RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-28 14:32:51 +01:00
Code Issues Projects Releases Wiki Activity
37,686 Commits 4 Branches 5 Tags 704 MiB
1d70509b55
Commit Graph

418 Commits

Author SHA1 Message Date
Chris Lattner
41e1fd13b2 My previous commit had an incomplete message, it should have been: 
make the 'fp return in ST(0)' optimization smart enough to
look through token factor nodes.  THis allows us to compile 
testcases like CodeGen/X86/fp-stack-retcopy.ll into:

_carg:
	subl	$12, %esp
	call	L_foo$stub
	fstpl	(%esp)
	fldl	(%esp)
	addl	$12, %esp
	ret

instead of:

_carg:
	subl	$28, %esp
	call	L_foo$stub
	fstpl	16(%esp)
	movsd	16(%esp), %xmm0
	movsd	%xmm0, 8(%esp)
	fldl	8(%esp)
	addl	$28, %esp
	ret

Still not optimal, but much better and this is a trivial patch.  Fixing 
the rest requires invasive surgery that is is not llvm 2.2 material.

llvm-svn: 46054
 2008-01-16 05:56:59 +00:00
Chris Lattner
afd4056065 verify x86 generates ud2 for llvm.trap 
llvm-svn: 46023
 2008-01-15 22:22:02 +00:00
Dale Johannesen
8ca78844b0 Disable for now. 
llvm-svn: 45881
 2008-01-11 20:47:33 +00:00
Duncan Sands
2c89976416 Output sinl for a long double FSIN node, not sin. 
Likewise fix up a bunch of other libcalls.  While
there I remove NEG_F32 and NEG_F64 since they are
not used anywhere.  This fixes 9 Ada ACATS failures.

llvm-svn: 45833
 2008-01-10 10:28:30 +00:00
Evan Cheng
0747381b13 Codegen improvement has reduced one spill. 
llvm-svn: 45814
 2008-01-10 02:54:40 +00:00
Evan Cheng
ba0214a6cb Special copy SUnit's do not have SDNode's. 
llvm-svn: 45787
 2008-01-09 23:01:55 +00:00
Evan Cheng
f91cfb435f Fix sse2.psrl.w and sse2.psrl.q definitions. 
llvm-svn: 45772
 2008-01-09 02:16:44 +00:00
Chris Lattner
c93ad7d569 Make load->store deletion a bit smarter. This allows us to compile this: 
void test(long long *P) { *P ^= 1; }

into just:

_test:
	movl	4(%esp), %eax
	xorl	$1, (%eax)
	ret

instead of code like this:

_test:
	movl	4(%esp), %ecx
        xorl    $1, (%ecx)
	movl	4(%ecx), %edx
	movl	%edx, 4(%ecx)
	ret

llvm-svn: 45762
 2008-01-08 23:08:06 +00:00
Duncan Sands
b3b1ae18ab Crashes llc when using Chris's new legalization logic. 
llvm-svn: 45758
 2008-01-08 21:51:53 +00:00
Nate Begeman
98dba4b0ce Update test to catch recent x86 insert regression and improvements 
llvm-svn: 45705
 2008-01-07 17:49:23 +00:00
Chris Lattner
7d567adef9 fix this to use a valid triple. 
llvm-svn: 45509
 2008-01-02 22:21:45 +00:00
Chris Lattner
fbd8cc03c8 verify that aligned common support doesn't break. 
llvm-svn: 45495
 2008-01-02 19:48:24 +00:00
Chris Lattner
d55e743cfe One readme entry is done, one is really easy (Evan, want to investigate 
eliminating the llvm.x86.sse2.loadl.pd intrinsic?), one shuffle optzn
may be done (if shufps is better than pinsw, Evan, please review), and
we already know about LICM of simple instructions.

llvm-svn: 45407
 2007-12-29 19:31:47 +00:00
Chris Lattner
ed55329cc9 upgrade this test 
llvm-svn: 45406
 2007-12-29 19:24:06 +00:00
Chris Lattner
cd147e5596 Fold comparisons against a constant nan, and optimize ORD/UNORD 
comparisons with a constant.  This allows us to compile isnan to:

_foo:
	fcmpu cr7, f1, f1
	mfcr r2
	rlwinm r3, r2, 0, 31, 31
	blr 

instead of:

LCPI1_0:					;  float
	.space	4
_foo:
	lis r2, ha16(LCPI1_0)
	lfs f0, lo16(LCPI1_0)(r2)
	fcmpu cr7, f1, f0
	mfcr r2
	rlwinm r3, r2, 0, 31, 31
	blr 

llvm-svn: 45405
 2007-12-29 08:37:08 +00:00
Chris Lattner
b36a4a7a84 this xform is implemented. 
llvm-svn: 45404
 2007-12-29 08:19:39 +00:00
Chris Lattner
f8e408b7b1 Codegen: 
as:

