RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-22 12:33:33 +02:00
Code Issues Projects Releases Wiki Activity
63,281 Commits 4 Branches 5 Tags 704 MiB
245a1faf76
Commit Graph

12018 Commits

Author SHA1 Message Date
Jakob Stoklund Olesen
245a1faf76 Implement register class inflation. 
When splitting a live range, the new registers have fewer uses and the
permissible register class may be less constrained. Recompute the register class
constraint from the uses of new registers created for a split. This may let them
be allocated from a larger set, possibly avoiding a spill.

llvm-svn: 110703
 2010-08-10 18:37:40 +00:00
Jakob Stoklund Olesen
e51a747336 Recalculate the spill weight and allocation hint for virtual registers created 
during live range splitting.

llvm-svn: 110686
 2010-08-10 17:07:22 +00:00
Mikhail Glushenkov
ba7b4da6de Trailing whitespace. 
llvm-svn: 110679
 2010-08-10 14:49:24 +00:00
Bruno Cardoso Lopes
460703a738 Remove replicate intrinsics, clang will generate shufflevector for those. The shuffles can't be matched by x86 codegen yet, but will soon 
llvm-svn: 110647
 2010-08-10 02:25:35 +00:00
Bruno Cardoso Lopes
7afe424c59 Use i32 instead of i8 for dot product intrinsic 
llvm-svn: 110643
 2010-08-10 01:40:05 +00:00
Jakob Stoklund Olesen
1ab2fab3af Transpose the calculation of spill weights such that we are calculating one 
register at a time. This turns out to be slightly faster than iterating over
instructions, but more importantly, it allows us to compute spill weights for
new registers created after the spill weight pass has run.

Also compute the allocation hint at the same time as the spill weight. This
allows us to use the spill weight as a cost metric for copies, and choose the
most profitable hint if there is more than one possibility.

The new hints provide a very small (< 0.1%) but universal code size improvement.

llvm-svn: 110631
 2010-08-10 00:02:26 +00:00
Bruno Cardoso Lopes
7d73127002 Fix the last argument type of AVX vblend intrinsics 
llvm-svn: 110628
 2010-08-10 00:00:22 +00:00
Bill Wendling
8a7a43a1cb Merge the OptimizeExts and OptimizeCmps passes into one PeepholeOptimizer 
pass. This pass should expand with all of the small, fine-grained optimization
passes to reduce compile time and increase happiment.

llvm-svn: 110627
 2010-08-09 23:59:04 +00:00
Kevin Enderby
8303a64b0a Next bit of support for the dwarf .file directive. This patch takes the 
previously collected info from the .file directives and outputs the encoded
bytes for it.  For now this is only in the Mach-O streamer but at some point
will move to a more generic place.

llvm-svn: 110617
 2010-08-09 22:52:14 +00:00
Devang Patel
8d0c79694e Refactor. 
llvm-svn: 110607
 2010-08-09 21:39:24 +00:00
Duncan Sands
6eec2f9421 Remove the ValueMap operator=, which was wrong (it did't correct the 
Map pointers of any contained ValueMapCallbackVH's) and unused.

llvm-svn: 110577
 2010-08-09 16:44:56 +00:00
Zhongxing Xu
d955bbe69b Change the Interval type to signed in ImmutableIntervalMap.h. 
llvm-svn: 110562
 2010-08-09 03:43:39 +00:00
Benjamin Kramer
bf4f7b3665 Remove unused variable. 
llvm-svn: 110551
 2010-08-08 19:54:10 +00:00
Duncan Sands
fa440174ee Remove the ValueMap copy constructor. It's not used anywhere, 
and removing it catches the mistake of passing a ValueMap by
copy rather than by reference.

llvm-svn: 110549
 2010-08-08 12:57:48 +00:00
Eric Christopher
c788c82027 Add a bit of a hack to avoid multiple defines of variables in 
config.h and llvm-config.h.  This could probably be improved.

llvm-svn: 110547
 2010-08-08 09:18:29 +00:00
Bill Wendling
39c49e3e17 Use the "isCompare" machine instruction attribute instead of calling the 
relatively expensive comparison analyzer on each instruction. Also rename the
comparison analyzer method to something more in line with what it actually does.

