RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-26 04:32:44 +01:00
Code Issues Projects Releases Wiki Activity
48,499 Commits 4 Branches 5 Tags 704 MiB
7bdb20f6a2
Commit Graph

7264 Commits

Author SHA1 Message Date
Evan Cheng
4f87295872 Targets sometimes assign fixed stack object to spill certain callee-saved 
registers based on dynamic conditions. For example, X86 EBP/RBP, when used as
frame register has to be spilled in the first fixed object. It should inform
PEI this so it doesn't get allocated another stack object. Also, it should not
be spilled as other callee-saved registers but rather its spilling and restoring
are being handled by emitPrologue and emitEpilogue. Avoid spilling it twice.

llvm-svn: 75116
 2009-07-09 06:53:48 +00:00
Lang Hames
ceb80b14d3 Improved tracking of value number kills. VN kills are now represented 
as an (index,bool) pair. The bool flag records whether the kill is a
PHI kill or not. This code will be used to enable splitting of live
intervals containing PHI-kills.

A slight change to live interval weights introduced an extra spill
into lsr-code-insertion (outside the critical sections). The test 
condition has been updated to reflect this.

llvm-svn: 75097
 2009-07-09 03:57:02 +00:00
Dan Gohman
01912bc556 Make SelectionDAG::getVectorShuffle work properly for VECTOR_SHUFFLE 
nodes with operand types that differ from the result type. (This
doesn't normally happen right now, because
SelectionDAGLowering::visitShuffleVector normalizes vector shuffles.)

llvm-svn: 75081
 2009-07-09 00:46:33 +00:00
David Goodwin
49fbd8d6b7 Use common code for both ARM and Thumb-2 instruction and register info. 
llvm-svn: 75067
 2009-07-08 23:10:31 +00:00
Duncan Sands
888e420637 Nowadays vectors are only split if they have an even 
number of elements.  Make some simplifications based
on this (in particular SplitVecRes_SETCC).  Tighten
up some checking while there.

llvm-svn: 75050
 2009-07-08 21:34:03 +00:00
Bill Wendling
8538bf59a8 Use interators instead of counters for loops. 
llvm-svn: 75046
 2009-07-08 20:57:27 +00:00
Owen Anderson
9331034838 Push LLVMContext _back_ through IRBuilder. 
llvm-svn: 75040
 2009-07-08 20:50:47 +00:00
Owen Anderson
332aae685b Switch GlobalVariable ctors to a sane API, where *either* a context or a module is required. 
llvm-svn: 75025
 2009-07-08 19:03:57 +00:00
Duncan Sands
f3d675f3dd Remove trailing whitespace. Reorder some methods 
and cases alphabetically.  No functionality change.

llvm-svn: 75001
 2009-07-08 11:36:39 +00:00
Nick Lewycky
d46a7b2d22 Remove the vicmp and vfcmp instructions. Because we never had a release with 
these instructions, no autoupgrade or backwards compatibility support is
provided.

llvm-svn: 74991
 2009-07-08 03:04:38 +00:00
Owen Anderson
7a1f69e433 Push LLVMContext through GlobalVariables and IRBuilder. 
llvm-svn: 74985
 2009-07-08 01:26:06 +00:00
Chris Lattner
ea7bd9b484 dag combine sext(setcc) -> vsetcc before legalize. To make this safe, 
VSETCC must define all bits, which is different than it was documented
to before.  Since all targets that implement VSETCC already have this
behavior, and we don't optimize based on this, just change the 
documentation.  We now get nice code for vec_compare.ll

llvm-svn: 74978
 2009-07-08 00:31:33 +00:00
Chris Lattner
0e88c6a9b4 SelectionDAG::SignBitIsZero doesn't work right for vectors, 
for now, conservatively return false.

