13:07 +00:00
e6214557e7
Change lower atomic pass to use IntrinsicInst to simplify it a bit.
...
llvm-svn: 113114
2010-09-05 20:10:47 +00:00
05ef361b5e
eliminate some non-obvious casts. UndefValue isa Constant.
...
llvm-svn: 113113
2010-09-05 20:03:09 +00:00
e3ac69eca3
Fix warning reported by MSVC++ builder.
...
llvm-svn: 113106
2010-09-05 09:11:38 +00:00
f3a07ec394
Switch FnSet to containing the ComparableFunction instead of a pointer to one.
...
This reduces malloc traffic (yay!) and removes MergeFunctionsEqualityInfo.
llvm-svn: 113105
2010-09-05 09:00:32 +00:00
0095937b13
Fix many bugs when merging weak-strong and weak-weak pairs. We now merge all
...
strong functions first to make sure they're the canonical definitions and then
do a second pass looking only for weak functions.
llvm-svn: 113104
2010-09-05 08:22:49 +00:00
65b48b5dfc
zap dead code.
...
llvm-svn: 113073
2010-09-04 18:12:00 +00:00
487e250109
Fix LoopSimplify to notify ScalarEvolution when splitting a loop backedge
...
into an inner loop, as the new loop iteration may differ substantially.
This fixes PR8078.
llvm-svn: 113057
2010-09-04 02:42:48 +00:00
50506787d1
fix a bug in my licm rewrite when a load from the promoted memory
...
location is being re-stored to the memory location. We would get
a dangling pointer from the SSAUpdate data structure and miss a
use. This fixes PR8068
llvm-svn: 113042
2010-09-04 00:12:30 +00:00
c91c1a205a
Propagate non-local comparisons. Fixes PR1757.
...
llvm-svn: 113025
2010-09-03 22:47:08 +00:00
c725462245
Add support for simplifying a load from a computed value to a load from a global when it
...
is provable that they're equivalent. This fixes PR4855.
llvm-svn: 112994
2010-09-03 19:08:37 +00:00
affc0e42f0
fix more AST updating bugs, correcting miscompilation in PR8041
...
llvm-svn: 112878
2010-09-02 22:19:10 +00:00
6778149f7e
Reapply commit 112699, speculatively reverted by echristo, since
...
I'm sure it is harmless. Original commit message:
If PrototypeValue is erased in the middle of using the SSAUpdator
then the SSAUpdator may access freed memory. Instead, simply pass
in the type and name explicitly, which is all that was used anyway.
llvm-svn: 112810
2010-09-02 08:14:03 +00:00
8af45a889d
deepen my MMX/SRoA hack to avoid hurting non-x86 codegen.
...
llvm-svn: 112763
2010-09-01 23:09:27 +00:00
0ad7d9c24e
Fix loop unswitching's assumption that a code path which either
...
infinite loops or exits will eventually exit. This fixes PR5373.
llvm-svn: 112745
2010-09-01 21:46:45 +00:00
73f988cafa
JumpThreading keeps LazyValueInfo up to date, so we don't need to rerun it
...
if we schedule another LVI-using pass afterwards.
llvm-svn: 112722
2010-09-01 18:27:22 +00:00
a5d315c665
Speculatively revert 112699 and 112702, they seem to be causing
...
self host errors on clang-x86-64.
llvm-svn: 112719
2010-09-01 17:29:10 +00:00
f7b18437b5
If PrototypeValue is erased in the middle of using the SSAUpdator
...
then the SSAUpdator may access freed memory. Instead, simply pass
in the type and name explicitly, which is all that was used anyway.
llvm-svn: 112699
2010-09-01 10:29:33 +00:00
34e5361eb5
add a gross hack to work around a problem that Argiris reported
...
on llvmdev: SRoA is introducing MMX datatypes like <1 x i64>,
which then cause random problems because the X86 backend is
producing mmx stuff without inserting proper emms calls.
In the short term, force off MMX datatypes. In the long term,
the X86 backend should not select generic vector types to MMX
registers. This is being worked on, but won't be done in time
for 2.8. rdar://8380055
llvm-svn: 112696
2010-09-01 05:14:33 +00:00
110ed64fbb
Revert 112442 and 112440 until the compile time problems introduced
...
by 112440 are resolved.
llvm-svn: 112692
2010-09-01 01:45:53 +00:00
030f02021b
licm is wasting time hoisting constant foldable operations,
...
instead of hoisting them, just fold them away. This occurs in the
testcase for PR8041, for example.
llvm-svn: 112669
2010-08-31 23:00:16 +00:00
daca6f3483
tidy up
...
llvm-svn: 112643
2010-08-31 21:21:25 +00:00
3c84ecb067
More cleanups of my JumpThreading transforms, including extracting some duplicated code into a helper function.
