Commit Graph

15218 Commits

Author SHA1 Message Date
Andrew Trick 9c17eab761 MI Sched: Track live-thru registers.
When registers must be live throughout the scheduling region, increase
the limit for the register class. Once we exceed the original limit,
they will be spilled, and there's no point further reducing pressure.

This isn't a perfect heuristics but avoids a situation where the
scheduler could become trapped by trying to achieve the impossible.

llvm-svn: 187436
2013-07-30 19:59:12 +00:00
Andrew Trick d9761776bc MI Sched fix: assert "Disconnected LRG within the scheduling region."
llvm-svn: 187435
2013-07-30 19:59:08 +00:00
Quentin Colombet 6bf4baa408 [DAGCombiner] insert_vector_elt: Avoid building a vector twice.
This patch prevents the following combine when the input vector is used more
than once.
insert_vector_elt (build_vector elt0, ..., eltN), NewEltIdx, idx
=>
build_vector elt0, ..., NewEltIdx, ..., eltN 

The reasons are:
- Building a vector may be expensive, so try to reuse the existing part of a
  vector instead of creating a new one (think big vectors).
- elt0 to eltN now have two users instead of one. This may prevent some other
  optimizations.

llvm-svn: 187396
2013-07-30 00:24:09 +00:00
Eric Christopher e414ece79a Fix a truly egregious thinko in anonymous namespace check,
update testcase to make sure we generate debug info for walrus
by adding a non-trivial constructor and verify that we don't
emit an ODR signature for the type.

llvm-svn: 187393
2013-07-29 23:53:08 +00:00
Eric Christopher d853ea3142 Make sure we don't emit an ODR hash for types with no name and make
sure the comments for each testcase are a bit easier to distinguish.

llvm-svn: 187392
2013-07-29 23:53:05 +00:00
Eric Christopher f8542ec305 Elaborate a bit on the type unit and ODR conditional code.
llvm-svn: 187385
2013-07-29 22:24:32 +00:00
Nico Rieck 7fdaee8f15 Use proper section suffix for COFF weak symbols
32-bit symbols have "_" as global prefix, but when forming the name of
COMDAT sections this prefix is ignored. The current behavior assumes that
this prefix is always present which is not the case for 64-bit and names
are truncated.

llvm-svn: 187356
2013-07-29 13:58:39 +00:00
Benjamin Kramer 409afcf174 DwarfDebug: MD5 is always little endian, bswap on big endian platforms.
This makes LLVM emit the same signature regardless of host and target endianess.

llvm-svn: 187304
2013-07-27 14:14:43 +00:00
Chandler Carruth 2a1c0d2c03 Fix a memory leak in the debug emission by simply not allocating memory.
There doesn't appear to be any reason to put this variable on the heap.
I'm suspicious of the LexicalScope above that we stuff in a map and then
delete afterward, but I'm just trying to get the valgrind bot clean.

llvm-svn: 187301
2013-07-27 11:09:58 +00:00
Nick Lewycky 0b68245ec8 Reimplement isPotentiallyReachable to make nocapture deduction much stronger.
Adds unit tests for it too.

Split BasicBlockUtils into an analysis-half and a transforms-half, and put the
analysis bits into a new Analysis/CFG.{h,cpp}. Promote isPotentiallyReachable
into llvm::isPotentiallyReachable and move it into Analysis/CFG.

llvm-svn: 187283
2013-07-27 01:24:00 +00:00
Tom Stellard 8b1e021e85 SimplifyCFG: Use parallel-and and parallel-or mode to consolidate branch conditions
Merge consecutive if-regions if they contain identical statements.
Both transformations reduce number of branches.  The transformation
is guarded by a target-hook, and is currently enabled only for +R600,
but the correctness has been tested on X86 target using a variety of
CPU benchmarks.

Patch by: Mei Ye

llvm-svn: 187278
2013-07-27 00:01:07 +00:00
Eric Christopher 219fb91499 Remove addLetterToHash, no functional change.
llvm-svn: 187245
2013-07-26 21:07:18 +00:00
Eric Christopher 67646438c9 Add preliminary support for hashing DIEs and breaking them into
type units.

