Duncan Sands [Tue, 25 Oct 2011 12:28:52 +0000 (12:28 +0000)]
Restore commits 142790 and 142843 - they weren't breaking the build
bots. Original commit messages:
- Reapply r142781 with fix. Original message:
Enhance SCEV's brute force loop analysis to handle multiple PHI nodes in the
loop header when computing the trip count.
With this, we now constant evaluate:
struct ListNode { const struct ListNode *next; int i; };
static const struct ListNode node1 = {0, 1};
static const struct ListNode node2 = {&node1, 2};
static const struct ListNode node3 = {&node2, 3};
int test() {
int sum = 0;
for (const struct ListNode *n = &node3; n != 0; n = n->next)
sum += n->i;
return sum;
}
- Now that we look at all the header PHIs, we need to consider all the header PHIs
when deciding that the loop has stopped evolving. Fixes miscompile in the gcc
torture testsuite!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142919
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Tue, 25 Oct 2011 09:47:41 +0000 (09:47 +0000)]
Fix the API usage in loop probability heuristics. It was incorrectly
classifying many edges as exiting which were in fact not. These mainly
formed edges into sub-loops. It was also not correctly classifying all
returning edges out of loops as leaving the loop. With this match most
of the loop heuristics are more rational.
Several serious regressions on loop-intesive benchmarks like perlbench's
loop tests when built with -enable-block-placement are fixed by these
updated heuristics. Unfortunately they in turn uncover some other
regressions. There are still several improvemenst that should be made to
loop heuristics including trip-count, and early back-edge management.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142917
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan Sands [Tue, 25 Oct 2011 09:26:43 +0000 (09:26 +0000)]
Speculatively revert commits 142790 and 142843 to see if it fixes
the dragonegg and llvm-gcc self-host buildbots. Original commit
messages:
- Reapply r142781 with fix. Original message:
Enhance SCEV's brute force loop analysis to handle multiple PHI nodes in the
loop header when computing the trip count.
With this, we now constant evaluate:
struct ListNode { const struct ListNode *next; int i; };
static const struct ListNode node1 = {0, 1};
static const struct ListNode node2 = {&node1, 2};
static const struct ListNode node3 = {&node2, 3};
int test() {
int sum = 0;
for (const struct ListNode *n = &node3; n != 0; n = n->next)
sum += n->i;
return sum;
}
- Now that we look at all the header PHIs, we need to consider all the header PHIs
when deciding that the loop has stopped evolving. Fixes miscompile in the gcc
torture testsuite!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142916
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Tue, 25 Oct 2011 07:05:26 +0000 (07:05 +0000)]
Remove dead enum value. There is no DIESectionOffset.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142912
91177308-0d34-0410-b5e6-
96231b3b80d8
Chad Rosier [Tue, 25 Oct 2011 01:22:20 +0000 (01:22 +0000)]
Fix these test cases to not use .bc files. Otherwise, we run into issues with
bitcode reader/writer backward compatibility.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142896
91177308-0d34-0410-b5e6-
96231b3b80d8
Bill Wendling [Tue, 25 Oct 2011 01:01:42 +0000 (01:01 +0000)]
Comment out external projects so that they can be refreshed for 3.0.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142893
91177308-0d34-0410-b5e6-
96231b3b80d8
Eric Christopher [Tue, 25 Oct 2011 00:55:35 +0000 (00:55 +0000)]
Remove unused forward decl.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142892
91177308-0d34-0410-b5e6-
96231b3b80d8
Bill Wendling [Tue, 25 Oct 2011 00:54:05 +0000 (00:54 +0000)]
An MBB which branches to an EH landing pad shouldn't be considered for tail merging.
In SjLj EH, the jump to the landing pad is not done explicitly through a branch
statement. The EH landing pad is added as a successor to the throwing
BB. Because of that however, the branch folding pass could mistakenly think that
it could merge the throwing BB with another BB. This isn't safe to do.