...
llvm-svn: 112634
2010-08-31 20:26:04 +00:00
6fdcb172a9
Add an RAII helper to make cleanup of the RecursionSet more fool-proof.
...
llvm-svn: 112628
2010-08-31 19:24:27 +00:00
048efbe225
Only try to clean up the current block if we changed that block already.
...
llvm-svn: 112625
2010-08-31 18:55:52 +00:00
cd4de7f399
Refactor my fix for PR5652 to terminate the predecessor lookups after the first failure.
...
llvm-svn: 112620
2010-08-31 18:48:48 +00:00
68984ede5c
Fix an infinite loop; merging two functions will create a new function (if the
...
two are weak, we make them thunks to a new strong function) so don't iterate
through the function list as we're modifying it.
Also add back the outermost loop which got removed during the cleanups.
llvm-svn: 112595
2010-08-31 08:29:37 +00:00
ce401be792
Don't perform an extra traversal of the function just to do cleanup. We can safely simplify instructions after each block has been processed without worrying about iterator invalidation.
...
llvm-svn: 112594
2010-08-31 07:55:56 +00:00
48d58ad64c
Rename ValuePropagation to a more descriptive CorrelatedValuePropagation.
...
llvm-svn: 112591
2010-08-31 07:48:34 +00:00
d2918a07bd
Rename file to something more descriptive.
...
llvm-svn: 112590
2010-08-31 07:41:39 +00:00
3997a07fb9
More Chris-inspired JumpThreading fixes: use ConstantExpr to correctly constant-fold undef, and be more careful with its return value.
...
This actually exposed an infinite recursion bug in ComputeValueKnownInPredecessors which theoretically already existed (in JumpThreading's
handling of and/or of i1's), but never manifested before. This patch adds a tracking set to prevent this case.
llvm-svn: 112589
2010-08-31 07:36:34 +00:00
0464d1d7ec
Switch to DenseSet, simplifying much more code. We now have a single iteration
...
where we hash, compare and fold, instead of one iteration where we build up
the hash buckets and a second one to fold.
llvm-svn: 112582
2010-08-31 05:53:05 +00:00
376597c13e
Remove r111665, which implemented store-narrowing in InstCombine. Chris discovered a miscompilation in it, and it's not easily
...
fixable at the optimizer level. I'll investigate reimplementing it in DAGCombine.
llvm-svn: 112575
2010-08-31 04:41:06 +00:00
b58b3c0dda
Fix a typo.
...
llvm-svn: 112560
2010-08-30 23:59:30 +00:00
b974dbbdd7
Cleanups suggested by Chris.
...
llvm-svn: 112553
2010-08-30 23:34:17 +00:00
c910acb54a
Re-apply r112539, being more careful to respect the return values of the constant folding methods. Additionally,
...
use the ConstantExpr::get*() methods to simplify some constant folding.
llvm-svn: 112550
2010-08-30 23:22:36 +00:00
30bacbdfdf
Add statistics to evaluate this pass.
...
llvm-svn: 112545
2010-08-30 22:45:55 +00:00
1ddcbbe49c
Revert r112539. It accidentally introduced a miscompilation.
...
llvm-svn: 112543
2010-08-30 22:33:41 +00:00
75f6037c7c
Fixes and cleanups pointed out by Chris. In general, be careful to handle 0 results from ComputeValueKnownInPredecessors
...
(indicating undef), and re-use existing constant folding APIs.
llvm-svn: 112539
2010-08-30 22:07:52 +00:00
c843fca2fd
rewrite DwarfEHPrepare to use SSAUpdater to promote its allocas
...
instead of PromoteMemToReg. This allows it to stop using DF and DT,
eliminating a computation of DT and DF from clang -O3. Clang is now
down to 2 runs of DomFrontier.
llvm-svn: 112457
2010-08-29 19:54:28 +00:00
f58382ed87
two changes: 1) make AliasSet hold the list of call sites with an
...
assertingvh so we get a violent explosion if the pointer dangles.