Initially this support is used in the computation of an ODR checker
for C++. For now we're attaching it to the DIE, but in the future
it will be attached to the type unit.

This also starts breaking out types into the separation for type
units, but without actually splitting the DIEs.

In preparation for hashing the DIEs this adds a DIEString type
that contains a StringRef with the string contained at the label.

llvm-svn: 187213
2013-07-26 17:02:41 +00:00
Justin Holewinski d3f2035a3c Add a target legalize hook for SplitVectorOperand (again)
CustomLowerNode was not being called during SplitVectorOperand,
meaning custom legalization could not be used by targets.

This also adds a test case for NVPTX that depends on this custom
legalization.

Differential Revision: http://llvm-reviews.chandlerc.com/D1195

Attempt to fix the buildbots by making the X86 test I just added platform independent

llvm-svn: 187202
2013-07-26 13:28:29 +00:00
Rafael Espindola 1d812728cc Revert "Add a target legalize hook for SplitVectorOperand"
This reverts commit 187198. It broke the bots.

The soft float test probably needs a -triple because of name differences.
On the hard float test I am getting a "roundss $1, %xmm0, %xmm0", instead of
"vroundss $1, %xmm0, %xmm0, %xmm0".

llvm-svn: 187201
2013-07-26 13:18:16 +00:00
Justin Holewinski f848a24e50 Add a target legalize hook for SplitVectorOperand
CustomLowerNode was not being called during SplitVectorOperand,
meaning custom legalization could not be used by targets.

This also adds a test case for NVPTX that depends on this custom
legalization.

Differential Revision: http://llvm-reviews.chandlerc.com/D1195

llvm-svn: 187198
2013-07-26 12:46:39 +00:00
Andrew Trick f4b1ee3492 RegAllocGreedy comment.
llvm-svn: 187141
2013-07-25 18:35:22 +00:00
Andrew Trick 8bb0a251fd Evict local live ranges if they can be reassigned.
The previous change to local live range allocation also suppressed
eviction of local ranges. In rare cases, this could result in more
expensive register choices. This commit actually revives a feature
that I added long ago: check if live ranges can be reassigned before
eviction. But now it only happens in rare cases of evicting a local
live range because another local live range wants a cheaper register.

The benefit is improved code size for some benchmarks on x86 and armv7.

I measured no significant compile time increase and performance
changes are noise.

llvm-svn: 187140
2013-07-25 18:35:19 +00:00
Andrew Trick 8485257d6d Allocate local registers in order for optimal coloring.
Also avoid locals evicting locals just because they want a cheaper register.

Problem: MI Sched knows exactly how many registers we have and assumes
they can be colored. In cases where we have large blocks, usually from
unrolled loops, greedy coloring fails. This is a source of
"regressions" from the MI Scheduler on x86. I noticed this issue on
x86 where we have long chains of two-address defs in the same live
range. It's easy to see this in matrix multiplication benchmarks like
IRSmk and even the unit test misched-matmul.ll.

A fundamental difference between the LLVM register allocator and
conventional graph coloring is that in our model a live range can't
discover its neighbors, it can only verify its neighbors. That's why
we initially went for greedy coloring and added eviction to deal with
the hard cases. However, for singly defined and two-address live
ranges, we can optimally color without visiting neighbors simply by
processing the live ranges in instruction order.

Other beneficial side effects:

It is much easier to understand and debug regalloc for large blocks
when the live ranges are allocated in order. Yes, global allocation is
still very confusing, but it's nice to be able to comprehend what
happened locally.

Heuristics could be added to bias register assignment based on
instruction locality (think late register pairing, banks...).

