Commit Graph

26855 Commits

Author SHA1 Message Date
Rafael Espindola e04df41ca2 implement a ArgumentLayout class to factor code common to LowerFORMAL_ARGUMENTS and LowerCALL
implement FMDRR
add support for f64 function arguments

llvm-svn: 30754
2006-10-05 16:48:49 +00:00
Jim Laskey 6549d22ef9 Alias analysis code clean ups.
llvm-svn: 30753
2006-10-05 15:07:25 +00:00
Chris Lattner 2deeaeaca7 add a new SimplifyDemandedVectorElts method, which works similarly to
SimplifyDemandedBits.  The idea is that some operations can be simplified if
not all of the computed elements are needed.  Some targets (like x86) have a
large number of intrinsics that operate on a single element, but pass other
elts through unmodified.  If those other elements are not needed, the
intrinsics can be simplified to scalar operations, and insertelement ops can
be removed.

This turns (f.e.):

ushort %Convert_sse(float %f) {
        %tmp = insertelement <4 x float> undef, float %f, uint 0                ; <<4 x float>> [#uses=1]
        %tmp10 = insertelement <4 x float> %tmp, float 0.000000e+00, uint 1             ; <<4 x float>> [#uses=1]
        %tmp11 = insertelement <4 x float> %tmp10, float 0.000000e+00, uint 2           ; <<4 x float>> [#uses=1]
        %tmp12 = insertelement <4 x float> %tmp11, float 0.000000e+00, uint 3           ; <<4 x float>> [#uses=1]
        %tmp28 = tail call <4 x float> %llvm.x86.sse.sub.ss( <4 x float> %tmp12, <4 x float> < float 1.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00 > )               ; <<4 x float>> [#uses=1]
        %tmp37 = tail call <4 x float> %llvm.x86.sse.mul.ss( <4 x float> %tmp28, <4 x float> < float 5.000000e-01, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00 > )               ; <<4 x float>> [#uses=1]
        %tmp48 = tail call <4 x float> %llvm.x86.sse.min.ss( <4 x float> %tmp37, <4 x float> < float 6.553500e+04, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00 > )               ; <<4 x float>> [#uses=1]
        %tmp59 = tail call <4 x float> %llvm.x86.sse.max.ss( <4 x float> %tmp48, <4 x float> zeroinitializer )          ; <<4 x float>> [#uses=1]
        %tmp = tail call int %llvm.x86.sse.cvttss2si( <4 x float> %tmp59 )              ; <int> [#uses=1]
        %tmp69 = cast int %tmp to ushort                ; <ushort> [#uses=1]
        ret ushort %tmp69
}

into:

ushort %Convert_sse(float %f) {
entry:
        %tmp28 = sub float %f, 1.000000e+00             ; <float> [#uses=1]
        %tmp37 = mul float %tmp28, 5.000000e-01         ; <float> [#uses=1]
        %tmp375 = insertelement <4 x float> undef, float %tmp37, uint 0         ; <<4 x float>> [#uses=1]
        %tmp48 = tail call <4 x float> %llvm.x86.sse.min.ss( <4 x float> %tmp375, <4 x float> < float 6.553500e+04, float undef, float undef, float undef > )           ; <<4 x float>> [#uses=1]
        %tmp59 = tail call <4 x float> %llvm.x86.sse.max.ss( <4 x float> %tmp48, <4 x float> < float 0.000000e+00, float undef, float undef, float undef > )            ; <<4 x float>> [#uses=1]
        %tmp = tail call int %llvm.x86.sse.cvttss2si( <4 x float> %tmp59 )              ; <int> [#uses=1]
        %tmp69 = cast int %tmp to ushort                ; <ushort> [#uses=1]
        ret ushort %tmp69
}

which improves codegen from:

_Convert_sse:
        movss LCPI1_0, %xmm0
        movss 4(%esp), %xmm1
        subss %xmm0, %xmm1
        movss LCPI1_1, %xmm0
        mulss %xmm0, %xmm1
        movss LCPI1_2, %xmm0
        minss %xmm0, %xmm1
        xorps %xmm0, %xmm0
        maxss %xmm0, %xmm1
        cvttss2si %xmm1, %eax
        andl $65535, %eax
        ret

to:

_Convert_sse:
        movss 4(%esp), %xmm0
        subss LCPI1_0, %xmm0
        mulss LCPI1_1, %xmm0
        movss LCPI1_2, %xmm1
        minss %xmm1, %xmm0
        xorps %xmm1, %xmm1
        maxss %xmm1, %xmm0
        cvttss2si %xmm0, %eax
        andl $65535, %eax
        ret