<rdar://problem/
10334833>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142891
91177308-0d34-0410-b5e6-
96231b3b80d8
Dan Gohman [Tue, 25 Oct 2011 00:36:31 +0000 (00:36 +0000)]
Really remove Blackfin from the configure script.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142889
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Tue, 25 Oct 2011 00:14:01 +0000 (00:14 +0000)]
ARM assembly parsing and encoding for VLD1 with writeback.
Four entry register lists.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142882
91177308-0d34-0410-b5e6-
96231b3b80d8
Dan Gohman [Tue, 25 Oct 2011 00:05:42 +0000 (00:05 +0000)]
Remove the Blackfin backend.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142880
91177308-0d34-0410-b5e6-
96231b3b80d8
Dan Gohman [Mon, 24 Oct 2011 23:48:32 +0000 (23:48 +0000)]
Remove the SystemZ backend.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142878
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Mon, 24 Oct 2011 23:40:46 +0000 (23:40 +0000)]
Nuke dead code. Nothing generates the VLD1d64QPseudo_UPD instruction.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142877
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Mon, 24 Oct 2011 23:26:05 +0000 (23:26 +0000)]
ARM assembly parsing and encoding for VLD1 w/ writeback.
Three entry register list variation.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142876
91177308-0d34-0410-b5e6-
96231b3b80d8
Owen Anderson [Mon, 24 Oct 2011 23:20:07 +0000 (23:20 +0000)]
More fixes and improvements to MachO relocation pretty-printing, particular for x86 and x86_64 relocations with addends.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142875
91177308-0d34-0410-b5e6-
96231b3b80d8
Eli Friedman [Mon, 24 Oct 2011 23:08:52 +0000 (23:08 +0000)]
Don't crash on variable insertelement on ARM. PR10258.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142871
91177308-0d34-0410-b5e6-
96231b3b80d8
Bill Wendling [Mon, 24 Oct 2011 23:05:43 +0000 (23:05 +0000)]
Check the visibility of the global variable before placing it into the stubs
table. A hidden variable could potentially end up in both lists.
<rdar://problem/
10336715>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142869
91177308-0d34-0410-b5e6-
96231b3b80d8
Evan Cheng [Mon, 24 Oct 2011 23:01:03 +0000 (23:01 +0000)]
ARMConstantPoolMBB::print should print BB number.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142867
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Mon, 24 Oct 2011 22:16:58 +0000 (22:16 +0000)]
ARM assembly parsing and encoding for VLD1 w/ writeback.
One and two length register list variants.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142861
91177308-0d34-0410-b5e6-
96231b3b80d8
Chad Rosier [Mon, 24 Oct 2011 21:56:50 +0000 (21:56 +0000)]
Add options to enable each individual level for the show-diagnostics tool.
rdar://
9683410
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142856
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Mon, 24 Oct 2011 21:45:13 +0000 (21:45 +0000)]
ARM refactor am6offset usage for VLD1.
Split am6offset into fixed and register offset variants so the instruction
encodings are explicit rather than relying an a magic reg0 marker.
