RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-24 19:52:54 +01:00
Code Issues Projects Releases Wiki Activity
42,768 Commits 4 Branches 5 Tags 704 MiB
ff89aeda21
Commit Graph

25864 Commits

Author SHA1 Message Date
Evan Cheng
c6f581034a ARM / Mac OS X also wants to invalidate icache after jitting. 
llvm-svn: 59291
 2008-11-14 02:33:17 +00:00
Dan Gohman
bab393fdff For post-regalloc scheduling, remove the instructions from the block 
before re-inserting them.

llvm-svn: 59281
 2008-11-14 00:33:17 +00:00
Dan Gohman
5b7b7b416b Check in the correct version of the patch in r59279. 
llvm-svn: 59280
 2008-11-14 00:32:34 +00:00
Dan Gohman
ad452993b0 Debug printing for SUnits that carry MachineInstrs. 
llvm-svn: 59279
 2008-11-14 00:28:56 +00:00
Dan Gohman
f92f2d0c97 Initial support for carrying MachineInstrs in SUnits. 
llvm-svn: 59278
 2008-11-14 00:06:09 +00:00
Dan Gohman
2e0edf0617 Change DOTGraphTraits<ScheduleDAG*>::getGraphName how to find the name of 
the current function on its own, rather than relying on the SelectionDAG.

llvm-svn: 59277
 2008-11-13 23:45:55 +00:00
Evan Cheng
f7c1da41a1 Handle the rest of pseudo instructions. 
llvm-svn: 59275
 2008-11-13 23:36:57 +00:00
Evan Cheng
feb7615798 Lazy compilation callback save / restore VFP registers. 
llvm-svn: 59274
 2008-11-13 23:28:54 +00:00
Dan Gohman
0a3ae5c0f2 Remove the FlaggedNodes member from SUnit. Instead of requiring each SUnit 
to carry a SmallVector of flagged nodes, just calculate the flagged nodes
dynamically when they are needed.

The local-liveness change is due to a trivial scheduling change where
the scheduler arbitrary decision differently.

llvm-svn: 59273
 2008-11-13 23:24:17 +00:00
Owen Anderson
aa9f1836d9 Don't allow the restore point to be placed after terminators. With this change, 
MultiSource/Applications is clean with the prealloc splitter.  Some failures
remain in SPEC.

llvm-svn: 59267
 2008-11-13 21:53:14 +00:00
Dale Johannesen
de9fdff884 Extend InlineAsm::C_Register to allow multiple specific registers 
(actually, code already all worked, only the comment
changed).  Use this to implement 'A' constraint on x86.
Fixes PR 1779.

llvm-svn: 59266
 2008-11-13 21:52:36 +00:00
Evan Cheng
f8ba919d76 Always emit a function pointer as a pointer to the function stub (if there is one). This makes it possible to compare function pointer values in lazy compilation mode. This fixes PR3043. 
llvm-svn: 59265
 2008-11-13 21:50:50 +00:00
Dan Gohman
0cefa8f2ff Make the Node member of SUnit private, and add accessors. 
llvm-svn: 59264
 2008-11-13 21:36:12 +00:00
Dan Gohman
df9f75e981 Change ScheduleDAG's DAG member from a reference to a pointer, to prepare 
for the possibility of scheduling without a SelectionDAG being present.

llvm-svn: 59263
 2008-11-13 21:21:28 +00:00
Evan Cheng
4c8297e12b Don't forget to emit stubs for function GV's emitted in CONSTPOOL_ENTRY's. 
llvm-svn: 59258
 2008-11-13 19:22:28 +00:00
Dan Gohman
f63a02ae6c Use find_first/find_next to iterate through all the set bits in a 
BitVector, instead of manually testing each bit.

llvm-svn: 59246
 2008-11-13 16:31:27 +00:00
Bill Wendling
9dc2bd7973 Modify the intrinsics pattern to separate out the "return" types from the 
"parameter" types. An intrinsic can now return a multiple return values like
this:

  def add_with_overflow : Intrinsic<[llvm_i32_ty, llvm_i1_ty],
                                    [LLVMMatchType<0>, LLVMMatchType<0>]>;

llvm-svn: 59237
 2008-11-13 09:08:33 +00:00
Evan Cheng
997eb67c20 fsub{d|s} encoding bugs. 
llvm-svn: 59234
 2008-11-13 07:59:48 +00:00
Evan Cheng
d67d6c4571 Missed a break statement. 
llvm-svn: 59231
 2008-11-13 07:46:59 +00:00
Evan Cheng
3c0113820b Fix pre- and post-indexed load / store encoding bugs. 
llvm-svn: 59230
 2008-11-13 07:34:59 +00:00
Bill Wendling
0eb23711fb Refactor the code that does the type checking for intrinsics. 
llvm-svn: 59228
 2008-11-13 07:11:27 +00:00
Daniel Dunbar
a830cb14f3 Add Binary flag to raw_fd_ostream constructor. 
Document raw_fd_ostream's treatment of "-".

llvm-svn: 59219
 2008-11-13 05:01:07 +00:00
Devang Patel
2876af73db Really remove all debug information. 
llvm-svn: 59208
 2008-11-13 01:28:40 +00:00
Bill Wendling
dedf359e39 Regenerated 
llvm-svn: 59204
 2008-11-13 01:03:00 +00:00
Bill Wendling
f5b264a64b Implement stack protectors as function attributes: "ssp" and "sspreq". 
llvm-svn: 59202
 2008-11-13 01:02:14 +00:00
Dan Gohman
229c65c05b Move the code that inserts X87 FP_REG_KILL instructions from a 
special-purpose hook to a new pass. Also, add check to see if any
x87 virtual registers are used, to avoid doing any work in the
common case that no x87 code is needed.

llvm-svn: 59190
 2008-11-12 22:55:05 +00:00
Evan Cheng
4768468ec1 Remove the incorrect assertion. We don't have enough information before relocation to set U bit. 
llvm-svn: 59170
 2008-11-12 21:37:59 +00:00
Oscar Fuentes
dc2008cb1a CMake: when bulding shared libraries on non-WIN32 systems, link dl to 
LLVMSystem.

llvm-svn: 59159
 2008-11-12 20:40:56 +00:00
Dale Johannesen
5ad71e113e Fix unsigned char->ppcf128 conversion. 
llvm-svn: 59150
 2008-11-12 18:38:44 +00:00
Dan Gohman
5f3e5a75b4 Do the LiveVariables update before printing the instruction in 
the debug output, so that the updated liveness flags are
reflected in the debug output.

llvm-svn: 59147
 2008-11-12 17:15:19 +00:00
Dan Gohman
ceb5b7477b Remove some debugging code made redundant by the change to do 
coalescing as a separate pass rather than inside of
LiveIntervalAnalysis.