This is just a first step, it can be extended in many ways.  Testcase here:
Transforms/InstCombine/vec_demanded_elts.ll

llvm-svn: 30752
2006-10-05 06:55:50 +00:00
Chris Lattner 3d5e9818bd new testcase
llvm-svn: 30751
2006-10-05 06:51:54 +00:00
Chris Lattner 65511ff69d Add insertelement/extractelement helper ctors.
llvm-svn: 30750
2006-10-05 06:24:58 +00:00
Chris Lattner f2ef243580 Lower some min/max idioms to minss/maxss when unsafe fp math is enabled.
llvm-svn: 30748
2006-10-05 04:11:26 +00:00
Andrew Lenharth 16b8f95831 Check that jump tables wind up in the rodata section
llvm-svn: 30747
2006-10-05 03:27:52 +00:00
Chris Lattner 40a95dd347 remove JumpTableTextSection
llvm-svn: 30746
2006-10-05 03:14:23 +00:00
Chris Lattner 8cfd10eff3 Don't bother setting JumpTableTextSection, it is about to disappear
llvm-svn: 30745
2006-10-05 03:13:59 +00:00
Chris Lattner 66c1625a37 Emit pic jumptables to the same section that the function is emitted to,
allowing label differences to work.  This fixes CodeGen/X86/pic_jumptable.ll

llvm-svn: 30744
2006-10-05 03:13:28 +00:00
Chris Lattner bfe59e87e5 Verify that jump tables are emitted to the same section as the function is,
when codegen'ing in pic mode.  This fixes a miscompilation of a switch stmt
in a template, as the template goes to a non-.text section.

llvm-svn: 30743
2006-10-05 03:12:36 +00:00
Chris Lattner a6a570e02f Pass the MachineFunction into EmitJumpTableInfo.
llvm-svn: 30742
2006-10-05 03:01:21 +00:00
Chris Lattner 38e2c8a0a2 implement and use getSectionForFunction
llvm-svn: 30741
2006-10-05 02:51:36 +00:00
Chris Lattner 4431699187 Use getSectionForFunction.
llvm-svn: 30740
2006-10-05 02:49:23 +00:00
Chris Lattner d4d255a408 Use getSectionForFunction
llvm-svn: 30739
2006-10-05 02:48:40 +00:00
Chris Lattner c8c78982d4 use getSectionForFunction to decide which section to emit code into
llvm-svn: 30738
2006-10-05 02:47:13 +00:00
Chris Lattner b82247b168 Implement getSectionForFunction, use it when printing function body.
llvm-svn: 30737
2006-10-05 02:43:52 +00:00
Chris Lattner dc82241182 move getSectionForFunction to AsmPrinter
llvm-svn: 30736
2006-10-05 02:42:47 +00:00
Chris Lattner 028d663ee6 Move getSectionForFunction to AsmPrinter, change it to return a string.
llvm-svn: 30735
2006-10-05 02:42:20 +00:00
Chris Lattner 0dca927148 move getSectionForFunction to AsmPrinter.
llvm-svn: 30734
2006-10-05 02:41:43 +00:00
Chris Lattner 0d236450aa implement DarwinTargetAsmInfo::getSectionForFunction, use it when outputting
function bodies

llvm-svn: 30733
2006-10-05 00:35:50 +00:00
Chris Lattner afe6d7a179 Give TargetAsmInfo a virtual dtor, add a new getSectionForFunction method.
llvm-svn: 30732
2006-10-05 00:35:16 +00:00
Chris Lattner 41e22a5419 emit jump table before debug info
llvm-svn: 30731
2006-10-05 00:26:05 +00:00
Chris Lattner aad26a19f0 Always emit the jump table after the function so it's part of the same 'atom'
as the function body.

llvm-svn: 30730
2006-10-05 00:24:46 +00:00
Chris Lattner 19721e8749 getFilename/getDirectory shouldn't abort if the global has no init. This
can happen on bugpoint reduced testcases f.e..

llvm-svn: 30729
2006-10-04 23:06:26 +00:00
Evan Cheng f80dfa83a0 Fix some typos that can cause a flag value to have more than one use.
llvm-svn: 30727
2006-10-04 22:23:53 +00:00
Chris Lattner c374ec43b5 Fix a static dtor issue
llvm-svn: 30726
2006-10-04 22:13:11 +00:00
Chris Lattner 8111c59279 Fix more static dtor issues
llvm-svn: 30725
2006-10-04 21:52:35 +00:00
Chris Lattner 538c6eb05c Fix some more static dtor issues.
llvm-svn: 30724
2006-10-04 21:49:37 +00:00
Evan Cheng 8c5766ef3f Added option -disable-x86-shuffle-opti to disable X86 specific vector shuffle optimizations.
llvm-svn: 30723
2006-10-04 18:33:38 +00:00
Evan Cheng 412aaabcbe Formating.
llvm-svn: 30722
2006-10-04 18:33:00 +00:00
Jim Laskey 708d0db2d8 More extensive alias analysis.
llvm-svn: 30721
2006-10-04 16:53:27 +00:00
Jim Laskey 0d5a0eae57 More long term solution
llvm-svn: 30720
2006-10-04 10:40:15 +00:00
Chris Lattner 9259b1efb6 Pattern match min/max nodes when we have sse. This implements
CodeGen/X86/scalar_sse_minmax.ll

llvm-svn: 30719
2006-10-04 06:57:07 +00:00
Chris Lattner 1e21d3a5ae pattern match min/max nodes
llvm-svn: 30718
2006-10-04 06:56:02 +00:00
Chris Lattner 3e11d99a0a add a note :(
llvm-svn: 30717
2006-10-04 05:52:13 +00:00
Chris Lattner 52886e72d7 This case isn't implemented yet. It seems unlikely to be needed, but if it
ever is, we want to get an assert instead of silent bad codegen.

