Commit Graph

85006 Commits

Author SHA1 Message Date
Chandler Carruth
ad41dcfd87 Be a bit more aggressive in bailing out of this routine. Spotted by
inspection by Duncan during review. My suspicion is that we would still
have returned 0 anyways in this case, but doing it sooner is better.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163895 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 10:30:42 +00:00
Chandler Carruth
8ed1ed8dfd Add some comments clarifying that the GEP analysis for vector GEPs is
deeply suspicious and likely to go away eventually. Also fix a bogus
comment about one of the checks in the vector GEP analysis. Based on
review from Duncan.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163894 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 10:30:40 +00:00
Chandler Carruth
8615cd236e Move an instance variable to a local variable based on review by Duncan.
Originally I had anticipated needing to thread this through more bits of
the SROA pass itself, but that ended up not happening. In the end, this
is a much simpler way to manange the variable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163893 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 10:26:38 +00:00
Chandler Carruth
50754f0892 Add a comment about debug intrinsics that I *really* don't want to
forget from Duncan's review as a FIXME.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163892 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 10:26:36 +00:00
Chandler Carruth
b3dd9a13e8 Add two asserts that Duncan thought would help ensure things don't rot
unexpectedly in the future. More fixes from his code review.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163891 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 10:26:34 +00:00
Chandler Carruth
63db8bebe6 Actually keep the flag default-off for now. =/ That's what I get for
being busy testing this...

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163890 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 10:18:54 +00:00
Chandler Carruth
1c2fe28109 Remove some dead, commented out code Duncan spotted in review.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163889 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 10:18:53 +00:00
Chandler Carruth
ba13d2efba Wrap the dumping and printing routines in NDEBUG and LLVM_ENABLE_DUMP macros.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163888 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 10:18:51 +00:00
Chandler Carruth
7f5bedec74 Lots of comment fixes and cleanups from Duncan's review.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163887 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 10:18:49 +00:00
NAKAMURA Takumi
ad9f5b8967 SROA.cpp: Unbreak gcc, sorry!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163886 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 10:06:10 +00:00
NAKAMURA Takumi
9b1e62454a SROA.cpp: Appease msvc. LLVM_ATTRIBUTE(s) should come front of "const".
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163885 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 09:55:22 +00:00
Chandler Carruth
2a9bf25f60 Speculative change to try to fix older GCC versions that can't handle
the injected class name of a dependent base class here.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163884 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 09:30:33 +00:00
Chandler Carruth
713aa9431d Introduce a new SROA implementation.
This is essentially a ground up re-think of the SROA pass in LLVM. It
was initially inspired by a few problems with the existing pass:
- It is subject to the bane of my existence in optimizations: arbitrary
  thresholds.
- It is overly conservative about which constructs can be split and
  promoted.
- The vector value replacement aspect is separated from the splitting
  logic, missing many opportunities where splitting and vector value
  formation can work together.
- The splitting is entirely based around the underlying type of the
  alloca, despite this type often having little to do with the reality
  of how that memory is used. This is especially prevelant with unions
  and base classes where we tail-pack derived members.
- When splitting fails (often due to the thresholds), the vector value
  replacement (again because it is separate) can kick in for
  preposterous cases where we simply should have split the value. This
  results in forming i1024 and i2048 integer "bit vectors" that
  tremendously slow down subsequnet IR optimizations (due to large
  APInts) and impede the backend's lowering.

The new design takes an approach that fundamentally is not susceptible
to many of these problems. It is the result of a discusison between
myself and Duncan Sands over IRC about how to premptively avoid these
types of problems and how to do SROA in a more principled way. Since
then, it has evolved and grown, but this remains an important aspect: it
fixes real world problems with the SROA process today.

First, the transform of SROA actually has little to do with replacement.
It has more to do with splitting. The goal is to take an aggregate
alloca and form a composition of scalar allocas which can replace it and
will be most suitable to the eventual replacement by scalar SSA values.
The actual replacement is performed by mem2reg (and in the future
SSAUpdater).

The splitting is divided into four phases. The first phase is an
analysis of the uses of the alloca. This phase recursively walks uses,
building up a dense datastructure representing the ranges of the
alloca's memory actually used and checking for uses which inhibit any
aspects of the transform such as the escape of a pointer.

Once we have a mapping of the ranges of the alloca used by individual
operations, we compute a partitioning of the used ranges. Some uses are
inherently splittable (such as memcpy and memset), while scalar uses are
not splittable. The goal is to build a partitioning that has the minimum
number of splits while placing each unsplittable use in its own
partition. Overlapping unsplittable uses belong to the same partition.
This is the target split of the aggregate alloca, and it maximizes the
number of scalar accesses which become accesses to their own alloca and
candidates for promotion.