Needed to being able to parse these.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142853
91177308-0d34-0410-b5e6-
96231b3b80d8
Owen Anderson [Mon, 24 Oct 2011 21:44:00 +0000 (21:44 +0000)]
Get relocation parsing/dumping to a mostly-working state for MachO files.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142852
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Mon, 24 Oct 2011 21:02:38 +0000 (21:02 +0000)]
Now that we look at all the header PHIs, we need to consider all the header PHIs
when deciding that the loop has stopped evolving. Fixes miscompile in the gcc
torture testsuite!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142843
91177308-0d34-0410-b5e6-
96231b3b80d8
Eli Friedman [Mon, 24 Oct 2011 20:24:21 +0000 (20:24 +0000)]
Add support to the old JIT for acquire/release loads and stores on x86. PR11207.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142841
91177308-0d34-0410-b5e6-
96231b3b80d8
Owen Anderson [Mon, 24 Oct 2011 20:19:18 +0000 (20:19 +0000)]
Stub out some of the MachO relocation decoding hooks.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142840
91177308-0d34-0410-b5e6-
96231b3b80d8
Douglas Gregor [Mon, 24 Oct 2011 18:10:52 +0000 (18:10 +0000)]
Really unbreak CMake build
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142822
91177308-0d34-0410-b5e6-
96231b3b80d8
Douglas Gregor [Mon, 24 Oct 2011 18:09:23 +0000 (18:09 +0000)]
Unbreak CMake build
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142821
91177308-0d34-0410-b5e6-
96231b3b80d8
Owen Anderson [Mon, 24 Oct 2011 18:04:29 +0000 (18:04 +0000)]
Fix a NEON disassembly case that was broken in the recent refactorings. As more of this code gets refactored, a lot of these manual decoding hooks should get smaller and/or go away entirely.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142817
91177308-0d34-0410-b5e6-
96231b3b80d8
Dan Gohman [Mon, 24 Oct 2011 18:01:06 +0000 (18:01 +0000)]
Delete the top-down "Latency" scheduler. Top-down scheduling doesn't handle
physreg dependencies, and upcoming codegen changes will require proper
physreg dependence handling.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142816
91177308-0d34-0410-b5e6-
96231b3b80d8
Dan Gohman [Mon, 24 Oct 2011 17:56:48 +0000 (17:56 +0000)]
Delete the Latency scheduling preference.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142815
91177308-0d34-0410-b5e6-
96231b3b80d8
Dan Gohman [Mon, 24 Oct 2011 17:55:11 +0000 (17:55 +0000)]
Change this overloaded use of Sched::Latency to be an overloaded
use of Sched::ILP instead, as Sched::Latency is going away.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142813
91177308-0d34-0410-b5e6-
96231b3b80d8
Dan Gohman [Mon, 24 Oct 2011 17:53:16 +0000 (17:53 +0000)]
Remove the explicit request for "Latency" scheduling from MSP430,
as the Latency scheduler is going away.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142811
91177308-0d34-0410-b5e6-
96231b3b80d8
Dan Gohman [Mon, 24 Oct 2011 17:45:02 +0000 (17:45 +0000)]
Change the default scheduler from Latency to ILP, since Latency
is going away.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142810
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Mon, 24 Oct 2011 17:26:26 +0000 (17:26 +0000)]
Update test for r142801.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142806
91177308-0d34-0410-b5e6-
96231b3b80d8
Benjamin Kramer [Mon, 24 Oct 2011 17:24:05 +0000 (17:24 +0000)]
XFAIL test on leak checkers.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142804
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Mon, 24 Oct 2011 17:16:24 +0000 (17:16 +0000)]
Thumb2 LDM instructions can target PC. Make sure to encode it.
PR11220
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142801
91177308-0d34-0410-b5e6-
96231b3b80d8
Bill Wendling [Mon, 24 Oct 2011 17:12:36 +0000 (17:12 +0000)]
Cleanup. Get rid of the old SjLj EH lowering code. No functionality change.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142800
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Mon, 24 Oct 2011 16:51:55 +0000 (16:51 +0000)]
Sink an otherwise unused variable's initializer into the asserts that
used it. Fixes an unused variable warning from GCC on release builds.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142799
91177308-0d34-0410-b5e6-
96231b3b80d8
Benjamin Kramer [Mon, 24 Oct 2011 13:50:56 +0000 (13:50 +0000)]
Implement comparison operators for BranchProbability in a way that can't overflow INT64_MAX.
Add a test case for the edge case that triggers this. Thanks to Chandler for bringing this to my attention.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142794
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Mon, 24 Oct 2011 12:01:08 +0000 (12:01 +0000)]
Remove return heuristics from the static branch probabilities, and
introduce no-return or unreachable heuristics.