llvm-svn: 30716
2006-10-04 04:58:58 +00:00
Jim Laskey 66b0b55816 Work around for some problems with templates.
llvm-svn: 30715
2006-10-04 01:43:13 +00:00
Evan Cheng 5d9fd977d3 Combine ISD::EXTLOAD, ISD::SEXTLOAD, ISD::ZEXTLOAD into ISD::LOADX. Add an
extra operand to LOADX to specify the exact value extension type.

llvm-svn: 30714
2006-10-04 00:56:09 +00:00
Chris Lattner fc416faaa0 Use $( $| $) to represent alternatives in asm blocks instead of {|}. This
is needed to support targets where {|} aren't special symbols.

llvm-svn: 30712
2006-10-03 23:27:09 +00:00
Evan Cheng 91d76cb27f Fix an obvious typo.
llvm-svn: 30711
2006-10-03 23:08:27 +00:00
Chris Lattner bd7286e606 Bugfixes
llvm-svn: 30709
2006-10-03 20:19:23 +00:00
Chris Lattner de154b1737 Print the MBB ID # along with the bb tag in the -print-machine-instrs output.
llvm-svn: 30708
2006-10-03 20:17:24 +00:00
Chris Lattner 9744b68d57 ADd a method
llvm-svn: 30707
2006-10-03 20:16:45 +00:00
Chris Lattner 64fd9487bd Provide a function that ensures MBB numbering is dense and inorder. This
can be used by MachineFunctionPasses who need this property.

llvm-svn: 30706
2006-10-03 19:18:57 +00:00
Nick Lewycky 58a910dff5 Simplify logic further.
Ensure that we copy KnownProperties before calling visitBasicBlock, else
we may leak properties into blocks where they don't belong.

llvm-svn: 30705
2006-10-03 17:36:01 +00:00
Rafael Espindola 68d238801c Implement floating point constants
llvm-svn: 30704
2006-10-03 17:27:58 +00:00
Chris Lattner f598d73142 Fix PR933 and CodeGen/X86/2006-10-02-BoolRetCrash.ll
llvm-svn: 30703
2006-10-03 17:18:42 +00:00
Chris Lattner 70a12ec8f8 testcase for PR933
llvm-svn: 30702
2006-10-03 17:17:54 +00:00
Chris Lattner e385fcf098 clean up use of 'explicit'. This is PR934.
Patch contributed by Kevin Sopp!

llvm-svn: 30701
2006-10-03 17:09:12 +00:00
Nick Lewycky 1d00f3e144 Simplify, now that predsimplify depends on break-crit-edges.
Fix SwitchInst where dest-block is the same as one of the cases.

llvm-svn: 30700
2006-10-03 15:19:11 +00:00
Nick Lewycky 755f801adc Move break-crit-edges before the predicate simplifier. Allows us to
optimize in more cases.

llvm-svn: 30699
2006-10-03 14:52:23 +00:00
Evan Cheng ff510a58c2 Revert previous patch. Still breaking things.
llvm-svn: 30698
2006-10-03 07:26:07 +00:00
Bill Wendling 984f0ce06b Fix for PR929. The PHI nodes were being gone through for each instruction
in a successor block for every block...resulting in some O(N^k) algorithm
which wasn't very good for performance. Calculating this information up
front and keeping it in a map made it much faster.

llvm-svn: 30697
2006-10-03 07:20:20 +00:00
Chris Lattner 8aca0ee8c3 Fix PR932 and Analysis/Dominators/2006-10-02-BreakCritEdges.ll:
The critical edge block dominates the dest block if the destblock dominates
all edges other than the one incoming from the critical edge.

llvm-svn: 30696
2006-10-03 07:02:02 +00:00
Chris Lattner 2a7e3c29a1 New testcase for PR932
llvm-svn: 30695
2006-10-03 07:00:13 +00:00
Evan Cheng a36e6cf44f These don't have immediate operands.
llvm-svn: 30694
2006-10-03 06:55:11 +00:00
Chris Lattner 1faa3f913c Move DominatorTree to immediately follow DominatorTreeBase
llvm-svn: 30693
2006-10-03 05:24:56 +00:00
Rafael Espindola d55c0a41df fix the names of the 64bit fp register
initial support for returning 64bit floating point numbers

llvm-svn: 30692
2006-10-02 19:30:56 +00:00
Reid Spencer 62ad409c2b Don't build the runtime library if LLVMGCC is not configured.
llvm-svn: 30691
2006-10-02 19:10:56 +00:00
Chris Lattner fb07457db0 fix misleading warning
llvm-svn: 30690
2006-10-02 17:24:55 +00:00
Chris Lattner 7a7081f3d0 Fix misleading warning message
llvm-svn: 30689
2006-10-02 17:23:42 +00:00
Jim Laskey e73a22514d Debugging kruft
llvm-svn: 30688
2006-10-02 13:01:17 +00:00
Jim Laskey 37c2c6c860 Update documentaion of node annotation (coloring) in viewGraph.
llvm-svn: 30687
2006-10-02 12:28:07 +00:00
Jim Laskey 1368c265da Add ability to annotate (color) nodes in a viewGraph.
llvm-svn: 30686
2006-10-02 12:26:53 +00:00
Chris Lattner 134ff353f9 Fixme is already done
llvm-svn: 30685
2006-10-01 22:46:33 +00:00
Chris Lattner bfdd19bc01 Teach globalsmodref-aa to track scalar pointer global variables which point
to unaliased allocations.  Use this information to disambiguate pointers loaded
from them.  This is a very common case, so it's worthwhile to handle efficiently.