llvm-svn: 74969
 2009-07-07 23:28:46 +00:00
Dale Johannesen
f77b78ed70 Operand of asm("call") (the callee function) is represented 
as "X" constraint and "P" modifier on x86.  Make this work.
(Change may not be sufficient to fix it for non-Darwin, but
I'm pretty sure it won't break anything.)
gcc.apple/asm-block-32.c
gcc.apple/asm-block-33.c

llvm-svn: 74967
 2009-07-07 23:26:33 +00:00
Chris Lattner
e2435c4f6f add support for legalizing an icmp where the result is illegal (4xi1) but 
the input is legal (4 x i32)

llvm-svn: 74964
 2009-07-07 23:03:54 +00:00
Chris Lattner
8604f71f91 random code cleanups. 
llvm-svn: 74962
 2009-07-07 22:49:15 +00:00
Chris Lattner
a754f344b2 implement support for spliting and scalarizing vector setcc's. This 
finishes off enough support for vector compares to get the icmp/fcmp
version of 2008-07-23-VSetCC.ll passing.

llvm-svn: 74961
 2009-07-07 22:47:46 +00:00
Chris Lattner
440ad90c83 lower vector icmp/fcmp to ICMP/FCMP nodes with the right result 
(vector of bool).

llvm-svn: 74960
 2009-07-07 22:41:32 +00:00
Chris Lattner
8361a0c867 ScalarizeVecRes_ShiftOp and ScalarizeVecRes_BinOp are the same, 
eliminate the former.

llvm-svn: 74959
 2009-07-07 22:28:41 +00:00
Chris Lattner
7e1482b246 add support for vector legalizing of *_EXTEND. 
llvm-svn: 74957
 2009-07-07 22:27:17 +00:00
Devang Patel
e416111eeb Accidently dropped this while removing dead code in previous commit. 
llvm-svn: 74953
 2009-07-07 21:55:14 +00:00
Devang Patel
4fdc3e6c1d Remove dead code. 
llvm-svn: 74949
 2009-07-07 21:12:32 +00:00
Owen Anderson
75f0f1cc6c Have scoped mutexes take referenes instead of pointers. 
llvm-svn: 74931
 2009-07-07 18:33:04 +00:00
Evan Cheng
c0a72563c9 Revert 74898. It broke several tests. 
llvm-svn: 74925
 2009-07-07 17:50:43 +00:00
Sanjiv Gupta
83f715997c if the terminator is a branch depending upon the side effects of a 
previous cmp; a copy can not be inserted here if the copy insn also has
 side effects. We don't have access to the attributes of copy insn here;
 so just play safe by finding a safe locations for branch terminators.

llvm-svn: 74898
 2009-07-07 08:04:51 +00:00
Evan Cheng
29ce3bfbb8 Avoid adding a duplicate def. This fixes PR4478. 
llvm-svn: 74857
 2009-07-06 21:34:05 +00:00
Duncan Sands
79c4498f99 Fix the cmake build - patch by Xerxes Rånby. 
llvm-svn: 74825
 2009-07-06 14:28:32 +00:00
Bruno Cardoso Lopes
f03970cf37 Changed ELFCodeEmitter to inherit from ObjectCodeEmitter 
llvm-svn: 74821
 2009-07-06 09:26:48 +00:00
Bruno Cardoso Lopes
fe9ea88db1 Cleanup MachO writer and code emitter. Fix 80 cols problems, remove extra spaces, shrink down includes and move some methods out-of-line 
llvm-svn: 74817
 2009-07-06 06:40:51 +00:00
Bruno Cardoso Lopes
c28df0f408 Just forgot to include the two new files 
llvm-svn: 74814
 2009-07-06 05:16:40 +00:00
Bruno Cardoso Lopes
38373542a1 Add the Object Code Emitter class. Original patch by Aaron Gray, I did some 
cleanup, removed some #includes and moved Object Code Emitter out-of-line.

llvm-svn: 74813
 2009-07-06 05:09:34 +00:00
Owen Anderson
977aa11bc6 More LLVMContext-ification. 
llvm-svn: 74807
 2009-07-05 22:41:43 +00:00
Tilmann Scheller
cea3c16aa5 Add NumFixedArgs attribute to CallSDNode which indicates the number of fixed arguments in a vararg call. 
With the SVR4 ABI on PowerPC, vector arguments for vararg calls are passed differently depending on whether they are a fixed or a variable argument. Variable vector arguments always go into memory, fixed vector arguments are put 
into vector registers. If there are no free vector registers available, fixed vector arguments are put on the stack.