2) Fix AliasSetTracker::deleteValue to remove call sites with
by-pointer comparisons instead of by-alias queries. Using
findAliasSetForCallSite can cause alias sets to get merged
when they shouldn't, and can also miss alias sets when the
call is readonly.
#2 fixes PR6889, which only repros with a .c file :(
llvm-svn: 112452
2010-08-29 18:42:23 +00:00
263f804699
LICM does get dead instructions input to it. Instead of sinking them
...
out of loops, just delete them.
llvm-svn: 112451
2010-08-29 18:22:25 +00:00
6ac0659a1c
use moveBefore instead of remove+insert, it avoids some
...
symtab manipulation, so its faster (in addition to being
more elegant)
llvm-svn: 112450
2010-08-29 18:18:40 +00:00
f03b4eac48
revert 112448 for now.
...
llvm-svn: 112449
2010-08-29 18:11:16 +00:00
11f8ad8211
optimize LICM::hoist to use moveBefore. Correct its updating
...
of AST to remove the hoisted instruction from the AST, since it
is no longer in the loop.
llvm-svn: 112448
2010-08-29 18:03:33 +00:00
1a1ed69435
fix some bugs (found by inspection) where LICM would not update
...
LICM correctly. When sinking an instruction, it should not add
entries for the sunk instruction to the AST, it should remove
the entry for the sunk instruction. The blocks being sunk to
are not in the loop, so their instructions shouldn't be in the
AST (yet)!
llvm-svn: 112447
2010-08-29 18:00:00 +00:00
cc9cbc66a3
rework the ownership of subloop alias information: instead of
...
keeping them around until the pass is destroyed, keep them
around a) just when useful (not for outer loops) and b) destroy
them right after we use them. This should reduce memory use
and fixes potential bugs where a loop is deleted and another
loop gets allocated to the same address.
llvm-svn: 112446
2010-08-29 17:46:00 +00:00
bc1a65ac6c
apparently unswitch had the same "Feature". Stop its
...
claims that it preserves domfrontier if it doesn't really.
llvm-svn: 112445
2010-08-29 17:23:19 +00:00
d6f46b8af8
now that loop passes don't use DomFrontier, there is no reason
...
for the unroller to pretend it supports updating it. It still
has a horrible hack for DomTree.
llvm-svn: 112444
2010-08-29 17:21:35 +00:00
002ff89cbd
Optionally rerun dedicated-register filtering after applying
...
other filtering techniques, as those may allow it to filter
out more obviously unprofitable candidates.
llvm-svn: 112441
2010-08-29 16:39:22 +00:00
f031792cc6
Fix several areas in LSR to do a better job keeping the main
...
LSRInstance data structures up to date. This fixes some
pessimizations caused by stale data which will be exposed
in an upcoming change.
llvm-svn: 112440
2010-08-29 16:32:54 +00:00
e9e0873b08
Refactor the three main groups of code out of
...
NarrowSearchSpaceUsingHeuristics into separate functions.
llvm-svn: 112439
2010-08-29 16:09:42 +00:00
37a0f68036
Delete a bogus check.
...
llvm-svn: 112438
2010-08-29 15:30:29 +00:00
b6a520d63c
Add some comments.
...
llvm-svn: 112437
2010-08-29 15:27:08 +00:00
bf673e0652
Move this debug output into GenerateAllReuseFormula, to declutter
...
the high-level logic.
llvm-svn: 112436
2010-08-29 15:21:38 +00:00
d366b6d5c8
Delete an unused declaration.
...
llvm-svn: 112435
2010-08-29 15:19:11 +00:00
4f13bbfefc
Do one lookup instead of two.
...
llvm-svn: 112434
2010-08-29 15:18:49 +00:00
f94f6bb0ba
licm preserves the cfg, it doesn't have to explicitly say it
...
preserves domfrontier. It does preserve AA though.
llvm-svn: 112419
2010-08-29 07:02:56 +00:00
abe61ef3b4
now that it doesn't use the PromoteMemToReg function, LICM doesn't
...
require DomFrontier. Dropping this doesn't actually save any runs
of the pass though.
llvm-svn: 112418
2010-08-29 06:49:44 +00:00
1dc98b47b5
completely rewrite the memory promotion algorithm in LICM.
...
Among other things, this uses SSAUpdater instead of
PromoteMemToReg.
llvm-svn: 112417
2010-08-29 06:43:52 +00:00
9c3931a544
use getUniqueExitBlocks instead of a manual set.