This implements Analysis/GlobalsModRef/indirect-global.ll

llvm-svn: 30684
2006-10-01 22:36:45 +00:00
Chris Lattner bfc049ae3f New testcase
llvm-svn: 30683
2006-10-01 22:35:45 +00:00
Chris Lattner 7d19067c42 Fix a bug from r1.391 of this file, where we checked the size instead of
the alignment when promoting allocations.  This implements
InstCombine/cast.ll:test32

llvm-svn: 30682
2006-10-01 19:40:58 +00:00
Chris Lattner c82477f3a8 new testcase, malloc should be promoted to [2 x double].
llvm-svn: 30681
2006-10-01 19:39:45 +00:00
Chris Lattner 4797c891c0 Fix debug output
llvm-svn: 30680
2006-09-30 23:32:50 +00:00
Chris Lattner 24d3d4280a Implement SRA of heap allocations.
llvm-svn: 30679
2006-09-30 23:32:09 +00:00
Chris Lattner 456a806692 Override use_back in instruction/basicblock to provide more type information.
llvm-svn: 30678
2006-09-30 22:20:34 +00:00
Chris Lattner 87732cfb71 Add a version of the globalvariable ctor that inserts at a specific location.
llvm-svn: 30677
2006-09-30 21:31:26 +00:00
Chris Lattner 80a01ef6f0 Add some ifdef'd out debug info
llvm-svn: 30676
2006-09-30 19:40:30 +00:00
Chris Lattner eb45775347 Stacker doesn't build universal right.
llvm-svn: 30675
2006-09-30 00:24:20 +00:00
Evan Cheng 1212b4d249 Not needed.
llvm-svn: 30674
2006-09-29 22:05:10 +00:00
Rafael Espindola 53f78be49e add floating point registers
implement SINT_TO_FP

llvm-svn: 30673
2006-09-29 21:20:16 +00:00
Chris Lattner 008b2de1fc Move CompileCommonOpts to the end of the list so that EXTRA_OPTIONS=-O0 will
override the optimization level.

llvm-svn: 30672
2006-09-29 18:47:13 +00:00
Chris Lattner adf4095338 Define this in the correct n/s
llvm-svn: 30671
2006-09-29 18:43:14 +00:00
Chris Lattner 754c8cf702 Adjust this to the wonky syntax that GCC expects.
llvm-svn: 30670
2006-09-29 17:34:56 +00:00
Jim Laskey 05a1947a31 Wrong directory.
llvm-svn: 30669
2006-09-29 17:31:45 +00:00
Chris Lattner b10b72e362 Minor cleanups
llvm-svn: 30668
2006-09-28 23:45:00 +00:00
Chris Lattner befaee91d3 Now that ConstantBool::True/False are gone, we can modify Type.cpp to
eliminate its static dtors, without having code that depends on order of
initialization.  Eliminate static ctors/dtors from Type.cpp.

llvm-svn: 30667
2006-09-28 23:38:07 +00:00
Chris Lattner a84df0a2f1 Eliminate ConstantBool::True and ConstantBool::False. Instead, provide
ConstantBool::getTrue() and ConstantBool::getFalse().

llvm-svn: 30666
2006-09-28 23:36:21 +00:00
Chris Lattner 6ab03f6a08 Eliminate ConstantBool::True and ConstantBool::False. Instead, provide
ConstantBool::getTrue() and ConstantBool::getFalse().

llvm-svn: 30665
2006-09-28 23:35:22 +00:00
Chris Lattner 7843066e15 Eliminate ConstantBool::True and ConstantBool::False. Instead, provideConstantBool::getTrue() and ConstantBool::getFalse().
llvm-svn: 30664
2006-09-28 23:34:49 +00:00
Chris Lattner d9e4bf5285 update comments
llvm-svn: 30663
2006-09-28 23:33:12 +00:00
Chris Lattner d32cb5ee7f wrap long lines
llvm-svn: 30662
2006-09-28 23:32:43 +00:00
Chris Lattner ddfe6998de simplify code
llvm-svn: 30661
2006-09-28 23:24:48 +00:00
Chris Lattner 0045295c18 Simplify some code, reformat break's
llvm-svn: 30660
2006-09-28 23:19:29 +00:00
Chris Lattner eea983eb5d simplify code
llvm-svn: 30659
2006-09-28 23:17:41 +00:00
Chris Lattner a7b5664318 Simplify some code
llvm-svn: 30658
2006-09-28 23:14:29 +00:00
Owen Anderson 7cb6809c25 Another attempt at making ArgPromotion smarter. This patch no longer breaks Burg.
llvm-svn: 30657
2006-09-28 23:02:22 +00:00
Chris Lattner 525804f31e simplify code
llvm-svn: 30656
2006-09-28 22:58:25 +00:00
Chris Lattner 94e39cbdfe simplify code
llvm-svn: 30655
2006-09-28 22:50:29 +00:00
Reid Spencer 5f2b79ec4c Use the -emit-llvm switch to generate LLVM assembly that can be parsed
by the test case.