llvm-svn: 59146
 2008-11-12 17:09:23 +00:00
Duncan Sands
4e70a21580 Simplify SplitVecRes_EXTRACT_SUBVECTOR. This means 
that it no longer handles non-power-of-two vectors.
However it previously only handled them sometimes,
depending on obscure numerical relationships between
the index and vector type.  For example, for a vector
of length 6, it would succeed if and only if the
index was an even multiple of 6.  I consider this
more confusing than useful.

llvm-svn: 59122
 2008-11-12 08:37:57 +00:00
Duncan Sands
117397c8dd Correct some thinkos in the expansion of ADD/SUB 
when the target does not support ADDC/SUBC.  This
fixes PR3044.

llvm-svn: 59120
 2008-11-12 08:23:26 +00:00
Evan Cheng
b6c760349e Change binary dump format. 
llvm-svn: 59119
 2008-11-12 08:22:43 +00:00
Evan Cheng
9cb6f40f8f Address mode immediate offset has already been divided by 4. 
llvm-svn: 59117
 2008-11-12 08:21:12 +00:00
Evan Cheng
80d0d5cfa8 Fix a VFP binary arithmetic instruction encoding bug. 
llvm-svn: 59116
 2008-11-12 08:14:21 +00:00
Evan Cheng
a4dc0e304e Fix address mode 3 immediate offset mode encoding. 
llvm-svn: 59109
 2008-11-12 07:34:37 +00:00
Evan Cheng
7fef8cfe75 Consolidate formats; fix FCMPED etc. encodings. 
llvm-svn: 59107
 2008-11-12 07:18:38 +00:00
Evan Cheng
0966883093 Fix VFP conversion instruction encodings. 
llvm-svn: 59104
 2008-11-12 06:41:41 +00:00
Evan Cheng
39d8bf295d Fix encoding of single-precision VFP registers. 
llvm-svn: 59102
 2008-11-12 02:19:38 +00:00
Dale Johannesen
a2cd0724ea Fix the testb optimization so x86 also bootstraps. 
Reenable test.

llvm-svn: 59101
 2008-11-12 02:00:35 +00:00
Evan Cheng
cc9323ac73 VFP fld / fst immediate field is multiplied by 4. 
llvm-svn: 59100
 2008-11-12 01:02:24 +00:00
Oscar Fuentes
444fd3cc2d CMake: Remove removed source file. 
llvm-svn: 59098
 2008-11-12 00:14:12 +00:00
Devang Patel
0a73799659 Remove 
llvm-svn: 59093
 2008-11-11 23:58:15 +00:00
Devang Patel
0616706919 Undo previous check-in. 
llvm-svn: 59092
 2008-11-11 23:57:33 +00:00
Andrew Lenharth
0e7b77369c This shouldn't be necessary 
llvm-svn: 59090
 2008-11-11 23:19:51 +00:00
Evan Cheng
8b7fefc126 Fix FMDRR encoding. 
llvm-svn: 59088
 2008-11-11 22:46:12 +00:00
Evan Cheng
95bad0db94 Handle floating point constpool_entry's. 
llvm-svn: 59087
 2008-11-11 22:19:31 +00:00
Owen Anderson
3ef1e1ff3f Don't walk into predecessors in which the vreg is not live when doing shrinkwrapping. 
This lets several failing tests get farther along, but doesn't completely fix any of them.

llvm-svn: 59086
 2008-11-11 22:11:27 +00:00
Evan Cheng
35f1477c5a Encode VFP load / store instructions. 
llvm-svn: 59084
 2008-11-11 21:48:44 +00:00
Dan Gohman
291ea92cc4 In ScheduleDAGRRList::CopyAndMoveSuccessors, create the SUnit for the load 
before creating the SUnit for the operation that it was unfolded from. This
allows each SUnit to have all of its predecessor SUnits available at the time
it is created. I don't know yet if this will be absolutely required, but it
is a little tidier to do it this way.

llvm-svn: 59083
 2008-11-11 21:34:44 +00:00
Dan Gohman
51db591eac Avoid relying on the SelectionDAG for initializing the MachineFunction and 
TargetLoweringInfo variables for the scheduler.

llvm-svn: 59082
 2008-11-11 21:31:56 +00:00
Oscar Fuentes
ac819d0394 CMake: Updated list of source files for lib/Transforms/Utils. 
llvm-svn: 59077
 2008-11-11 19:51:36 +00:00
Evan Cheng
91e4ef91c5 Encode VFP conversion instructions. 
llvm-svn: 59074
 2008-11-11 19:40:26 +00:00
Devang Patel
c9fdfd2f0b Add utility pass to remove dbg info. 
llvm-svn: 59068
 2008-11-11 19:33:39 +00:00
Devang Patel
9e6295b55c Silence unused variable warning. 
llvm-svn: 59064
 2008-11-11 19:17:41 +00:00
Devang Patel
ef6a48b56f Use actual function name in comments. 
llvm-svn: 59063
 2008-11-11 19:16:41 +00:00
Dan Gohman
ec3b31e7eb Change the scheduler accessor methods to accept an explicit TargetMachine 
argument instead of taking the SelectionDAG's TargetMachine. This is
needed for some upcoming scheduler changes.

llvm-svn: 59055
 2008-11-11 17:50:47 +00:00
Evan Cheng
82b17df2b0 Add a README entry. 
llvm-svn: 59052
 2008-11-11 17:35:52 +00:00
Oscar Fuentes
e4e9531262 CMake: corrected split of Alpha and Sparc AsmPrinters. 
llvm-svn: 59050
 2008-11-11 17:10:13 +00:00
Anton Korobeynikov
fe80f44b35 Separate sparc asmprinter. This should unbreak the native build 
llvm-svn: 59047
 2008-11-11 16:42:57 +00:00
Anton Korobeynikov
e3e227b40b Separate alpha asmprinter. This should unbreak native build. 
llvm-svn: 59046
 2008-11-11 16:42:17 +00:00
Dan Gohman
fd00e20872 The 32-bit displacement field in an x86 address is signed. Arrange for it 
to be sign-extended when it is promoted to 64 bits for intermediate
offset calculations. The offset calculations are done as uint64_t so that
overflow conditions are well defined.

