Commit Graph

1733 Commits

Author SHA1 Message Date
Evan Cheng b9e3db67fb Estimate a cost using the possible number of scratch registers required and use
it as a late BURR scheduling tie-breaker.
Intuitively, it's good to push down instructions whose results are liveout so
their long live ranges won't conflict with other values which are needed inside
the BB. Further prioritize liveout instructions by the number of operands which
are calculated within the BB.

llvm-svn: 35109
2007-03-14 22:43:40 +00:00
Evan Cheng 2874855302 Try schedule def + use closer whne Sethi-Ullman numbers are the same.
e.g.
t1 = op t2, c1
t3 = op t4, c2
and the following instructions are both ready.
t2 = op c3
t4 = op c4

Then schedule t2 = op first.
i.e.
t4 = op c4
t2 = op c3
t1 = op t2, c1
t3 = op t4, c2

This creates more short live intervals which work better with the register
allocator.

llvm-svn: 35089
2007-03-13 23:25:11 +00:00
Evan Cheng b7004fd889 More flexible TargetLowering LSR hooks for testing whether an immediate is a legal target address immediate or scale.
llvm-svn: 35076
2007-03-12 23:37:10 +00:00
Chris Lattner ce8aba03ee implement support for floating point constants used as inline asm memory operands.
llvm-svn: 35033
2007-03-08 22:29:47 +00:00
Chris Lattner b7bc3f2d30 make this fail even in non-assert builds.
llvm-svn: 35025
2007-03-08 07:07:03 +00:00
Anton Korobeynikov ed4b303c10 Refactoring of formal parameter flags. Enable properly use of
zext/sext/aext stuff.

llvm-svn: 35008
2007-03-07 16:25:09 +00:00
Evan Cheng 8a1d09d079 Avoid combining indexed load further.
llvm-svn: 35005
2007-03-07 08:07:03 +00:00
Chris Lattner 13780ac7db big endian 32-bit systems (e.g. ppc32) want to return the high reg first, not
the lo-reg first.  This is fallout from my ppc calling conv change yesterday,
it fixes test/ExecutionEngine/2003-05-06-LivenessClobber.llx

llvm-svn: 34983
2007-03-06 20:01:06 +00:00
Anton Korobeynikov f0b9316552 Enumerate SDISel formal parameter attributes. Make use of new
enumeration.

llvm-svn: 34960
2007-03-06 06:10:33 +00:00
Jeff Cohen b622c11f77 Unbreak VC++ build.
llvm-svn: 34917
2007-03-05 00:00:42 +00:00
Chris Lattner 47206667c0 fold away addc nodes when we know there cannot be a carry-out.
llvm-svn: 34913
2007-03-04 20:40:38 +00:00
Chris Lattner 2dcc6e7f58 generalize
llvm-svn: 34910
2007-03-04 20:08:45 +00:00
Chris Lattner e2e13caeb2 canonicalize constants to the RHS of addc/adde. If nothing uses the carry out of
addc, turn it into add.

This allows us to compile:

long long test(long long A, unsigned B) {
  return (A + ((long long)B << 32)) & 123;
}

into:

_test:
        movl $123, %eax
        andl 4(%esp), %eax
        xorl %edx, %edx
        ret

instead of:
_test:
        xorl %edx, %edx
        movl %edx, %eax
        addl 4(%esp), %eax   ;; add of zero
        andl $123, %eax
        ret

llvm-svn: 34909
2007-03-04 20:03:15 +00:00
Chris Lattner 362621c7ae eliminate some ops if they have an undef RHS
llvm-svn: 34908
2007-03-04 20:01:46 +00:00
Chris Lattner ca401aac31 Fix CodeGen/Generic/fpowi-promote.ll and PR1239
llvm-svn: 34893
2007-03-03 23:43:21 +00:00
Chris Lattner 567b9254cd Add an expand action for ISD label which just deletes the label.
This "fixes" PR1238.

llvm-svn: 34890
2007-03-03 19:21:38 +00:00
Jim Laskey d5453d7b56 Lower eh filter intrinsic.
llvm-svn: 34802
2007-03-01 20:24:30 +00:00
Jim Laskey 644af6b68f Chain is on second operand.
llvm-svn: 34759
2007-02-28 20:43:58 +00:00
Jim Laskey cf465fcebc MERGE_VALUES unnecessary.
llvm-svn: 34750
2007-02-28 18:37:04 +00:00
Chris Lattner 74bb92902e add methods for analysis of call results and return nodes.
llvm-svn: 34738
2007-02-28 07:09:40 +00:00
Chris Lattner e74744143f add methods to analyze calls and formals.
llvm-svn: 34736
2007-02-28 06:56:37 +00:00
Chris Lattner 9f059194a7 Minor refactoring of CC Lowering interfaces
llvm-svn: 34656
2007-02-27 05:13:54 +00:00
Chris Lattner dc3adc83e7 move CC Lowering stuff to its own public interface
llvm-svn: 34655
2007-02-27 04:43:02 +00:00
Chris Lattner fce448f856 Fold (sext (truncate x)) more aggressively, by avoiding creation of a
sextinreg if not needed.   This is useful in two cases: before legalize,
it avoids creating a sextinreg that will be trivially removed.  After legalize
if the target doesn't support sextinreg, the trunc/sext would not have been
removed before.

llvm-svn: 34621
2007-02-26 03:13:59 +00:00
Chris Lattner ab5d0ac02c track signedness of formal argument, though we have a fixme here.
llvm-svn: 34620
2007-02-26 02:56:58 +00:00
Jim Laskey 14059d958a Fix for PR1224.
llvm-svn: 34610
2007-02-25 21:43:59 +00:00
Chris Lattner 8c504cf9a0 optimize duplicate ValueMap lookups
llvm-svn: 34599
2007-02-25 18:40:32 +00:00
Chris Lattner 387f464121 fold trivial token factor nodes. This allows us to compile
test/CodeGen/X86/fp-stack-ret.ll into:

        movl 4(%esp), %eax
        fldl (%eax)
        ret

instead of:

        subl $12, %esp
        movl 16(%esp), %eax
        movsd (%eax), %xmm0
        movsd %xmm0, (%esp)
        fldl (%esp)
        addl $12, %esp
        ret

by eliminating a token factor that blocked a check.