_bar:
	pushl	%esi
	subl	$8, %esp
	movl	16(%esp), %esi
	call	L_foo$stub
	fstps	(%esi)
	addl	$8, %esp
	popl	%esi
	#FP_REG_KILL
	ret

instead of:

_bar:
	pushl	%esi
	subl	$8, %esp
	movl	16(%esp), %esi
	call	L_foo$stub
	fstpl	(%esi)
	cvtsd2ss	(%esi), %xmm0
	movss	%xmm0, (%esi)
	addl	$8, %esp
	popl	%esi
	#FP_REG_KILL
	ret

llvm-svn: 45401
 2007-12-29 06:57:38 +00:00
Chris Lattner
e3515220d2 avoid going through a stack slot to convert from fpstack to xmm reg 
if we are just going to store it back anyway.  This improves things 
like:
double foo();
void bar(double *P) { *P = foo(); }

llvm-svn: 45399
 2007-12-29 06:41:28 +00:00
Chris Lattner
a432f12b76 one fewer uncond branch with my codegenprepare hack for single-mbb backedges. 
llvm-svn: 45360
 2007-12-26 17:23:47 +00:00
Evan Cheng
8824950e8f Fix PR1872: SrcValue and SrcValueOffset should not be used to compute load / store node id. 
llvm-svn: 45167
 2007-12-18 19:38:14 +00:00
Evan Cheng
36bfae49e3 FIX for PR1799: When a load is unfolded from an instruction, check if it is a new node. If not, do not create a new SUnit. 
llvm-svn: 45157
 2007-12-18 08:42:10 +00:00
Evan Cheng
1d95b669b6 Make better use of instructions that clear high bits; fix various 2-wide shuffle bugs. 
llvm-svn: 45058
 2007-12-15 03:00:47 +00:00
Evan Cheng
6909ff8c4b Fix ctlz and cttz. llvm definition requires them to return number of bits in of the src type when value is zero. 
llvm-svn: 45029
 2007-12-14 08:30:15 +00:00
Evan Cheng
51cf86ded0 Implement ctlz and cttz with bsr and bsf. 
llvm-svn: 45024
 2007-12-14 02:13:44 +00:00
Evan Cheng
a152909956 Be extra careful with extension use optimation. Now turned on by default. 
llvm-svn: 44981
 2007-12-13 03:32:53 +00:00
Evan Cheng
343929c773 Fold some and + shift in x86 addressing mode. 
llvm-svn: 44970
 2007-12-13 00:43:27 +00:00
Evan Cheng
64a1febf9a Implicit def instructions, e.g. X86::IMPLICIT_DEF_GR32, are always re-materializable and they should not be spilled. 
llvm-svn: 44960
 2007-12-12 23:12:09 +00:00
Dan Gohman
0075ea1f5f Allow vector integer constants to be created with 
SelectionDAG::getConstant, in the same way as vector floating-point
constants. This allows the legalize expansion code for @llvm.ctpop and
friends to be usable with vector types.

llvm-svn: 44954
 2007-12-12 22:21:26 +00:00
Evan Cheng
ad3e7f3286 Use shuffles to implement insert_vector_elt for i32, i64, f32, and f64. 
llvm-svn: 44929
 2007-12-12 07:55:34 +00:00
Evan Cheng
af6ba4dfd4 Add a test case for -optimize-ext-uses. 
llvm-svn: 44928
 2007-12-12 07:54:08 +00:00
Evan Cheng
d36d69fe92 Lower a build_vector with all constants into a constpool load unless it can be done with a move to low part. 
llvm-svn: 44921
 2007-12-12 06:45:40 +00:00
Evan Cheng
f6c2838f36 - Improved v8i16 shuffle lowering. It now uses pshuflw and pshufhw as much as 
possible before resorting to pextrw and pinsrw.
- Better codegen for v4i32 shuffles masquerading as v8i16 or v16i8 shuffles.
- Improves (i16 extract_vector_element 0) codegen by recognizing
  (i32 extract_vector_element 0) does not require a pextrw.

