Commit Graph

2558 Commits

Author SHA1 Message Date
Chris Lattner
967948b4e2 "Allow LICM to sink or lift loads from constant memory. Also add a test
case for this.

This allows instructions like loads from global variables declared to
be constant to be moved out of loops."

Patch by Stefanus Du Toit!



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53945 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-23 05:06:28 +00:00
Dan Gohman
fc74abfba5 Enable first-class aggregates support.
Remove the GetResultInst instruction. It is still accepted in LLVM assembly
and bitcode, where it is now auto-upgraded to ExtractValueInst. Also, remove
support for return instructions with multiple values. These are auto-upgraded
to use InsertValueInst instructions.

The IRBuilder still accepts multiple-value returns, and auto-upgrades them
to InsertValueInst instructions.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53941 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-23 00:34:11 +00:00
Dan Gohman
7375fbcbf2 Fix a bug in LSR's dead-PHI cleanup. If a PHI has a def-use chain that
leads into a cycle involving a different PHI, LSR got stuck running
around that cycle looking for the original PHI. To avoid this, keep
track of visited PHIs and stop searching if we see one more than once.
This fixes PR2570.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53879 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-21 21:45:02 +00:00
Duncan Sands
6c4d8e3ee4 Supress a gcc-4.3 warning.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53771 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-18 21:06:02 +00:00
Owen Anderson
e98c54c47c Make PRE actually handle critical edges (by splitting them). Confirmed that bootstrap passes with this change.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53762 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-18 18:03:38 +00:00
Owen Anderson
c2b856e5b4 Enable PRE. My last batch of changes fixed the miscompile.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53730 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-17 19:41:00 +00:00
Chris Lattner
d137ab4783 Fix PR2553
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53715 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-17 06:07:20 +00:00
Owen Anderson
b31b06d04b Factor MergeBlockIntoPredecessor out into BasicBlockUtils.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53705 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-17 00:01:40 +00:00
Owen Anderson
5d0af038a7 There's no need to iterate block merging and PRE. In fact, iterating the latter
could cause problems for memdep when it breaks critical edges.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53691 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-16 17:52:31 +00:00
Matthijs Kooijman
780ae5e179 Redo InstCombiner::visitExtractValueInst. Instead of using the (complicate)
FindInsertedValue, it now performs a number of simple transformations that
should result in the same effect when applied iteratively.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53673 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-16 12:55:45 +00:00
Evan Cheng
be8efa54b6 Fix PR2296. Do not transform x86_sse2_storel_dq into a full-width store.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53666 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-16 07:28:14 +00:00
Owen Anderson
2b4f7bbd49 Revert this, as it seems to still be broken.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53627 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-15 17:59:02 +00:00
Owen Anderson
b33f7cc7e6 Enable local PRE by default.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53616 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-15 16:28:23 +00:00
Owen Anderson
961edc8be6 Have GVN do a pre-pass over the CFG that folds away unconditional branches where possible. This allows local PRE to be more aggressive.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53615 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-15 16:28:06 +00:00
Dan Gohman
3fea643fb4 Fix uninitialized use of the Changed variable.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53564 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-14 17:55:01 +00:00
Chris Lattner
fdb19e5887 Fix PR2506 by being a bit more careful about reverse fact propagation when
disproving a condition.  This actually compiles the existing testcase
(udiv_select_to_select_shift) to:

define i64 @test(i64 %X, i1 %Cond) {
entry:
	%divisor1.t = lshr i64 %X, 3		; <i64> [#uses=1]
	%quotient2 = lshr i64 %X, 3		; <i64> [#uses=1]
	%sum = add i64 %divisor1.t, %quotient2		; <i64> [#uses=1]
	ret i64 %sum
}

instead of:

define i64 @test(i64 %X, i1 %Cond) {
entry:
	%quotient1.v = select i1 %Cond, i64 3, i64 4		; <i64> [#uses=1]
	%quotient1 = lshr i64 %X, %quotient1.v		; <i64> [#uses=1]
	%quotient2 = lshr i64 %X, 3		; <i64> [#uses=1]
	%sum = add i64 %quotient1, %quotient2		; <i64> [#uses=1]
	ret i64 %sum
}



