Commit Graph

92 Commits

Author SHA1 Message Date
Jakob Stoklund Olesen
8a3eab9b20 Remove the local register allocator.
Please use the fast allocator instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106051 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-15 21:58:33 +00:00
Jakob Stoklund Olesen
700bfada63 Add a -regalloc=default option that chooses a register allocator based on the -O
optimization level.

This only really affects llc for now because both the llvm-gcc and clang front
ends override the default register allocator. I intend to remove that code later.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104904 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-27 23:57:25 +00:00
Jakob Stoklund Olesen
1fa1c7b23b Remove ancient prototype.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104903 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-27 23:57:19 +00:00
Jakob Stoklund Olesen
00207237dd Add fast register allocator, enabled with -regalloc=fast.
So far this is just a clone of -regalloc=local that has been lobotomized to run
25% faster. It drops the least-recently-used calculations, and is just plain
stupid when it runs out of registers.

The plan is to make this go even faster for -O0 by taking advantage of the short
live intervals in unoptimized code. It should not be necessary to calculate
liveness when most virtual registers are killed 2-3 instructions after they are
born.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102006 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 18:02:42 +00:00
Dan Gohman
55e59c1863 Code that needs a TargetMachine should have access to one directly, rather
than just getting one through a TargetLowering.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101802 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-19 19:05:59 +00:00
Evan Cheng
d94671a25e Post regalloc LICM. Work in progress.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100592 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-07 00:41:17 +00:00
David Greene
5c8aa950fe Ok, third time's the charm. No changes from last time except the CMake
source addition.  Apparently the buildbots were wrong about failures.

---

Add some switches helpful for debugging:

-print-before=<Pass Name>

Dump IR before running pass <Pass Name>.

-print-before-all

Dump IR before running each pass.

-print-after-all

Dump IR after running each pass.

These are helpful when tracking down a miscompilation.  It is easy to
get IR dumps and do diffs on them, etc.

To make this work well, add a new getPrinterPass API to Pass so that
each kind of pass (ModulePass, FunctionPass, etc.) can create a Pass
suitable for dumping out the kind of object the Pass works on.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100249 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-02 23:17:14 +00:00
Evan Cheng
cf5862d8ac Revert 100204. It broke a bunch of tests and apparently changed what passes are run during codegen.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100207 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-02 19:29:15 +00:00
David Greene
434bd8551d Let's try this again. Re-apply 100143 including an apparent missing
<string> include.  For some reason the buildbot choked on this while my
builds did not.  It's probably due to a difference in system headers.

---

Add some switches helpful for debugging:

-print-before=<Pass Name>

Dump IR before running pass <Pass Name>.

-print-before-all

Dump IR before running each pass.

-print-after-all

Dump IR after running each pass.

These are helpful when tracking down a miscompilation.  It is easy to
get IR dumps and do diffs on them, etc.

To make this work well, add a new getPrinterPass API to Pass so that
each kind of pass (ModulePass, FunctionPass, etc.) can create a Pass
suitable for dumping out the kind of object the Pass works on.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100204 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-02 18:46:26 +00:00
Eric Christopher
1d8f83d0a0 Revert r100143.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100146 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-01 22:54:42 +00:00
David Greene
8ef3acba00 Add some switches helpful for debugging:
-print-before=<Pass Name>

Dump IR before running pass <Pass Name>.

-print-before-all

Dump IR before running each pass.

-print-after-all

Dump IR after running each pass.

These are helpful when tracking down a miscompilation.  It is easy to
get IR dumps and do diffs on them, etc.