llvm-svn: 44836
 2007-12-11 01:46:18 +00:00
Christopher Lamb
5c577eb543 Improve branch folding by recgonizing that explict successor relationships impact the value of fall-through choices. 
llvm-svn: 44785
 2007-12-10 07:24:06 +00:00
Evan Cheng
34c7b35135 Much improved v8i16 shuffles. (Step 1). 
llvm-svn: 44676
 2007-12-07 08:07:39 +00:00
Evan Cheng
8d1f8b2f27 New test case. 
llvm-svn: 44672
 2007-12-07 01:48:46 +00:00
Evan Cheng
cab253ba13 Fix a bogus test case. 
llvm-svn: 44668
 2007-12-06 22:12:45 +00:00
Evan Cheng
d53f72dfb1 Turning simple splitting on. Start testing new coalescer heuristics as new llcbeta. 
llvm-svn: 44660
 2007-12-06 08:54:31 +00:00
Chris Lattner
64a1a9f502 third time around: instead of disabling this completely, 
only disable it if we don't know it will be obviously profitable.
Still fixme, but less so. :)

llvm-svn: 44658
 2007-12-06 07:47:55 +00:00
Chris Lattner
bb5fb18af8 Actually, disable this code for now. More analysis and improvements to 
the X86 backend are needed before this should be enabled by default.

llvm-svn: 44657
 2007-12-06 07:44:31 +00:00
Chris Lattner
c467b49c96 implement a readme entry, compiling the code into: 
_foo:
	movl	$12, %eax
	andl	4(%esp), %eax
	movl	_array(%eax), %eax
	ret

instead of:

_foo:
	movl	4(%esp), %eax
	shrl	$2, %eax
	andl	$3, %eax
	movl	_array(,%eax,4), %eax
	ret

As it turns out, this triggers all the time, in a wide variety of
situations, for example, I see diffs like this in various programs:

-       movl    8(%eax), %eax
-       shll    $2, %eax
-       andl    $1020, %eax
-       movl    (%esi,%eax), %eax
+       movzbl  8(%eax), %eax
+       movl    (%esi,%eax,4), %eax


-       shll    $2, %edx
-       andl    $1020, %edx
-       movl    (%edi,%edx), %edx
+       andl    $255, %edx
+       movl    (%edi,%edx,4), %edx

Unfortunately, I also see stuff like this, which can be fixed in the
X86 backend:

-       andl    $85, %ebx
-       addl    _bit_count(,%ebx,4), %ebp
+       shll    $2, %ebx
+       andl    $340, %ebx
+       addl    _bit_count(%ebx), %ebp

llvm-svn: 44656
 2007-12-06 07:33:36 +00:00
Chris Lattner
1de4db0446 fix this when run on non x86 hosts. 
llvm-svn: 44645
 2007-12-06 01:05:52 +00:00
Evan Cheng
79e8b92dc3 Allow some reloads to be folded in multi-use cases. Specifically testl r, r -> cmpl [mem], 0. 
llvm-svn: 44479
 2007-12-01 02:07:52 +00:00
Evan Cheng
90c548af8e Do not fold reload into an instruction with multiple uses. It issues one extra load. 
llvm-svn: 44467
 2007-11-30 21:23:43 +00:00
Dan Gohman
d12155d8c8 Remove unnecessary && from the RUN lines of this test. 
llvm-svn: 44342
 2007-11-27 00:03:38 +00:00
Dan Gohman
a9f8208852 Don't lower srem/urem X%C to X-X/C*C unless the division is actually 
optimized. This avoids creating illegal divisions when the combiner is
running after legalize; this fixes PR1815. Also, it produces better
code in the included testcase by avoiding the subtract and multiply
when the division isn't optimized.

llvm-svn: 44341
 2007-11-26 23:46:11 +00:00
Chris Lattner
be0c5a0500 Fix a long standing deficiency in the X86 backend: we would 
sometimes emit "zero" and "all one" vectors multiple times,
for example:

_test2:
	pcmpeqd	%mm0, %mm0
	movq	%mm0, _M1
	pcmpeqd	%mm0, %mm0
	movq	%mm0, _M2
	ret

instead of:

_test2:
	pcmpeqd	%mm0, %mm0
	movq	%mm0, _M1
	movq	%mm0, _M2
	ret

This patch fixes this by always arranging for zero/one vectors
to be defined as v4i32 or v2i32 (SSE/MMX) instead of letting them be
any random type.  This ensures they get trivially CSE'd on the dag.
This fix is also important for LegalizeDAGTypes, as it gets unhappy
when the x86 backend wants BUILD_VECTOR(i64 0) to be legal even when
'i64' isn't legal.

This patch makes the following changes:

1) X86TargetLowering::LowerBUILD_VECTOR now lowers 0/1 vectors into
   their canonical types.
2) The now-dead patterns are removed from the SSE/MMX .td files.
3) All the patterns in the .td file that referred to immAllOnesV or
   immAllZerosV in the wrong form now use *_bc to match them with a
   bitcast wrapped around them.
4) X86DAGToDAGISel::SelectScalarSSELoad is generalized to handle 
   bitcast'd zero vectors, which simplifies the code actually.
5) getShuffleVectorZeroOrUndef is updated to generate a shuffle that
   is legal, instead of generating one that is illegal and expecting
   a later legalize pass to clean it up.
6) isZeroShuffle is generalized to handle bitcast of zeros.
7) several other minor tweaks.