This pass is will eventually be folded into the Machine CSE pass.

llvm-svn: 110539
 2010-08-08 05:04:59 +00:00
Oscar Fuentes
38ae8f5381 CMake: set configure macro LLVM_PREFIX 
llvm-svn: 110536
 2010-08-08 04:32:21 +00:00
Oscar Fuentes
8a58d72172 CMake: configure header llvm-config.h 
This is the cmake equivalent of r110532, which fixed bug #7809.

llvm-svn: 110535
 2010-08-08 04:17:18 +00:00
Eric Christopher
33fd5d63b3 Fix PR7809 by creating a header for just llvm variables that can be 
included in exported interfaces.  Update a couple of exported interfaces.

llvm-svn: 110532
 2010-08-08 02:44:17 +00:00
Bill Wendling
f4d90d9196 Add back in r109901, which adds a Compare flag to the target instructions. It's 
useful after all.

llvm-svn: 110531
 2010-08-08 01:49:35 +00:00
Benjamin Kramer
a7febd985f Roll back my last two commits, valgrind complains. 
llvm-svn: 110518
 2010-08-07 13:27:41 +00:00
Benjamin Kramer
bbdbd2be19 A reference to the Timer's name is safe, it outlives the contents of the vector. 
llvm-svn: 110517
 2010-08-07 13:07:57 +00:00
Benjamin Kramer
ffdaca28a5 Shrink PassNameParser's binary size with array_pod_sort. 
llvm-svn: 110512
 2010-08-07 11:45:42 +00:00
Owen Anderson
d17104afda Remove layering violation. 
llvm-svn: 110505
 2010-08-07 06:01:13 +00:00
Owen Anderson
e59c4411b3 Add an inverse() method to ConstantRange. 
llvm-svn: 110504
 2010-08-07 05:47:46 +00:00
Nick Lewycky
b37e4d1eb2 Fix typo. 
llvm-svn: 110502
 2010-08-07 05:25:29 +00:00
Dan Gohman
0c6a626902 Remove assignPassManager's default arguments. It's really 
confusing to have different arguments for the same virtual
function at different levels of the class hierarchy.

llvm-svn: 110500
 2010-08-07 01:25:32 +00:00
Dan Gohman
48a98809af More #include cleanups. 
llvm-svn: 110499
 2010-08-07 01:18:18 +00:00
Dan Gohman
caf690af8e Delete this explicit assignment operator; it's equivalent to 
the implicit one.

llvm-svn: 110498
 2010-08-07 01:17:47 +00:00
Dan Gohman
a581d6d202 Tidy up PMStack. Add a bunch of consts, use std::vector instead of 
std::deque, since this is a stack and only supports push/pop on
one end, and remove an unimplemented declaration.

llvm-svn: 110495
 2010-08-07 00:53:01 +00:00
Dan Gohman
1d48a4b1d7 Tidy some #includes and forward-declarations, and move the C binding code 
out of PassManager.cpp and into Core.cpp with the rest of the C binding code.

llvm-svn: 110494
 2010-08-07 00:43:20 +00:00
Owen Anderson
31ced50119 Add a convenience constructor. 
llvm-svn: 110493
 2010-08-07 00:42:06 +00:00
Dan Gohman
1637747481 Make AnalysisImpls private. 
llvm-svn: 110492
 2010-08-07 00:34:52 +00:00
Owen Anderson
c61b8d0eb2 Add a predicate to determine if a call is an inline asm statement. 
llvm-svn: 110488
 2010-08-07 00:19:59 +00:00
Rafael Espindola
2145c31de3 Some cleanup. Use a class (OptionInfo) instead of a pair of a pair and remove 
some default values that are not used.