This fixes a problem which is currently hidden by the x86 AsmPrinter but
which was exposed by r58917 (which is temporarily reverted).  See PR3027
for details.

llvm-svn: 59044
 2008-11-11 15:52:29 +00:00
Cedric Venet
49a701ae17 Update CMakeLists.txt 
llvm-svn: 59039
 2008-11-11 09:55:48 +00:00
Bill Wendling
9da4535062 Fix for PR3040: 
The CC was changed, but wasn't checked to see if it was legal if the DAG
combiner was being run after legalization. Threw in a couple of checks just to
make sure that it's okay. As far as the PR is concerned, no back-end target
actually exhibited this problem, so there isn't an associated testcase.

llvm-svn: 59035
 2008-11-11 08:25:46 +00:00
Andrew Lenharth
d096adcb5f fix another libgcc blocker 
llvm-svn: 59026
 2008-11-11 06:06:07 +00:00
Mon P Wang
c4bf9b94d5 Cleaned up and fix bugs in convert_rndsat node 
llvm-svn: 59025
 2008-11-11 05:40:06 +00:00
Scott Michel
2ab73cf730 Unbreak the buildbot and back out (inadvertant) casting edits in CellSPU 
backend.

llvm-svn: 59018
 2008-11-11 03:06:06 +00:00
Evan Cheng
3c62bb7471 Encode VFP arithmetic instructions. 
llvm-svn: 59016
 2008-11-11 02:11:05 +00:00
Devang Patel
fbb29cdaee Cleanup debug info. assocated with deleted instructions. 
llvm-svn: 59012
 2008-11-11 00:54:10 +00:00
Devang Patel
ec4ae29ae4 Add utility routines to remove dead debug info. 
llvm-svn: 59011
 2008-11-11 00:53:02 +00:00
Scott Michel
ecc5cc6238 CellSPU: Fix bug 3606, as well as some ongoing work. 
llvm-svn: 59009
 2008-11-10 23:43:06 +00:00
Bill Wendling
d3c149eb90 Small simplification. Use the iterator already present as the insertion point. 
llvm-svn: 59008
 2008-11-10 23:38:59 +00:00
Evan Cheng
b9dbae83d7 Comments and indentation. 
llvm-svn: 59007
 2008-11-10 23:26:16 +00:00
Evan Cheng
bc61b6009b Correct PIC function stub codegen. 
llvm-svn: 59006
 2008-11-10 23:14:47 +00:00
Dan Gohman
b5857c2859 Fix indentation. 
llvm-svn: 59004
 2008-11-10 22:09:58 +00:00
Bill Wendling
891f177dd0 Temporarily revert r58979 and related patch. It's causing a failure in X86 bootstrap: 
Comparing stages 2 and 3
warning: ./cc1-checksum.o differs
warning: ./cc1obj-checksum.o differs
warning: ./cc1objplus-checksum.o differs
warning: ./cc1plus-checksum.o differs
Bootstrap comparison failure!
./alias.o differs
./alloc-pool.o differs
./attribs.o differs
./bb-reorder.o differs
./bitmap.o differs
./build/errors.o differs
./build/genattrtab.o differs
./build/genautomata.o differs
./build/genemit.o differs
./build/genextract.o differs
...

-bw

llvm-svn: 59003
 2008-11-10 21:22:06 +00:00
Bill Wendling
0421da9e9e - Make sure that we don't over-increment the iterator when going through the 
basic blocks.
- Minor code clean-up.

llvm-svn: 59002
 2008-11-10 21:13:10 +00:00
Mon P Wang
6792115592 Added CONVERT_RNDSAT (conversion with rounding and saturation) SDNode to 
support targets that support these conversions. Users should avoid using
this node as the current targets don't generating code for it.

llvm-svn: 59001
 2008-11-10 20:54:11 +00:00
Devang Patel
f0d6bd18d5 If the sign of exit condition and split condition does not match 
then do not split loop index.

llvm-svn: 58995
 2008-11-10 19:48:34 +00:00
Duncan Sands
22e8a45a01 Fix PR2667: add soft float support for sint_to_fp/uint_to_fp 
where the argument is an apint, or smaller than the minimum
size for which there is a libcall (i32). 

llvm-svn: 58994
 2008-11-10 17:36:26 +00:00
Duncan Sands
eca6e696ca Tweak some comments. 
llvm-svn: 58993
 2008-11-10 17:31:56 +00:00
Duncan Sands
b6c3634c90 Small cleanups. No functionality change intended! 
llvm-svn: 58992
 2008-11-10 17:29:56 +00:00
Duncan Sands
1d0b7dccf7 When promoting the result of fp_to_uint/fp_to_sint, 
inform the optimizers that the result must be zero/
sign extended from the smaller type.  For example,
if a fp to unsigned i16 is promoted to fp to i32,
then we are allowed to assume that the extra 16 bits
are zero (because the result of fp to i16 is undefined
if the result does not fit in an i16).  This is
quite aggressive, but should help the optimizers
produce better code.  This requires correcting a
test which thought that fp_to_uint is some kind
of truncation, which it is not: in the testcase
(which does fp to i1), either the fp value converts
to 0 or 1 or the result is undefined, which is
quite different to truncation.

llvm-svn: 58991
 2008-11-10 17:28:30 +00:00
Dale Johannesen
8a43172ff1 Really fix testb optimization on big-endian. 
Fixes ppc32 bootstrap.

llvm-svn: 58979
 2008-11-10 07:16:42 +00:00
Bill Wendling
b7d5ca543e Third time's a charm. 
The previous patches didn't match correctly. Also, we need to make sure that
the conditional is the same before doing the transformation.

llvm-svn: 58978
 2008-11-10 06:59:06 +00:00
Mon P Wang
911ee5bf8b Added support for the following definition of shufflevector 
<result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> 

llvm-svn: 58964
 2008-11-10 04:46:22 +00:00
Chris Lattner
ed6022f676 eliminate a couple more uses of utohexstr. 
llvm-svn: 58963
 2008-11-10 04:35:24 +00:00
Chris Lattner
1185c9a72c Use utohex_buffer instead of utohexstr to avoid creating a temporary 
string in the .ll and .s printers.

llvm-svn: 58962
 2008-11-10 04:30:26 +00:00
Chris Lattner
694acd0d73 Move getCastToEmpty out of DIDescriptor into DIFactory. It is an 
implementation detail of DIFactory anyway, and this allows it to avoid
recomputing the same type over and over.

llvm-svn: 58960
 2008-11-10 04:10:34 +00:00
Chris Lattner
2fefd3dd69 Fix an over aggressive assumption that struct decls will have a pointer to 
an array that is empty.  Instead of requiring this array, allow a null pointer.
This shrinks all forward references of structs.

llvm-svn: 58959
 2008-11-10 03:48:55 +00:00
Chris Lattner
0d1da3eeaa add new file 
llvm-svn: 58955
 2008-11-10 02:58:28 +00:00
Chris Lattner
83d19fba9b Add a new set of helper classes for creating and reading debug 
information.  This logically replaces the "Desc" classes in
MachineModuleInfo.  Nice features of these classes are that they:

1. Are much more efficient than MMI because they don't create a 
   temporary parallel data structure for debug info that has to be
   'serialized' and 'deserialized' into/out of the module.
2. These provide a much cleaner abstraction for debug info than 
   MMI, which will make it easier to change the implementation in 
   the future (to be MDNode-based).
3. These are much easier to use than the MMI interfaces, requiring
   a lot less code in the front-ends.
4. These can be used to both create (for frontends) and read (for
   codegen) debug information.  DebugInfoBuilder can only be used
   to create the nodes.

So far, this is implemented just enough to support the debug info
generation needs of clang.  This can and should be extended to 
support the full set of debug info constructs, and we should switch
llvm-gcc and llc over to using this in the near future.