Third, we re-walk the uses of the alloca and assign each specific memory
access to all the partitions touched so that we have dense use-lists for
each partition.

Finally, we build a new, smaller alloca for each partition and rewrite
each use of that partition to use the new alloca. During this phase the
pass will also work very hard to transform uses of an alloca into a form
suitable for promotion, including forming vector operations, speculating
loads throguh PHI nodes and selects, etc.

After splitting is complete, each newly refined alloca that is
a candidate for promotion to a scalar SSA value is run through mem2reg.

There are lots of reasonably detailed comments in the source code about
the design and algorithms, and I'm going to be trying to improve them in
subsequent commits to ensure this is well documented, as the new pass is
in many ways more complex than the old one.

Some of this is still a WIP, but the current state is reasonbly stable.
It has passed bootstrap, the nightly test suite, and Duncan has run it
successfully through the ACATS and DragonEgg test suites. That said, it
remains behind a default-off flag until the last few pieces are in
place, and full testing can be done.

Specific areas I'm looking at next:
- Improved comments and some code cleanup from reviews.
- SSAUpdater and enabling this pass inside the CGSCC pass manager.
- Some datastructure tuning and compile-time measurements.
- More aggressive FCA splitting and vector formation.

Many thanks to Duncan Sands for the thorough final review, as well as
Benjamin Kramer for lots of review during the process of writing this
pass, and Daniel Berlin for reviewing the data structures and algorithms
and general theory of the pass. Also, several other people on IRC, over
lunch tables, etc for lots of feedback and advice.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163883 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 09:22:59 +00:00
Duncan Sands
74b4762234 Remove silly dead store. Patch by Ettl Martin.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163882 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 09:00:11 +00:00
Craig Topper
f4d78241ea Allow the second opcode info table to be 8, 16, or 32-bits as needed to represent additional fragments. This recovers some space on ATT X86 syntax and PowerPC which only need 40-bits instead of 48-bits. This also increases ARM to 64-bits to fully encode all of its operands.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163880 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 08:33:11 +00:00
Craig Topper
4e5babe4f3 Reduce size of register name index tables by using uint16_t for all in tree targets. If more than 16-bits are needed for any out of tree targets, code will detect and use uint32_t instead.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163878 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 06:37:49 +00:00
Andrew Trick
a2ed0e8e8b misched: Generic tablegen classes for the new machine model.
This is mostly documentation for the new machine model. It is designed
to be flexible, easy to incrementally refine for a subtarget, and
provide all the information that MachineScheduler will need.

If all goes well, I will follow up with an example of the new model in
use for ARM.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163877 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 06:18:55 +00:00
Andrew Trick
1008f8c789 comment
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163876 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 06:18:52 +00:00
Andrew Trick
8fa5319504 comment
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163875 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 06:18:50 +00:00
Akira Hatanaka
99258f6755 mips16 fixes.
1. Add MoveR3216
2. Correct spelling for Move32R16

Patch by Reed Kotler.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163869 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-14 03:21:56 +00:00
Galina Kistanova
e30ba2ebcf Patch by Sean Silva!
The patch converts the "How to add a builder" document over to reStructuredText..


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163860 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 23:51:08 +00:00
Eric Christopher
ffaf69b8b1 Fix both the test for zero and what we do if we have a zero for
umulo legalization.

Fixes PR13839

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163856 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 23:24:02 +00:00
Eric Christopher
331f64cd0d Reformat, remove a couple unused variables and move some variables
closer to where they're needed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163855 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 23:23:58 +00:00
Jim Grosbach
3f90a4c42d Assembler: Darwin variables defined via .set are no-dead-strip.
For gas compatibility.

rdar://12219394

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163854 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 23:11:31 +00:00
Jim Grosbach
45d81bdde8 MachO: Correctly mark symbol-difference variables as N_ABS.
.set a, b - c + CONSTANT
d = b - c + CONSTANT

Both 'a' and 'd' should be marked as absolute symbols (N_ABS).