llvm-svn: 30654
2006-09-28 21:36:21 +00:00
Reid Spencer 4fde7faeb0 Add a test case for PR902.
llvm-svn: 30653
2006-09-28 21:20:05 +00:00
Chris Lattner 78370606d0 Shift amounts are always 32-bits, even in 64-bit mode. This fixes
CodeGen/PowerPC/2006-09-28-shift_64.ll

llvm-svn: 30652
2006-09-28 20:48:45 +00:00
Chris Lattner a3542a82be new testcase
llvm-svn: 30651
2006-09-28 20:48:17 +00:00
Reid Spencer 309080abbc Fix PR902:
Errors are generated with the YYERROR macro which can only be called from
a production (inside yyparse) because of the goto statement in the macro.
This lead to several situations where GEN_ERROR was not called but
GenerateError was used instead (because it doesn't use YYERROR). However,
in such situations, catching the error much later (e.g. at the end of
the production) is not sufficient because LLVM can assert on invalid data
before the end of the production is reached. The solution is to ensure that
the CHECK_FOR_ERROR macro (which invokes YYERROR if there's an error) is
used as soon as possible after a call to GenerateError has been made.

llvm-svn: 30650
2006-09-28 19:28:24 +00:00
Chris Lattner 83fb402291 Testcase for PR924
llvm-svn: 30649
2006-09-28 18:58:02 +00:00
Evan Cheng bcf1b4120f TargetRegisterClass specifies the desired spill alignment. However, it cannot be honored if stack alignment is smaller.
llvm-svn: 30648
2006-09-28 18:52:32 +00:00
Jim Laskey 8f1510cfd3 Fix search file for -release.
llvm-svn: 30647
2006-09-28 18:45:11 +00:00
Jim Laskey 68722b7e08 Add support for -release-asserts.
llvm-svn: 30646
2006-09-28 17:49:20 +00:00
Reid Spencer 48cf6a6f21 Provide a gdb usage fix provided by Zhongzing Xu.
llvm-svn: 30645
2006-09-28 16:53:47 +00:00
Bill Wendling 5d409822d6 "Once more into the breach, dear friends, once more, or fill the wall up
with our English dead."

No! Really! Serious this time...It was how the vreg uses were being
adjusted that was causing hte Olden tests to fail. I corrected this and
the Olden and Regression tests all passed.

llvm-svn: 30644
2006-09-28 07:10:24 +00:00
Chris Lattner a9caf95591 refactor critical edge breaking out into the SplitCritEdgesForPHIConstants method.
This is a baby step towards fixing PR925.

llvm-svn: 30643
2006-09-28 06:17:10 +00:00
Chris Lattner 3cbd1604f8 add a note about a general improvement to the code generator
llvm-svn: 30642
2006-09-28 06:01:17 +00:00
Chris Lattner c7d3195254 remove reference to dead method
llvm-svn: 30641
2006-09-28 00:38:19 +00:00
Chris Lattner b044ab2b82 remove dead method
llvm-svn: 30640
2006-09-28 00:37:43 +00:00
Chris Lattner 69edc98b74 Use the new ManagedStatic class to explicitly manage static variables, eliminating static ctors/dtors
llvm-svn: 30639
2006-09-28 00:35:06 +00:00
Chris Lattner 9daff49ebb new helper class to provide more explicit management of static ctor/dtors.
llvm-svn: 30638
2006-09-28 00:31:55 +00:00
Chris Lattner 764caf6129 re-re-revert this, back to the right revision. It currently breaks bisort/mst
in olden among others.

llvm-svn: 30637
2006-09-28 00:11:54 +00:00
Evan Cheng d31f55c236 PEI now place callee save spills closest to the address pointed to by the
incoming stack. This allows X86 backend to use push / pop in epilogue /
prologue.

llvm-svn: 30636
2006-09-28 00:10:27 +00:00
Evan Cheng 17a8d08dba - Added a hook processFunctionBeforeCalleeSaveScn(). This is called by PEI just
before it determines which callee-save registers are to be spilled. This allows
the target to make changes such as forcing certain physical registers to be
spilled.
- Modified comments. It's important to note the order of registers in the array
returns by getCalleeSaveRegs() determines the order of callee save spill code.

llvm-svn: 30635
2006-09-28 00:07:19 +00:00
Chris Lattner 9f7d0d1dc5 re-revert this patch, bisort and mst are still broken in Olden.
llvm-svn: 30634
2006-09-28 00:04:21 +00:00
Bill Wendling 1f29e6c5c5 Reapplying this patch. With the newest commits, the error in Olden/bisort
has disappeared.

llvm-svn: 30633
2006-09-27 22:37:35 +00:00
Chris Lattner fc36039f86 silence warnings in release build
llvm-svn: 30631
2006-09-27 18:29:38 +00:00
Devang Patel 0a79537341 Fix DFS walk.
Fix http://llvm.org/bugs/show_bug.cgi?id=923

llvm-svn: 30630
2006-09-27 17:18:05 +00:00
Chris Lattner 5f6c93701b Temporarily revert this. This breaks Olden/bisort on PPC
llvm-svn: 30628
2006-09-27 16:59:16 +00:00
Jim Laskey 3bbf884855 Make sure C++ protection shows up in debug info
llvm-svn: 30626
2006-09-27 16:55:19 +00:00
Chris Lattner 5d751b4a3a This:
AggregateString += "\0\0";

Doesn't add two nuls to the AggregateString (for obvious reasons), which
broke the asmprinter when the first character of an asm string was not
literal text.