To make this work well, add a new getPrinterPass API to Pass so that
each kind of pass (ModulePass, FunctionPass, etc.) can create a Pass
suitable for dumping out the kind of object the Pass works on.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100143 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-01 22:43:57 +00:00
Evan Cheng
c6fe333688 Add skeleton of a machine level cse pass.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@97543 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-02 02:38:24 +00:00
Dan Gohman
fb76fe0929 Fix various doxygen warnings.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@96779 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-22 04:10:52 +00:00
Bob Wilson
fe61fb1e10 Add a new pass on machine instructions to optimize away PHI cycles that
reduce down to a single value.  InstCombine already does this transformation
but DAG legalization may introduce new opportunities.  This has turned out to
be important for ARM where 64-bit values are split up during type legalization:
InstCombine is not able to remove the PHI cycles on the 64-bit values but
the separate 32-bit values can be optimized.  I measured the compile time 
impact of this (running llc on 176.gcc) and it was not significant.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@95951 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-12 01:30:21 +00:00
Jim Grosbach
e45ab8a0a9 For aligned load/store instructions, it's only required to know whether a
function can support dynamic stack realignment. That's a much easier question
to answer at instruction selection stage than whether the function actually
will have dynamic alignment prologue. This allows the removal of the
stack alignment heuristic pass, and improves code quality for cases where
the heuristic would result in dynamic alignment code being generated when
it was not strictly necessary.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@93885 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-19 18:31:11 +00:00
Evan Cheng
7da9ecf967 Add a quick pass to optimize sign / zero extension instructions. For targets where the pre-extension values are available in the subreg of the result of the extension, replace the uses of the pre-extension value with the result + extract_subreg.
For now, this pass is fairly conservative. It only perform the replacement when both the pre- and post- extension values are used in the block. It will miss cases where the post-extension values are live, but not used.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@93278 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-13 00:30:23 +00:00
Evan Cheng
79fc6f44b6 Add a pre-regalloc tail duplication pass.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90567 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-04 09:42:45 +00:00
Jim Grosbach
e27d205d5d Factor the stack alignment calculations out into a target independent pass.
No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90336 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-02 19:30:24 +00:00
Bob Wilson
2d521e51ac Rename new TailDuplicationPass to avoid name conflict with the old one.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@89968 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-26 21:38:41 +00:00
Bob Wilson
15acadde5f Split tail duplication into a separate pass. This is needed to avoid
running tail duplication when doing branch folding for if-conversion, and
we also want to be able to run tail duplication earlier to fix some
reg alloc problems.  Move the CanFallThrough function from BranchFolding
to MachineBasicBlock so that it can be shared by TailDuplication.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@89904 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-26 00:32:21 +00:00
Devang Patel
f5a7a2c84a Remove DebugLabelFolder pass. It is not used by dwarf writer anymore.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@89790 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-24 19:37:07 +00:00
Bill Wendling
8bedf97f68 Don't put in these EH changes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85460 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-29 00:37:35 +00:00
Bill Wendling
f58898fd41 Reverting r85338 for now. It's causing a bootstrap failure on PPC darwin9.
--- Reverse-merging r85338 into '.':
U    lib/CodeGen/SimpleRegisterCoalescing.cpp
U    lib/CodeGen/SimpleRegisterCoalescing.h



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85454 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-29 00:22:16 +00:00
Bob Wilson
a597103c32 Revert r85346 change to control tail merging by CodeGenOpt::Level.
I'm going to redo this using the OptimizeForSize function attribute.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85426 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-28 20:46:46 +00:00
Bob Wilson
cd4f04d6bc Record CodeGen optimization level in the BranchFolding pass so that we can
use it to control tail merging when there is a tradeoff between performance
and code size.  When there is only 1 instruction in the common tail, we have
been merging.  That can be good for code size but is a definite loss for
performance.  Now we will avoid tail merging in that case when the
optimization level is "Aggressive", i.e., "-O3".  Radar 7338114.

Since the IfConversion pass invokes BranchFolding, it too needs to know
the optimization level.  Note that I removed the RegisterPass instantiation
for IfConversion because it required a default constructor.  If someone
wants to keep that for some reason, we can add a default constructor with
a hard-wired optimization level.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85346 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-27 23:49:38 +00:00
Evan Cheng
fa16354e03 Change createPostRAScheduler so it can be turned off at llc -O1.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84273 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-16 21:06:15 +00:00
Evan Cheng
f436463dbd Remove simple regalloc. It has bit rotted.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82127 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-17 05:48:07 +00:00
Chris Lattner
cf143a4d91 remove std::ostream versions of printing stuff for MBB and MF,
upgrading a few things to use raw_ostream


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@79811 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-23 03:13:20 +00:00
Jim Grosbach
8b818d7e98 Move the sjlj exception handling conversions to a back-end pass where they
more properly belong. This allows removing the front-end conditionalized
SJLJ code, and cleans up the generated IR considerably. All of the
infrastructure code (calling _Unwind_SjLj_Register/Unregister, etc) is
added by the SjLjEHPrepare pass.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@79250 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-17 16:41:22 +00:00
Daniel Dunbar
275872e799 Fix some comments referring to std::cerr.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@77931 91177308-0d34-0410-b5e6-96231b3b80d8
2009-08-03 01:02:24 +00:00
Dan Gohman
ad2afc2a42 Reapply r77654 with a fix: MachineFunctionPass's getAnalysisUsage
shouldn't do AU.setPreservesCFG(), because even though CodeGen passes
don't modify the LLVM IR CFG, they may modify the MachineFunction CFG,
and passes like MachineLoop are registered with isCFGOnly set to true.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@77691 91177308-0d34-0410-b5e6-96231b3b80d8
2009-07-31 18:16:33 +00:00
Daniel Dunbar
c5b19b21d8 Revert r77654, it appears to be causing llvm-gcc bootstrap failures, and many
failures when building assorted projects with clang.