...
llvm-svn: 112412
2010-08-29 05:12:21 +00:00
85bf5421e1
reimplement LICM::sink to use SSAUpdater instead of PromoteMemToReg.
...
This leads to much simpler code.
llvm-svn: 112410
2010-08-29 04:55:06 +00:00
c3fb03e289
implement SSAUpdater::RewriteUseAfterInsertions, a helpful form of RewriteUse.
...
llvm-svn: 112409
2010-08-29 04:54:06 +00:00
b50407f104
remove dead proto
...
llvm-svn: 112408
2010-08-29 04:53:24 +00:00
cd96b4df56
reduce indentation in LICM::sink by using early exits, use
...
getUniqueExitBlocks instead of getExitBlocks and a manual
set to eliminate dupes.
llvm-svn: 112405
2010-08-29 04:28:20 +00:00
188cc5a0fc
modernize this pass a bit: use efficient set/map and reduce indentation.
...
llvm-svn: 112404
2010-08-29 04:23:04 +00:00
13ee795c42
remove unions from LLVM IR. They are severely buggy and not
...
being actively maintained, improved, or extended.
llvm-svn: 112356
2010-08-28 04:09:24 +00:00
504e5100d3
remove the ABCD and SSI passes. They don't have any clients that
...
I'm aware of, aren't maintained, and LVI will be replacing their value.
nlewycky approved this on irc.
llvm-svn: 112355
2010-08-28 03:51:24 +00:00
50df36ac0a
for completeness, allow undef also.
...
llvm-svn: 112351
2010-08-28 03:36:51 +00:00
95bb297c26
squish dead code.
...
llvm-svn: 112350
2010-08-28 03:21:03 +00:00
d0214f3efe
handle the constant case of vector insertion. For something
...
like this:
struct S { float A, B, C, D; };
struct S g;
struct S bar() {
struct S A = g;
++A.B;
A.A = 42;
return A;
}
we now generate:
_bar: ## @bar
## BB#0: ## %entry
movq _g@GOTPCREL(%rip), %rax
movss 12(%rax), %xmm0
pshufd $16, %xmm0, %xmm0
movss 4(%rax), %xmm2
movss 8(%rax), %xmm1
pshufd $16, %xmm1, %xmm1
unpcklps %xmm0, %xmm1
addss LCPI1_0(%rip), %xmm2
pshufd $16, %xmm2, %xmm2
movss LCPI1_1(%rip), %xmm0
pshufd $16, %xmm0, %xmm0
unpcklps %xmm2, %xmm0
ret
instead of:
_bar: ## @bar
## BB#0: ## %entry
movq _g@GOTPCREL(%rip), %rax
movss 12(%rax), %xmm0
pshufd $16, %xmm0, %xmm0
movss 4(%rax), %xmm2
movss 8(%rax), %xmm1
pshufd $16, %xmm1, %xmm1
unpcklps %xmm0, %xmm1
addss LCPI1_0(%rip), %xmm2
movd %xmm2, %eax
shlq $32, %rax
addq $1109917696, %rax ## imm = 0x42280000
movd %rax, %xmm0
ret
llvm-svn: 112345
2010-08-28 01:50:57 +00:00
dd6601048e
optimize bitcasts from large integers to vector into vector
...
element insertion from the pieces that feed into the vector.
This handles a pattern that occurs frequently due to code
generated for the x86-64 abi. We now compile something like
this:
struct S { float A, B, C, D; };
struct S g;
struct S bar() {
struct S A = g;
++A.A;
++A.C;
return A;
}
into all nice vector operations:
_bar: ## @bar
## BB#0: ## %entry
movq _g@GOTPCREL(%rip), %rax
movss LCPI1_0(%rip), %xmm1
movss (%rax), %xmm0
addss %xmm1, %xmm0
pshufd $16, %xmm0, %xmm0
movss 4(%rax), %xmm2
movss 12(%rax), %xmm3
pshufd $16, %xmm2, %xmm2
unpcklps %xmm2, %xmm0
addss 8(%rax), %xmm1
pshufd $16, %xmm1, %xmm1
pshufd $16, %xmm3, %xmm2
unpcklps %xmm2, %xmm1
ret
instead of icky integer operations:
_bar: ## @bar
movq _g@GOTPCREL(%rip), %rax
movss LCPI1_0(%rip), %xmm1
movss (%rax), %xmm0
addss %xmm1, %xmm0
movd %xmm0, %ecx
movl 4(%rax), %edx
movl 12(%rax), %esi
shlq $32, %rdx
addq %rcx, %rdx
movd %rdx, %xmm0
addss 8(%rax), %xmm1
movd %xmm1, %eax
shlq $32, %rsi
addq %rax, %rsi
movd %rsi, %xmm1
ret
This resolves rdar://8360454
llvm-svn: 112343
2010-08-28 01:20:38 +00:00
83f9ff0452
Update CMake build. Add newline at end of file.