llvm-svn: 34584
2007-02-25 08:24:27 +00:00
Chris Lattner 168c5856bf initialize a instance variable
llvm-svn: 34567
2007-02-25 01:28:05 +00:00
Jim Laskey e1d1c0590f Deal with cases when MMI is not requested.
llvm-svn: 34556
2007-02-24 09:45:44 +00:00
Jim Laskey b869ab6f31 Drop unused operand.
llvm-svn: 34555
2007-02-24 09:44:17 +00:00
Chris Lattner d7ef3f804d Fix CodeGen/Generic/2007-02-23-DAGCombine-Miscompile.ll and PR1219
llvm-svn: 34551
2007-02-24 02:09:29 +00:00
Jim Laskey 31fef788eb Handle improper cast.
llvm-svn: 34535
2007-02-23 21:45:01 +00:00
Jim Laskey 3e3a65b764 Need to init.
llvm-svn: 34499
2007-02-22 18:04:49 +00:00
Jim Laskey 44c37e7dbf Tighten up error checking of args.
llvm-svn: 34493
2007-02-22 16:10:05 +00:00
Jim Laskey 504e99479c Handle lowering invoke to call correctly.
llvm-svn: 34492
2007-02-22 15:38:06 +00:00
Jim Laskey 7f5872c455 Simplify lowering and selection of exception ops.
llvm-svn: 34491
2007-02-22 15:37:19 +00:00
Jim Laskey 4b37a4c712 Selection and lowering for exception handling.
llvm-svn: 34481
2007-02-21 22:53:45 +00:00
Chris Lattner 56e5fea163 print target nodes nicely
llvm-svn: 34369
2007-02-17 06:38:37 +00:00
Chris Lattner a9f917af59 Implement i/n/s constraints correctly. This fixes
test/CodeGen/PowerPC/2007-02-16-InlineAsmNConstraint.ll

llvm-svn: 34368
2007-02-17 06:00:35 +00:00
Chris Lattner 68dcec6fea fix indentation
llvm-svn: 34307
2007-02-15 18:19:15 +00:00
Chris Lattner 21ebae3394 Apply B Scott Michel's patch for PR1184, which improves diagnostics in an
abort case.

llvm-svn: 34306
2007-02-15 18:17:56 +00:00
Reid Spencer 09575bac2e For PR1195:
Change use of "packed" term to "vector" in comments, strings, variable
names, etc.

llvm-svn: 34300
2007-02-15 03:39:18 +00:00
Reid Spencer d84d35ba70 For PR1195:
Rename PackedType -> VectorType, ConstantPacked -> ConstantVector, and
PackedTyID -> VectorTyID. No functional changes.

llvm-svn: 34293
2007-02-15 02:26:10 +00:00
Chris Lattner ab1812f806 fix a warning
llvm-svn: 34272
2007-02-14 07:34:56 +00:00
Chris Lattner 1cf84d2745 Refix CodeGen/Generic/switch-lower.ll. In contrast to my previous patch,
this doesn't miscompile lots of programs :)

llvm-svn: 34268
2007-02-14 07:18:16 +00:00
Chris Lattner 945e437c65 Generalize TargetData strings, to support more interesting forms of data.
Patch by Scott Michel.

llvm-svn: 34266
2007-02-14 05:52:17 +00:00
Chris Lattner 59b27fa371 implement expand of truncate. This allows truncates from i128 to i64 to
be supported on 32-bit hosts.

llvm-svn: 34257
2007-02-13 23:55:16 +00:00
Chris Lattner d08d31f68a Fix PR1198, by adding initial i128 support. Patch by Dan Gohman.
llvm-svn: 34256
2007-02-13 23:41:38 +00:00
Chris Lattner 2fbff4d2dc revert my previous switch lowering change, which miscompiles a few programs.
This will break a dj test until I have time to investigate.

llvm-svn: 34247
2007-02-13 20:09:07 +00:00
Lauro Ramos Venancio abde3cc16c Add a space between // and the comment.
llvm-svn: 34244
2007-02-13 18:10:13 +00:00
Lauro Ramos Venancio 9956dcffbe Add "original alignment" to function arguments flags.
llvm-svn: 34240
2007-02-13 13:50:08 +00:00
Chris Lattner 9056bae3be Fix switch lowering to order cases in zext order, which is how we emit the
comparisons.  This fixes an infinite loop on CodeGen/Generic/switch-lower.ll
and PR1197

llvm-svn: 34216
2007-02-13 01:05:56 +00:00
Chris Lattner c473d8e431 Privatize StructLayout::MemberOffsets, adding an accessor
llvm-svn: 34156
2007-02-10 19:55:17 +00:00
Evan Cheng 276b44b0f9 Add function live-ins to entry block live-in set.
llvm-svn: 34112
2007-02-10 02:43:39 +00:00
Evan Cheng de6083463d Rename some variables to avoid confusion with SelectionDAGISel::BB.
llvm-svn: 34110
2007-02-10 01:08:18 +00:00
Evan Cheng 93049457ee Make use of TLI.SimplifySetCC() in LegalizeSetCCOperands().
llvm-svn: 34066
2007-02-08 22:16:19 +00:00
Evan Cheng 92658d5648 Move SimplifySetCC to TargetLowering and allow it to be shared with legalizer.
llvm-svn: 34065
2007-02-08 22:13:59 +00:00
Chris Lattner 19083a4671 switch the VRBaseMap in the scheduler from an std::map to a DenseMap. This
speeds up the isel pass from 2.5570s to 2.4722s on kc++ (3.4%).

llvm-svn: 33879
2007-02-04 08:47:20 +00:00
Chris Lattner 9af2c86bc8 Introduce new UnarySDNode/BinarySDNode/TernarySDNode nodes, which coallocate
their operands with the node itself.  This reduces malloc traffic for operand
lists.  This reduces isel time on kc++ from 2.6164 to 2.5570s, about 2.3%.

llvm-svn: 33878
2007-02-04 08:35:21 +00:00
Chris Lattner 22639f3d90 eliminate the SDNode::setValueTypes method.
llvm-svn: 33876
2007-02-04 07:37:24 +00:00
Chris Lattner f17b4222e2 eliminate a bunch of duplicate ctors and helper functions.
llvm-svn: 33875
2007-02-04 07:28:00 +00:00
Chris Lattner edfc7e5fa2 move MorphNode to out of line and merge setNodeOperands into it. There is
no behavior or performance change here.

llvm-svn: 33869
2007-02-04 02:49:29 +00:00
Chris Lattner 3bf17b6fa5 simplify MorphNodeTo to take a VTList operand.
llvm-svn: 33868
2007-02-04 02:41:42 +00:00
Chris Lattner 486edfbc6f eliminate some extraneous methods in SDNode
llvm-svn: 33867
2007-02-04 02:32:44 +00:00
Chris Lattner 20754cc579 Give each selectiondag node class a home for it's vtable and rtti info
llvm-svn: 33866
2007-02-04 02:23:32 +00:00
Chris Lattner 289aa4495c Switch VAlueMap from std::map to DenseMap.
llvm-svn: 33863
2007-02-04 01:35:11 +00:00
Chris Lattner 79084305ee Switch NodeMap from std::map to DenseMap, this speeds up isel by 2.3%
llvm-svn: 33862
2007-02-04 01:31:47 +00:00
Chris Lattner 94c44c96d3 swtich vector-> smallvector, speeding up selectiondag stuff 1%
llvm-svn: 33861
2007-02-04 01:20:02 +00:00
Chris Lattner 4b0ddb22e9 Switch promoted/expanded ops over to using a DenseMap. Vector related maps
aren't worth it.