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53534 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-14 00:15:52 +00:00
Nick Lewycky
3ac9e109b2 Enhance analysis of srem.
Remove dead code analyzing urem. 'urem' of power-of-2 is canonicalized to an
'and' instruction.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53506 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-12 05:04:38 +00:00
Owen Anderson
c45996bf74 Don't call lookupNumber more than we have to.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53470 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 20:05:13 +00:00
Nick Lewycky
7d9843fc37 Document 'mask' in this calculation.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53454 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 08:16:26 +00:00
Nick Lewycky
1f26c188f8 Remove misleading constant from comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53452 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 07:36:19 +00:00
Nick Lewycky
4bf1e59819 Add another optimization from PR2330. Also catch some missing cases that are
similar.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53451 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 07:20:53 +00:00
Chris Lattner
d01bee7170 a missed optimization that Eli spotted
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53449 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 06:40:29 +00:00
Chris Lattner
f9685ac68e another bug in the same line.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53448 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 06:38:16 +00:00
Chris Lattner
a8ff4a88bb fix a bug spotted by Eli's eagle eyes
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53447 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 06:36:01 +00:00
Chris Lattner
183661e43a simplify and merge a bunch of code. Instead of comparing against
the min/max values for an integer type, compare against the min/max
values we can prove contain the input.  This might be a tighter bound,
so this is general goodness.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53446 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 05:40:05 +00:00
Chris Lattner
84dff672a4 fold away (x <= cst) earlier, allowing us to not have to
handle them in some code.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53445 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 05:08:55 +00:00
Chris Lattner
85b5eb0505 Fix folding of icmp's of i1 where the comparison is signed. The code
was using the algorithm for folding unsigned comparisons which is
completely wrong.  This has been broken since the signless types change.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53444 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 04:20:58 +00:00
Chris Lattner
f299184565 Fix a bogus optimization: folding (slt (zext i1 A to i32), 1) -> (slt i1 A, true)
This cause a regression in InstCombine/JavaCompare, which was doing the right
thing on accident.  To handle the missed case, generalize the comparisons based
on masked bits a little bit to handle comparisons against the max value. For 
example, we can now xform (slt i32 (and X, 4), 4) -> (setne i32 (and X, 4), 4)




git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53443 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-11 04:09:09 +00:00
Nick Lewycky
5dcc41f5b3 Fix overzealous optimization. Thanks to Duncan Sands for pointing out my error!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53393 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-10 05:51:40 +00:00
Nick Lewycky
d8ad492238 Simplify, suggested by Chris Lattner.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53283 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-09 07:35:26 +00:00
Nick Lewycky
9ee863ecc0 Fold (a < 8) && (b < 8) into (a|b) < 8 for unsigned less or greater than.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53282 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-09 07:29:11 +00:00
Nick Lewycky
d1f77bf931 Fold ((1 << a) & 1) to (a == 0).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53276 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-09 05:20:13 +00:00
Nick Lewycky
b4d1bc989e Reduce x - y to -y when we know the 'x' part will get masked off anyways.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53271 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-09 04:32:37 +00:00
Devang Patel
a5e27f8028 If loop induction variable's start value is less then its exit value then do not split the loop.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53265 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-09 00:12:01 +00:00
Chris Lattner
710429645a Fix PR2496, a really nasty bug which involved sinking volatile loads
into phis.  This is actually the same bug as PR2262 /
2008-04-29-VolatileLoadDontMerge.ll, but I missed checking the first 
predecessor for multiple successors.  Testcase here:
InstCombine/2008-07-08-VolatileLoadMerge.ll



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53240 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-08 17:18:32 +00:00
Evan Cheng
1ce75dcbbc Fix two serious LSR bugs.
1. LSR runOnLoop is always returning false regardless if any transformation is made.
2. AddUsersIfInteresting can create new instructions that are added to DeadInsts. But there is a later early exit which prevents them from being freed.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53193 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-07 19:51:32 +00:00
Dan Gohman
6b345ee9b2 Make DenseMap's insert return a pair, to more closely resemble std::map.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53177 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-07 17:46:23 +00:00
Nick Lewycky
b8cd6a49b5 Fix missed optimization opportunity when analyzing cast of mul and select.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53151 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-05 21:19:34 +00:00
Owen Anderson
0ae33ef202 Use information already present in the ValueTable to fast-fail when we know there won't be a value number match. This speeds up GVN on a case where there are very few redundancies by ~25%.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53108 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-03 17:44:33 +00:00
Devang Patel
77a01136ff Do not try to update dominator info while manipulating CFG. This code does not handle all cases and keeps invalid dom info around some cases, which misleads other passes down stream.
Right now, dom info is recaluclated in the end if the loop is switched.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53106 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-03 17:37:52 +00:00
Owen Anderson
9da02b844f Remove the ability for ADCE to remove unreachable blocks in loop nests, because, as Eli pointed out, SimplifyCFG already does this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53104 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-03 17:21:41 +00:00
Bill Wendling
38976dc394 Remove unused function.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53090 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-03 07:10:03 +00:00
Devang Patel
98260a474c Preserve dom info.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53089 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-03 07:04:22 +00:00
Devang Patel
f31b3bbd10 Remove extra FIXME
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53087 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-03 06:50:04 +00:00
Devang Patel
4be7d294f8 Reconstruct dom info, if loop is unswitched.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53086 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-03 06:48:21 +00:00
Devang Patel
b76247b8b1 LoopUnswitch does not preserve dominator info in all cases.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53085 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-03 05:55:03 +00:00
Devang Patel
825cb98d9a Undo previous patch. It is not that simple to fix dom info here.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53062 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-03 00:08:13 +00:00
Devang Patel
501e763dfe Preserve dom info while simplifing loop after the unswitch.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53052 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-02 22:58:54 +00:00
Owen Anderson
ea6462bfd0 Use df_ext_iterator to capture the reachable set without allocating an extra set.
Also, move large sets and vectors out of instance variables and onto the stack,
and give them more reasonable sizes.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53044 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-02 18:41:09 +00:00
Owen Anderson
cb29a4f66a Avoid a redundant call.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53040 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-02 18:15:31 +00:00