The return heuristics from the Ball and Larus paper don't work well in
practice as they pessimize early return paths. The only good hitrate
return heuristics are those for:
- NULL return
- Constant return
- negative integer return
Only the last of these three can possibly require significant code for
the returning block, and even the last is fairly rare and usually also
a constant. As a consequence, even for the cold return paths, there is
little code on that return path, and so little code density to be gained
by sinking it. The places where sinking these blocks is valuable (inner
loops) will already be weighted appropriately as the edge is a loop-exit
branch.
All of this aside, early returns are nearly as common as all three of
these return categories, and should actually be predicted as taken!
Rather than muddy the waters of the static predictions, just remain
silent on returns and let the CFG itself dictate any layout or other
issues.
However, the return heuristic was flagging one very important case:
unreachable. Unfortunately it still gave a 1/4 chance of the
branch-to-unreachable occuring. It also didn't do a rigorous job of
finding those blocks which post-dominate an unreachable block.
This patch builds a more powerful analysis that should flag all branches
to blocks known to then reach unreachable. It also has better worst-case
runtime complexity by not looping through successors for each block. The
previous code would perform an N^2 walk in the event of a single entry
block branching to N successors with a switch where each successor falls
through to the next and they finally fall through to a return.
Test case added for noreturn heuristics. Also doxygen comments improved
along the way.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142793
91177308-0d34-0410-b5e6-
96231b3b80d8
NAKAMURA Takumi [Mon, 24 Oct 2011 10:03:25 +0000 (10:03 +0000)]
Revert "Test commit"
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142792
91177308-0d34-0410-b5e6-
96231b3b80d8
NAKAMURA Takumi [Mon, 24 Oct 2011 10:02:59 +0000 (10:02 +0000)]
Test commit
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142791
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Mon, 24 Oct 2011 06:57:05 +0000 (06:57 +0000)]
Reapply r142781 with fix. Original message:
Enhance SCEV's brute force loop analysis to handle multiple PHI nodes in the
loop header when computing the trip count.
With this, we now constant evaluate:
struct ListNode { const struct ListNode *next; int i; };
static const struct ListNode node1 = {0, 1};
static const struct ListNode node2 = {&node1, 2};
static const struct ListNode node3 = {&node2, 3};
int test() {
int sum = 0;
for (const struct ListNode *n = &node3; n != 0; n = n->next)
sum += n->i;
return sum;
}
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142790
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Mon, 24 Oct 2011 05:55:58 +0000 (05:55 +0000)]
Doxygen-ify the comments on the public interface for BPI. Also, move the
two more subtle routines to the bottom and expand on their cautionary
comments a bit. No functionality or actual interface change here.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142789
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Mon, 24 Oct 2011 05:51:01 +0000 (05:51 +0000)]
PHI nodes not in the loop header aren't part of the loop iteration initial
state. Furthermore, they might not have two operands. This fixes the underlying
issue behind the crashes introduced in r142781.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142788
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Mon, 24 Oct 2011 04:35:36 +0000 (04:35 +0000)]
A dead malloc, a free(NULL) and a free(undef) are all trivially dead
instructions.
This doesn't introduce any optimizations we weren't doing before (except
potentially due to pass ordering issues), now passes will eliminate them sooner
as part of their own cleanups.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142787
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Mon, 24 Oct 2011 04:00:25 +0000 (04:00 +0000)]
Speculatively revert r142781. Bots are showing
Assertion `i_nocapture < OperandTraits<PHINode>::operands(this) && "getOperand() out of range!"' failed.
coming out of indvars.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142786
91177308-0d34-0410-b5e6-
96231b3b80d8
NAKAMURA Takumi [Mon, 24 Oct 2011 03:27:19 +0000 (03:27 +0000)]
Windows/Path.inc: [PR8460] Get rid of ScopedNullTerminator. Thanks to Zvi Rackover!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142785
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Mon, 24 Oct 2011 01:40:45 +0000 (01:40 +0000)]
Simplify the design of BranchProbabilityInfo by collapsing it into
a single class. Previously it was split between two classes, one
internal and one external. The concern seemed to center around exposing
the weights used, but those can remain confined to the implementation
file.