llvm-svn: 33860
2007-02-04 01:17:38 +00:00
Chris Lattner ed39c86176 switch LegalizedNodes from std::map to a DenseMap. This speeds up isel
time as a whole on kc++ by 11%.

llvm-svn: 33857
2007-02-04 00:50:02 +00:00
Chris Lattner ebeb48d4bc Eliminate some malloc traffic from LegalizeAllNodesNotLeadingTo, speeding
up isel on kimwitu by 0.7%.

llvm-svn: 33853
2007-02-04 00:27:56 +00:00
Chris Lattner cba058ce51 Eliminate some std::sets. This speeds up isel of kimwitu by about 0.9%
llvm-svn: 33852
2007-02-04 00:24:41 +00:00
Chris Lattner feec7137ce Switch SelectionDAG::ReplaceAllUsesOfValueWith to use a SmallSetVector for
the users set (most nodes have 1 or 2 users).  This speeds up the isel pass
3.2% on kimwitu.

llvm-svn: 33849
2007-02-04 00:14:31 +00:00
Chris Lattner 0a30b1f00f switch the sched unit map over to use a DenseMap instead of std::map. This
speeds up isel as a whole time by 2.6%.

llvm-svn: 33810
2007-02-03 01:34:13 +00:00
Chris Lattner e83030b9c8 Switch ComputeTopDownOrdering over to using a densemap. This speeds up
isel as a whole by 3.3%.

llvm-svn: 33809
2007-02-03 01:12:36 +00:00
Evan Cheng f309d13677 Pasto
llvm-svn: 33806
2007-02-03 00:43:46 +00:00
Reid Spencer 2341c22ec7 Changes to support making the shift instructions be true BinaryOperators.
This feature is needed in order to support shifts of more than 255 bits
on large integer types.  This changes the syntax for llvm assembly to
make shl, ashr and lshr instructions look like a binary operator:
   shl i32 %X, 1
instead of
   shl i32 %X, i8 1
Additionally, this should help a few passes perform additional optimizations.

llvm-svn: 33776
2007-02-02 02:16:23 +00:00
Anton Korobeynikov 1b4e6015b4 Fixed uninitialized stuff inside LegalizeDAG. Fortunately, the only
affected part is codegen of "memove" inside x86 backend. This fixes
PR1144

llvm-svn: 33752
2007-02-01 08:39:52 +00:00
Chris Lattner 296a83cefb Fit in 80 columns
llvm-svn: 33745
2007-02-01 04:55:59 +00:00
Chris Lattner e3eeb24a86 Emit a better assertion message for PR1133
llvm-svn: 33736
2007-02-01 01:21:12 +00:00
Evan Cheng 53026f1d5a Allow the target to override the ISD::CondCode that's to be used to test the
result of the comparison libcall against zero.

llvm-svn: 33701
2007-01-31 09:29:11 +00:00
Reid Spencer 5301e7c605 For PR1136: Rename GlobalVariable::isExternal as isDeclaration to avoid
confusion with external linkage types.

llvm-svn: 33663
2007-01-30 20:08:39 +00:00
Chris Lattner d27f95e08d add initial support for handling inline asms with multiple constraints.
This doesn't do the "right thing" but will probably work in most cases.

This implements CodeGen/PowerPC/2007-01-29-lbrx-asm.ll.

llvm-svn: 33643
2007-01-29 23:45:14 +00:00
Nate Begeman eda5997cc8 Finish off bug 680, allowing targets to custom lower frame and return
address nodes.

llvm-svn: 33636
2007-01-29 22:58:52 +00:00
Anton Korobeynikov 06f7d4bec7 Arguments are counting from 1. not from 0. Maybe we should change
numbering somehow? E.g. make return argument the last?

llvm-svn: 33606
2007-01-28 18:01:49 +00:00
Anton Korobeynikov 9fa3839d29 More cleanup
llvm-svn: 33605
2007-01-28 16:04:40 +00:00
Anton Korobeynikov 037c867b54 Propagate changes from my local tree. This patch includes:
1. New parameter attribute called 'inreg'. It has meaning "place this
parameter in registers, if possible". This is some generalization of
gcc's regparm(n) attribute. It's currently used only in X86-32 backend.
2. Completely rewritten CC handling/lowering code inside X86 backend.
Merged stdcall + c CCs and fastcall + fast CC.
3. Dropped CSRET CC. We cannot add struct return variant for each
target-specific CC (e.g. stdcall + csretcc and so on).
4. Instead of CSRET CC introduced 'sret' parameter attribute. Setting in
on first attribute has meaning 'This is hidden pointer to structure
return. Handle it gently'.
5. Fixed small bug in llvm-extract + add new feature to
FunctionExtraction pass, which relinks all internal-linkaged callees
from deleted function to external linkage. This will allow further
linking everything together.

NOTEs: 1. Documentation will be updated soon.
       2. llvm-upgrade should be improved to translate csret => sret.
          Before this, there will be some unexpected test fails.
llvm-svn: 33597
2007-01-28 13:31:35 +00:00
Jim Laskey c56315c2b5 Change the MachineDebugInfo to MachineModuleInfo to better reflect usage
for debugging and exception handling.

llvm-svn: 33550
2007-01-26 21:22:28 +00:00
Jim Laskey f9e5445ed4 Make LABEL a builtin opcode.
llvm-svn: 33537
2007-01-26 14:34:52 +00:00
Evan Cheng f5c96fabf9 Renamed getTypeAlignmentShift() to getPreferredTypeAlignmentShift().
llvm-svn: 33482
2007-01-24 07:03:39 +00:00
Evan Cheng be48a47d9d Remove the DoubleTy special case.
llvm-svn: 33449
2007-01-22 23:13:55 +00:00
Reid Spencer 2eadb5310d For PR970:
Clean up handling of isFloatingPoint() and dealing with PackedType.
Patch by Gordon Henriksen!

llvm-svn: 33415
2007-01-21 00:29:26 +00:00
Chris Lattner 50ee0e40e5 Teach TargetData to handle 'preferred' alignment for each target, and use
these alignment amounts to align scalars when we can.  Patch by Scott Michel!

llvm-svn: 33409
2007-01-20 22:35:55 +00:00
Evan Cheng 00a640dbe0 Fix for PR1108: type of insert_vector_elt index operand is PtrVT, not MVT::i32.
llvm-svn: 33398
2007-01-20 10:10:26 +00:00
Reid Spencer a94d394ad2 For PR1043:
This is the final patch for this PR. It implements some minor cleanup
in the use of IntegerType, to wit:
1. Type::getIntegerTypeMask -> IntegerType::getBitMask
2. Type::Int*Ty changed to IntegerType* from Type*
3. ConstantInt::getType() returns IntegerType* now, not Type*

This also fixes PR1120.

