1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-02 00:42:52 +01:00
Commit Graph

8923 Commits

Author SHA1 Message Date
Bill Wendling
2aa666c419 Fix to get it to compile.
llvm-svn: 95840
2010-02-11 01:15:27 +00:00
Bill Wendling
123c791ff5 Don't print out a default newline when emitting the section offset. There are
almost always comments afterwards that need printing.

llvm-svn: 95839
2010-02-11 01:13:02 +00:00
Jakob Stoklund Olesen
3b5abdefc3 Reuse operand location when updating PHI instructions.
Calling RemoveOperand is very expensive on huge PHI instructions. This makes
early tail duplication run twice as fast on the Firefox JavaScript
interpreter.

llvm-svn: 95832
2010-02-11 00:34:33 +00:00
Jakob Stoklund Olesen
215d9f3898 Remove duplicate successors from indirectbr instructions before building the machine CFG.
This makes early tail duplication run 60 times faster when compiling the Firefox
JavaScript interpreter, see PR6186.

llvm-svn: 95831
2010-02-11 00:34:18 +00:00
Mon P Wang
c17e781f35 The previous fix of widening divides that trap was too fragile as it depends on custom
lowering and requires that certain types exist in ValueTypes.h.  Modified widening to
check if an op can trap and if so, the widening algorithm will apply only the op on
the defined elements.  It is safer to do this in widening because the optimizer can't
guarantee removing unused ops in some cases.

llvm-svn: 95823
2010-02-10 23:37:45 +00:00
Dale Johannesen
5f337b582f Ignore debug info one more place during coalescing.
llvm-svn: 95819
2010-02-10 23:04:09 +00:00
Bob Wilson
82d5534acc Delete dead PHI machine instructions. These can be created due to type
legalization even when the IR-level optimizer has removed dead phis, such
as when the high half of an i64 value is unused on a 32-bit target.
I had to adjust a few test cases that had dead phis.
This is a partial fix for Radar 7627077.

llvm-svn: 95816
2010-02-10 22:58:57 +00:00
Dale Johannesen
9d9478a53a Skip debug info in a couple of places.
llvm-svn: 95814
2010-02-10 21:47:48 +00:00
Bill Wendling
b288ec36ae Use an index instead of pointers into the vector. If the vector resizes, then
the pointer values could be invalid.

llvm-svn: 95813
2010-02-10 21:41:57 +00:00
Dale Johannesen
99090c9fe0 When I rewrote this loop per Chris' preference I
changed its behavior.  Oops.

llvm-svn: 95811
2010-02-10 21:41:41 +00:00
Dan Gohman
efb42fa7c7 Fix several comments which had previously been "the the" where a
different word was intended.

llvm-svn: 95795
2010-02-10 20:04:19 +00:00
Dan Gohman
92b6122204 Fix "the the" and similar typos.
llvm-svn: 95781
2010-02-10 16:03:48 +00:00
Evan Cheng
8bee7fb61d Now that ShrinkDemandedOps() is separated out from DAG combine. It sometimes leave some obvious nops which dag combine used to clean up afterwards e.g. (trunk (ext n)) -> n. Look for them and squash them.
llvm-svn: 95757
2010-02-10 02:17:34 +00:00
Dale Johannesen
12262a957f Rewrite loop to suit Chris' preference.
llvm-svn: 95749
2010-02-10 01:31:26 +00:00
Chris Lattner
72156036b9 fix a layering violation: VirtRegRewriter.cpp shouldn't use AsmPrinter.h.
llvm-svn: 95748
2010-02-10 01:23:18 +00:00
Evan Cheng
e0e620c006 Remove duplicated #include.
llvm-svn: 95747
2010-02-10 01:22:57 +00:00
Evan Cheng
8c2662f96a Emit an error for illegal inline asm constraint (which uses illegal type) rather than asserting.
llvm-svn: 95746
2010-02-10 01:21:02 +00:00
Chris Lattner
750bf54fc6 fix missing #includes.
llvm-svn: 95745
2010-02-10 01:17:36 +00:00
Bill Wendling
270a5c4426 Improve comments a even more.
llvm-svn: 95740
2010-02-10 00:59:47 +00:00
Dale Johannesen
9b0d228b5f Skip DBG_VALUE many places in live intervals and
register coalescing.  This fixes many crashes and
places where debug info affects codegen (when
dbg.value is lowered to machine instructions, which
it isn't yet in TOT).