--- Reverse-merging r77654 into '.':
U    include/llvm/CodeGen/Passes.h
U    include/llvm/CodeGen/MachineFunctionPass.h
U    include/llvm/CodeGen/MachineFunction.h
U    include/llvm/CodeGen/LazyLiveness.h
U    include/llvm/CodeGen/SelectionDAGISel.h
D    include/llvm/CodeGen/MachineFunctionAnalysis.h
U    include/llvm/Function.h
U    lib/Target/CellSPU/SPUISelDAGToDAG.cpp
U    lib/Target/PowerPC/PPCISelDAGToDAG.cpp
U    lib/CodeGen/LLVMTargetMachine.cpp
U    lib/CodeGen/MachineVerifier.cpp
U    lib/CodeGen/MachineFunction.cpp
U    lib/CodeGen/PrologEpilogInserter.cpp
U    lib/CodeGen/MachineLoopInfo.cpp
U    lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
D    lib/CodeGen/MachineFunctionAnalysis.cpp
D    lib/CodeGen/MachineFunctionPass.cpp
U    lib/CodeGen/LiveVariables.cpp


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@77661 91177308-0d34-0410-b5e6-96231b3b80d8
2009-07-31 03:02:41 +00:00
Dan Gohman
933c762371 Manage MachineFunctions with an analysis Pass instead of the Annotable
mechanism. To support this, make MachineFunctionPass a little more
complete.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@77654 91177308-0d34-0410-b5e6-96231b3b80d8
2009-07-31 01:52:50 +00:00
Lang Hames
2f75c727f2 Removed the BigBlock register allocator.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@77640 91177308-0d34-0410-b5e6-96231b3b80d8
2009-07-30 23:18:43 +00:00
Duncan Sands
b0f1e1780c Add a new codegen pass that normalizes dwarf exception handling
code in preparation for code generation.  The main thing it does
is handle the case when eh.exception calls (and, in a future
patch, eh.selector calls) are far away from landing pads.  Right
now in practice you only find eh.exception calls close to landing
pads: either in a landing pad (the common case) or in a landing
pad successor, due to loop passes shifting them about.  However
future exception handling improvements will result in calls far
from landing pads:
(1) Inlining of rewinds.  Consider the following case:
In function @f:
...
  invoke @g to label %normal unwind label %unwinds
...
unwinds:
  %ex = call i8* @llvm.eh.exception()
...

In function @g:
...
  invoke @something to label %continue unwind label %handler
...
handler:
  %ex = call i8* @llvm.eh.exception()
... perform cleanups ...
  "rethrow exception"

Now inline @g into @f.  Currently this is turned into:
In function @f:
...
  invoke @something to label %continue unwind label %handler
...
handler:
  %ex = call i8* @llvm.eh.exception()
... perform cleanups ...
  invoke "rethrow exception" to label %normal unwind label %unwinds
unwinds:
  %ex = call i8* @llvm.eh.exception()
...

However we would like to simplify invoke of "rethrow exception" into
a branch to the %unwinds label.  Then %unwinds is no longer a landing
pad, and the eh.exception call there is then far away from any landing
pads.

(2) Using the unwind instruction for cleanups.
It would be nice to have codegen handle the following case:
  invoke @something to label %continue unwind label %run_cleanups
...
handler:
... perform cleanups ...
  unwind

This requires turning "unwind" into a library call, which
necessarily takes a pointer to the exception as an argument
(this patch also does this unwind lowering).  But that means
you are using eh.exception again far from a landing pad.

(3) Bugpoint simplifications.  When bugpoint is simplifying
exception handling code it often generates eh.exception calls
far from a landing pad, which then causes codegen to assert.
Bugpoint then latches on to this assertion and loses sight
of the original problem.

Note that it is currently rare for this pass to actually do
anything.  And in fact it normally shouldn't do anything at
all given the code coming out of llvm-gcc!  But it does fire
a few times in the testsuite.  As far as I can see this is
almost always due to the LoopStrengthReduce codegen pass
introducing pointless loop preheader blocks which are landing
pads and only contain a branch to another block.  This other
block contains an eh.exception call.  So probably by tweaking
LoopStrengthReduce a bit this can be avoided.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72276 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-22 20:36:31 +00:00
Jakob Stoklund Olesen
48872e0d84 Pass to verify generated machine code.
The following is checked:

* Operand counts: All explicit operands must be present.