Patch by Sheng Zhou.

llvm-svn: 33370
2007-01-19 21:13:56 +00:00
Evan Cheng 9201100b29 Remove this xform:
(shl (add x, c1), c2) -> (add (shl x, c2), c1<<c2)
Replace it with:
(add (shl (add x, c1), c2), ) -> (add (add (shl x, c2), c1<<c2), )

This fixes test/CodeGen/ARM/smul.ll

llvm-svn: 33361
2007-01-19 17:51:44 +00:00
Chris Lattner 4dc4489286 Fix PR1114 and CodeGen/Generic/2007-01-15-LoadSelectCycle.ll by being
careful when folding "c ? load p : load q" that C doesn't reach either load.
If so, folding this into load (c ? p : q) will induce a cycle in the graph.

llvm-svn: 33251
2007-01-16 05:59:59 +00:00
Chris Lattner f70c5cd5db add options to view the dags before the first or second pass of dag combine.
llvm-svn: 33249
2007-01-16 04:55:25 +00:00
Reid Spencer a8a0f2cf68 Compensate for loss of DerivedTypes.h in TargetLowering.h
llvm-svn: 33159
2007-01-12 23:31:12 +00:00
Reid Spencer ddf1421b8e Move a function out of line.
llvm-svn: 33158
2007-01-12 23:30:31 +00:00
Evan Cheng 61a4be88b4 Minor fix.
llvm-svn: 33149
2007-01-12 22:51:10 +00:00
Evan Cheng 31cbddf28a Store default libgcc routine names and allow them to be redefined by target.
llvm-svn: 33105
2007-01-12 02:11:51 +00:00
Zhou Sheng 75b871fb1e For PR1043:
Merge ConstantIntegral and ConstantBool into ConstantInt.
Remove ConstantIntegral and ConstantBool from LLVM.

llvm-svn: 33073
2007-01-11 12:24:14 +00:00
Evan Cheng 6730f03370 Naming consistency.
llvm-svn: 33026
2007-01-08 23:55:53 +00:00
Evan Cheng 961bbd393b Fix for PR1075: bottom-up register-reduction scheduling actually increases register pressure.
- Fixed bugs in sethi-ullman number computation and priority queue comparison
functions.
- Separate code that handles priority computation special cases from SU number computation.

llvm-svn: 33025
2007-01-08 23:50:38 +00:00
Chris Lattner 0199fd6d59 Implement some trivial FP foldings when -enable-unsafe-fp-math is specified.
This implements CodeGen/PowerPC/unsafe-math.ll

llvm-svn: 33024
2007-01-08 23:04:05 +00:00
Chris Lattner 10cae15d8e remove support for llvm.isunordered
llvm-svn: 32992
2007-01-07 08:37:22 +00:00
Evan Cheng 5f80c450f3 Expand fcopysign to the bitwise sequence if select is marked as expensive.
llvm-svn: 32940
2007-01-05 23:33:44 +00:00
Evan Cheng 3b841ddbe0 Bug in ExpandFCOPYSIGNToBitwiseOps(). Clear the old sign bit of operand 0
before or'ing in the sign bit of operand 1.

llvm-svn: 32930
2007-01-05 21:31:51 +00:00
Evan Cheng 376c9c4c97 CopyToReg source operand can be a register as well. e.g. Copy from GlobalBaseReg.
llvm-svn: 32929
2007-01-05 20:59:06 +00:00
Evan Cheng 8ec5283dc4 GEP subscript is interpreted as a signed value.
llvm-svn: 32888
2007-01-05 01:46:20 +00:00
Chris Lattner 96035bed51 fix PowerPC/2007-01-04-ArgExtension.ll, a bug handling K&R prototypes with
the recent signless changes.

llvm-svn: 32884
2007-01-04 22:22:37 +00:00
Evan Cheng 003feb03d5 Expand fcopysign to a series of bitwise of operations when it's profitable to
do so.

llvm-svn: 32881
2007-01-04 21:56:39 +00:00
Reid Spencer e6f81876eb Legalizer doesn't do an ANY_EXTEND if we don't ask for one so make sure
that we default to an ANY_EXTEND if no parameter attribute is set on the
result value of a function.

llvm-svn: 32836
2007-01-03 16:49:33 +00:00
Reid Spencer 2a34b91666 Restore previous behavior of defaulting to ZEXT. This works around two
things: (1) preventing PR1071 and (2) working around missing parameter
attributes for bool type. (2) will be fixed shortly. When PR1071 is fixed,
this patch should be undone.

llvm-svn: 32831
2007-01-03 05:03:05 +00:00
Reid Spencer 0917adf614 Two changes:
1. Switch expression and cases are compared signed and are sign extended.
2. For function results needing extended, do SIGN_EXTEND if the SExtAttribute
   is set and ZERO_EXTEND if the ZExtAttribute is set, otherwise just let
   the Legalizer do ANY_EXTEND.
This fixes the recent regression in kimwitu++ and probably the llvm-gcc
bootstrap issue we had today.

llvm-svn: 32830
2007-01-03 04:25:33 +00:00
Reid Spencer 791864c6a5 Clean up from recent changes. Comment the new parameter to ExpandLibCall.
Consolidate some lines of code and remove duplication.

llvm-svn: 32829
2007-01-03 04:22:32 +00:00
Reid Spencer e63b6518fa For PR950:
Three changes:
1. Convert signed integer types to signless versions.
2. Implement the @sext and @zext parameter attributes. Previously the
   type of an function parameter was used to determine whether it should
   be sign extended or zero extended before the call. This information is
   now communicated via the function type's parameter attributes.
3. The interface to LowerCallTo had to be changed in order to accommodate
   the parameter attribute information. Although it would have been
   convenient to pass in the FunctionType itself, there isn't always one
   present in the caller. Consequently, a signedness indication for the
   result type and for each parameter was provided for in the interface
   to this method. All implementations were changed to make the adjustment
   necessary.

llvm-svn: 32788
2006-12-31 05:55:36 +00:00
Reid Spencer 266e42b312 For PR950:
This patch removes the SetCC instructions and replaces them with the ICmp
and FCmp instructions. The SetCondInst instruction has been removed and
been replaced with ICmpInst and FCmpInst.

llvm-svn: 32751
2006-12-23 06:05:41 +00:00
Evan Cheng 258657e64e getLoad() and getStore() calls missed SVOffset operand. Thanks to Dan Gohman
for pointing it out!

llvm-svn: 32712
2006-12-20 01:27:29 +00:00
Chris Lattner aee775a6b7 Eliminate static ctors from Statistics
llvm-svn: 32698
2006-12-19 22:41:21 +00:00
Evan Cheng 9ad6edf2ec May need to promote the operand (either sign_extend_inreg or and) before
expanding a {s|u}int_to_fp.

llvm-svn: 32665
2006-12-19 01:44:04 +00:00
Evan Cheng adc80f98cf LegalizeSetCCOperands() may end up inserting libcalls. They need to be
properly serialized. Do not clear LastCallSEQ_END until that is done.