The NumFixedArgs attribute allows to decide for an argument in a vararg call whether it belongs to the fixed or variable portion of the parameter list.

llvm-svn: 74764
 2009-07-03 06:44:53 +00:00
Bruno Cardoso Lopes
0dc4fd3fcc Factor some code out and support for Jump Table relocations 
llvm-svn: 74760
 2009-07-03 04:36:26 +00:00
Devang Patel
5f93a737f6 Simplify debug info intrisinc lowering. 
llvm-svn: 74733
 2009-07-02 22:43:26 +00:00
Douglas Gregor
b523a25108 CMake build fixes, from Xerxes Ranby 
llvm-svn: 74720
 2009-07-02 18:53:52 +00:00
Bruno Cardoso Lopes
810ef07890 shrinking down #includes 
llvm-svn: 74718
 2009-07-02 18:29:24 +00:00
Bruno Cardoso Lopes
bd4ae81317 Remove getFunctionAlignment from TargetELFInfo and use new MachineFunction alignment method 
llvm-svn: 74686
 2009-07-02 02:13:13 +00:00
Devang Patel
735f7d53ca Simplify. 
llvm-svn: 74677
 2009-07-02 00:28:03 +00:00
Devang Patel
2e373cdc9f Simplify. No intentional functionality change. 
llvm-svn: 74673
 2009-07-02 00:08:09 +00:00
Devang Patel
1fececd8c1 Refactor. No functionality change. 
llvm-svn: 74659
 2009-07-01 23:19:01 +00:00
Devang Patel
9149c185a4 llvm.dbg.declare is always used for local variable's debug info. 
llvm-svn: 74625
 2009-07-01 18:51:07 +00:00
Evan Cheng
e6989735a6 CommuteChangesDestination() should check if to-be-commuted instruction defines any register. Also teaches the default commuteInstruction() to commute instruction without definitions (e.g. X86::test / ARM::tsp). 
llvm-svn: 74602
 2009-07-01 08:29:08 +00:00
Evan Cheng
7d78cb531e Remove special handling of implicit_def. Fix a couple more bugs in liveintervalanalysis and coalescer handling of implicit_def. 
Note, isUndef marker must be placed even on implicit_def def operand or else the scavenger will not ignore it. This is necessary because -O0 path does not use liveintervalanalysis, it treats implicit_def just like any other def.

llvm-svn: 74601
 2009-07-01 08:19:36 +00:00
Evan Cheng
37503e9671 Handle IMPLICIT_DEF with isUndef operand marker, part 2. This patch moves the code to annotate machineoperands to LiveIntervalAnalysis. It also add markers for implicit_def that define physical registers. The rest, is just a lot of details. 
llvm-svn: 74580
 2009-07-01 01:59:31 +00:00
Daniel Dunbar
bc3d149e98 Remove unused AsmPrinter OptLevel argument, and propogate. 
- This more or less amounts to a revert of r65379. I'm curious to know what
   happened that caused this variable to become unused.

llvm-svn: 74579
 2009-07-01 01:48:54 +00:00
Bill Wendling
c0fb316bd3 Add an "alignment" field to the MachineFunction object. It makes more sense to 
have the alignment be calculated up front, and have the back-ends obey whatever
alignment is decided upon.