This patch is definite goodness, but has the potential to cause random
code quality regressions.  Please be on the lookout for these and let 
me know if they happen.

llvm-svn: 44310
 2007-11-25 00:24:49 +00:00
Chris Lattner
6304b1e16d upgrade this test 
llvm-svn: 44298
 2007-11-24 05:39:29 +00:00
Dan Gohman
0f62120b01 Add support in SplitVectorOp for remainder operators. 
llvm-svn: 44233
 2007-11-19 15:15:03 +00:00
Chris Lattner
bef568f3f8 fix bogus test that the more strict lexer is finding. 
llvm-svn: 44216
 2007-11-18 18:26:45 +00:00
Evan Cheng
121c50d5e3 Typo. 
llvm-svn: 44196
 2007-11-16 23:55:08 +00:00
Evan Cheng
c19506f69d Fix a thinko in post-allocation coalescer. 
llvm-svn: 44166
 2007-11-15 08:13:29 +00:00
Anton Korobeynikov
58298cb9cc Fix PIC jump table codegen on x86-32/linux. In fact, such thing should be applied 
to all targets uses GOT-relative offsets for PIC (Alpha?)

llvm-svn: 44108
 2007-11-14 09:18:41 +00:00
Arnold Schwaighofer
64ad6fa1fa Update tailcall code to include inline attribute operand for memcpy. 
llvm-svn: 43978
 2007-11-10 10:48:01 +00:00
Evan Cheng
ea1474bdf3 Fix tests. 
llvm-svn: 43961
 2007-11-09 20:46:00 +00:00
Evan Cheng
d9bab93a44 If both parts of smul_lohi, etc. are used, don't simplify. If only one part is used, try simplify it. 
llvm-svn: 43888
 2007-11-08 09:25:29 +00:00
Evan Cheng
3764ad2bac Add pseudo dependency to force two-address instruction to be scheduled after 
other uses. There was a overly restricted check that prevented some obvious
cases.

llvm-svn: 43762
 2007-11-06 08:44:59 +00:00
Dan Gohman
6255ce9f5d Add support for vector remainder operations. 
llvm-svn: 43744
 2007-11-05 23:35:22 +00:00
Dale Johannesen
1f70f86c7a Make labels work in asm blocks; allow labels as 
parameters.  Rename ValueRefList to ParamList
in AsmParser, since its only use is for parameters.

llvm-svn: 43734
 2007-11-05 21:20:28 +00:00
Evan Cheng
e5eac2c5ac Handle cases where a register and one of its super-register are both marked as 
defined on the same instruction. This fixes PR1767.

llvm-svn: 43699
 2007-11-05 03:11:55 +00:00
Evan Cheng
bc39e175c4 Fix test case. Chris didn't do make check. :-) 
llvm-svn: 43698
 2007-11-05 03:04:26 +00:00
Evan Cheng
947c271e37 Doh. PR1187 -> PR1766. 
llvm-svn: 43693
 2007-11-05 01:00:44 +00:00
Evan Cheng
13d79ab67a Fix PR1187. 
llvm-svn: 43692
 2007-11-05 00:59:10 +00:00
Chris Lattner
8fac63c8b5 Fix PR1761 by not printing (rip) suffix when in -static mode. 
Evan, please review this.

llvm-svn: 43680
 2007-11-04 19:23:28 +00:00
Chris Lattner
67cd357fb8 Fix PR1763 by allowing the 'q' constraint to work with 64-bit 
regs on x86-64.

llvm-svn: 43669
 2007-11-04 06:51:12 +00:00
Evan Cheng
1771f6da9c There are times when the coalescer would not coalesce away a copy but the copy 
can be eliminated by the allocator is the destination and source targets the
same register. The most common case is when the source and destination registers
are in different class. For example, on x86 mov32to32_ targets GR32_ which
contains a subset of the registers in GR32.

The allocator can do 2 things:
1. Set the preferred allocation for the destination of a copy to that of its source.
2. After allocation is done, change the allocation of a copy destination (if
   legal) so the copy can be eliminated.