llvm-svn: 30625
2006-09-27 16:44:09 +00:00
Bill Wendling 0a7f617a6c PR878: Instead of calculating the vreg to PHI use count everytime we get
a function, do it up front in linear time (going through all of the
instructions once). We create a map out of them. Then it's no problem to
use the information in it during elimination...

llvm-svn: 30624
2006-09-27 09:04:15 +00:00
Chris Lattner e03ca2ca4a set DEBUG_TYPE right
llvm-svn: 30623
2006-09-27 04:58:23 +00:00
Devang Patel ce73ac3089 Add http://llvm.org/bugs/show_bug.cgi?id=923 test case.
llvm-svn: 30622
2006-09-27 02:58:44 +00:00
Chris Lattner 67f8cc51f4 Use abstract private/comment directives, to increase portability to ppc/linux
llvm-svn: 30621
2006-09-27 02:55:21 +00:00
Chris Lattner 30b4708df2 Add support for ${:private} which prints "L" on darwin.
llvm-svn: 30620
2006-09-27 00:06:07 +00:00
Chris Lattner a32814b033 Add support for ${:comment}, which expands to the current target's comment
character, and ${:uid} which expands to a unique ID for the MachineInstr.
More can be added if/when they are needed.

llvm-svn: 30619
2006-09-26 23:59:50 +00:00
Chris Lattner 8d21485660 Actually, name the method PrintSpecial to match other stuff in AsmPrinter.h
llvm-svn: 30618
2006-09-26 23:47:10 +00:00
Chris Lattner 5fd0154523 Add support for ${:foo} syntax, where "foo" is passed into "printSpecial" and
has no associated operand.  This is useful for portably encoding stuff like
the comment character into an asm string.

llvm-svn: 30617
2006-09-26 23:45:08 +00:00
Evan Cheng 5405c06737 Rename function. It's determining which callee-save registers to save.
llvm-svn: 30616
2006-09-26 22:29:31 +00:00
Andrew Lenharth c19ef92403 Comments on JumpTableness
llvm-svn: 30615
2006-09-26 20:02:30 +00:00
Jim Laskey 60832693a7 Load chain check is not needed
llvm-svn: 30613
2006-09-26 17:44:58 +00:00
Jim Laskey dde51671e5 Chain can be any operand
llvm-svn: 30611
2006-09-26 09:32:41 +00:00
Jim Laskey 5f3e0af9d0 Wrong size for load
llvm-svn: 30610
2006-09-26 08:14:06 +00:00
Jim Laskey b4a864d533 Can't move a load node if it's chain is not used.
llvm-svn: 30609
2006-09-26 07:37:42 +00:00
Chris Lattner 104aa5dbc1 Various random and minor code cleanups.
llvm-svn: 30608
2006-09-26 03:57:53 +00:00
Chris Lattner ae7ff8b0c6 test that the no_dead_strip directive is emitted on darwin-x86
llvm-svn: 30607
2006-09-26 03:44:20 +00:00
Chris Lattner 9a1e91b107 print the preds of each MBB
llvm-svn: 30606
2006-09-26 03:41:59 +00:00
Chris Lattner ad4e7eb59e Compile:
int x __attribute__((used));

to:

        .data
.comm _x,4              ; 'x'
        .no_dead_strip  _x

on both x86 and ppc darwin targets.

llvm-svn: 30605
2006-09-26 03:39:53 +00:00
Chris Lattner 66af390631 Add support for targets that want to do something with the llvm.used list,
because they have an aggressive linker that does dead code stripping.

llvm-svn: 30604
2006-09-26 03:38:18 +00:00
Chris Lattner 991d3997cf order this properly to avoid warnings in TargetAsmInfo.cpp. Add a comment
in a format that matches every other ivars in this class.

llvm-svn: 30603
2006-09-25 22:38:36 +00:00
Jim Laskey 7aa0638aa9 Accidental enable of bad code
llvm-svn: 30601
2006-09-25 21:11:32 +00:00
Jim Laskey b5534e5c28 Fix chain dropping in load and drop unused stores in ret blocks.
llvm-svn: 30600
2006-09-25 19:32:58 +00:00
John Criswell 437d9a3fb2 Regression test for PR#922.
llvm-svn: 30599
2006-09-25 19:12:01 +00:00
Chris Lattner f11327daab more notes
llvm-svn: 30598
2006-09-25 17:12:14 +00:00
Jim Laskey d07be232ba Core antialiasing for load and store.
llvm-svn: 30597
2006-09-25 16:29:54 +00:00
Andrew Lenharth 6d8e227ac7 Fix
llvm-svn: 30596
2006-09-25 14:04:53 +00:00
Andrew Lenharth 5e2bacd6be Fix jump tables to match gcc (and the ABI and whatnot)
llvm-svn: 30594
2006-09-24 19:46:56 +00:00
Andrew Lenharth 783a4a9d86 Add support for other relocation bases to jump tables, as well as custom asm directives
llvm-svn: 30593
2006-09-24 19:45:58 +00:00
Andrew Lenharth 3cec4952ab basic jump table test
llvm-svn: 30592
2006-09-24 19:42:02 +00:00
Andrew Lenharth 68324f8f85 jump table note
llvm-svn: 30591
2006-09-24 13:13:10 +00:00
Evan Cheng 77c0757f8b PIC jump table entries are always 32-bit. This fixes PIC jump table support on X86-64.
llvm-svn: 30590
2006-09-24 05:22:38 +00:00
Nick Lewycky 25587182f3 Enable 'predsimplify' optimization.
llvm-svn: 30589
2006-09-24 00:08:16 +00:00
Nick Lewycky 059c79264f Style changes only. Remove dead code, fix a comment.
llvm-svn: 30588
2006-09-23 15:13:08 +00:00
Chris Lattner 6bd6da4097 Be far more careful when splitting a loop header, either to form a preheader
or when splitting loops with a common header into multiple loops.  In particular
the old code would always insert the preheader before the old loop header.  This
is disasterous in cases where the loop hasn't been rotated.  For example, it can
produce code like:

        .. outside the loop...
        jmp LBB1_2      #bb13.outer
LBB1_1: #bb1
        movsd 8(%esp,%esi,8), %xmm1
        mulsd (%edi), %xmm1
        addsd %xmm0, %xmm1
        addl $24, %edi
        incl %esi
        jmp LBB1_3      #bb13
LBB1_2: #bb13.outer
        leal (%edx,%eax,8), %edi
        pxor %xmm1, %xmm1
        xorl %esi, %esi
LBB1_3: #bb13
        movapd %xmm1, %xmm0
        cmpl $4, %esi
        jl LBB1_1       #bb1

Note that the loop body is actually LBB1_1 + LBB1_3, which means that the
loop now contains an uncond branch WITHIN it to jump around the inserted
loop header (LBB1_2).  Doh.

This patch changes the preheader insertion code to insert it in the right
spot, producing this code:

        ... outside the loop, fall into the header ...
LBB1_1: #bb13.outer
        leal (%edx,%eax,8), %esi
        pxor %xmm0, %xmm0
        xorl %edi, %edi
        jmp LBB1_3      #bb13
LBB1_2: #bb1
        movsd 8(%esp,%edi,8), %xmm0
        mulsd (%esi), %xmm0
        addsd %xmm1, %xmm0
        addl $24, %esi
        incl %edi
LBB1_3: #bb13
        movapd %xmm0, %xmm1
        cmpl $4, %edi
        jl LBB1_2       #bb1

Totally crazy, no branch in the loop! :)

llvm-svn: 30587
2006-09-23 08:19:21 +00:00
Chris Lattner 608cd05e3f Teach UpdateDomInfoForRevectoredPreds to handle revectored preds that are not
reachable, making it general purpose enough for use by InsertPreheaderForLoop.
Eliminate custom dominfo updating code in InsertPreheaderForLoop, using
UpdateDomInfoForRevectoredPreds instead.

llvm-svn: 30586
2006-09-23 07:40:52 +00:00
Chris Lattner e4583e8c2c remove misleading comment
llvm-svn: 30585
2006-09-23 06:09:45 +00:00
Chris Lattner 4091f4690a add method, correct comment
llvm-svn: 30584
2006-09-23 04:03:45 +00:00
Evan Cheng 1da0ab2f58 Delete dead code; fix 80 col violations.
llvm-svn: 30583
2006-09-22 21:43:59 +00:00
Rafael Espindola 72d4c070c0 add a note
llvm-svn: 30581
2006-09-22 11:36:17 +00:00
Bill Wendling fc0e612b65 Use "%llvmgxx" instead of "llvm-g++".
llvm-svn: 30580
2006-09-22 07:09:01 +00:00
Nate Begeman 9007f5a48d Testcase for better rotate left and mask support
llvm-svn: 30579
2006-09-22 05:49:57 +00:00
Bill Wendling fe95123f58 Regression testcase for PR855.
llvm-svn: 30578
2006-09-22 05:34:25 +00:00
Nate Begeman d31efd190f Fold AND and ROTL more often
llvm-svn: 30577
2006-09-22 05:01:56 +00:00
Devang Patel 81c9e42bea remove extra white spaces.
llvm-svn: 30576
2006-09-22 01:07:57 +00:00
Devang Patel 0c4e730c9c Use iterative algorith to assign DFS number. This reduces
call stack depth.

llvm-svn: 30575
2006-09-22 01:05:33 +00:00
Chris Lattner 7d3fd4f888 don't allow 'imm' or specific imms, like '1' on the LHS of a binop.
This shrinks X86GenDAGISel by ~330 lines.