This allows for future work that would allow for precise no-op placement and the
like.

llvm-svn: 74564
 2009-06-30 22:38:32 +00:00
Evan Cheng
28b9e77f19 Temporarily restore the scavenger implicit_def checking code. MachineOperand isUndef mark is not being put on implicit_def of physical registers (created for parameter passing, etc.). 
llvm-svn: 74519
 2009-06-30 09:19:42 +00:00
Evan Cheng
c6c942b70f Add a bit IsUndef to MachineOperand. This indicates the def / use register operand is defined by an implicit_def. That means it can def / use any register and passes (e.g. register scavenger) can feel free to ignore them. 
The register allocator, when it allocates a register to a virtual register defined by an implicit_def, can allocate any physical register without worrying about overlapping live ranges. It should mark all of operands of the said virtual register so later passes will do the right thing.

This is not the best solution. But it should be a lot less fragile to having the scavenger try to track what is defined by implicit_def.

llvm-svn: 74518
 2009-06-30 08:49:04 +00:00
Devang Patel
e35cd91347 Struct types are described using field types only. 
llvm-svn: 74477
 2009-06-29 23:46:50 +00:00
Devang Patel
bff451b95b s/MainCU/ModuleCU/g 
llvm-svn: 74452
 2009-06-29 20:45:18 +00:00
Devang Patel
4ba8cbff55 Multiple DW_TAG_compile_unit is not used, afaict, on any target. 
Update dwarf writer to only emit one DW_TAG_compile_unit per .o file. 

llvm-svn: 74449
 2009-06-29 20:38:13 +00:00
Dan Gohman
e25e17d91e Eliminate a layer of indirection in LoopInfo and MachineLoopInfo. 
llvm-svn: 74394
 2009-06-27 21:22:48 +00:00
Chris Lattner
0ce83b0e95 When doing remat, don't consider uses of non-allocatable physregs. Patch 
by Evan.

llvm-svn: 74370
 2009-06-27 04:06:41 +00:00
Chris Lattner
ce7f3c052e fix a typo that GCC should have caught that causes crashes with -view-*-dags 
llvm-svn: 74364
 2009-06-27 00:57:02 +00:00
Chris Lattner
ca9aefcc93 fix a really subtle bug in the cross section of aliases and TLS: 
the SelectionDAG::getGlobalAddress function properly looks through
aliases to determine thread-localness, but then passes the GV* down
to GlobalAddressSDNode::GlobalAddressSDNode which does not.  Instead
of passing down isTarget, just pass down the predetermined node
opcode.  This fixes some assertions with out of tree changes I'm 
working on.

llvm-svn: 74325
 2009-06-26 21:14:05 +00:00
Owen Anderson
0cf6b34d5d Get rid of these cache variables, which are a holdover from the days when 
we had multiple type planes and these lookups were expensive.

llvm-svn: 74319
 2009-06-26 20:33:47 +00:00
Chris Lattner
0fee902c2d implement DOTGraphTraits<SelectionDAG*>::getNodeLabel in terms of 
SDNode::print_details to eliminate a ton of near-duplicate code.

llvm-svn: 74311
 2009-06-26 19:06:10 +00:00
Douglas Gregor
6d9b0e8c19 Fix linking of llvm-ld and lli with CMake, from Xerxes Rånby 
llvm-svn: 74285
 2009-06-26 15:37:00 +00:00
Chris Lattner
05eabf4f95 dot graph viewing is apparently not using SDNode::print_details, this is bad, 
but in the meantime lets print targetflags on node labels.

llvm-svn: 74274
 2009-06-26 05:55:43 +00:00
Chris Lattner
3aef0c897e propagate target operand flags from dag nodes into MachineOperands. 
llvm-svn: 74273
 2009-06-26 05:52:14 +00:00
Chris Lattner
181d139bc3 fit in 80 cols 
llvm-svn: 74270
 2009-06-26 05:39:02 +00:00
Devang Patel
84a8914a4a Remove debug info anchors - llvm.dbg.compile_units, llvm.dbg.subprograms 
and llvm.dbg.global_variables.