This code also has a ton of FIXMEs in it, because the way we 
currently represent debug info in LLVM IR is basically insane in a
variety of details.  This sort of issue should be fixed when we
eventually reimplement debug info on top of MDNodes.

llvm-svn: 58954
 2008-11-10 02:56:27 +00:00
Evan Cheng
e2bed5dc3b Forgot these. 
llvm-svn: 58952
 2008-11-10 01:52:24 +00:00
Evan Cheng
080c1a9fac Rename isGVNonLazyPtr to isIndirectSym to reflect how it will be used. 
llvm-svn: 58949
 2008-11-10 01:08:07 +00:00
Bill Wendling
137550d34d Correction for the last patch. Should match the conditional in the first part 
of the select match, not the select instruction itself.

llvm-svn: 58947
 2008-11-09 23:37:53 +00:00
Bill Wendling
3b91357ef0 The method of doing the matching with a 'select' instruction was wrong. The 
original code was matching like this:

	if (match(A, m_Not(m_Value(B))))

B was already matched as a 'select' instruction. However, this isn't matching
what we think it's matching. It would match B as a 'Value', so basically
anything would match to it. In this case, a Constant matched. B was replaced
with a constant representation. And then the wrong value would be used in the
SelectInst::Create statement, causing a crash.

After thinking on this for a moment, and after Nick L. told me how the pattern
matching stuff was supposed to work, the solution was to match NOT an m_Value,
but an m_Select.

llvm-svn: 58946
 2008-11-09 23:17:42 +00:00
Nuno Lopes
6dfcc8cf45 fix leakage of ValueNumbering 
llvm-svn: 58933
 2008-11-09 12:45:23 +00:00
Dale Johannesen
27c03be35e Temporarily revert 58825, which breaks PPC bootstrap. 
xs

llvm-svn: 58930
 2008-11-09 06:48:10 +00:00
Bill Wendling
436d4cce83 If the LHS of the FCMP is coming from a UIToFP instruction, then we don't want 
to generate signed ICMP instructions to replace the FCMP. This would violate
the following:

define i1 @test1(i32 %val) {
  %1 = uitofp i32 %val to double
  %2 = fcmp ole double %1, 0.000000e+00
  ret i1 %2
}

would be transformed into:

define i1 @test1(i32 %val) {
  %1 = icmp slt i33 %val, 1
  ret i1 %1
}

which is obviously wrong. This patch modifes InstCombiner::FoldFCmp_IntToFP_Cst
to handle when the LHS comes from UIToFP.

llvm-svn: 58929
 2008-11-09 04:26:50 +00:00
Anton Korobeynikov
cc2b340d11 Typo fix 
llvm-svn: 58928
 2008-11-09 02:54:13 +00:00
Anton Korobeynikov
a4933e408d Temporary revert my last commit: it seems it's triggering some subtle bug in backend 
and breaks llvm-gcc

llvm-svn: 58926
 2008-11-08 23:05:05 +00:00
Oscar Fuentes
4d83014ea6 CMake: corrected library target name for dependency: LLVMCellSPU -> 
LLVMCellSPUCodeGen.

llvm-svn: 58925
 2008-11-08 21:23:15 +00:00
Oscar Fuentes
a2b285ef48 CMake: Reflected changes on the CellSPU target build. May require a 
clean start.

llvm-svn: 58924
 2008-11-08 20:37:19 +00:00
Oscar Fuentes
486fd29325 Fixed a pasto. 
llvm-svn: 58923
 2008-11-08 20:34:18 +00:00
Scott Michel
4a683f438d CellSPU: Bring SPU's assembly printer more in-line with current LLVM code 
structure. Assembly printer now outputs the correct section for strings.

llvm-svn: 58921
 2008-11-08 18:59:02 +00:00
Duncan Sands
da4e03de04 Try to produce better code when scalarizing VSETCC. 
llvm-svn: 58920
 2008-11-08 18:26:48 +00:00
Anton Korobeynikov
e3b9284fa8 Factor out offset printing code into generic AsmPrinter. 
FIXME: it seems, that most of targets don't support
offsets wrt CPI/GlobalAddress', was it intentional?

llvm-svn: 58917
 2008-11-08 17:21:38 +00:00
Nicolas Geoffray
76100fd4de The Index field of an AttributeWithIndex is of type unsigned, not uint16_t. 
llvm-svn: 58908
 2008-11-08 15:36:01 +00:00
Anton Korobeynikov
3af4269671 StoreInst does not produce any result thus it's useless to create new 
variable for it. This greatly reduces amount of unused variables in
llvm2cpp-generated code

llvm-svn: 58905
 2008-11-08 12:58:07 +00:00
Evan Cheng
e5b6b3f81f Moved InvalidateInstructionCache to ARMJITInfo::emitFunctionStub which knows size of stub. 
llvm-svn: 58899
 2008-11-08 08:16:49 +00:00
Evan Cheng
7f1eececb4 Remove a InvalidateInstructionCache call with incorrect size. 
llvm-svn: 58898
 2008-11-08 08:15:39 +00:00
Evan Cheng
3d9c134136 Rename startFunctionStub to startGVStub since it's also used for GV non-lazy ptr. 
llvm-svn: 58897
 2008-11-08 08:02:53 +00:00
Evan Cheng
cfa57662d1 Tell ARMJITInfo if codegen relocation is PIC. It changes how function stubs are generated. 
llvm-svn: 58896
 2008-11-08 07:38:22 +00:00
Evan Cheng
fe6df6ae18 Rename isString -> isExternalSymbol; getString -> getExternalSymbol since these work on externsym machine relocations. 
llvm-svn: 58895
 2008-11-08 07:37:34 +00:00
Evan Cheng
8dfc67205a More debug output. 
llvm-svn: 58894
 2008-11-08 07:22:53 +00:00
Evan Cheng
f2ebaa6a64 Fix relocation for calls to external symbols. 
llvm-svn: 58893
 2008-11-08 07:22:33 +00:00
Scott Michel
70cc5f61b2 CellSPU: Fix prologue/epilogue emission when function contains calls but 
theframe size is 0; the prologue and epilogue should be emitted in this case.

llvm-svn: 58890
 2008-11-08 05:16:20 +00:00
Daniel Dunbar
ce947e0039 Rework r58829, allowing removal of dbg info intrinsics during alloca 
promotion.
 - Eliminate uses after free and simplify tests.