...
llvm-svn: 112332
2010-08-28 00:11:12 +00:00
cf7f941121
Add a prototype of a new peephole optimizing pass that uses LazyValue info to simplify PHIs and select's.
...
This pass addresses the missed optimizations from PR2581 and PR4420.
llvm-svn: 112325
2010-08-27 23:31:36 +00:00
6c1395f62a
Enhance the shift propagator to handle the case when you have:
...
A = shl x, 42
...
B = lshr ..., 38
which can be transformed into:
A = shl x, 4
...
iff we can prove that the would-be-shifted-in bits
are already zero. This eliminates two shifts in the testcase
and allows eliminate of the whole i128 chain in the real example.
llvm-svn: 112314
2010-08-27 22:53:44 +00:00
18d7fc8fc6
Implement a pretty general logical shift propagation
...
framework, which is good at ripping through bitfield
operations. This generalize a bunch of the existing
xforms that instcombine does, such as
(x << c) >> c -> and
to handle intermediate logical nodes. This is useful for
ripping up the "promote to large integer" code produced by
SRoA.
llvm-svn: 112304
2010-08-27 22:24:38 +00:00
25a198e72b
remove some special shift cases that have been subsumed into the
...
more general simplify demanded bits logic.
llvm-svn: 112291
2010-08-27 21:04:34 +00:00
99d4cb861b
Fix typos in comments.
...
llvm-svn: 112286
2010-08-27 20:32:56 +00:00
7398434675
teach the truncation optimization that an entire chain of
...
computation can be truncated if it is fed by a sext/zext that doesn't
have to be exactly equal to the truncation result type.
llvm-svn: 112285
2010-08-27 20:32:06 +00:00
90cd746e63
Add an instcombine to clean up a common pattern produced
...
by the SRoA "promote to large integer" code, eliminating
some type conversions like this:
%94 = zext i16 %93 to i32 ; <i32> [#uses=2]
%96 = lshr i32 %94, 8 ; <i32> [#uses=1]
%101 = trunc i32 %96 to i8 ; <i8> [#uses=1]
This also unblocks other xforms from happening, now clang is able to compile:
struct S { float A, B, C, D; };
float foo(struct S A) { return A.A + A.B+A.C+A.D; }
into:
_foo: ## @foo
## BB#0: ## %entry
pshufd $1, %xmm0, %xmm2
addss %xmm0, %xmm2
movdqa %xmm1, %xmm3
addss %xmm2, %xmm3
pshufd $1, %xmm1, %xmm0
addss %xmm3, %xmm0
ret
on x86-64, instead of:
_foo: ## @foo
## BB#0: ## %entry
movd %xmm0, %rax
shrq $32, %rax
movd %eax, %xmm2
addss %xmm0, %xmm2
movapd %xmm1, %xmm3
addss %xmm2, %xmm3
movd %xmm1, %rax
shrq $32, %rax
movd %eax, %xmm0
addss %xmm3, %xmm0
ret
This seems pretty close to optimal to me, at least without
using horizontal adds. This also triggers in lots of other
code, including SPEC.
llvm-svn: 112278
2010-08-27 18:31:05 +00:00
6ebbd92380
Use LVI to eliminate conditional branches where we've tested a related condition previously. Update tests for this change.
...