This eliminates 443 extra moves from 403.gcc.

llvm-svn: 43662
 2007-11-03 07:20:12 +00:00
Evan Cheng
8d473f667d Add run line. 
llvm-svn: 43645
 2007-11-02 17:36:58 +00:00
Evan Cheng
65a07e73e2 One more extract_subreg coalescing bug. 
llvm-svn: 43644
 2007-11-02 17:35:08 +00:00
Evan Cheng
b50cc64eb0 Missing a getNumOperands check. 
llvm-svn: 43630
 2007-11-02 01:26:22 +00:00
Dale Johannesen
c125d9b4e8 Test that expand_vector_elt(v2i64) works in 32-bit mode. 
llvm-svn: 43598
 2007-11-01 02:38:24 +00:00
Evan Cheng
5e058e94b5 It's not safe to tell SplitCriticalEdge to merge identical edges. It may delete the phi instruction that's being processed. 
llvm-svn: 43524
 2007-10-30 22:27:26 +00:00
Evan Cheng
633cd3e84d - Bug fixes. 
- Allow icmp rewrite using an iv / stride of a smaller integer type.

llvm-svn: 43480
 2007-10-29 22:07:18 +00:00
Dan Gohman
02b8beff5f Fix a DAGCombiner abort on a bitcast from a scalar to a vector. 
llvm-svn: 43470
 2007-10-29 20:44:42 +00:00
Evan Cheng
5fe81cf64e Enable more fold (sext (load x)) -> (sext (truncate (sextload x))) 
transformation. Previously, it's restricted by ensuring the number of load uses
is one. Now the restriction is loosened up by allowing setcc uses to be
"extended" (e.g. setcc x, c, eq -> setcc sext(x), sext(c), eq).

llvm-svn: 43465
 2007-10-29 19:58:20 +00:00
Chris Lattner
1503362624 Add support for the x86-64 'q' regigster modifier, and add support for the 
b/h/w/k/q inline asm memory modifiers, which are just ignored.  This fixes
PR1748 and CodeGen/X86/2007-10-28-inlineasm-q-modifier.ll

llvm-svn: 43430
 2007-10-29 03:09:07 +00:00
Evan Cheng
53696b7e9f Loosen up iv reuse to allow reuse of the same stride but a larger type when truncating from the larger type to smaller type is free. 
e.g.
Turns this loop:
LBB1_1: # entry.bb_crit_edge
        xorl    %ecx, %ecx
        xorw    %dx, %dx
        movw    %dx, %si
LBB1_2: # bb
        movl    L_X$non_lazy_ptr, %edi
        movw    %si, (%edi)
        movl    L_Y$non_lazy_ptr, %edi
        movw    %dx, (%edi)
		addw    $4, %dx
		incw    %si
		incl    %ecx
		cmpl    %eax, %ecx
		jne     LBB1_2  # bb
	
into

LBB1_1: # entry.bb_crit_edge
        xorl    %ecx, %ecx
        xorw    %dx, %dx
LBB1_2: # bb
        movl    L_X$non_lazy_ptr, %esi
        movw    %cx, (%esi)
        movl    L_Y$non_lazy_ptr, %esi
        movw    %dx, (%esi)
        addw    $4, %dx
		incl    %ecx
        cmpl    %eax, %ecx
        jne     LBB1_2  # bb

llvm-svn: 43375
 2007-10-26 01:56:11 +00:00
Evan Cheng
66cbf54030 If a loop termination compare instruction is the only use of its stride, 
and the compaison is against a constant value, try eliminate the stride
by moving the compare instruction to another stride and change its
constant operand accordingly. e.g.

loop:
...
v1 = v1 + 3
v2 = v2 + 1
if (v2 < 10) goto loop
=>
loop:
...
v1 = v1 + 3
if (v1 < 30) goto loop

llvm-svn: 43336
 2007-10-25 09:11:16 +00:00
Dale Johannesen
402c11966a This was failing on Darwin, which defaults to PIC; 
no lea was generated.  I think this follows the intent.

llvm-svn: 43312
 2007-10-24 20:58:14 +00:00
Dan Gohman
df1f166e4a Strength reduction improvements. 
- Avoid attempting stride-reuse in the case that there are users that
   aren't addresses. In that case, there will be places where the
   multiplications won't be folded away, so it's better to try to
   strength-reduce them.

 - Several SSE intrinsics have operands that strength-reduction can
   treat as addresses. The previous item makes this more visible, as
   any non-address use of an IV can inhibit stride-reuse.

 - Make ValidStride aware of whether there's likely to be a base
   register in the address computation. This prevents it from thinking
   that things like stride 9 are valid on x86 when the base register is
   already occupied.