llvm-svn: 74251
 2009-06-26 01:49:18 +00:00
Devang Patel
32bbebe82f Simplify. 
llvm-svn: 74215
 2009-06-25 22:36:02 +00:00
Chris Lattner
6a77e841e8 add targetflags to jump tables and constant pool entries. 
llvm-svn: 74204
 2009-06-25 21:35:31 +00:00
Chris Lattner
f94959f1a3 allow setting target operand flags on TargetGlobalAddress nodes. 
llvm-svn: 74203
 2009-06-25 21:21:14 +00:00
Chris Lattner
55fdaaf6e7 start bringing targetoperand flags into isel, first up, ExternalSymbol. 
llvm-svn: 74199
 2009-06-25 18:45:50 +00:00
Owen Anderson
76d92148a7 Provide guards for this shared structure. I'm not sure this actually needs 
to be shared, but how/where to privatize it is not immediately clear to me.

If any SelectionDAG experts see a better solution, please share!

llvm-svn: 74180
 2009-06-25 17:09:00 +00:00
Owen Anderson
7bc2987c5c Privatize some more debug-related static data. 
llvm-svn: 74179
 2009-06-25 16:55:32 +00:00
Bruno Cardoso Lopes
5e0be0b9a1 Support Constant Pool Sections 
Add section symbols to the symbol table

llvm-svn: 74170
 2009-06-25 07:36:24 +00:00
Bill Wendling
edfdb32218 My guess is that RegInfo should only call the Allocator.Deallocator if it's not 
null.

llvm-svn: 74147
 2009-06-25 00:32:48 +00:00
Owen Anderson
696fd9b29e Now with EVEN FEWER statics! 
llvm-svn: 74143
 2009-06-25 00:04:15 +00:00
Owen Anderson
a21f94d6ba Fewer static variables, part 3 of many. 
llvm-svn: 74140
 2009-06-24 23:41:44 +00:00
Owen Anderson
e194f62a43 Down with _even more_ statics! 
llvm-svn: 74137
 2009-06-24 23:13:56 +00:00
Owen Anderson
7b32a5b971 Down with statics! 
llvm-svn: 74134
 2009-06-24 22:53:20 +00:00
Owen Anderson
9832095b30 Move local statics to per-instance variables. 
llvm-svn: 74132
 2009-06-24 22:28:12 +00:00
Lang Hames
b9f524d565 Completed basic intra block split implementation. 
llvm-svn: 74114
 2009-06-24 20:46:24 +00:00
David Greene
08b3d807e8 This increases the maximum for MVT::LAST_VALUETYPE 
This change doubles the allowable value for MVT::LAST_VALUETYPE. It does
this by doing several things.

1. Introduces MVT::MAX_ALLOWED_LAST_VALUETYPE which in this change has a
value of 64.  This value contains the current maximum for the
MVT::LAST_VALUETYPE.

2. Instead of checking "MVT::LAST_VALUETYPE <= 32", all of those uses
now become "MVT::LAST_VALUETYPE <= MVT::MAX_ALLOWED_LAST_VALUETYPE"

3. Changes the dimension of the ValueTypeActions from 2 elements to four
elements and adds comments ahead of the declaration indicating the it is
"(MVT::MAX_ALLOWED_LAST_VALUETYPE/32) * 2".  This at least lets us find
what is affected if and when MVT::MAX_ALLOWED_LAST_VALUETYPE gets
changed.

4. Adds initializers for the new elements of ValueTypeActions.

This does NOT add any types in MVT. That would be done separately.

This doubles the size of ValueTypeActions from 64 bits to 128 bits and
gives us the freedom to add more types for AVX.

llvm-svn: 74110
 2009-06-24 19:41:55 +00:00
Chris Lattner
ee30c14d3b sink management of DwarfWriter & MachineModuleInfo into the AsmPrinter base class. 
llvm-svn: 74101
 2009-06-24 19:09:55 +00:00
Chris Lattner
d59a5b9f8b sink dwarf finalization out of each target into AsmPrinter::doFinalization 
llvm-svn: 74097
 2009-06-24 18:54:37 +00:00
Chris Lattner
0690755320 eliminate the ExtWeakSymbols set from AsmPrinter. This eliminates 
a bunch of code from all the targets, and eliminates nondeterministic
ordering of directives being emitted in the output.