This fixes PR5652.
llvm-svn: 112270
2010-08-27 17:12:29 +00:00
bfd2228182
optimize "integer extraction out of the middle of a vector" as produced
...
by SRoA. This is part of rdar://7892780, but needs another xform to
expose this.
llvm-svn: 112232
2010-08-26 22:14:59 +00:00
d4ebd6df5a
optimize bitcast(trunc(bitcast(x))) where the result is a float and 'x'
...
is a vector to be a vector element extraction. This allows clang to
compile:
struct S { float A, B, C, D; };
float foo(struct S A) { return A.A + A.B+A.C+A.D; }
into:
_foo: ## @foo
## BB#0: ## %entry
movd %xmm0, %rax
shrq $32, %rax
movd %eax, %xmm2
addss %xmm0, %xmm2
movapd %xmm1, %xmm3
addss %xmm2, %xmm3
movd %xmm1, %rax
shrq $32, %rax
movd %eax, %xmm0
addss %xmm3, %xmm0
ret
instead of:
_foo: ## @foo
## BB#0: ## %entry
movd %xmm0, %rax
movd %eax, %xmm0
shrq $32, %rax
movd %eax, %xmm2
addss %xmm0, %xmm2
movd %xmm1, %rax
movd %eax, %xmm1
addss %xmm2, %xmm1
shrq $32, %rax
movd %eax, %xmm0
addss %xmm1, %xmm0
ret
... eliminating half of the horribleness.
llvm-svn: 112227
2010-08-26 21:55:42 +00:00
bd2ecc7e68
Make JumpThreading smart enough to properly thread StrSwitch when it's compiled with clang++.
...
llvm-svn: 112198
2010-08-26 17:40:24 +00:00
ca26f79051
Reapply r112091 and r111922, support for metadata linking, with a
...
fix: add a flag to MapValue and friends which indicates whether
any module-level mappings are being made. In the common case of
inlining, no module-level mappings are needed, so MapValue doesn't
need to examine non-function-local metadata, which can be very
expensive in the case of a large module with really deep metadata
(e.g. a large C++ program compiled with -g).
This flag is a little awkward; perhaps eventually it can be moved
into the ClonedCodeInfo class.
llvm-svn: 112190
2010-08-26 15:41:53 +00:00
ce45863f0d
Revert r111922, "MapValue support for MDNodes. This is similar to r109117,
...
except ...", it is causing *massive* performance regressions when building Clang
with itself (-O3 -g).
llvm-svn: 112158
2010-08-26 03:48:11 +00:00
95fe13c720
Revert r112091, "Remap metadata attached to instructions when remapping
...
individual ...", which depends on r111922, which I am reverting.
llvm-svn: 112157
2010-08-26 03:48:08 +00:00
07afbd5a08
zap dead code.
...
llvm-svn: 112130
2010-08-26 01:13:54 +00:00
8f292e7a6d
Rewrite ExtractGV, removing a bunch of stuff that didn't fully work,
...
and was over-complicated, and replacing it with a simple implementation.
llvm-svn: 112120
2010-08-26 00:22:55 +00:00
8df99b523e
remove some llvmcontext arguments that are now dead post-refactoring.
...
llvm-svn: 112104
2010-08-25 23:00:45 +00:00
fd824487a3
Remap metadata attached to instructions when remapping individual
...
instructions, not when remapping modules.
llvm-svn: 112091
2010-08-25 21:36:50 +00:00
01262e129e
DIGlobalVariable can be used to encode debug info for globals that are directly folded into a constant by FE.
...
llvm-svn: 112072
2010-08-25 18:52:02 +00:00
a209503467
Use MapValue in the Linker instead of having a private function
...
which does the same thing. This eliminates redundant code and
handles MDNodes better. MDNode linking still doesn't fully
work yet though.
llvm-svn: 111941
2010-08-24 18:50:07 +00:00
7c853e877e
Turn LVI on, previously detected failures should be fixed now.
...
llvm-svn: 111923
2010-08-24 17:21:18 +00:00
6901283544
MapValue support for MDNodes. This is similar to r109117, except
...
that it avoids a lot of unnecessary cloning by avoiding remapping
MDNode cycles when none of the nodes in the cycle actually need to
be remapped. Also it uses the new temporary MDNode mechanism.
llvm-svn: 111922
2010-08-24 17:10:10 +00:00
6ffa3f2aea
Turn LVI back off, I have a testcase now.
...
llvm-svn: 111834
2010-08-23 19:59:27 +00:00
630add39a6
Re-enable LazyValueInfo. Monitoring for failures.
...
llvm-svn: 111816
2010-08-23 18:12:23 +00:00
d31d82d75c
Now that PassInfo and Pass::ID have been separated, move the rest of the passes over to the new registration API.
...
llvm-svn: 111815
2010-08-23 17:52:01 +00:00
84c29a096b
Re-apply r111568 with a fix for the clang self-host.