Having a single class to maintain the state and analyses in use will
also simplify several of the enhancements I want to make to our static
heuristics.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142783
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Sun, 23 Oct 2011 23:43:14 +0000 (23:43 +0000)]
Enhance SCEV's brute force loop analysis to handle multiple PHI nodes in the
loop header when computing the trip count.
With this, we now constant evaluate:
struct ListNode { const struct ListNode *next; int i; };
static const struct ListNode node1 = {0, 1};
static const struct ListNode node2 = {&node1, 2};
static const struct ListNode node3 = {&node2, 3};
int test() {
int sum = 0;
for (const struct ListNode *n = &node3; n != 0; n = n->next)
sum += n->i;
return sum;
}
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142781
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Sun, 23 Oct 2011 22:40:13 +0000 (22:40 +0000)]
Tidy up a loop to be more idiomatic for LLVM's codebase, and remove some
extraneous whitespace. Trying to clean-up this pass as much as I can
before I start making functional changes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142780
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Sun, 23 Oct 2011 22:18:24 +0000 (22:18 +0000)]
Add X86 SARX, SHRX, and SHLX instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142779
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Sun, 23 Oct 2011 21:21:50 +0000 (21:21 +0000)]
Teach the BranchProbabilityInfo pass to print its results, and use that
to bring it under direct test instead of merely indirectly testing it in
the BlockFrequencyInfo pass.
The next step is to start adding tests for the various heuristics
employed, and to start fixing those heuristics once they're under test.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142778
91177308-0d34-0410-b5e6-
96231b3b80d8
Bill Wendling [Sun, 23 Oct 2011 20:14:06 +0000 (20:14 +0000)]
Rename the script to indicate that this is for the TEST=simple tests.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142764
91177308-0d34-0410-b5e6-
96231b3b80d8
Bill Wendling [Sun, 23 Oct 2011 20:13:14 +0000 (20:13 +0000)]
Resurrect the 'find regressions for the TEST=nightly tests' script.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142763
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Sun, 23 Oct 2011 20:10:34 +0000 (20:10 +0000)]
Now that we have comparison on probabilities, add some static functions
to get important constant branch probabilities and use them for finding
the best branch out of a set of possibilities.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142762
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Sun, 23 Oct 2011 20:10:30 +0000 (20:10 +0000)]
Remove a commented out line of code that snuck by my auditing.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142761
91177308-0d34-0410-b5e6-
96231b3b80d8
Benjamin Kramer [Sun, 23 Oct 2011 11:32:54 +0000 (11:32 +0000)]
Print branch probabilities as percentages.
50% is much more readable than 5.
000000e-01.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142752
91177308-0d34-0410-b5e6-
96231b3b80d8
Benjamin Kramer [Sun, 23 Oct 2011 11:19:14 +0000 (11:19 +0000)]
Add compare operators to BranchProbability and use it to determine if an edge is hot.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142751
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Sun, 23 Oct 2011 09:18:45 +0000 (09:18 +0000)]
Completely re-write the algorithm behind MachineBlockPlacement based on
discussions with Andy. Fundamentally, the previous algorithm is both
counter productive on several fronts and prioritizing things which
aren't necessarily the most important: static branch prediction.
The new algorithm uses the existing loop CFG structure information to
walk through the CFG itself to layout blocks. It coalesces adjacent
blocks within the loop where the CFG allows based on the most likely
path taken. Finally, it topologically orders the block chains that have
been formed. This allows it to choose a (mostly) topologically valid
ordering which still priorizes fallthrough within the structural
constraints.