Devang: Please check that this is still doing what you intended.
llvm-svn: 58887
 2008-11-08 04:12:17 +00:00
Evan Cheng
064befa37b Skip over two-address use operands. 
llvm-svn: 58883
 2008-11-08 01:44:13 +00:00
Evan Cheng
bc946b09aa Handle ARM machine constantpool entry with non-lazy ptr. 
llvm-svn: 58882
 2008-11-08 01:31:27 +00:00
Evan Cheng
2ea87890bb Use ARMFunctionInfo to track number of constpool entries and jumptables. 
llvm-svn: 58877
 2008-11-08 00:51:41 +00:00
Dale Johannesen
e0608af6a4 Make testb optimization work on big-endian targets. 
llvm-svn: 58874
 2008-11-08 00:01:16 +00:00
Evan Cheng
bf8a1ef40f More code clean up. 
llvm-svn: 58872
 2008-11-07 22:57:53 +00:00
Dale Johannesen
bc914a7cf9 Make FP tests requiring two compares work on PPC (PR 642). 
This is Chris' patch from the PR, modified to realize that
SETUGT/SETULT occur legitimately with integers, plus
two fixes in LegalizeDAG to pass a valid result type into
LegalizeSetCC.  The argument of TLI.getSetCCResultType is
ignored on PPC, but I think I'm following usage elsewhere.

llvm-svn: 58871
 2008-11-07 22:54:33 +00:00
Evan Cheng
3e00dcfebb Get PIC jump table working. 
llvm-svn: 58869
 2008-11-07 22:30:53 +00:00
Evan Cheng
237175e1ab More debug output. 
llvm-svn: 58868
 2008-11-07 22:30:29 +00:00
Duncan Sands
60220e7127 Sign-extend rather than zero-extend when promoting 
the condition for a BRCOND, according to what is
returned by getSetCCResultContents.  Since all
targets return the same thing (ZeroOrOneSetCCResult),
this should be harmless!  The point is that all over
the place the result of SETCC is fed directly into
BRCOND.  On machines for which getSetCCResultContents
returns ZeroOrNegativeOneSetCCResult, this is a
sign-extended boolean.  So it seems dangerous to
also feed BRCOND zero-extended booleans in some
circumstances - for example, when promoting the
condition.

llvm-svn: 58861
 2008-11-07 20:13:04 +00:00
Dan Gohman
1418c52e1a Flush the raw_ostream after emitting the assembly for a function. 
This is a temporary fix for the -print-emitted-asm option, where
errs() is used as the stream, in the case where other code is
using stderr without using errs()' buffer. Hopefully soon we'll
fix errs() to be non-buffered instead. Patch by Preston Gurd.

llvm-svn: 58859
 2008-11-07 19:49:17 +00:00
Dale Johannesen
4510149fd2 Fix unsigned->ppcf128 conversion. 
llvm-svn: 58856
 2008-11-07 19:11:43 +00:00
Richard Osborne
224268f88b Fix compile warnings. 
llvm-svn: 58840
 2008-11-07 11:21:09 +00:00
Scott Michel
b5f73899a2 CellSPU: Ensure that C strings are always put in the .rodata section 
llvm-svn: 58839
 2008-11-07 11:06:44 +00:00
Richard Osborne
acc7a27e24 Add XCore backend. 
llvm-svn: 58838
 2008-11-07 10:59:00 +00:00
Evan Cheng
21df9f3b4f Jump table JIT support. Work in progress. 
llvm-svn: 58836
 2008-11-07 09:06:08 +00:00
Evan Cheng
6e23b1e651 Jump tables may be emitted by target. 
llvm-svn: 58835
 2008-11-07 09:02:17 +00:00
Scott Michel
828731190c Teach CellSPU about ELF sections and new section emitter classes. 
NB: This is likely to need more work.
llvm-svn: 58832
 2008-11-07 04:36:25 +00:00
Bill Wendling
a5eb92b137 BCUI + 1 doesn't work. Use next instead. 
llvm-svn: 58830
 2008-11-07 01:59:41 +00:00
Bill Wendling
dea20bc52f Refactor code that adjusts the offsets of stack objects. 
llvm-svn: 58829
 2008-11-07 01:48:58 +00:00
Evan Cheng
88726d85eb Encode misc arithmetic instructions. 
llvm-svn: 58828
 2008-11-07 01:41:35 +00:00
Devang Patel
0958cd6437 Handle (delete) dbg intrinsics while promoting alloca. 
llvm-svn: 58826
 2008-11-07 01:30:07 +00:00
Dale Johannesen
89da8440e1 When we're doing a compare of load-AND-constant to 0 
(e.g. a bitfield test) narrow the load as much as possible.
The has the potential to avoid unnecessary partial-word
load-after-store conflicts, which cause stalls on several targets.
Also a size win on x86 (testb vs testl).

llvm-svn: 58825
 2008-11-07 01:28:02 +00:00
Bill Wendling
3fe9fef0da - Modify the stack protector algorithm so that the stack slot is allocated in 
LLVM IR code and not in the selection DAG ISel. This is a cleaner solution.

- Fix the heuristic for determining if protectors are necessary. The previous
  one wasn't checking the proper type size.

llvm-svn: 58824
 2008-11-07 01:23:58 +00:00
Bill Wendling
54a5c87823 Remove unneeded header file. 
llvm-svn: 58823
 2008-11-06 23:56:59 +00:00
Bill Wendling
70e06738b4 Don't build a vector of returns. Just modify the Function in the loop. 
llvm-svn: 58822
 2008-11-06 23:55:49 +00:00
Mon P Wang
888f4e6fb0 Fixed scalarizing an extract subvector and prevent an infinite loop 
when simplify a vector. 

llvm-svn: 58820
 2008-11-06 22:52:21 +00:00
Bill Wendling
1e6576e1e0 The size limit is for individual arrays. So if any array has more than 8 bytes 
in it, then emit stack protectors.

llvm-svn: 58819
 2008-11-06 22:18:44 +00:00
Evan Cheng
3bcb71912f Encode extend instructions; more clean up. 
llvm-svn: 58818
 2008-11-06 22:15:19 +00:00
Dan Gohman
1130e8c23f Fix a use of an invalid iterator when -debug-pass=Details is used. 
llvm-svn: 58816
 2008-11-06 21:57:17 +00:00
Bill Wendling
26eadae94d Don't recalculate the stack position of the stack protector. 
llvm-svn: 58815
 2008-11-06 21:37:09 +00:00
Devang Patel
8640fd500a Emit label for llvm.dbg.func.start of the inlined function. 
llvm-svn: 58814
 2008-11-06 21:28:20 +00:00
Evan Cheng
af54e4ed18 - Improve naming consistency: Branch -> BrFrm, BranchMisc -> BrMiscFrm. 
- Consolidate instruction formats.
- Other clean up.

llvm-svn: 58808
 2008-11-06 17:48:05 +00:00
Evan Cheng
1ea366a419 Improve JIT debugging outputs format consistency. 
llvm-svn: 58807
 2008-11-06 17:46:04 +00:00
Nuno Lopes
e5fb2e4d84 plug leakage of mutex data. pthread_mutex_destroy() doesnt free our malloc'ed memory. 
llvm-svn: 58805
 2008-11-06 16:21:49 +00:00
Duncan Sands
50acaf2367 Formating/comment changes - no functionality change. 
llvm-svn: 58801
 2008-11-06 08:51:32 +00:00
Evan Cheng
aa24d19533 Remove opcode from instruction TS flags; add MOVCC support; fix addrmode3 encoding bug. 
llvm-svn: 58800
 2008-11-06 08:47:38 +00:00
Bill Wendling
b6e2d60e7a - Rename stackprotector_{prologue,epilogue} to stackprotector_{create,check}. 
- Get rid of "HasStackProtector" in MachineFrameInfo.
- Modify intrinsics to tell which are doing what with memory.