Also, XFAIL the 2007-08-10-LEA16Use32.ll test; the new logic to avoid
stride-reuse elimintes the LEA in the loop, so the test is no longer
testing what it was intended to test.

llvm-svn: 43231
 2007-10-22 20:40:42 +00:00
Dan Gohman
76e104c8ad Fix the folding of multiplication into addresses on x86, which was broken 
by the recent {U,S}MUL_LOHI changes.

llvm-svn: 43230
 2007-10-22 20:22:24 +00:00
Evan Cheng
2d53c3f15e New test case. 
llvm-svn: 43193
 2007-10-19 22:05:00 +00:00
Rafael Espindola
2b5b200b9f Test byval with a 8 bit aligned struct 
llvm-svn: 43173
 2007-10-19 11:29:21 +00:00
Rafael Espindola
d8d4372845 Add support for byval function whose argument is not 32 bit aligned. 
To do this it is necessary to add a "always inline" argument to the
memcpy node. For completeness I have also added this node to memmove
and memset.  I have also added getMem* functions, because the extra
argument makes it cumbersome to use getNode and because I get confused
by it :-)

llvm-svn: 43172
 2007-10-19 10:41:11 +00:00
Evan Cheng
f6d1c7be14 Really fix PR1734. Carefully track which register uses are sub-register uses by 
traversing inverse register coalescing map.

llvm-svn: 43118
 2007-10-18 07:49:59 +00:00
Dan Gohman
2903f7fc26 Add support for ISD::SELECT in SplitVectorOp. 
llvm-svn: 43072
 2007-10-17 14:48:28 +00:00
Evan Cheng
524c0e6c3d Yet another test case for extract_subreg coalescing crash. 
llvm-svn: 43063
 2007-10-17 02:15:06 +00:00
Evan Cheng
09fa6ed483 Fix PR1734. 
llvm-svn: 43035
 2007-10-16 19:29:47 +00:00
Dale Johannesen
dd254c4efa New test for svn rev 43033, radar 5538745. 
llvm-svn: 43034
 2007-10-16 18:10:14 +00:00
Evan Cheng
f5bcd3d737 LowerFP_TO_SINT must not create a stack object if it's not needed. 
llvm-svn: 43004
 2007-10-15 20:11:21 +00:00
Dan Gohman
2dc6099def Reapply the fix in 42908 for this file. This changes the function names 
from "test" to "foo" so that they don't match the grep -i ST.

llvm-svn: 43001
 2007-10-15 19:22:17 +00:00
Evan Cheng
43887d3714 Fix PR1729: watch out for val# with no def. 
llvm-svn: 42996
 2007-10-15 18:33:50 +00:00
Tanya Lattner
3a64752342 Fix run line. 
llvm-svn: 42990
 2007-10-15 16:35:13 +00:00
Evan Cheng
850c8739dc New test case. 
llvm-svn: 42963
 2007-10-14 10:15:03 +00:00
Evan Cheng
33df6a6bed Revert 42908 for now. 
llvm-svn: 42960
 2007-10-14 05:57:21 +00:00
Evan Cheng
f5ed18f7d3 Fix test case. 
llvm-svn: 42949
 2007-10-13 03:14:06 +00:00
Evan Cheng
6101e4ffdf New tests. 
llvm-svn: 42948
 2007-10-13 03:10:54 +00:00
Dan Gohman
c96f2809ca Fix this test to not depend on the assembly output containing something 
that includes the string "st". This probably fixes the regression on
Darwin.

llvm-svn: 42932
 2007-10-12 20:42:14 +00:00
Dan Gohman
a75e4a62e6 Change the names used for internal labels to use the current 
function symbol name instead of a codegen-assigned function
number.

Thanks Evan! :-)

llvm-svn: 42908
 2007-10-12 14:53:36 +00:00
Evan Cheng
51791564b0 Doh. 
llvm-svn: 42901
 2007-10-12 09:10:27 +00:00
Evan Cheng
947b4a6c3d EXTRACT_SUBREG test case. 
llvm-svn: 42900
 2007-10-12 09:03:31 +00:00
Arnold Schwaighofer
f1e49dd41d Added missing -march=x86 flag. 
llvm-svn: 42893
 2007-10-12 07:49:48 +00:00
Dan Gohman
ab5c3ed0d1 Add intrinsics for sin, cos, and pow. These use llvm_anyfloat_ty, and so 
may be overloaded with vector types. And add a testcase for codegen for
these.

llvm-svn: 42885
 2007-10-12 00:01:22 +00:00
Dan Gohman
9a70be46f1 Add an explicit target triple to make this test behave as expected on 
non-Apple hosts. And use the count script instead of wc + grep.

llvm-svn: 42878
 2007-10-11 23:04:36 +00:00
Arnold Schwaighofer
d47210011e Added tail call optimization to the x86 back end. It can be 
enabled by passing -tailcallopt to llc.  The optimization is
performed if the following conditions are satisfied:
* caller/callee are fastcc
* elf/pic is disabled OR
  elf/pic enabled + callee is in module + callee has
  visibility protected or hidden

llvm-svn: 42870
 2007-10-11 19:40:01 +00:00
Dan Gohman
708e76e663 These two tests now require only two multiply instructions, 
instead of four.

llvm-svn: 42784
 2007-10-09 15:39:37 +00:00
Evan Cheng
25b65542d9 Update test. 
llvm-svn: 42775
 2007-10-08 22:20:32 +00:00
Dan Gohman
9da3bddf43 These two tests now require only three multiply instructions, 
instead of four.