llvm-svn: 95739
2010-02-10 00:55:42 +00:00
Chris Lattner
db4b4d819d Move verbose asm instruction comments to using MCStreamer.
The major win of this is that the code is simpler and they 
print on the same line as the instruction again:

        movl    %eax, 96(%esp)          ## 4-byte Spill
        movl    96(%esp), %eax          ## 4-byte Reload
        cmpl    92(%esp), %eax          ## 4-byte Folded Reload
        jl      LBB7_86

llvm-svn: 95738
2010-02-10 00:47:53 +00:00
Bill Wendling
076bf7e91c Improve comments a bit more.
llvm-svn: 95737
2010-02-10 00:45:28 +00:00
Dale Johannesen
da728faa59 more comment updates
llvm-svn: 95736
2010-02-10 00:44:23 +00:00
Dale Johannesen
59679b7966 Add isDebug argument to ChangeToRegister; this prevents
the field from being used uninitialized later in some cases.

llvm-svn: 95735
2010-02-10 00:41:49 +00:00
Chris Lattner
6d7856867a print all the newlines at the end of instructions with
OutStreamer.AddBlankLine instead of textually.

llvm-svn: 95734
2010-02-10 00:36:00 +00:00
Dale Johannesen
c9f253214e Fix comments to reflect renaming elsewhere.
llvm-svn: 95730
2010-02-10 00:11:11 +00:00
David Greene
d2c34ce826 Only dump output in debug mode.
llvm-svn: 95711
2010-02-09 23:03:05 +00:00
Bill Wendling
fe743bc4d0 Improve comments in the LSDA somewhat. They can be improved much more.
llvm-svn: 95707
2010-02-09 22:49:16 +00:00
Chris Lattner
7acf9be6c4 move target-independent opcodes out of TargetInstrInfo
into TargetOpcodes.h.  #include the new TargetOpcodes.h
into MachineInstr.  Add new inline accessors (like isPHI())
to MachineInstr, and start using them throughout the 
codebase.

llvm-svn: 95687
2010-02-09 19:54:29 +00:00
Jakob Stoklund Olesen
70aa88d977 Remember to update live-in lists when coalescing physregs.
Patch by M Wahab!

llvm-svn: 95668
2010-02-09 17:20:11 +00:00
Dale Johannesen
cfb0894dba Skip DEBUG_VALUE in some places where it was affecting codegen.
llvm-svn: 95647
2010-02-09 02:01:46 +00:00
Devang Patel
d381c20ee7 Add declaration attribute to a variable DIE, if there is a separate DIE for the definition.
llvm-svn: 95646
2010-02-09 01:58:33 +00:00
Lang Hames
796e4c2b4a Fixed a bug in the PBQP allocator's findCoalesces method.
Previously spill registers, whose def indexes are not defined, would sometimes be improperly marked as coalescable with conflicting registers. The new findCoalesces routine conservatively assumes that any register with at least one undefined def is not coalescable with any register it interferes with.

llvm-svn: 95636
2010-02-09 00:50:27 +00:00
Lang Hames
47fb92c5e7 Added copy sensible construction & assignment to PBQP graphs and fixed a memory access bug in the heuristic solver.
llvm-svn: 95633
2010-02-09 00:45:48 +00:00
Dan Gohman
d19999d757 Implement AsmPrinter support for several more operators which have
direct MCExpr equivalents. Don't use MCExpr::Shr because it isn't
consistent between targets.