Intuituvely this will make some test cases that are on the threshold
of register pressure more stable.

llvm-svn: 187139
2013-07-25 18:35:14 +00:00
Adrian Prantl e4daf52a63 typo.
llvm-svn: 187135
2013-07-25 17:52:30 +00:00
Andrew Trick 401b6959ae MI Sched: Register pressure heuristics.
Consider which set is being increased or decreased before comparing.

llvm-svn: 187110
2013-07-25 07:26:35 +00:00
Andrew Trick 27e5fea665 MI Sched: track register pressure by importance of the set, not weight of the units.
llvm-svn: 187109
2013-07-25 07:26:32 +00:00
Andrew Trick 9706496b0d Dump LIS before regalloc. MI sched changes them.
llvm-svn: 187107
2013-07-25 07:26:26 +00:00
Bill Wendling 440e9d81bf Replace the "NoFramePointerElimNonLeaf" target option with a function attribute.
There's no need to specify a flag to omit frame pointer elimination on non-leaf
nodes...(Honestly, I can't parse that option out.) Use the function attribute
stuff instead.

llvm-svn: 187093
2013-07-25 00:34:29 +00:00
Quentin Colombet bdab227e53 Fix a bug in IfConverter with nested predicates.
Prior to this patch, IfConverter may widen the cases where a sequence of
instructions were executed because of the way it uses nested predicates. This
result in incorrect execution.

For instance, Let A be a basic block that flows conditionally into B and B be a
predicated block.
B can be predicated with A.BrToBPredicate into A iff B.Predicate is less
"permissive" than A.BrToBPredicate, i.e., iff A.BrToBPredicate subsumes
B.Predicate.

The IfConverter was checking the opposite: B.Predicate subsumes
A.BrToBPredicate.

<rdar://problem/14379453>

llvm-svn: 187071
2013-07-24 20:20:37 +00:00
Tom Stellard c54731aa9d DAGCombiner: Pass the correct type to TargetLowering::isF(Abs|Neg)Free
This commit also implements these functions for R600 and removes a test
case that was relying on the buggy behavior.

llvm-svn: 187007
2013-07-23 23:55:03 +00:00
Eric Christopher 7f2b551d4e Reformat options.
llvm-svn: 186994
2013-07-23 22:16:41 +00:00
Michael Gottesman c0659fad7f [stackprotector] Changed isNoopBitcast/sameNoopInput to take TargetLoweringBase instead of TargetLowering.
Both functions only use functionality from TargetLoweringBase.

rdar://13935163

llvm-svn: 186874
2013-07-22 21:05:47 +00:00
Michael Gottesman a6188f9fcd [stackprotector] Refactored ssp prologue creation code into its own helper function.
No functionality change.

rdar://13935163

llvm-svn: 186868
2013-07-22 20:44:11 +00:00
Bill Wendling c02a0aabb5 Recommit r186217 with testcase fix:
Use the function attributes to pass along the stack protector buffer size.

 Now that we have robust function attributes, don't use a command line option to
 specify the stack protecto buffer size.

llvm-svn: 186863
2013-07-22 20:15:21 +00:00
Eric Christopher 19d153261f Formatting.
llvm-svn: 186851
2013-07-22 18:26:15 +00:00
Michael Gottesman da6365f4ed Added missing - in the header of PrologEpilogInserter.h so that editors properly realize it is a c++ header and not a c header.
llvm-svn: 186801
2013-07-22 00:52:55 +00:00
Alexey Samsonov 64c391dbe4 Fix uninitialized memory read found by MemorySanitizer: always set output parameter of ConvergingScheduler::SchedBoundary::getOtherResourceCount
llvm-svn: 186658
2013-07-19 08:55:18 +00:00
Andrew Trick b13ef17a14 MI Sched: Update the way resources are tracked so the current heuristics make more sense.
llvm-svn: 186632
2013-07-19 00:20:07 +00:00
Michael Gottesman f87a6ae65f Add -*- C++ -*- to InstrEmitter.h.
llvm-svn: 186527
2013-07-17 18:53:29 +00:00
Craig Topper 9fdc70e846 Make constant string pointer into an array to remove a pointer lookup for every access.
llvm-svn: 186482
2013-07-17 03:11:32 +00:00
Manman Ren 8bfde8917e Add getModuleFlag(StringRef Key) to query a module flag given Key.
No functionality change.

llvm-svn: 186470
2013-07-16 23:21:16 +00:00
Rafael Espindola 6d35481c94 Add a wrapper for open.
This centralizes the handling of O_BINARY and opens the way for hiding more
differences (like how open behaves with directories).

llvm-svn: 186447
2013-07-16 19:44:17 +00:00
Jakob Stoklund Olesen efeb3a1969 Remove floats from live range splitting costs.
These floats all represented block frequencies anyway, so just use the
BlockFrequency class directly.