llvm-svn: 30574
2006-09-21 20:46:13 +00:00
Evan Cheng 449a0c7e33 Make it work for DAG combine of multi-value nodes.
llvm-svn: 30573
2006-09-21 19:04:05 +00:00
Chris Lattner 99ae714459 Fit to 80 columns.
llvm-svn: 30572
2006-09-21 18:28:27 +00:00
Jim Laskey 35f7eebb49 core corrections
llvm-svn: 30570
2006-09-21 17:35:47 +00:00
Devang Patel 5af1651f7d Use abstract class to facilitate dlopen() interface.
llvm-svn: 30569
2006-09-21 17:22:55 +00:00
Jim Laskey 5d19d59017 Basic "in frame" alias analysis.
llvm-svn: 30568
2006-09-21 16:28:59 +00:00
Rafael Espindola 7b700e517a more condition codes
llvm-svn: 30567
2006-09-21 13:06:26 +00:00
Rafael Espindola 0c71a5adc8 if a constant can't be an immediate, add it to the constant pool
llvm-svn: 30566
2006-09-21 11:29:52 +00:00
Chris Lattner 082db3f9aa fold (aext (and (trunc x), cst)) -> (and x, cst).
llvm-svn: 30561
2006-09-21 06:40:43 +00:00
Chris Lattner fa9f92cf65 Check the right value type. This fixes 186.crafty on x86
llvm-svn: 30560
2006-09-21 06:17:39 +00:00
Chris Lattner 08a8ccaaf1 implemented
llvm-svn: 30559
2006-09-21 06:14:54 +00:00
Chris Lattner 8d8a3bf9c9 Compile:
int %test(ulong *%tmp) {
        %tmp = load ulong* %tmp         ; <ulong> [#uses=1]
        %tmp.mask = shr ulong %tmp, ubyte 50            ; <ulong> [#uses=1]
        %tmp.mask = cast ulong %tmp.mask to ubyte
        %tmp2 = and ubyte %tmp.mask, 3          ; <ubyte> [#uses=1]
        %tmp2 = cast ubyte %tmp2 to int         ; <int> [#uses=1]
        ret int %tmp2
}

to:

_test:
        movl 4(%esp), %eax
        movl 4(%eax), %eax
        shrl $18, %eax
        andl $3, %eax
        ret

instead of:

_test:
        movl 4(%esp), %eax
        movl 4(%eax), %eax
        shrl $18, %eax
        # TRUNCATE movb %al, %al
        andb $3, %al
        movzbl %al, %eax
        ret

llvm-svn: 30558
2006-09-21 06:14:31 +00:00
Chris Lattner a31f0a622b Generalize (zext (truncate x)) and (sext (truncate x)) folding to work when
the src/dst are not the same size.  This catches things like "truncate
32-bit X to 8 bits, then zext to 16", which happens a bit on X86.

llvm-svn: 30557
2006-09-21 06:00:20 +00:00
Chris Lattner 1c18c0db79 Fit in 80-cols
llvm-svn: 30556
2006-09-21 05:46:00 +00:00
Chris Lattner 51c95cdd82 Fix Transforms/IndVarsSimplify/2006-09-20-LFTR-Crash.ll
llvm-svn: 30555
2006-09-21 05:12:20 +00:00
Chris Lattner 6597e053e1 new testcase
llvm-svn: 30554
2006-09-21 05:11:01 +00:00
Nick Lewycky c68bbef874 Fix compile error.
llvm-svn: 30553
2006-09-21 02:08:31 +00:00
Nick Lewycky fde9c308b2 Don't rewrite ConstantExpr::get.
llvm-svn: 30552
2006-09-21 01:05:35 +00:00
Nick Lewycky d74c55f483 Once we're down to "setcc type constant1, constant2", at least come up
with the right answer.

llvm-svn: 30550
2006-09-20 23:02:24 +00:00
Anton Korobeynikov 3c5b3df6a0 Adding codegeneration for StdCall & FastCall calling conventions
llvm-svn: 30549
2006-09-20 22:03:51 +00:00
Andrew Lenharth ccdaecc448 Account for pseudo-ops correctly
llvm-svn: 30548
2006-09-20 20:08:52 +00:00
Chris Lattner a81a75c390 The DarwinAsmPrinter need not check for isDarwin. createPPCAsmPrinterPass
should create the right asmprinter subclass.

llvm-svn: 30542
2006-09-20 17:12:19 +00:00
Chris Lattner 8597a2fc4e Wrap some darwin'isms with isDarwin checks.
llvm-svn: 30541
2006-09-20 17:07:15 +00:00
Nick Lewycky cfff1c3f86 Use a total ordering to compare instructions.
Fixes infinite loop in resolve().

llvm-svn: 30540
2006-09-20 17:04:01 +00:00
Andrew Lenharth 44cb67af5c simplify
llvm-svn: 30535
2006-09-20 15:37:57 +00:00
Andrew Lenharth f007f21c8a catch constants more often
llvm-svn: 30534
2006-09-20 15:05:49 +00:00
Andrew Lenharth b04e899bb4 catch another constant
llvm-svn: 30533
2006-09-20 15:04:55 +00:00
Andrew Lenharth 97a4e99aff clarify with test case
llvm-svn: 30531
2006-09-20 14:48:00 +00:00
Andrew Lenharth e2d138a462 Add Note
llvm-svn: 30530
2006-09-20 14:40:01 +00:00
Jim Laskey fbeab72759 Trim the home directory from the dejagnu test
llvm-svn: 30519
2006-09-20 09:20:22 +00:00
Chris Lattner fba9e8f422 item done
llvm-svn: 30518
2006-09-20 06:41:56 +00:00
Chris Lattner c8cd62d381 Compile:
int test3(int a, int b) { return (a < 0) ? a : 0; }

to:

_test3:
        srawi r2, r3, 31
        and r3, r2, r3
        blr

instead of:

_test3:
        cmpwi cr0, r3, 1
        li r2, 0
        blt cr0, LBB2_2 ;entry
LBB2_1: ;entry
        mr r3, r2
LBB2_2: ;entry
        blr


This implements: PowerPC/select_lt0.ll:seli32_a_a

llvm-svn: 30517
2006-09-20 06:41:35 +00:00