rdar://12219394

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163853 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 23:11:25 +00:00
Dan Gohman
b998913ff4 Handle the new !tbaa.struct metadata tags when converting a memcpy into scalar
loads and stores.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163844 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 21:51:01 +00:00
Jim Grosbach
e940c1bb6c Better const handling for RuntimeDyld and MCJIT.
mapSectionAddress() wasn't consistent.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163843 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 21:50:06 +00:00
Richard Smith
0067567927 Fix some code which is invalid in C++11: an expression of enumeration type
can't be used as a non-type template argument of type bool.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163840 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 21:18:18 +00:00
Manman Ren
785a41d138 AsmWriterEmitter: OpInfo2 should be unsigned 16-bit.
Fix an issue in r163814.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163837 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 20:47:48 +00:00
Michael Liao
9aba7ea472 Fix comment
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163835 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 20:30:16 +00:00
Dmitri Gribenko
065c0f800c Fix documentation: parameter being documented was removed in r98220.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163834 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 20:28:31 +00:00
Michael Liao
f966e4e5b3 Add wider vector/integer support for PR12312
- Enhance the fix to PR12312 to support wider integer, such as 256-bit
  integer. If more than 1 fully evaluated vectors are found, POR them
  first followed by the final PTEST.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163832 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 20:24:54 +00:00
Michael Liao
092122f124 Enhance type legalization on bitcast from vector to integer
- Find a legal vector type before casting and extracting element from it.
- As the new vector type may have more than 2 elements, build the final
  hi/lo pair by BFS pairing them from bottom to top.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163830 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 19:58:21 +00:00
Jakob Stoklund Olesen
da0e8219b7 Fix test case to avoid PIC magic.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163827 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 19:47:45 +00:00
Jakob Stoklund Olesen
7bba7d0efc Fix the TCRETURNmi64 bug differently.
Add a PatFrag to match X86tcret using 6 fixed registers or less. This
avoids folding loads into TCRETURNmi64 using 7 or more volatile
registers.

<rdar://problem/12282281>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163819 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 18:31:27 +00:00
Dan Gohman
ce52bc5353 Extract code for reducing a type to a single value type into a helper function.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163817 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 18:19:06 +00:00
Dan Gohman
b54834b7e6 Define an official slot for the new !tbaa.struct metadata tag.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163815 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 17:56:17 +00:00
Manman Ren
6579cf8df6 AsmWriterEmitter: increase the number of bits for OpcodeInfo from 32-bit to
48-bit if necessary, in order to reduce the generated code size.

We have 900 cases not covered by OpcodeInfo in ATT AsmWriter and more in Intel
AsmWriter and ARM AsmWriter.

This patch reduced the clang Release build size by 50k, running on a Mac Pro.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163814 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 17:43:46 +00:00
Akira Hatanaka
e6ac7d69d3 mips16: When copying operands in a conditional branch instruction, allow for
immediate operands to be copied.

Patch by Reed Kotler.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163811 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 17:12:37 +00:00
Jakob Stoklund Olesen
0767dc546e Revert r163761 "Don't fold indexed loads into TCRETURNmi64."
The patch caused "Wrong topological sorting" assertions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163810 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 16:52:17 +00:00
Benjamin Kramer
39acdb0200 MemCpyOpt: When forming a memset from stores also take GEP constexprs into account.
This is common when storing to global variables.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163809 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 16:29:49 +00:00
Nadav Rotem
aa8405811e Fix an 80 char line limit.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163808 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 16:27:32 +00:00
Nadav Rotem
18e2c29094 Rename the flag which protects from escaped allocas, which may come from bugs in user code or in the compiler. Also, dont assert if the protection is not enabled.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163807 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 15:46:30 +00:00
Micah Villmow
af40a5be77 The current implementation does not allow more than 32 types to be properly handled with target lowering. This doubles the size to 64bit types and easily allows extension to more types.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163806 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 15:24:43 +00:00
Micah Villmow
d3766dfd2c Unify the emission of the calling conventions into a single function to reduce code duplication.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163805 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 15:11:12 +00:00
Silviu Baranga
616471d4bf This patch introduces A15 as a target in LLVM.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163803 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 15:05:10 +00:00
Nadav Rotem
91a7e0184a Fix a dagcombine optimization. The optimization attempts to optimize a bitcast of fneg to integers
by xoring the high-bit. This fails if the source operand is a vector because we need to negate
each of the elements in the vector.

Fix rdar://12281066 PR13813.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163802 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 14:54:28 +00:00
Nadav Rotem
faf31d01db Fix a typo.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163801 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 14:51:00 +00:00
Bill Wendling
032dbee2a9 Use Nick's suggestion of storing a large NULL into the GV instead of memset, which requires TargetData.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163799 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 14:32:30 +00:00
Nadav Rotem
0cd19b9301 Stack Coloring: We have code that checks that all of the uses of allocas
are within the lifetime zone. Sometime legitimate usages of allocas are
hoisted outside of the lifetime zone. For example, GEPS may calculate the
address of a member of an allocated struct. This commit makes sure that
we only check (abort regions or assert) for instructions that read and write
memory using stack frames directly. Notice that by allowing legitimate
usages outside the lifetime zone we also stop checking for instructions
which use derivatives of allocas. We will catch less bugs in user code
and in the compiler itself.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163791 91177308-0d34-0410-b5e6-96231b3b80d8
2012-09-13 12:38:37 +00:00