Some floating point computations remain in tryLocalSplit(). They are
estimating spill weights which are still floats.

llvm-svn: 186435
2013-07-16 18:26:18 +00:00
Jakob Stoklund Olesen c5454ff046 Reapply r185393.
Original commit message:

Remove floating point computations from SpillPlacement.cpp.

Patch by Benjamin Kramer!

Use the BlockFrequency class instead of floats in the Hopfield network
computations. This rescales the node Bias field from a [-2;2] float
range to two block frequencies BiasN and BiasP pulling in opposite
directions. This construct has a more predictable behavior when block
frequencies saturate.

The per-node scaling factors are no longer necessary, assuming the block
frequencies around a bundle are consistent.

This patch can cause the register allocator to make different spilling
decisions. The differences should be small.

llvm-svn: 186434
2013-07-16 18:26:15 +00:00
Craig Topper d3a34f81f8 Add 'const' qualifiers to static const char* variables.
llvm-svn: 186371
2013-07-16 01:17:10 +00:00
Manman Ren b827123cf7 PEI: Support for non-zero SPAdj at beginning of a basic block.
We can have a FrameSetup in one basic block and the matching FrameDestroy
in a different basic block when we have struct byval. In that case, SPAdj
is not zero at beginning of the basic block.

Modify PEI to correctly set SPAdj at beginning of each basic block using
DFS traversal. We used to assume SPAdj is 0 at beginning of each basic block.

PEI had an assert SPAdjCount || SPAdj == 0.
If we have a Destroy <n> followed by a Setup <m>, PEI will assert failure.
We can add an extra condition to make sure the pairs are matched:
  The pairs start with a FrameSetup.
But since we are doing a much better job in the verifier, this patch removes
the check in PEI.

PR16393

llvm-svn: 186364
2013-07-15 23:47:29 +00:00
Manman Ren aa6875b1f9 Machine Verifier: verify FrameSetup and FrameDestroy
1> on every path through the CFG, a FrameSetup <n> is always followed by a
   FrameDestroy <n> and a FrameDestroy is always followed by a FrameSetup.
2> stack adjustments are identical on all CFG edges to a merge point.
3> frame is destroyed at end of a return block.

PR16393

llvm-svn: 186350
2013-07-15 21:26:31 +00:00
Hal Finkel 2f5e8e3d95 Remove invalid assert in DAGTypeLegalizer::RemapValue
There is a comment at the top of DAGTypeLegalizer::PerformExpensiveChecks
which, in part, says:

  // Note that these invariants may not hold momentarily when processing a node:
  // the node being processed may be put in a map before being marked Processed.

Unfortunately, this assert would be valid only if the above-mentioned invariant
held unconditionally. This was causing llc to assert when, in fact,
everything was fine.

Thanks to Richard Sandiford for investigating this issue!

Fixes PR16562.

llvm-svn: 186338
2013-07-15 18:57:05 +00:00
Craig Topper 06b3b6651e Add 'const' qualifier to some arrays.
llvm-svn: 186312
2013-07-15 08:02:13 +00:00
Eric Christopher 7980b957cc Clarify comments.
llvm-svn: 186297
2013-07-14 22:23:54 +00:00
Eric Christopher 666dc635c7 Collapse temporary variable into call.
llvm-svn: 186295
2013-07-14 21:46:51 +00:00
Tobias Grosser 84f34be98e Fix build by replacing '>>' with '> >'
llvm-svn: 186276
2013-07-14 06:12:01 +00:00
Craig Topper b94011fd28 Use SmallVectorImpl& instead of SmallVector to avoid repeating small vector size.
llvm-svn: 186274
2013-07-14 04:42:23 +00:00
Craig Topper e0b711864c Pass SmallVector by const reference instead of by value.
llvm-svn: 186243
2013-07-13 07:43:40 +00:00