As a final twist in the algorithm, it does violate the CFG when it
discovers a "hot" edge, that is an edge that is more than 4x hotter than
the competing edges in the CFG. These are forcibly merged into
a fallthrough chain.
Future transformations that need te be added are rotation of loop exit
conditions to be fallthrough, and better isolation of cold block chains.
I'm also planning on adding statistics to model how well the algorithm
does at laying out blocks based on the probabilities it receives.
The old tests mostly still pass, and I have some new tests to add, but
the nested loops are still behaving very strangely. This almost seems
like working-as-intended as it rotated the exit branch to be
fallthrough, but I'm not convinced this is actually the best layout. It
is well supported by the probabilities for loops we currently get, but
those are pretty broken for nested loops, so this may change later.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142743
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Sun, 23 Oct 2011 07:34:00 +0000 (07:34 +0000)]
Add X86 RORX instruction
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142741
91177308-0d34-0410-b5e6-
96231b3b80d8
Cameron Zwarich [Sun, 23 Oct 2011 07:02:10 +0000 (07:02 +0000)]
The element insertion code in scalar replacement doesn't handle incorrect
element types, even though the element extraction code does. It is surprising
that this bug has been here for so long. Fixes <rdar://problem/
10318778>.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142740
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Sun, 23 Oct 2011 00:33:32 +0000 (00:33 +0000)]
Add X86 MULX instruction for disassembler.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142738
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Sat, 22 Oct 2011 23:13:53 +0000 (23:13 +0000)]
Remove some duplicate specifying of neverHasSideEffects and mayLoad from X86 multiply instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142737
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Sat, 22 Oct 2011 22:07:31 +0000 (22:07 +0000)]
Oops! Fix test I forgot to submit as part of r142735.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142736
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Sat, 22 Oct 2011 21:59:35 +0000 (21:59 +0000)]
A non-escaping malloc in the entry block is not unlike an alloca. Do dead-store
elimination on them too.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142735
91177308-0d34-0410-b5e6-
96231b3b80d8
Nick Lewycky [Sat, 22 Oct 2011 19:58:20 +0000 (19:58 +0000)]
Make SCEV's brute force analysis stronger in two ways. Firstly, we should be
able to constant fold load instructions where the argument is a constant.
Second, we should be able to watch multiple PHI nodes through the loop; this
patch only supports PHIs in loop headers, more can be done here.
With this patch, we now constant evaluate:
static const int arr[] = {1, 2, 3, 4, 5};
int test() {
int sum = 0;
for (int i = 0; i < 5; ++i) sum += arr[i];
return sum;
}
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142731
91177308-0d34-0410-b5e6-
96231b3b80d8
Nadav Rotem [Sat, 22 Oct 2011 18:44:51 +0000 (18:44 +0000)]
Fix a typo.w
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142729
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Sat, 22 Oct 2011 18:17:32 +0000 (18:17 +0000)]
Minor updates.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142728
91177308-0d34-0410-b5e6-
96231b3b80d8
Nadav Rotem [Sat, 22 Oct 2011 17:51:04 +0000 (17:51 +0000)]
Added my name to CREDITS.TXT
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142727
91177308-0d34-0410-b5e6-
96231b3b80d8
Benjamin Kramer [Sat, 22 Oct 2011 16:50:00 +0000 (16:50 +0000)]
Move various generated tables into read-only memory, fixing up const correctness along the way.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142726
91177308-0d34-0410-b5e6-
96231b3b80d8
Nadav Rotem [Sat, 22 Oct 2011 12:39:25 +0000 (12:39 +0000)]
Fix pr11193.
SHL inserts zeros from the right, thus even when the original
sign_extend_inreg value was of 1-bit, we need to sra.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142724
91177308-0d34-0410-b5e6-
96231b3b80d8
Bill Wendling [Sat, 22 Oct 2011 00:29:28 +0000 (00:29 +0000)]
The different flavors of ARM have different valid subsets of registers. Check
that the set of callee-saved registers is correct for the specific platform.