llvm-svn: 95620
2010-02-09 00:02:37 +00:00
Chris Lattner
32fa6adff6 now that @GOTOFF is no longer represented as a suffix on a
MCSymbol, we can remove the 'suffix' argument of 
GetBlockAddressSymbol.  Do so.

llvm-svn: 95601
2010-02-08 23:10:08 +00:00
Dan Gohman
8b46c6a6cb ConstantFoldConstantExpression can theoretically return the original
expression; don't go into an infinite loop if it does.

llvm-svn: 95591
2010-02-08 22:19:11 +00:00
Dan Gohman
56b9ea088b When CodeGen'ing unoptimized code, there may be unfolded constant expressions
in global initializers. Instead of aborting, attempt to fold them on the
spot. If folding succeeds, emit the folded expression instead.

This fixes PR6255.

llvm-svn: 95583
2010-02-08 22:02:38 +00:00
Dale Johannesen
5ddd318b03 Apply the 95471 fix to SelectionDAGBuilder as well;
we can get in here if FastISel gives up in a block.
(Actually the two copies of this need to be unified.  Later.)

llvm-svn: 95579
2010-02-08 21:53:27 +00:00
Dan Gohman
f113e5466c In guaranteed tailcall mode, don't decline the tailcall optimization
for blocks ending in "unreachable".

llvm-svn: 95565
2010-02-08 20:34:14 +00:00
Duncan Sands
ea225349b0 Give DwarfPrinter a protected (but not virtual) destructor. Cppcheck
warns about this base class not having a virtual destructor, but since
this class has no virtual methods and neither it or the types derived
from it has a destructor, a protected trivial destructor will do (and
shuts cppcheck up) the trick without the cost of introducing a vtable.

llvm-svn: 95526
2010-02-07 21:09:22 +00:00
Evan Cheng
5541068ad3 Run codegen dce pass for all targets at all optimization levels. Previously it's
only run for x86 with fastisel. I've found it being very effective in
eliminating some obvious dead code as result of formal parameter lowering
especially when tail call optimization eliminated the need for some of the loads
from fixed frame objects. It also shrinks a number of the tests. A couple of
tests no longer make sense and are now eliminated.

llvm-svn: 95493
2010-02-06 09:07:11 +00:00
Bob Wilson
61a15581e7 Fix an uninitialized value. Radar 7609421.
llvm-svn: 95488
2010-02-06 05:55:20 +00:00
Dale Johannesen
83f31d3511 After Victor's latest commits I am seeing null
addresses in dbg.declare; ignore this for the
moment to prevent things from breaking.

llvm-svn: 95471
2010-02-06 02:26:02 +00:00
Devang Patel
102b8a3103 Set DW_AT_artificial only if argument is marked as artificial.
llvm-svn: 95461
2010-02-06 01:02:37 +00:00
Devang Patel
9eb9bb03c9 Do not generate specification DIE for nested functions.
llvm-svn: 95452
2010-02-05 23:09:20 +00:00
Evan Cheng
94fe5501b7 When the scheduler unfold a load folding instruction it move some of the predecessors to the unfolded load. It decides what gets moved to the load by checking whether the new load is using the predecessor as an operand. The check neglects the cases whether the predecessor is a flagged scheduling unit.
rdar://7604000

llvm-svn: 95339
2010-02-05 01:27:11 +00:00
Bill Wendling
9761f067f8 An empty global constant (one of size 0) may have a section immediately
following it. However, the EmitGlobalConstant method wasn't emitting a body for
the constant. The assembler doesn't like that. Before, we were generating this:

  .zerofill __DATA, __common, __cmd, 1, 3

This fix puts us back to that semantic.