llvm-svn: 32659
2006-12-18 22:55:34 +00:00
Bill Wendling e33ce528da Fixed so that it dereferences the ostream pointer.
llvm-svn: 32640
2006-12-17 11:15:53 +00:00
Bill Wendling a77f14265b Added an automatic cast to "std::ostream*" etc. from OStream. We then can
rework the hacks that had us passing OStream in. We pass in std::ostream*
instead, check for null, and then dispatch to the correct print() method.

llvm-svn: 32636
2006-12-17 05:15:13 +00:00
Chris Lattner 9bd5ed636c Fix PR1049 and CodeGen/Generic/2006-12-16-InlineAsmCrash.ll
by producing target constants instead of constants.  Constants can get
selected to li/movri instructions, which causes the scheduler to explode.

llvm-svn: 32633
2006-12-16 21:14:48 +00:00
Evan Cheng 28cf4277bb Cannot combine an indexed load / store any further.
llvm-svn: 32629
2006-12-16 06:25:23 +00:00
Evan Cheng 851e589eda Expand FP undef
llvm-svn: 32623
2006-12-16 02:20:50 +00:00
Evan Cheng 860004688a Allow promoted FP_TO_UINT / FP_TO_SINT to expand operand.
llvm-svn: 32621
2006-12-16 02:10:30 +00:00
Evan Cheng 388cbbf000 Expand fabs / fneg to and / xor.
llvm-svn: 32619
2006-12-16 00:52:40 +00:00
Evan Cheng 884bc09d10 Fix select_cc, select expansion to soft-fp bugs.
llvm-svn: 32616
2006-12-15 22:42:55 +00:00
Jim Laskey 26df19ace6 This code was usurping the sextload expand in teh legalizer. Just make
sure the right conditions are checked.

llvm-svn: 32611
2006-12-15 21:38:30 +00:00
Chris Lattner b1a9492ed7 silence a bogus warning
llvm-svn: 32597
2006-12-15 07:36:19 +00:00
Evan Cheng 35fdd5ffe1 Expand FP compares to soft-fp call(s)
llvm-svn: 32590
2006-12-15 02:59:56 +00:00
Jim Laskey 70323a8146 1. Tidy up jump table info.
2. Allow the jit to handle PIC relocable jump tables.

llvm-svn: 32581
2006-12-14 19:17:33 +00:00
Evan Cheng 22cf89967b More soft-fp work.
llvm-svn: 32559
2006-12-13 20:57:08 +00:00
Evan Cheng e370e0eb09 Expand (f64 extload f32) to (f64 fp_ext (load f32)) if f64 type action is expand.
llvm-svn: 32527
2006-12-13 03:19:57 +00:00
Evan Cheng f3a80c6235 Expand fsqrt, fsin, and fcos to libcalls.
llvm-svn: 32526
2006-12-13 02:38:13 +00:00
Evan Cheng 0a5b805f6d Expand f32 / f64 to i32 / i64 conversion to soft-fp library calls.
llvm-svn: 32523
2006-12-13 01:57:55 +00:00
Reid Spencer bfe26ffcfc Replace CastInst::createInferredCast calls with more accurate cast
creation calls.

llvm-svn: 32521
2006-12-13 00:50:17 +00:00
Evan Cheng 3766fc60da Expand FP constant to integers if FP types are not legal.
llvm-svn: 32497
2006-12-12 22:19:28 +00:00
Evan Cheng 97a750fc47 Soft fp FNEG, SINT_TO_FP, UINT_TO_FP libcall expansion.
llvm-svn: 32495
2006-12-12 21:51:17 +00:00
Evan Cheng 47833a1d28 Expand ConstantFP to load from CP if float types are being expanded.
llvm-svn: 32494
2006-12-12 21:32:44 +00:00
Evan Cheng 634885f71e Expand i32/i64 CopyToReg f32/f64 to BIT_CONVERT + CopyToReg.
llvm-svn: 32493
2006-12-12 21:21:32 +00:00
Evan Cheng 0076ca0da9 - When expanding a bit_convert whose src operand is also to be expanded and
its expansion result type is equal to the result type of the bit_convert,
e.g. (i64 bit_convert (f64 op)) if FP is not legal
returns the result of the expanded source operand.
- Store f32 / f64 may be expanded to a single store i32/i64.

llvm-svn: 32490
2006-12-12 19:53:13 +00:00
Evan Cheng 0c0b78c18e Expand formal arguments and call arguments recursively: e.g. f64 -> i64 -> 2 x i32.
llvm-svn: 32476
2006-12-12 07:27:38 +00:00
Chris Lattner 2f96e7d241 fit in 80 cols
llvm-svn: 32474
2006-12-12 05:22:21 +00:00
Chris Lattner 080881614d this can only be fptrunc.
llvm-svn: 32473
2006-12-12 05:21:51 +00:00
Chris Lattner 6ba11fbd75 Revert Nate's patch to fix X86/store-fp-constant.ll. With the dag combiner
and legalizer separated like they currently are, I don't see a way to handle
this xform.

llvm-svn: 32466
2006-12-12 04:18:56 +00:00
Chris Lattner b7524b6d0e make this code more aggressive about turning store fpimm into store int imm.
This is not sufficient to fix X86/store-fp-constant.ll

llvm-svn: 32465
2006-12-12 04:16:14 +00:00
Reid Spencer 3c49edcaa1 Change inferred cast creation calls to more specific cast creations.
llvm-svn: 32460
2006-12-12 01:17:41 +00:00
Evan Cheng 3432ab97c1 Re-apply changes that were backed out and fix a naughty typo.
llvm-svn: 32442
2006-12-11 19:27:14 +00:00
John Criswell b3b285185f It seems the llvm::OStream class does not handle stream manipulators.
For now, just grab the stream and perform the output on it directly.

llvm-svn: 32441
2006-12-11 19:15:36 +00:00
Chris Lattner e9a203c4e5 Revert changes that broke oggenc on ppc
llvm-svn: 32440
2006-12-11 18:53:38 +00:00
Evan Cheng 218369881f Don't convert store double C, Ptr to store long C, Ptr if i64 is not a legal type.
llvm-svn: 32434
2006-12-11 17:25:19 +00:00
Evan Cheng f4bec95b58 f32 / f64 node is expanded to one i32 / i64 node.
llvm-svn: 32433
2006-12-11 06:50:04 +00:00
Evan Cheng f6b01fdb48 Clean up some bad code.
llvm-svn: 32432
2006-12-11 06:25:26 +00:00
Nate Begeman 8e20c760fa Move something that should be in the dag combiner from the legalizer to the
dag combiner.

llvm-svn: 32431
2006-12-11 02:23:46 +00:00
Anton Korobeynikov 3b7c257cae Cleaned setjmp/longjmp lowering interfaces. Now we're producing right
code (both asm & cbe) for Mingw32 target.
Removed autoconf checks for underscored versions of setjmp/longjmp.