...
llvm-svn: 111665
2010-08-20 18:24:43 +00:00
43057cd56a
Revert r111568 to unbreak clang self-host.
...
llvm-svn: 111571
2010-08-19 23:25:16 +00:00
bb723b228a
When a set of bitmask operations, typically from a bitfield initialization, only modifies the low bytes of a value,
...
we can narrow the store to only over-write the affected bytes.
llvm-svn: 111568
2010-08-19 22:15:40 +00:00
aac8cbb261
Disable LVI while I evaluate a failure.
...
llvm-svn: 111551
2010-08-19 19:47:08 +00:00
5c87dd55d3
Tentatively enabled LVI by default. I'll be monitoring for any failures.
...
llvm-svn: 111543
2010-08-19 19:04:40 +00:00
129a816ee6
Process the step before the start, because it's usually the simpler
...
of the two.
llvm-svn: 111495
2010-08-19 01:02:31 +00:00
208636fa33
Inform LazyValueInfo whenever a block is deleted, to avoid dangling pointer issues.
...
llvm-svn: 111382
2010-08-18 18:39:01 +00:00
3c603024bb
Fix PR7755: knowing something about an inval for a pred
...
from the LHS should disable reconsidering that pred on the
RHS. However, knowing something about the pred on the RHS
shouldn't disable subsequent additions on the RHS from
happening.
llvm-svn: 111349
2010-08-18 03:14:36 +00:00
f0b5b67ba5
fit in 80 cols
...
llvm-svn: 111348
2010-08-18 03:13:35 +00:00
b45de95345
remove some dead code.
...
llvm-svn: 111344
2010-08-18 02:41:56 +00:00
6aabb66139
remove dead prototype.
...
llvm-svn: 111342
2010-08-18 02:37:06 +00:00
51edc7b7e1
Temporarily revert r110987 as it's causing some miscompares in
...
vector heavy code. I'll re-enable when we've tracked down the problem.
llvm-svn: 111318
2010-08-17 22:55:27 +00:00
5047ca0c02
When rotating loops, put the original header at the bottom of the
...
loop, making the resulting loop significantly less ugly. Also, zap
its trivial PHI nodes, since it's easy.
llvm-svn: 111255
2010-08-17 17:39:21 +00:00
941020ed72
Use the getUniquePredecessor() utility function, instead of doing
...
what it does manually.
llvm-svn: 111248
2010-08-17 17:07:02 +00:00
8b637b177c
Add an option to disable codegen prepare critical edge splitting. In theory, PHI elimination is already doing all (most?) of the splitting needed. But machine-licm and machine-sink seem to miss some important optimizations when splitting is disabled.
...
llvm-svn: 111224
2010-08-17 01:34:49 +00:00
89fdbaf99a
Instead of having CollectSubexpr's categorize operands as interesting or
...
uninteresting, just put all the operands on one list and make
GenerateReassociations make the decision about what's interesting.
This is simpler, and it avoids an extra ScalarEvolution::getAddExpr call.
llvm-svn: 111133
2010-08-16 15:50:00 +00:00
9b7632df26
Put add operands in ScalarEvolution-canonical order, when convenient.
...
This isn't necessary, because ScalarEvolution sorts them anyway,
but it's tidier this way.
llvm-svn: 111132
2010-08-16 15:39:27 +00:00
6e964c7fb4
Avoid #include <ScalarEvolution.h> in LoopSimplify.cpp, which doesn't
...
actually use ScalarEvolution.
llvm-svn: 111124
2010-08-16 14:44:03 +00:00
250b754428
Instead, teach SimplifyCFG to trim non-address-taken blocks from
...
indirectbr destination lists.
llvm-svn: 111122
2010-08-16 14:41:14 +00:00
aa445c0751
LoopSimplify shouldn't split loop backedges that use indirectbr. PR7867.
...
llvm-svn: 111061
2010-08-14 00:43:09 +00:00
4a63fad976
Teach SimplifyCFG how to simplify indirectbr instructions.
...
- Eliminate redundant successors.
- Convert an indirectbr with one successor into a direct branch.
Also, generalize SimplifyCFG to be able to be run on a function entry block.
It knows quite a few simplifications which are applicable to the entry
block, and it only needs a few checks to avoid trouble with the entry block.
llvm-svn: 111060
2010-08-14 00:29:42 +00:00
Bob Wilson