* Register classes: All physical and virtual register operands must be
  compatible with the register class required by the instruction descriptor.

* Register live intervals: Registers must be defined only once, and must be
  defined before use.

The machine code verifier is enabled with the command-line option
'-verify-machineinstrs', or by defining the environment variable
LLVM_VERIFY_MACHINEINSTRS to the name of a file that will receive all the
verifier errors.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@71918 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-16 00:33:53 +00:00
Evan Cheng
bbf1db7213 Rename "loop aligner" pass to "code placement optimization" pass.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@71150 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-07 05:42:24 +00:00
Bill Wendling
2c1d7726f2 Just turn aggressive stack coloring off at -O3.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@71140 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-07 01:33:38 +00:00
Bill Wendling
d0c1f9c932 Temporarily revert r71010. It was causing massive failures during self-hosting.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@71138 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-07 01:27:25 +00:00
Evan Cheng
8c87ae229c Forgot this in the last commit.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@71014 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-05 20:54:11 +00:00
Bill Wendling
e9e6bdf27f Implement stack protectors as function attributes: "ssp" and "sspreq".
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@59202 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-13 01:02:14 +00:00
Bill Wendling
80a320d974 Update in response to feedback from Chris:
- Use enums instead of magic numbers.

- Rework algorithm to use the bytes size from the target to determine when to
  emit stack protectors.

- Get rid of "propolice" in any comments.

- Renamed an option to its expanded form.

- Other miscellanenous changes.

More changes will come after this.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@58723 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-04 21:53:09 +00:00
Bill Wendling
2b58ce5ab4 Initial checkin for stack protectors. Here's what it does:
* The prologue is modified to read the __stack_chk_guard global and insert it
  onto the stack.

* The epilogue is modified to read the stored guard from the stack and compare
  it to the original __stack_chk_guard value. If they differ, then the
  __stack_chk_fail() function is called.

* The stack protector needs to be first on the stack (after the parameters) to
  catch any stack-smashing activities.

Front-end support will follow after a round of beta testing.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@58673 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-04 02:10:20 +00:00
Evan Cheng
09e8ca8a58 Add skeleton for the pre-register allocation live interval splitting pass.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@57847 91177308-0d34-0410-b5e6-96231b3b80d8
2008-10-20 21:44:59 +00:00
Evan Cheng
b1290a6cc4 A Partitioned Boolean Quadratic Programming (PBQP) based register allocator.
Contributed by Lang Hames.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@56959 91177308-0d34-0410-b5e6-96231b3b80d8
2008-10-02 18:29:27 +00:00
Dan Gohman
d3ead4329e Add a new MachineInstr-level DCE pass. It is very simple, and is intended to
be used with fast-isel.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@56268 91177308-0d34-0410-b5e6-96231b3b80d8
2008-09-17 00:43:24 +00:00
Gordon Henriksen
5eca075b74 Rename some GC classes so that their roll will hopefully be clearer.
In particular, Collector was confusing to implementors. Several
thought that this compile-time class was the place to implement
their runtime GC heap. Of course, it doesn't even exist at runtime.
Specifically, the renames are:

  Collector               -> GCStrategy
  CollectorMetadata       -> GCFunctionInfo
  CollectorModuleMetadata -> GCModuleInfo
  CollectorRegistry       -> GCRegistry
  Function::getCollector  -> getGC (setGC, hasGC, clearGC)

Several accessors and nested types have also been renamed to be
consistent. These changes should be obvious.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54899 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-17 18:44:35 +00:00
Owen Anderson
bd3ba461eb - Fix SelectionDAG to generate correct CFGs.
- Add a basic machine-level dead block eliminator.

These two have to go together, since many other parts of the code generator are unable to handle the unreachable blocks otherwise created.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54333 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-04 23:54:43 +00:00
Evan Cheng
3f32d65912 Add a stack slot coloring pass. Not yet enabled.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51934 91177308-0d34-0410-b5e6-96231b3b80d8
2008-06-04 09:18:41 +00:00
Dan Gohman
6ddba2b933 Change class' public PassInfo variables to by initialized with the
address of the PassInfo directly instead of calling getPassInfo.
This eliminates a bunch of dynamic initializations of static data.

Also, fold RegisterPassBase into PassInfo, make a bunch of its
data members const, and rearrange some code to initialize data
members in constructors instead of using setter member functions.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51022 91177308-0d34-0410-b5e6-96231b3b80d8
2008-05-13 02:05:11 +00:00