llvm-svn: 74096
 2009-06-24 18:52:01 +00:00
Chris Lattner
cabe904f79 Rearrange some stuff in MachineOperand and add a new TargetFlags field. 
llvm-svn: 74087
 2009-06-24 17:54:48 +00:00
Owen Anderson
90f7655f57 Get rid of the global CFGOnly flag by threading a ShortNames parameters through the GraphViz rendering code. 
Update other uses in the codebase for this change.

llvm-svn: 74084
 2009-06-24 17:37:09 +00:00
Dale Johannesen
5bad20bda8 Rewrite 73900 per Duncan's suggestion. 
llvm-svn: 74082
 2009-06-24 17:11:31 +00:00
Chris Lattner
07463b0a0a remove dead makefile flags. 
llvm-svn: 74065
 2009-06-24 05:29:56 +00:00
Lang Hames
38bf7cf4e2 Fixed a bug in LiveInterval scaling (failure to scale VNI defs correctly), removed old TODO comments. 
llvm-svn: 74054
 2009-06-24 02:17:32 +00:00
Evan Cheng
7292cadf06 Fix support for inline asm input / output operand tying when operand spans across multiple registers (e.g. two i64 operands in 32-bit mode). 
llvm-svn: 74053
 2009-06-24 02:05:51 +00:00
Bruno Cardoso Lopes
bd023b8ebe Use a default alignment for data and bss sections. 
Only pad when the section size > 0 and move the code that deals
with globals initializers to a place we know for sure the global
is initialized.

llvm-svn: 73944
 2009-06-23 04:39:27 +00:00
Dale Johannesen
e577db0efd Fix memcpy expansion so it won't generate invalid 
types for the target (I think).  This was breaking
the PPC32 calling sequence.

llvm-svn: 73900
 2009-06-22 20:59:07 +00:00
Evan Cheng
2410955c62 Fix another register coalescer crash: forgot to check if the instruction being updated has already been coalesced. 
llvm-svn: 73898
 2009-06-22 20:49:32 +00:00
Bruno Cardoso Lopes
4615b9cdf9 Use different functions to emit the string and symbol tables. 
llvm-svn: 73895
 2009-06-22 19:29:56 +00:00
Bruno Cardoso Lopes
7b196ba30e Add more methods to gather target specific elf stuff 
Support for .text relocations, implementing TargetELFWriter overloaded methods for x86/x86_64.
Use a map to track global values to their symbol table indexes
Code cleanup and small fixes

llvm-svn: 73894
 2009-06-22 19:16:16 +00:00
Evan Cheng
b45918e5bb Fix PR4419: handle defs of partial uses. 
llvm-svn: 73816
 2009-06-20 04:34:51 +00:00
Devang Patel
e2b1c9a530 mv CodeGen/DebugLoc.h Support/DebugLoc.h 
llvm-svn: 73786
 2009-06-19 22:08:58 +00:00
Devang Patel
35810cd07b Move up dwarf writer initialization in common AsmPrinter class. 
llvm-svn: 73784
 2009-06-19 21:54:26 +00:00
Eli Friedman
eb83635a7e Minor cleanup; fixes review comments for a previous patch. Sorry for 
taking so long to get to this!

llvm-svn: 73757
 2009-06-19 06:01:55 +00:00
Lang Hames
1bade5ff09 More VNInfo tweaking, plus a little progress on intra-block splitting. 
llvm-svn: 73750
 2009-06-19 02:17:53 +00:00
Chris Lattner
6cd267dcc5 move mangler quote handling from asm printers to TargetAsmInfo. 
llvm-svn: 73738
 2009-06-18 23:41:35 +00:00
Chris Lattner
ea0ea2f4bc simplify macro debug info directive handling. 
llvm-svn: 73736
 2009-06-18 23:31:37 +00:00
Lang Hames
7f288c29af Improved PHI def marking, replaced some gotos with breaks. 
llvm-svn: 73727
 2009-06-18 22:01:47 +00:00