llvm-svn: 58799
 2008-11-06 07:23:03 +00:00
Mon P Wang
41f90a3ee5 Widening cleanup 
llvm-svn: 58796
 2008-11-06 05:31:54 +00:00
Evan Cheng
078361bddc Handle smul<x><y>, smulw<y>, smla<x><y>, smlaw<y>. 
llvm-svn: 58793
 2008-11-06 03:35:07 +00:00
Bill Wendling
489791a127 Adjust the stack protector heuristic to care about only arrays or calls to 
"alloca".

llvm-svn: 58792
 2008-11-06 02:38:58 +00:00
Bill Wendling
08905ed703 Implement the stack protector stack accesses via intrinsics: 
- stackprotector_prologue creates a stack object and stores the guard there.

- stackprotector_epilogue reads the stack guard from the stack position created
  by stackprotector_prologue.

- The PrologEpilogInserter was changed to make sure that the stack guard is
  first on the stack frame.

llvm-svn: 58791
 2008-11-06 02:29:10 +00:00
Evan Cheng
058721d10b Fix so_imm encoding bug; add support for MOVi2pieces. 
llvm-svn: 58790
 2008-11-06 02:25:39 +00:00
Evan Cheng
ca6759021b Fix encoding of multiple instructions with 3 src operands; also handle smmul, smmla, and smmls. 
llvm-svn: 58789
 2008-11-06 01:21:28 +00:00
Evan Cheng
3e727721d0 Need a \n. 
llvm-svn: 58788
 2008-11-06 01:18:29 +00:00
Devang Patel
cea9dfa11f InstructionNamer preserves everything. 
llvm-svn: 58787
 2008-11-06 01:00:16 +00:00
Devang Patel
ec135e1f33 Emit label for llvm.dbg.func.start of the inlined function. 
llvm-svn: 58786
 2008-11-06 00:30:09 +00:00
Evan Cheng
7390b46342 Undo 58778 but makes the binary dump prettier. 
llvm-svn: 58782
 2008-11-05 23:44:08 +00:00
Bill Wendling
538cbf1e09 Add comments to function. 
llvm-svn: 58781
 2008-11-05 23:42:27 +00:00
Evan Cheng
ce97712aa6 Encode pic load / store instructions; fix some encoding bugs. 
llvm-svn: 58780
 2008-11-05 23:22:34 +00:00
Evan Cheng
575f3150cf Remove debug output that's not really useful. 
llvm-svn: 58778
 2008-11-05 23:21:11 +00:00
Evan Cheng
9970c31dcf Restructure ARM code emitter to use instruction formats instead of addressing modes to determine how to encode instructions. 
llvm-svn: 58764
 2008-11-05 18:35:52 +00:00
Dan Gohman
1db84e57c5 Reintroduce a comment that was removed with the AddToISelQueue 
changes.

llvm-svn: 58760
 2008-11-05 17:16:24 +00:00
Richard Osborne
efd7edc731 Test commit, add Makefile for XCore target, more to follow. 
llvm-svn: 58755
 2008-11-05 09:53:58 +00:00
Duncan Sands
f56e2fb5c2 Fix thinko in ppcf128 expansion of truncating store. 
llvm-svn: 58753
 2008-11-05 07:17:27 +00:00
Evan Cheng
1378d6c7a9 Add more vector move low and zero-extend patterns. 
llvm-svn: 58752
 2008-11-05 06:04:51 +00:00
Evan Cheng
1bde698192 Type of shuffle mask has changed. 
llvm-svn: 58751
 2008-11-05 06:04:18 +00:00
Evan Cheng
fdd6d65e39 Indentation. 
llvm-svn: 58750
 2008-11-05 06:03:38 +00:00
Dan Gohman
cd4b68bee9 Eliminate the ISel priority queue, which used the topological order for a 
priority function. Instead, just iterate over the AllNodes list, which is
already in topological order. This eliminates a fair amount of bookkeeping,
and speeds up the isel phase by about 15% on many testcases.

The impact on most targets is that AddToISelQueue calls can be simply removed.

In the x86 target, there are two additional notable changes.

The rule-bending AND+SHIFT optimization in MatchAddress that creates new
pre-isel nodes during isel is now a little more verbose, but more robust.
Instead of either creating an invalid DAG or creating an invalid topological
sort, as it has historically done, it can now just insert the new nodes into
the node list at a position where they will be consistent with the topological
ordering.

Also, the address-matching code has logic that checked to see if a node was
"already selected". However, when a node is selected, it has all its uses
taken away via ReplaceAllUsesWith or equivalent, so it won't recieve any
further visits from MatchAddress. This code is now removed.

llvm-svn: 58748
 2008-11-05 04:14:16 +00:00
Dan Gohman
215587186e Use getTargetConstant instead of getConstant for nodes that should not be visited 
by isel and potentially forced into registers.

llvm-svn: 58747
 2008-11-05 02:06:09 +00:00
Evan Cheng
2702e22b83 Rename isGVLazyPtr to isGVNonLazyPtr relocation. This represents Mac OS X 
indirect gv reference. Please don't call it lazy.

llvm-svn: 58746
 2008-11-05 01:50:32 +00:00
Devang Patel
de9b95965a Do now allow InlineAlways pass to remove dead functions. 
llvm-svn: 58744
 2008-11-05 01:39:16 +00:00
Devang Patel
b55a78e553 Silence unused variable warnings. 
llvm-svn: 58743
 2008-11-05 01:37:40 +00:00
Devang Patel
bc6ef4c191 Check Attribute::NoInline. 
llvm-svn: 58742
 2008-11-05 01:37:05 +00:00
Bill Wendling
986e386794 Remove dead variable. 
llvm-svn: 58741
 2008-11-05 00:56:35 +00:00
Bill Wendling
8c86d20576 Simplify the allocated size calculation. 
llvm-svn: 58740
 2008-11-05 00:54:27 +00:00
Bill Wendling
2461aaa183 Fix comment 
llvm-svn: 58739
 2008-11-05 00:46:15 +00:00
Owen Anderson
df29b0d7b2 Use the new predicate to control when we do prealloc splitting. Fix a small bug. 
llvm-svn: 58738
 2008-11-05 00:32:13 +00:00
Evan Cheng
8a5c3c5cc3 Debugging output tweak. 
llvm-svn: 58737
 2008-11-05 00:22:28 +00:00
Oscar Fuentes
afb65a6c2a CMake: updated list of source files. 
llvm-svn: 58736
 2008-11-05 00:11:22 +00:00
Bill Wendling
e73f31f526 Some code simplification. It now doesn't generate a prologue if the epilogue 
isn't going to be generated.