llvm-svn: 42765
 2007-10-08 20:48:12 +00:00
Dale Johannesen
b600202c68 Make test work on non-x86 hosts. 
llvm-svn: 42671
 2007-10-06 01:22:39 +00:00
Evan Cheng
0a642eaa62 Test case for 3-address conversion. 
llvm-svn: 42664
 2007-10-05 23:33:09 +00:00
Evan Cheng
dc467c6323 Enable convertToThreeAddress for X86 by default. 
llvm-svn: 42655
 2007-10-05 22:31:10 +00:00
Evan Cheng
6fd2606ff5 New test case. 
llvm-svn: 42628
 2007-10-05 01:44:22 +00:00
Evan Cheng
1d3c836933 -pre-RA-sched=none, simple, simple-noitin are gone. 
llvm-svn: 42505
 2007-10-01 22:17:20 +00:00
Dan Gohman
02f80006f8 Teach SplitVectorOp how to split INSERT_VECTOR_ELT. 
llvm-svn: 42457
 2007-09-28 23:53:40 +00:00
Rafael Espindola
01b306e575 Refactor the memcpy lowering for the x86 target. 
The only generated code difference is that now we call memcpy when
the size of the array is unknown. This matches GCC behavior and is
better since the run time value can be arbitrarily large.

llvm-svn: 42433
 2007-09-28 12:53:01 +00:00
Dale Johannesen
e61886cee4 Add sqrt and powi intrinsics for long double. 
llvm-svn: 42423
 2007-09-28 01:08:20 +00:00
Dale Johannesen
57b6470e8b Modernize fabs.ll, add long double. Add tests 
for direct codegen of fsin/fcos.

llvm-svn: 42369
 2007-09-26 21:12:10 +00:00
Dan Gohman
1bb346f9f1 When both x/y and x%y are needed (x and y both scalar integer), compute 
both results with a single div or idiv instruction. This uses new X86ISD
nodes for DIV and IDIV which are introduced during the legalize phase
so that the SelectionDAG's CSE can automatically eliminate redundant
computations.

llvm-svn: 42308
 2007-09-25 18:23:27 +00:00
Dale Johannesen
fe773726a9 Some tests for APFloat conversions. 
llvm-svn: 42303
 2007-09-25 17:50:55 +00:00
Evan Cheng
dab211c87b Forgot to check in the changes. Fix test case so it doesn't break with any scheduling changes. 
llvm-svn: 42302
 2007-09-25 17:47:38 +00:00
Dan Gohman
dd675a5064 Use the correct result value type instead of using getValueType(0) 
in ExpandEXTRACT_VECTOR_ELT and SplitVectorOp. This fixes an abort
in the included testcase.

llvm-svn: 42264
 2007-09-24 15:54:53 +00:00
Dale Johannesen
a928e8b8b2 Implementation of +sse -sse2 has changed; add 
-sse to preserve intent of this test.

llvm-svn: 42247
 2007-09-23 14:58:14 +00:00
Rafael Espindola
11ee0898b9 Don't add a default STACK_ALIGN (use the generic ABI alignment) 
Implement calls to functions with byval arguments on X86

llvm-svn: 42192
 2007-09-21 15:50:22 +00:00
Dan Gohman
eb622df2ef Fix several more entries in the x86 reload/remat folding tables. 
llvm-svn: 42162
 2007-09-20 14:17:21 +00:00
Evan Cheng
28a7839505 Clean up. 
llvm-svn: 42112
 2007-09-18 22:56:31 +00:00
Evan Cheng
2716b97b13 Fix a bogus splat xform: 
shuffle <undef, undef, x, undef>, <undef, undef, undef, undef>, <2, 2, 2, 2>
!=
<undef, undef, x, undef>

llvm-svn: 42111
 2007-09-18 21:54:37 +00:00
Bill Wendling
803e0d9970 Objective-C was generating EH frame info like this: 
"_-[NSString(local) isNullOrNil]".eh = 0
        .no_dead_strip  "_-[NSString(local) isNullOrNil]".eh

The ".eh" should be inside the quotes.

llvm-svn: 42074
 2007-09-18 01:47:22 +00:00
Dan Gohman
1aeaeec570 Emit integer x<1 as x<=0, as comparisons with zero (now includeing 
64-bit) can use test instead of cmp with an immediate.

llvm-svn: 42026
 2007-09-17 14:49:27 +00:00
Dan Gohman
b9449c9118 Use "test reg,reg" in place of "cmp reg,0" for 64-bit operands. This was 
previously only done for 32-bit and smaller operands.