llvm-svn: 95336
2010-02-05 00:17:02 +00:00
Evan Cheng
ce5962aaf9 Fix typo Duncan noticed.
llvm-svn: 95322
2010-02-04 19:07:06 +00:00
Jakob Stoklund Olesen
d72e82107d Fix small bug in handling instructions with more than one implicitly defined operand.
ProcessImplicitDefs would only mark one operand per instruction with <undef>.
This fixed PR6086.

llvm-svn: 95319
2010-02-04 18:46:28 +00:00
Evan Cheng
3c93245a64 It's too risky to eliminate sext / zext of call results for tail call optimization even if the caller / callee attributes completely match. The callee may have been bitcast'ed (or otherwise lied about what it's doing).
llvm-svn: 95282
2010-02-04 02:45:02 +00:00
Devang Patel
32eb7508e4 Emit appropriate expression to find virtual base offset.
llvm-svn: 95242
2010-02-03 20:08:48 +00:00
Devang Patel
83d905f82b Provide interface to identifiy artificial methods.
llvm-svn: 95240
2010-02-03 19:57:19 +00:00
Chris Lattner
837d242070 make MachineModuleInfoMachO hold non-const MCSymbol*'s instead
of const ones.  non-const ones aren't very useful, because you can't
even, say, emit them.

llvm-svn: 95205
2010-02-03 06:18:30 +00:00
Chris Lattner
e8811c3222 change addPassesToEmitFile to return true on failure instead of its input,
add -filetype=null for performance testing and remove -filetype=dynlib,
which isn't planned to be implemented.

llvm-svn: 95202
2010-02-03 05:55:08 +00:00
Evan Cheng
e273e42195 Revert 94937 and move the noreturn check to codegen.
llvm-svn: 95198
2010-02-03 03:55:59 +00:00
Evan Cheng
d9cf09b0d6 Allow all types of callee's to be tail called. But avoid automatic tailcall if the callee is a result of bitcast to avoid losing necessary zext / sext etc.
llvm-svn: 95195
2010-02-03 03:28:02 +00:00
Chris Lattner
eb6d6da2e4 don't emit \n's at the start of X86AsmPrinter::runOnMachineFunction,
.o files don't like that.

llvm-svn: 95187
2010-02-03 01:49:49 +00:00
Chris Lattner
a64be6d665 privatize a bunch of methods and move \n printing into them.
llvm-svn: 95186
2010-02-03 01:46:05 +00:00
Dale Johannesen
1e9d147461 Reapply 95050 with a tweak to check the register class.
llvm-svn: 95183
2010-02-03 01:40:33 +00:00
Chris Lattner
edf1e280a2 rejigger the world so that EmitInstruction prints the \n at
the end of the instruction instead of expecting the caller to
do it.  This currently causes the asm-verbose instruction 
comments to be on the next line.

llvm-svn: 95178
2010-02-03 01:09:55 +00:00
Chris Lattner
cd07a3a0b7 sink handling of target-independent machine instrs (other
than DEBUG_VALUE :(  ) into the target indep AsmPrinter.cpp
file.   This allows elimination of the 
NO_ASM_WRITER_BOILERPLATE hack among other things.

llvm-svn: 95177
2010-02-03 01:00:52 +00:00
Chris Lattner
7e4c2b183b pass an instprinter into the AsmPrinter if it is available.
llvm-svn: 95168
2010-02-03 00:29:55 +00:00
Chris Lattner
bdd4ae564a make any use of the "O" stream in asmprinter print to
stderr if in filetype=obj mode.  This is a hack, and will
live until dwarf emission and other random stuff that is
not yet going through MCStreamer is upgraded.  It only
impacts filetype=obj mode.

llvm-svn: 95166
2010-02-03 00:22:02 +00:00
Chris Lattner
78f57ac8dc Hook up -filetype=obj through the MachO streamer. Here's a demo:
$ cat t.ll 
@g = global i32 42
$ llc t.ll -o t.o -filetype=obj
$ nm t.o
00000000 D _g