llvm-svn: 58734
 2008-11-05 00:00:21 +00:00
Dan Gohman
c162a200ad Add a new pass to simplify specific half_powr function calls. This is 
a specialized pass that it not likely to be generally useful.

llvm-svn: 58732
 2008-11-04 23:41:45 +00:00
Nuno Lopes
a5d766c285 fix memory leak in pass manager when adding an analysis pass that already existed. as pass manager takes ownership of the added passes, it has to delete the pass if it isnt added to the pass list 
tweak the opt tool so that it doesnt access a Pass after the ownership was taken by the pass manager

llvm-svn: 58730
 2008-11-04 23:03:58 +00:00
Bill Wendling
214f515922 Small simplification of the stack guard type. 
llvm-svn: 58728
 2008-11-04 22:54:43 +00:00
Bill Wendling
79a8798e07 - Add a "getOrInsertGlobal" method to the Module class. This acts similarly to 
"getOrInsertFunction" in that it either adds a new declaration of the global
  and returns it, or returns the current one -- optionally casting it to the
  correct type.
- Use the new getOrInsertGlobal in the stack protector code.
- Use "splitBasicBlock" in the stack protector code.

llvm-svn: 58727
 2008-11-04 22:51:24 +00:00
Owen Anderson
2ed3bc9016 First pass at checking for the creation of a new join point when doing pre-alloc splitting. This is not turned on yet. 
llvm-svn: 58726
 2008-11-04 22:22:41 +00:00
Evan Cheng
59112bc108 Actually ARM / Mac OS X does have UINTTOFP_I64_F{64|32} libcalls. 
llvm-svn: 58725
 2008-11-04 22:19:55 +00:00
Bill Wendling
ae168b2c83 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.

llvm-svn: 58723
 2008-11-04 21:53:09 +00:00
Dale Johannesen
e51c5296a6 Allow SROA of vectors. Removing this caused a 
huge performance regression in something we care
about.  This may not be final fix.

llvm-svn: 58718
 2008-11-04 20:54:03 +00:00
Dale Johannesen
eee3a8a2e0 80 columns 
llvm-svn: 58717
 2008-11-04 20:52:49 +00:00
Evan Cheng
45496b349f Custom lower bit_convert i64 -> f64 into FMDRR. This is now happening with legalizetypes. 
llvm-svn: 58714
 2008-11-04 19:57:48 +00:00
Duncan Sands
aed2dfe3f6 Fix typo. Patch by nlewycky. 
llvm-svn: 58709
 2008-11-04 18:05:30 +00:00
Evan Cheng
d63b7563b7 Debug output tweak. 
llvm-svn: 58708
 2008-11-04 17:58:53 +00:00
Evan Cheng
088f7c51a4 LDM_RET restores pc, do not set 's' bit which would restore CPSR from SPSR. 
llvm-svn: 58707
 2008-11-04 17:57:07 +00:00
Duncan Sands
58ebf09772 Fix PR3011: LegalizeTypes support for scalarizing 
SELECT_CC.

llvm-svn: 58706
 2008-11-04 17:31:08 +00:00
Dan Gohman
50183113b9 Duncan pointed out that the Extended case in getTypeForMVT could 
be considerably simplified.

llvm-svn: 58703
 2008-11-04 16:19:44 +00:00
Dan Gohman
13237dda2d Add a return statement to suppress warnings in NDEBUG builds. 
llvm-svn: 58702
 2008-11-04 16:08:57 +00:00
Dan Gohman
2dff90b025 Add some asserts to verify MVT invariant assumptions. 
llvm-svn: 58701
 2008-11-04 16:03:56 +00:00
Nuno Lopes
b16d9bd0c3 regenerate 
llvm-svn: 58697
 2008-11-04 14:43:20 +00:00
Nuno Lopes
064bb29c9d fix leakage of APFloats in getExistingVal() 
llvm-svn: 58696
 2008-11-04 14:42:19 +00:00
Nuno Lopes
8499bc5e1d regenerate 
llvm-svn: 58694
 2008-11-04 14:28:33 +00:00
Nuno Lopes
5caf466051 fix leakage of APSInt in getVal() 
llvm-svn: 58693
 2008-11-04 14:26:58 +00:00
Nuno Lopes
0995eae6b8 fix leakage of IfcvtTokens 
llvm-svn: 58690
 2008-11-04 13:02:59 +00:00
Evan Cheng
28e234a959 For some targets, it's not possible to place GVs in the same memory buffer as the MachineCodeEmitter allocated memory. Code and data has different read / write / execution privilege requirements. 
This is a short term workaround. The current solution is for the JIT memory manager to manage code and data memory separately.

llvm-svn: 58688
 2008-11-04 09:30:48 +00:00
Evan Cheng
7847f4f030 80 col violation. 
llvm-svn: 58684
 2008-11-04 06:10:31 +00:00
Evan Cheng
f15a9cfb31 Stylistic change. 
llvm-svn: 58683
 2008-11-04 06:10:06 +00:00
Evan Cheng
0369860450 hasDisassembler should return false if disassembler isn't available. 
llvm-svn: 58682
 2008-11-04 06:09:38 +00:00
Oscar Fuentes
9f6d4e7fb0 CMake: Updated list of source files. 
llvm-svn: 58676
 2008-11-04 03:24:04 +00:00
Bill Wendling
0f3f36688b 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.

llvm-svn: 58673
 2008-11-04 02:10:20 +00:00
Evan Cheng
f117632c3f Handle ARM machine constantpool entries. 
llvm-svn: 58671
 2008-11-04 00:50:32 +00:00
Dan Gohman
0ba8aad1af The ANDMask node folds to a constant, and isn't the node that needs to 
have its node id set. The new and and shift nodes are the nodes that need
the IDs. This fixes PR2982.

llvm-svn: 58655
 2008-11-03 23:43:55 +00:00
Devang Patel
09aea7e93e Fix unused variable warnings. 
llvm-svn: 58653
 2008-11-03 23:20:04 +00:00
Devang Patel
40cfcf7ce2 Fix unused variable warnings. 
llvm-svn: 58651
 2008-11-03 23:14:09 +00:00
Dan Gohman
3258c45d0f Add C bindings for extractvalue and insertvalue. Patch by Frits van Bommel! 
llvm-svn: 58650
 2008-11-03 22:55:43 +00:00
Evan Cheng
999398c004 Remove a dead switch statement. 
llvm-svn: 58644
 2008-11-03 21:26:52 +00:00
Evan Cheng
b3fd30ed7c Minor code restructuring. No functionality change. 
llvm-svn: 58643
 2008-11-03 21:02:39 +00:00
Dale Johannesen
d9906b90d0 Fix some ppcf128 regressions: make ExpandFloatRes_LOAD 
work correctly, and bring over a late change to ppcf128
SetCC handling.