llvm-svn: 32415
2006-12-10 23:12:42 +00:00
Evan Cheng 4eee72471c Preliminary soft float support.
llvm-svn: 32394
2006-12-09 02:42:38 +00:00
Chris Lattner d9f04e4875 Fix CodeGen/PowerPC/2006-12-07-SelectCrash.ll on PPC64
llvm-svn: 32336
2006-12-07 22:36:47 +00:00
Bill Wendling 355fc5ad50 Removed more <iostream> includes
llvm-svn: 32321
2006-12-07 20:28:15 +00:00
Bill Wendling 22e978a736 Removing even more <iostream> includes.
llvm-svn: 32320
2006-12-07 20:04:42 +00:00
Chris Lattner 700b873130 Detemplatize the Statistic class. The only type it is instantiated with
is 'unsigned'.

llvm-svn: 32279
2006-12-06 17:46:33 +00:00
Jeff Cohen cc08c83186 Unbreak VC++ build.
llvm-svn: 32113
2006-12-02 02:22:01 +00:00
Evan Cheng 67fc141db5 Match TargetInstrInfo changes.
llvm-svn: 32098
2006-12-01 21:52:58 +00:00
Evan Cheng a743fada65 Avoid inifinite looping if READCYCLECOUNTER isn't custom lowered.
llvm-svn: 32022
2006-11-29 19:13:47 +00:00
Evan Cheng 6973993e9c Allow target to custom lower READCYCLECOUNTER (when it doesn't have to be expanded).
llvm-svn: 32016
2006-11-29 08:26:18 +00:00
Evan Cheng feba507a97 Fix for PR1023 by Dan Gohman.
llvm-svn: 32003
2006-11-29 01:58:12 +00:00
Evan Cheng 6e12a052ff Fix for PR1022 (folding loads of static initializers) by Dan Gohman.
llvm-svn: 32000
2006-11-29 01:38:07 +00:00
Chris Lattner 90f4238c38 add a hook to allow targets to hack on inline asms to lower them to llvm
when they want to.

llvm-svn: 31997
2006-11-29 01:12:32 +00:00
Chris Lattner 3abb63651b Fix PR1016
llvm-svn: 31950
2006-11-28 01:03:30 +00:00
Evan Cheng 20350c4025 Change MachineInstr ctor's to take a TargetInstrDescriptor reference instead
of opcode and number of operands.

llvm-svn: 31947
2006-11-27 23:37:22 +00:00
Chris Lattner 5d5916b4d1 Fix the dag combiner bug corresponding to PR1014.
llvm-svn: 31943
2006-11-27 21:50:02 +00:00
Chris Lattner 3da631f29a For better or worse, load from i1 is assumed to be zero extended. Do not
form a load from i1 from larger loads that may not be zext'd.

llvm-svn: 31933
2006-11-27 04:40:53 +00:00
Chris Lattner db18938355 If a brcond condition is promoted, make sure to zero extend it, even if not
expanded into BR_CC.

llvm-svn: 31932
2006-11-27 04:39:56 +00:00
Reid Spencer 6c38f0bb07 For PR950:
The long awaited CAST patch. This introduces 12 new instructions into LLVM
to replace the cast instruction. Corresponding changes throughout LLVM are
provided. This passes llvm-test, llvm/test, and SPEC CPUINT2000 with the
exception of 175.vpr which fails only on a slight floating point output
difference.

llvm-svn: 31931
2006-11-27 01:05:10 +00:00
Chris Lattner 3676a994ca Fix PR1011 and CodeGen/Generic/2006-11-20-DAGCombineCrash.ll
llvm-svn: 31878
2006-11-20 18:05:46 +00:00
Reid Spencer d9436b6837 For PR950:
First in a series of patches to convert SetCondInst into ICmpInst and
FCmpInst using only two opcodes and having the instructions contain their
predicate value. Nothing uses these classes yet. More patches to follow.

llvm-svn: 31867
2006-11-20 01:22:35 +00:00
Jim Laskey da0add3fd0 Fixing the ENABLE_OPTIMIZED=1 DISABLE_ASSERTIONS=1 build.
llvm-svn: 31822
2006-11-17 13:07:55 +00:00
Evan Cheng f64da389f8 Fix an incorrectly inverted condition.
llvm-svn: 31773
2006-11-16 00:08:20 +00:00
Chris Lattner 30d08801ef remove dead #include
llvm-svn: 31753
2006-11-15 17:51:15 +00:00
Evan Cheng dbd3d294e6 Matches MachineInstr changes.
llvm-svn: 31712
2006-11-13 23:36:35 +00:00
Reid Spencer 2230144a75 Make an assert comment match the tested assertion.
llvm-svn: 31686
2006-11-11 20:07:59 +00:00
Evan Cheng 979bbf48d5 Add methods to add implicit def use operands to a MI.
llvm-svn: 31675
2006-11-11 10:20:02 +00:00
Chris Lattner a0a8003f59 disallow preinc of a frameindex. This is not profitable and causes 2-addr
pass to explode.  This fixes a bunch of llc-beta failures on ppc last night.

llvm-svn: 31661
2006-11-11 01:00:15 +00:00
Chris Lattner eabc15c1d8 reduce indentation by using early exits. No functionality change.
llvm-svn: 31660
2006-11-11 00:56:29 +00:00
Chris Lattner ffad2166e1 move big chunks of code out-of-line, no functionality change.
llvm-svn: 31658
2006-11-11 00:39:41 +00:00
Chris Lattner 4eac5f59e6 Fix a dag combiner bug exposed by my recent instcombine patch. This fixes
CodeGen/Generic/2006-11-10-DAGCombineMiscompile.ll and PPC gsm/toast

llvm-svn: 31644
2006-11-10 21:37:15 +00:00
Evan Cheng 8c9c6d71ed Add implicit def / use operands to MachineInstr.
llvm-svn: 31633
2006-11-10 08:43:01 +00:00
Evan Cheng 13440b025c When forming a pre-indexed store, make sure ptr isn't the same or is a pred of value being stored. It would cause a cycle.
llvm-svn: 31631
2006-11-10 08:28:11 +00:00
Chris Lattner d5e604dbb2 commentate
llvm-svn: 31627
2006-11-10 04:41:34 +00:00
Evan Cheng 6878378390 Don't attempt expensive pre-/post- indexed dag combine if target does not support them.
llvm-svn: 31598
2006-11-09 19:10:46 +00:00
Evan Cheng d550248f2c Add a mechanism to specify whether a target supports a particular indexed load / store.
llvm-svn: 31597
2006-11-09 18:56:43 +00:00
Evan Cheng c034f14fbe Rename ISD::MemOpAddrMode to ISD::MemIndexedMode
llvm-svn: 31596
2006-11-09 18:44:21 +00:00
Evan Cheng b15000736c Rename ISD::MemOpAddrMode to ISD::MemIndexedMode
llvm-svn: 31595
2006-11-09 17:55:04 +00:00
Evan Cheng b58e06bc9e getPostIndexedAddressParts change: passes in load/store instead of its loaded / stored VT.
llvm-svn: 31584
2006-11-09 04:29:46 +00:00
Evan Cheng 85e54223cd Match more post-indexed ops.
llvm-svn: 31569
2006-11-08 20:27:27 +00:00
Jim Laskey 61feeb90f9 Remove redundant <cmath>.
llvm-svn: 31561
2006-11-08 19:16:44 +00:00
Evan Cheng 0303cb9b33 - When performing pre-/post- indexed load/store transformation, do not worry
about whether the new base ptr would be live below the load/store. Let two
  address pass split it back to non-indexed ops.
- Minor tweaks / fixes.