<rdar://problem/
10313708> & ctor_dtor_count & ctor_dtor_count-2
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142706
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Fri, 21 Oct 2011 23:58:57 +0000 (23:58 +0000)]
Assembly parsing for 4-register sequential variant of VLD2.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142704
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Fri, 21 Oct 2011 22:21:10 +0000 (22:21 +0000)]
Assembly parsing for 2-register sequential variant of VLD2.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142691
91177308-0d34-0410-b5e6-
96231b3b80d8
Bill Wendling [Fri, 21 Oct 2011 22:08:56 +0000 (22:08 +0000)]
Make sure that the landing pads themselves have no PHI instructions in them.
The assumption in the back-end is that PHIs are not allowed at the start of the
landing pad block for SjLj exceptions.
<rdar://problem/
10313708>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142689
91177308-0d34-0410-b5e6-
96231b3b80d8
Benjamin Kramer [Fri, 21 Oct 2011 21:13:47 +0000 (21:13 +0000)]
Extend the floating point heuristic to consider NaN checks unlikely.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142687
91177308-0d34-0410-b5e6-
96231b3b80d8
Tanya Lattner [Fri, 21 Oct 2011 20:51:54 +0000 (20:51 +0000)]
Revert r141657 for now. This has broken css and changed links on llvm.org. I'd like to understand exactly why the links have changed and if a newer doxygen is required. This may be reapplied once we upgrade on llvm.org and it is fully tested.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142686
91177308-0d34-0410-b5e6-
96231b3b80d8
Eli Friedman [Fri, 21 Oct 2011 20:45:19 +0000 (20:45 +0000)]
Remap blockaddress correctly when inlining a function. Fixes PR10162.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142684
91177308-0d34-0410-b5e6-
96231b3b80d8
Owen Anderson [Fri, 21 Oct 2011 20:35:58 +0000 (20:35 +0000)]
Use LLVMBool for a function that logically returns a boolean value.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142683
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Fri, 21 Oct 2011 20:35:01 +0000 (20:35 +0000)]
Assembly parsing for 4-register variant of VLD1.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142682
91177308-0d34-0410-b5e6-
96231b3b80d8
Owen Anderson [Fri, 21 Oct 2011 20:28:19 +0000 (20:28 +0000)]
Fix typo.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142681
91177308-0d34-0410-b5e6-
96231b3b80d8
Benjamin Kramer [Fri, 21 Oct 2011 20:12:47 +0000 (20:12 +0000)]
BranchProbabilityInfo: floating point equality is unlikely.
This is from the same paper from Ball and Larus as the rest of the currently implemented heuristics.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142677
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Fri, 21 Oct 2011 20:02:19 +0000 (20:02 +0000)]
Assembly parsing for 3-register variant of VLD1.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142675
91177308-0d34-0410-b5e6-
96231b3b80d8
Owen Anderson [Fri, 21 Oct 2011 19:26:54 +0000 (19:26 +0000)]
STABS symbols are debug symbols.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142673
91177308-0d34-0410-b5e6-
96231b3b80d8
Eli Friedman [Fri, 21 Oct 2011 19:11:34 +0000 (19:11 +0000)]
Minor simplification: use ShuffleVectorInst::getMaskValue instead of a more expensive helper.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142672
91177308-0d34-0410-b5e6-
96231b3b80d8
Eli Friedman [Fri, 21 Oct 2011 19:06:29 +0000 (19:06 +0000)]
Extend instcombine's shufflevector simplification to handle more cases where the input and output vectors have different sizes. Patch by Xiaoyi Guo.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142671
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Fri, 21 Oct 2011 18:54:25 +0000 (18:54 +0000)]
ARM VLD parsing and encoding.