llvm-svn: 42024
 2007-09-17 14:35:24 +00:00
Rafael Espindola
5d8b225881 Add support for functions with byval arguments on x86 
llvm-svn: 41953
 2007-09-14 15:48:13 +00:00
Dan Gohman
3bc1bc2590 Avoid storing and reloading zeros and other constants from stack slots 
by flagging the associated instructions as being trivially rematerializable.

llvm-svn: 41775
 2007-09-07 21:32:51 +00:00
Rafael Espindola
8c57e70f93 Add support for having different alignment for objects on call frames. 
The x86-64 ABI states that objects passed on the stack have
8 byte alignment. Implement that.

llvm-svn: 41768
 2007-09-07 14:52:14 +00:00
Anton Korobeynikov
899c0c9c8d Split eh.select / eh.typeid.for intrinsics into i32/i64 versions. This is needed, because they just "mark" register 
liveins and we let frontend solve type issue, not lowering code :)

llvm-svn: 41763
 2007-09-07 11:39:35 +00:00
Anton Korobeynikov
0e3789f07a Proper handle case, when aliasee is external weak symbol referenced only by alias itself. 
Also, fix a case, when target doesn't have weak symbols supported.

llvm-svn: 41746
 2007-09-06 17:21:48 +00:00
Evan Cheng
896c1ed385 Fix a bug in X86InstrInfo::convertToThreeAddress that caused it to codegen: 
leal    (,%rcx,8), %rcx
It should be
leal    (,%rcx,8), %ecx

llvm-svn: 41735
 2007-09-06 00:14:41 +00:00
Dale Johannesen
f9ca7b6094 Change all floating constants that are not exactly 
representable to use hex format.

llvm-svn: 41722
 2007-09-05 17:50:36 +00:00
Evan Cheng
bb21883dd3 Fix for PR1632. EHSELECTION always produces a i32 value. 
llvm-svn: 41712
 2007-09-04 20:39:26 +00:00
Rafael Espindola
4ddaad4de0 Initial support for calling functions with byval arguments on x86-64 
llvm-svn: 41643
 2007-08-31 15:06:30 +00:00
Evan Cheng
064691b876 Update test case to reflect Dale's change. 
llvm-svn: 41639
 2007-08-31 06:29:32 +00:00
Tanya Lattner
0b279ff814 Do not run on darwin. 
llvm-svn: 41608
 2007-08-30 16:07:20 +00:00
Evan Cheng
cb317912b2 Added support to fold X86 load / store instructions. This allow rematerialized loads to be folded into their uses. 
llvm-svn: 41599
 2007-08-30 05:54:07 +00:00
Dan Gohman
c6e88b9bc2 Add explicit triples to avoid default behavior that varies by host. 
llvm-svn: 41510
 2007-08-27 20:54:48 +00:00
Dan Gohman
2e7e251f24 If the source and destination pointers in an llvm.memmove are known 
to not alias each other, it can be translated as an llvm.memcpy.

llvm-svn: 41489
 2007-08-27 16:26:13 +00:00
Rafael Espindola
3d52fe3ef3 call libc memcpy/memset if array size is bigger then threshold. 
Coping 100MB array (after a warmup) shows that glibc 2.6.1 implementation on
x86-64 (core 2) is 30% faster (from 0.270917s to 0.188079s)

llvm-svn: 41479
 2007-08-27 10:18:20 +00:00
Evan Cheng
703bafa177 Test dag xform: Fold C ? 0 : 1 to ~C or zext(~C) or trunc(~C) 
llvm-svn: 41164
 2007-08-18 06:11:57 +00:00
Evan Cheng
97e0a167ec New test. Make sure dynamic_stackalloc size is rounded up. 
llvm-svn: 41135
 2007-08-16 23:52:23 +00:00
Evan Cheng
625e712911 Update test: dynamic_stackalloc size *must* be rounded to ensure stack ptr be left in a valid state. 
llvm-svn: 41134
 2007-08-16 23:51:28 +00:00
Rafael Espindola
817adb6532 add byval test 
llvm-svn: 41123
 2007-08-16 13:09:02 +00:00
Dan Gohman
34263074cb Convert tests using "grep -c ... | grep ..." to use the count script. 
llvm-svn: 41100
 2007-08-15 13:49:33 +00:00
Dan Gohman
10c3892fde Delete extraneous uses of wc -l. 
llvm-svn: 41099
 2007-08-15 13:45:35 +00:00
Dan Gohman
f259c770df Convert another test to use the count script. This one didn't fit the 
regex used to convert all the others because the first '|' was on a
separate line.

llvm-svn: 41098
 2007-08-15 13:42:36 +00:00
Dan Gohman
794fa1f8f7 Convert tests using "| wc -l | grep ..." to use the count script. 
llvm-svn: 41097
 2007-08-15 13:36:28 +00:00