llvm-svn: 31544
2006-11-08 08:30:28 +00:00
Evan Cheng 6072435756 Fixed a minor bug preventing some pre-indexed load / store transformation.
llvm-svn: 31543
2006-11-08 06:56:05 +00:00
Reid Spencer fdff938a7e For PR950:
This patch converts the old SHR instruction into two instructions,
AShr (Arithmetic) and LShr (Logical). The Shr instructions now are not
dependent on the sign of their operands.

llvm-svn: 31542
2006-11-08 06:47:33 +00:00
Evan Cheng d48f7dd250 Fix a obscure post-indexed load / store dag combine bug.
llvm-svn: 31537
2006-11-08 02:38:55 +00:00
Evan Cheng 60c6846d21 Add post-indexed load / store transformations.
llvm-svn: 31498
2006-11-07 09:03:05 +00:00
Chris Lattner 94c231f453 Fix PR988 and CodeGen/Generic/2006-11-06-MemIntrinsicExpand.ll.
The low part goes in the first operand of expandop, not the second one.

llvm-svn: 31487
2006-11-07 04:11:44 +00:00
Evan Cheng f24d15f969 Remove dead code; added a missing null ptr check.
llvm-svn: 31478
2006-11-06 21:33:46 +00:00
Evan Cheng eb99bd736a Add comment.
llvm-svn: 31473
2006-11-06 08:14:30 +00:00
Jeff Cohen 7d6f3db3e2 Unbreak VC++ build.
llvm-svn: 31464
2006-11-05 19:31:28 +00:00
Evan Cheng 33157700d9 Added pre-indexed store support.
llvm-svn: 31459
2006-11-05 09:31:14 +00:00
Evan Cheng 1a1e23eff7 Added getIndexedStore.
llvm-svn: 31458
2006-11-05 09:30:09 +00:00
Evan Cheng fd2c5dd806 Changes to use operand constraints to process two-address instructions.
llvm-svn: 31453
2006-11-04 09:44:31 +00:00
Evan Cheng 9456dd8b81 Fix comments.
llvm-svn: 31414
2006-11-03 07:31:32 +00:00
Evan Cheng 1dfd26a151 Rename
llvm-svn: 31413
2006-11-03 07:21:16 +00:00
Reid Spencer 52f958741a Remove dead variable. Fix 80 column violations.
llvm-svn: 31412
2006-11-03 03:30:34 +00:00
Evan Cheng 357017f4a9 Added DAG combiner transformation to generate pre-indexed loads.
llvm-svn: 31410
2006-11-03 03:06:21 +00:00
Evan Cheng c176f038b9 Added isPredecessor.
llvm-svn: 31409
2006-11-03 03:05:24 +00:00
Chris Lattner cd7b92251d silence warning
llvm-svn: 31397
2006-11-03 01:28:29 +00:00
Reid Spencer de46e48420 For PR786:
Turn on -Wunused and -Wno-unused-parameter. Clean up most of the resulting
fall out by removing unused variables. Remaining warnings have to do with
unused functions (I didn't want to delete code without review) and unused
variables in generated code. Maintainers should clean up the remaining
issues when they see them. All changes pass DejaGnu tests and Olden.

llvm-svn: 31380
2006-11-02 20:25:50 +00:00
Reid Spencer 7eb55b395f For PR950:
Replace the REM instruction with UREM, SREM and FREM.

llvm-svn: 31369
2006-11-02 01:53:59 +00:00
Chris Lattner 55402d4403 Allow the getRegForInlineAsmConstraint method to return a register class with
no fixes physreg.  Treat this as permission to use any register in the register
class.  When this happens and it is safe, allow the llvm register allcoator to
allocate the register instead of doing it at isel time.  This eliminates a ton
of copies around common inline asms.  For example:

int test2(int Y, int X) {
  asm("foo %0, %1" : "=r"(X): "r"(X));
  return X;
}

now compiles to:

_test2:
        foo r3, r4
        blr

instead of:

_test2:
        mr r2, r4
        foo r2, r2
        mr r3, r2
        blr

GCC produces:

_test2:
        foo r4, r4
        mr r3,r4
        blr

llvm-svn: 31366
2006-11-02 01:41:49 +00:00
Evan Cheng 1359196c4e Clean up.
llvm-svn: 31359
2006-11-01 22:39:30 +00:00
Evan Cheng 47218fab42 CopyFromReg starts a live range so its use should not be considered a floater.
llvm-svn: 31356
2006-11-01 22:17:06 +00:00
Evan Cheng 415f365e5c Print jumptable index.
llvm-svn: 31340
2006-11-01 04:48:30 +00:00
Chris Lattner fe43befeda Compile CodeGen/PowerPC/fp-branch.ll to:
_intcoord_cond_next55:
LBB1_3: ;cond_next55
        lis r2, ha16(LCPI1_0)
        lfs f0, lo16(LCPI1_0)(r2)
        fcmpu cr0, f1, f0
        blt cr0, LBB1_2 ;cond_next62.exitStub
LBB1_1: ;bb72.exitStub
        li r3, 1
        blr
LBB1_2: ;cond_next62.exitStub
        li r3, 0
        blr

instead of:

_intcoord_cond_next55:
LBB1_3: ;cond_next55
        lis r2, ha16(LCPI1_0)
        lfs f0, lo16(LCPI1_0)(r2)
        fcmpu cr0, f1, f0
        bge cr0, LBB1_1 ;bb72.exitStub
LBB1_4: ;cond_next55
        lis r2, ha16(LCPI1_0)
        lfs f0, lo16(LCPI1_0)(r2)
        fcmpu cr0, f1, f0
        bnu cr0, LBB1_2 ;cond_next62.exitStub
LBB1_1: ;bb72.exitStub
        li r3, 1
        blr
LBB1_2: ;cond_next62.exitStub
        li r3, 0
        blr

llvm-svn: 31330
2006-10-31 23:06:00 +00:00
Chris Lattner 427301fdae look through isunordered to inline it into branch blocks.
llvm-svn: 31328
2006-10-31 22:37:42 +00:00
Chris Lattner 1fd360e13a handle global address constant sdnodes
llvm-svn: 31323
2006-10-31 20:01:56 +00:00
Chris Lattner 6f043b90ea TargetLowering::isOperandValidForConstraint
llvm-svn: 31319
2006-10-31 19:41:18 +00:00
Chris Lattner 8c6949e5b2 Change the prototype for TargetLowering::isOperandValidForConstraint
llvm-svn: 31318
2006-10-31 19:40:43 +00:00
Chris Lattner 968f803928 Turn an assert into an error message. This is commonly triggered when
we don't support a specific constraint yet.  When this happens, print the
unsupported constraint.