There is still a ton of work left.  Instructions are not being encoded
yet apparently.

llvm-svn: 95162
2010-02-02 23:57:42 +00:00
Evan Cheng
9057fea7ef Revert 95130.
llvm-svn: 95160
2010-02-02 23:55:14 +00:00
Chris Lattner
62ba09bb47 use OwningPtr and factor code better.
llvm-svn: 95156
2010-02-02 23:45:17 +00:00
Chris Lattner
643f2f9bc4 refactor code so that LLVMTargetMachine creates the asmstreamer and
mccontext instead of having AsmPrinter do it.  This allows other 
types of MCStreamer's to be passed in.

llvm-svn: 95155
2010-02-02 23:37:42 +00:00
Chris Lattner
9ba7b16a19 simplify getVerboseAsm
llvm-svn: 95153
2010-02-02 22:58:13 +00:00
Chris Lattner
501138c6ad move handling of asm-verbose out of AsmPrinter.cpp into LLVMTargetMachine.cpp with the rest of the command line options.
llvm-svn: 95152
2010-02-02 22:54:51 +00:00
Chris Lattner
665a764dbe remove dead #include, stupid symlinks.
llvm-svn: 95150
2010-02-02 22:37:42 +00:00
Chris Lattner
7dd4ad3ced Remove a bunch of stuff around the edges of the ELF writer.
Now the only use of the ELF writer is the JIT, which won't be
easy to fix in the short term. :( :(

llvm-svn: 95148
2010-02-02 22:31:11 +00:00
Evan Cheng
48375fbf4f Pass callsite return type to TargetLowering::LowerCall and use that to check sibcall eligibility.
llvm-svn: 95130
2010-02-02 21:29:10 +00:00
Chris Lattner
7d162688a9 eliminate FileModel::Model, just use CodeGenFileType. The client
of the code generator shouldn't care what object format a target
uses.

llvm-svn: 95124
2010-02-02 21:06:45 +00:00
Chris Lattner
ff8ba8fc79 eliminate all forms of addPassesToEmitMachineCode except
the one used by the JIT.  Remove all forms of
addPassesToEmitFileFinish except the one used by the static
code generator.  Inline the remaining version of
addPassesToEmitFileFinish into its only caller.

llvm-svn: 95109
2010-02-02 19:14:27 +00:00
Dale Johannesen
08ab638bdc Test revert 95050; there's a good chance it's causing
buildbot failure.

llvm-svn: 95103
2010-02-02 18:52:56 +00:00
Chris Lattner
fa04a6e968 Inline addAssemblyEmitter into its one real caller and delete
the -print-emitted-asm option.  The JIT shouldn't have to pull
in the asmprinter.

llvm-svn: 95100
2010-02-02 18:44:12 +00:00
Devang Patel
59e8b027b8 Apparently gdb is not amused by empty lines in pubtypes section.
llvm-svn: 95064
2010-02-02 03:47:27 +00:00
Devang Patel
38cf687622 NULL terminate name in pubtypes sections.
llvm-svn: 95062
2010-02-02 03:37:03 +00:00
Dale Johannesen
a20fc3d1a9 Make local RA smarter about reusing input register of a copy
as output.  Needed for (functional) correctness in inline asm,
and should be generally beneficial.  7361612.

llvm-svn: 95050
2010-02-02 02:08:02 +00:00
Daniel Dunbar
9aad068bba Update CMake.
llvm-svn: 95041
2010-02-02 01:12:20 +00:00
Nate Begeman
4d489ea053 Kill the Mach-O writer, and temporarily make filetype=obj an error.
The MCStreamer based assemblers will take over for this functionality.