llvm-svn: 110485
 2010-08-06 23:03:52 +00:00
Bruno Cardoso Lopes
5b602f8822 Patterns to match AVX 256-bit vzero intrinsics 
llvm-svn: 110480
 2010-08-06 22:10:01 +00:00
Dan Gohman
13c66ce161 Eliminate PromoteMemoryToRegisterID; just use addPreserved("mem2reg") 
instead, as an example of what this looks like.

llvm-svn: 110478
 2010-08-06 21:48:06 +00:00
Jim Grosbach
e4f646b03f tidy up 
llvm-svn: 110476
 2010-08-06 21:31:35 +00:00
Daniel Dunbar
49c592cd1b MC: Add default value for AddrSpace argument to EmitValue. 
llvm-svn: 110475
 2010-08-06 21:24:05 +00:00
Jakob Stoklund Olesen
a37c7509bf Add LiveInterval::RenumberValues - Garbage collection for VNInfos. 
After heavy editing of a live interval, it is much easier to simply renumber the
live values instead of trying to keep track of the unused ones.

llvm-svn: 110463
 2010-08-06 18:46:59 +00:00
Owen Anderson
f2fea95f2f Reapply r110396, with fixes to appease the Linux buildbot gods. 
llvm-svn: 110460
 2010-08-06 18:33:48 +00:00
Dan Gohman
da3f592fb3 Implement a proper getModRefInfo for va_arg. 
llvm-svn: 110458
 2010-08-06 18:24:38 +00:00
Jim Grosbach
a66c2222ff spelling 
llvm-svn: 110457
 2010-08-06 18:24:36 +00:00
Rafael Espindola
6d53fded19 Fix eabi calling convention when a 64 bit value shadows r3. 
Without this what was happening was:

* R3 is not marked as "used"
* ARM backend thinks it has to save it to the stack because of vaarg
* Offset computation correctly ignores it
* Offsets are wrong

llvm-svn: 110446
 2010-08-06 15:35:32 +00:00
Dan Gohman
66b637e4da Add a comment. 
llvm-svn: 110426
 2010-08-06 02:04:07 +00:00
Bill Wendling
0ec56431c5 Revert r109901. The implementation of <rdar://problem/7405933> (r110423) doesn't 
need the Compare flag after all.

--- Reverse-merging r109901 into '.':
U    include/llvm/Target/TargetInstrDesc.h
U    include/llvm/Target/Target.td
U    utils/TableGen/InstrInfoEmitter.cpp
U    utils/TableGen/CodeGenInstruction.cpp
U    utils/TableGen/CodeGenInstruction.h

llvm-svn: 110424
 2010-08-06 01:36:09 +00:00
Bill Wendling
0cd2ae5158 Add the Optimize Compares pass (disabled by default). 
This pass tries to remove comparison instructions when possible. For instance,
if you have this code:

   sub r1, 1
   cmp r1, 0
   bz  L1

and "sub" either sets the same flag as the "cmp" instruction or could be
converted to set the same flag, then we can eliminate the "cmp" instruction all
together. This is a important for ARM where the ALU instructions could set the
CPSR flag, but need a special suffix ('s') to do so.

llvm-svn: 110423
 2010-08-06 01:32:48 +00:00
Owen Anderson
aadd8a89ca Revert r110396 to fix buildbots. 
llvm-svn: 110410
 2010-08-06 00:23:35 +00:00
Bruno Cardoso Lopes
90d1c89488 Remove unused AVX intrinsics 
llvm-svn: 110407
 2010-08-06 00:04:07 +00:00
Dan Gohman
ecf79cf569 Make AA private, since subclasses shouldn't (aren't don't) access it directly. 
llvm-svn: 110398
 2010-08-05 23:44:45 +00:00