llvm-svn: 58642
 2008-11-03 20:47:45 +00:00
Duncan Sands
8a94be8c5b Make VAARG promotion work correctly with large funky 
sized integers like i129, and also reduce the number
of assumptions made about how vaarg is implemented.
This still doesn't work correctly for small integers
like (eg) i1 on x86, since x86 passes each of them
(essentially an i8) in a 4 byte stack slot, so the
pointer needs to be advanced by 4 bytes not by 1 byte
as now.  But this is no longer a LegalizeTypes problem
(it was also wrong in LT before): it is a bug in the
operation expansion in LegalizeDAG: now LegalizeTypes
turns an i1 vaarg into an i8 vaarg which would work
fine if only the i8 vaarg was turned into correct code
later.

llvm-svn: 58635
 2008-11-03 20:22:12 +00:00
Dan Gohman
1532a1621c Overload AddInteger on int/long/long long instead of on int/int64_t, 
to avoid overload ambiguities. This fixes build errors introduced
by r58623.

llvm-svn: 58632
 2008-11-03 19:40:18 +00:00
Devang Patel
c959b4dbd3 Ignore conditions that are outside the loop. 
llvm-svn: 58631
 2008-11-03 19:38:07 +00:00
Andrew Lenharth
97343f0e8b add a period at the end of the comment, ignoring the fact that the comment would be hard pressed to be considered a sentence, but if it makes Bill happy... 
llvm-svn: 58630
 2008-11-03 19:29:29 +00:00
Jim Grosbach
5262898365 Add binary encoding support for multiply instructions. Some blanks left to fill in, but the basics are there. 
llvm-svn: 58626
 2008-11-03 18:38:31 +00:00
Devang Patel
78b7de25d1 Turn floating point IVs into integer IVs where possible. 
This allows SCEV users to effectively calculate trip count.
LSR later on transforms back integer IVs to floating point IVs
later on to avoid int-to-float casts inside the loop.

llvm-svn: 58625
 2008-11-03 18:32:19 +00:00
Dan Gohman
155df8a79e Refactor various TargetAsmInfo subclasses' TargetMachine members away 
adding a TargetMachine member to the base TargetAsmInfo class instead.

llvm-svn: 58624
 2008-11-03 18:22:42 +00:00
Dan Gohman
edf3dc97c2 Change how extended types are represented in MVTs. Instead of fiddling 
bits, use a union of a SimpleValueType enum and a regular Type*.

This increases the size of MVT on 64-bit hosts from 32 bits to 64 bits.
In most cases, this doesn't add significant overhead. There are places
in codegen that use arrays of MVTs, so these are now larger, but
they're small in common cases.

This eliminates restrictions on the size of integer types and vector
types that can be represented in codegen. As the included testcase
demonstrates, it's now possible to codegen very large add operations.
There are still some complications with using very large types. PR2880
is still open so they can't be used as return values on normal targets,
there are no libcalls defined for very large integers so operations
like multiply and divide aren't supported.

This also introduces a minimal tablgen Type library, capable of
handling IntegerType and VectorType. This will allow parts of
TableGen that don't depend on using SimpleValueType values to handle
arbitrary integer and vector types.

llvm-svn: 58623
 2008-11-03 17:56:27 +00:00
Andrew Lenharth
312c00cdd4 Ensure that we are checking only calls to the function we are interested in specializing 
llvm-svn: 58615
 2008-11-03 16:05:35 +00:00
Duncan Sands
a9047944bc Make VAARG work with x86 long double (which is 
10 bytes long, but is passed in 12/16 bytes).

llvm-svn: 58608
 2008-11-03 11:51:11 +00:00
Matthijs Kooijman
a91b759ccf Make MachineFrameInfo::print not crash when no TargetFrameInfo is available. 
llvm-svn: 58606
 2008-11-03 11:16:43 +00:00
Evan Cheng
45516a94c1 Silence a compiler warning. 
llvm-svn: 58598
 2008-11-03 07:14:02 +00:00
Nick Lewycky
f393d56364 Replace explicit loop with utility function. 
llvm-svn: 58593
 2008-11-03 03:49:14 +00:00
Nick Lewycky
cc1b7622a5 Don't crash analyzing certain quadratics (addrec of {X,+,Y,+,1}). 
We're still waiting on code that actually analyzes them properly.

llvm-svn: 58592
 2008-11-03 02:43:49 +00:00
Owen Anderson
6d82fd0e8e Revert my last patch until I consult with Evan about it. 
llvm-svn: 58591
 2008-11-03 02:33:28 +00:00
Nick Lewycky
49abbde699 Changes from Duncan's review: 
* merge two weak functions by making them both alias a third non-weak fn
 * don't reimplement CallSite::hasArgument
 * whitelist the safe linkage types

llvm-svn: 58568
 2008-11-02 16:46:26 +00:00
Anton Korobeynikov
9e3342fcdf Silence a warning 
llvm-svn: 58563
 2008-11-02 11:47:11 +00:00
Duncan Sands
613d060c2e Get this building on 64 bit machines (error: 
cast from ‘const llvm::PointerType*’ to ‘unsigned int’
loses precision).

llvm-svn: 58561
 2008-11-02 09:00:33 +00:00
Owen Anderson
146d114669 Don't do pre-splitting if doing so would create a value join that did not 
exist before.  Updating the live intervals in that care is tricky in the general
case.

Evan, if you see a tighter guard condition for this, let me know.

llvm-svn: 58560
 2008-11-02 08:08:18 +00:00
Oscar Fuentes
45b63793cf CMake: added a source file. 
llvm-svn: 58559
 2008-11-02 06:01:39 +00:00
Nick Lewycky
15a23d029c Add a new MergeFunctions pass. It finds identical functions and merges them. 
This triggers only 60 times in llvm-test (look at .llvm.bc, not .linked.rbc)
and so it probably wont be turned on by default. Also, may of those are likely
to go away when PR2973 is fixed.

llvm-svn: 58557
 2008-11-02 05:52:50 +00:00
Nick Lewycky
bcadcbb1ec Fix demanded bits analysis with srem by negative number. Based on a patch 
by Richard Osborne.

llvm-svn: 58555
 2008-11-02 02:41:50 +00:00
Dan Gohman
1f1ebc5389 Fix this recently moved code to use the correct type. CI is now a 
ConstantInt, and SI is the original cast instruction. This fixes
PR2996.

llvm-svn: 58549
 2008-11-02 00:17:33 +00:00
Mon P Wang
0d137a1c51 Added interface to allow clients to create a MemIntrinsicNode for 
target intrinsics that touches memory

llvm-svn: 58548
 2008-11-01 20:24:53 +00:00
Sanjiv Gupta
945942c4ad Reverting back 58505. Will commit it once I have the bc reader/writer/docs 
ready.

llvm-svn: 58547
 2008-11-01 10:57:12 +00:00