Next step in the ongoing saga of NEON load/store assmebly parsing. Handle
VLD1 instructions that take a two-register register list.
Adjust the instruction definitions to only have the single encoded register
as an operand. The super-register from the pseudo is kept as an implicit def,
so passes which come after pseudo-expansion still know that the instruction
defines the other subregs.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142670
91177308-0d34-0410-b5e6-
96231b3b80d8
Owen Anderson [Fri, 21 Oct 2011 18:43:28 +0000 (18:43 +0000)]
Don't automatically set the "fc" bits on MSR instructions if the user didn't ask for them. This is a divergence from gas' behavior, but it is correct per the documentation and allows us to forge ahead with roundtrip testing.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142669
91177308-0d34-0410-b5e6-
96231b3b80d8
Owen Anderson [Fri, 21 Oct 2011 18:21:22 +0000 (18:21 +0000)]
Bind libObject API for obtaining the section containing a Symbol.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142667
91177308-0d34-0410-b5e6-
96231b3b80d8
Owen Anderson [Fri, 21 Oct 2011 17:50:59 +0000 (17:50 +0000)]
Expand the coverage of the libObject C bindings to include more SectionRef accessors as well as Symbol iterators.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142661
91177308-0d34-0410-b5e6-
96231b3b80d8
Nadav Rotem [Fri, 21 Oct 2011 17:35:19 +0000 (17:35 +0000)]
Fix pr11194. When promoting and splitting integers we need to use
ZExtPromotedInteger and SExtPromotedInteger based on the operation we legalize.
SetCC return type needs to be legalized via PromoteTargetBoolean.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142660
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Fri, 21 Oct 2011 16:59:08 +0000 (16:59 +0000)]
Nuke an #if0 that got accidentally left in.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142658
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Fri, 21 Oct 2011 16:56:40 +0000 (16:56 +0000)]
whitespace.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142657
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Fri, 21 Oct 2011 16:41:39 +0000 (16:41 +0000)]
Don't hard code the desired alignment for loops -- it isn't 16-bytes on
all x86 systems. Sorry for the breakage.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142656
91177308-0d34-0410-b5e6-
96231b3b80d8
Jim Grosbach [Fri, 21 Oct 2011 16:14:12 +0000 (16:14 +0000)]
Remove some outdated comments.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142653
91177308-0d34-0410-b5e6-
96231b3b80d8
Nadav Rotem [Fri, 21 Oct 2011 11:42:07 +0000 (11:42 +0000)]
1. Fix the widening of SETCC in WidenVecOp_SETCC. Use the correct return CC type.
2. Fix a typo in CONCAT_VECTORS which exposed the bug in #1.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142648
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan Sands [Fri, 21 Oct 2011 09:47:14 +0000 (09:47 +0000)]
Ensure timestamps are not embedded into files when doing a release build.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142647
91177308-0d34-0410-b5e6-
96231b3b80d8
Anton Korobeynikov [Fri, 21 Oct 2011 09:38:50 +0000 (09:38 +0000)]
Fix build on mingw-w64.
Patch by Ruben Van Boxem!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142646
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Fri, 21 Oct 2011 08:57:37 +0000 (08:57 +0000)]
Add loop aligning to MachineBlockPlacement based on review discussion so
it's a bit more plausible to use this instead of CodePlacementOpt. The
code for this was shamelessly stolen from CodePlacementOpt, and then
trimmed down a bit. There doesn't seem to be much utility in returning
true/false from this pass as we may or may not have rewritten all of the
blocks. Also, the statistic of counting how many loops were aligned
doesn't seem terribly important so I removed it. If folks would like it
to be included, I'm happy to add it back.
This was probably the most egregious of the missing features, and now
I'm going to start gathering some performance numbers and looking at
specific loop structures that have different layout between the two.
Test is updated to include both basic loop alignment and nested loop
alignment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@142645
91177308-0d34-0410-b5e6-
96231b3b80d8