llvm-svn: 95033
2010-02-01 23:56:58 +00:00
Mon P Wang
65b01b6ce7 Improve EXTRACT_VECTOR_ELT patch based on comments from Duncan
llvm-svn: 95012
2010-02-01 22:15:09 +00:00
Chris Lattner
af2e10ddef eliminate a bunch of pointless LLVMContext arguments.
llvm-svn: 95001
2010-02-01 20:48:08 +00:00
Dale Johannesen
1483e87700 fix PR 6157. Testcase pending.
llvm-svn: 94996
2010-02-01 19:54:53 +00:00
Mon P Wang
f9fa3aa2c0 Fixed a couple of optimization with EXTRACT_VECTOR_ELT that assumes the result
type is the same as the element type of the vector.  EXTRACT_VECTOR_ELT can
be used to extended the width of an integer type.  This fixes a bug for
Generic/vector-casts.ll on a ppc750.

llvm-svn: 94990
2010-02-01 19:03:18 +00:00
Devang Patel
70713c75da Emit declaration DIE for the class static variables.
llvm-svn: 94870
2010-01-30 01:08:30 +00:00
Dale Johannesen
9774a29dc0 Fix a case where debug_value could affect codegen.
llvm-svn: 94866
2010-01-30 00:57:47 +00:00
Devang Patel
022d4a7fc0 Add size and location info in DW_TAG_class_type descriptor.
llvm-svn: 94822
2010-01-29 18:34:58 +00:00
Duncan Sands
6b277c2823 Change the SREM case to match the logic in the IR version ComputeMaskedBits.
llvm-svn: 94805
2010-01-29 09:45:26 +00:00
Bill Wendling
d408e7be4e Assign the ordering of SDNodes in a much less intrusive fashion. After the
"visit*" method is called, take the newly created nodes, walk them in a DFS
fashion, and if they don't have an ordering set, then give it one.

llvm-svn: 94757
2010-01-28 21:51:40 +00:00
Chris Lattner
641dab84e4 add target hooks for emitting random gunk before and after the function body.
llvm-svn: 94732
2010-01-28 01:58:58 +00:00
Jim Grosbach
7474937004 Update of 94055 to track the IR level call site information via an intrinsic.
This allows code gen and the exception table writer to cooperate to make sure
landing pads are associated with the correct invoke locations.

llvm-svn: 94726
2010-01-28 01:45:32 +00:00
Chris Lattner
95118672e3 Give AsmPrinter the most common expected implementation of
runOnMachineFunction, and switch PPC to use EmitFunctionBody.
The two ppc asmprinters now don't heave to define 
runOnMachineFunction.

llvm-svn: 94722
2010-01-28 01:28:58 +00:00
Chris Lattner
df1662f0e6 emit a 0 byte instead of a noop if a function is empty on darwin.
"0" is nice and target independent.

llvm-svn: 94718
2010-01-28 01:06:32 +00:00
Chris Lattner
8411a9084e Remove the argument from EmitJumpTableInfo, because it doesn't need it.
Move the X86 implementation of function body emission up to 
AsmPrinter::EmitFunctionBody, which works by calling the virtual
EmitInstruction method.

llvm-svn: 94716
2010-01-28 01:02:27 +00:00
Chris Lattner
6a127932c1 Drop the argument to AsmPrinter::EmitConstantPool and make it virtual.
Overload it in the ARM backend to do nothing, since is does insane
constant pool emission.

llvm-svn: 94708
2010-01-28 00:19:24 +00:00
Chris Lattner
de4b09beb9 rename printVisibility to EmitVisibility and make it private,
constify EmitLinkage.

llvm-svn: 94705
2010-01-28 00:05:10 +00:00
Dale Johannesen
74ab14a99f If the only use of something is a DEBUG_VALUE, don't
let that stop it from being deleted, and change the
DEBUG_VALUE value to undef.

llvm-svn: 94694
2010-01-27 22:12:36 +00:00
Chandler Carruth
866dc92123 Silence GCC warnings with asserts turned off. No functionality change.
llvm-svn: 94673
2010-01-27 10:27:10 +00:00