llvm-svn: 31310
2006-10-31 07:33:13 +00:00
Evan Cheng e6d584765f Fix a typo which can break jumptables.
llvm-svn: 31305
2006-10-31 02:31:00 +00:00
Evan Cheng 84a28d4e76 Lower jumptable to BR_JT. The legalizer can lower it to a BRIND or let the target custom lower it.
llvm-svn: 31293
2006-10-30 08:00:44 +00:00
Evan Cheng c3e695137d Added a new SDNode type: BR_JT for jumptable branch.
llvm-svn: 31292
2006-10-30 07:59:36 +00:00
Chris Lattner e60ae823e8 fix Generic/2006-10-29-Crash.ll
llvm-svn: 31281
2006-10-29 21:01:20 +00:00
Chris Lattner f31b9ef458 Fix a load folding issue that Evan noticed: there is no need to export values
used by comparisons in the main block.

llvm-svn: 31279
2006-10-29 18:23:37 +00:00
Evan Cheng 7ab6123c42 VLOAD is not the LoadSDNode opcode.
llvm-svn: 31276
2006-10-29 06:14:47 +00:00
Nick Lewycky dc146a9fb9 Remove spurious case. EXTLOAD is not one of the node opcodes.
llvm-svn: 31275
2006-10-29 02:26:30 +00:00
Chris Lattner bba52191fa split critical edges more carefully and intelligently. In particular, critical
edges whose destinations are not phi nodes don't bother us.  Also, share
split edges, since the split edge can't have a phi.  This significantly
reduces the complexity of generated code in some cases.

llvm-svn: 31274
2006-10-28 19:22:10 +00:00
Jim Laskey eef273a16f Load and stores have not been uniqued properly.
llvm-svn: 31261
2006-10-28 17:25:28 +00:00
Chris Lattner 3e6b1c6157 Split *all* critical edges before isel. This resolves issues with spill code
being inserted on unsplit critical edges, which introduces (sometimes large
amounts of) partially dead spill code.

This also fixes PR925 + CodeGen/Generic/switch-crit-edge-constant.ll

llvm-svn: 31260
2006-10-28 17:04:37 +00:00
Chris Lattner b78eb6c8d1 Fix a serious bug that caused any x86 vector stuff to infinite loop
llvm-svn: 31254
2006-10-28 06:15:26 +00:00
Jim Laskey bd0f088743 Clean up.
llvm-svn: 31243
2006-10-27 23:52:51 +00:00
Chris Lattner 84a035056e Fix a bug in merged condition handling (CodeGen/Generic/2006-10-27-CondFolding.ll).
Add many fewer CFG edges and PHI node entries.  If there is a switch which has
the same block as multiple destinations, only add that block once as a successor/phi
node (in the jumptable case)

llvm-svn: 31242
2006-10-27 23:50:33 +00:00
Jim Laskey f576b42bb2 Switch over from SelectionNodeCSEMap to FoldingSet.
llvm-svn: 31240
2006-10-27 23:46:08 +00:00
Chris Lattner b9392fb635 remove debug code
llvm-svn: 31233
2006-10-27 21:58:03 +00:00
Chris Lattner f1b54fd7a5 Codegen cond&cond with two branches. This compiles (f.e.) PowerPC/and-branch.ll to:
cmpwi cr0, r4, 4
        bgt cr0, LBB1_2 ;UnifiedReturnBlock
LBB1_3: ;entry
        cmplwi cr0, r3, 0
        bne cr0, LBB1_2 ;UnifiedReturnBlock

instead of:

        cmpwi cr7, r4, 4
        mfcr r2
        addic r4, r3, -1
        subfe r3, r4, r3
        rlwinm r2, r2, 30, 31, 31
        or r2, r2, r3
        cmplwi cr0, r2, 0
        bne cr0, LBB1_2 ;UnifiedReturnBlock
LBB1_1: ;cond_true

llvm-svn: 31232
2006-10-27 21:54:23 +00:00
Chris Lattner ed0110b949 Turn conditions like x<Y|z==q into multiple blocks.
This compiles Regression/CodeGen/X86/or-branch.ll into:

_foo:
        subl $12, %esp
        call L_bar$stub
        movl 20(%esp), %eax
        movl 16(%esp), %ecx
        cmpl $5, %eax
        jl LBB1_1       #cond_true
LBB1_3: #entry
        testl %ecx, %ecx
        jne LBB1_2      #UnifiedReturnBlock
LBB1_1: #cond_true
        call L_bar$stub
        addl $12, %esp
        ret
LBB1_2: #UnifiedReturnBlock
        addl $12, %esp
        ret

instead of:

_foo:
        subl $12, %esp
        call L_bar$stub
        movl 20(%esp), %eax
        movl 16(%esp), %ecx
        cmpl $4, %eax
        setg %al
        testl %ecx, %ecx
        setne %cl
        testb %cl, %al
        jne LBB1_2      #UnifiedReturnBlock
LBB1_1: #cond_true
        call L_bar$stub
        addl $12, %esp
        ret
LBB1_2: #UnifiedReturnBlock
        addl $12, %esp
        ret

And on ppc to:

        cmpwi cr0, r29, 5
        blt cr0, LBB1_1 ;cond_true
LBB1_3: ;entry
        cmplwi cr0, r30, 0
        bne cr0, LBB1_2 ;UnifiedReturnBlock

instead of:

        cmpwi cr7, r4, 4
        mfcr r2
        addic r4, r3, -1
        subfe r30, r4, r3
        rlwinm r29, r2, 30, 31, 31
        and r2, r29, r30
        cmplwi cr0, r2, 0
        bne cr0, LBB1_2 ;UnifiedReturnBlock

llvm-svn: 31230
2006-10-27 21:36:01 +00:00
Evan Cheng 96d6bf50ae getPreIndexedLoad -> getIndexedLoad.
llvm-svn: 31209
2006-10-26 21:53:40 +00:00
Reid Spencer 7e80b0b31e For PR950:
Make necessary changes to support DIV -> [SUF]Div. This changes llvm to
have three division instructions: signed, unsigned, floating point. The
bytecode and assembler are bacwards compatible, however.

llvm-svn: 31195
2006-10-26 06:15:43 +00:00
Chris Lattner 61bcf9154d visitSwitchCase knows how to insert conditional branches well. Change
visitBr to just call visitSwitchCase, eliminating duplicate logic.

llvm-svn: 31167
2006-10-24 18:07:37 +00:00