RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-27 14:02:50 +01:00
Code Issues Projects Releases Wiki Activity
70,725 Commits 4 Branches 5 Tags 704 MiB
c5f50f7322
Commit Graph

1536 Commits

Author SHA1 Message Date
Evan Cheng
ae26b91353 Revert r122955. It seems using movups to lower memcpy can cause massive regression (even on Nehalem) in edge cases. I also didn't see any real performance benefit. 
llvm-svn: 123015
 2011-01-07 19:35:30 +00:00
Evan Cheng
1a1771584e Use movups to lower memcpy and memset even if it's not fast (like corei7). 
The theory is it's still faster than a pair of movq / a quad of movl. This
will probably hurt older chips like P4 but should run faster on current
and future Intel processors. rdar://8817010

llvm-svn: 122955
 2011-01-06 07:58:36 +00:00
Evan Cheng
cb39cc2164 Re-implement r122936 with proper target hooks. Now getMaxStoresPerMemcpy 
etc. takes an option OptSize. If OptSize is true, it would return
the inline limit for functions with attribute OptSize.

llvm-svn: 122952
 2011-01-06 06:52:41 +00:00
Benjamin Kramer
d8387aa9bd X86: Lower a select directly to a setcc_carry if possible. 
int test(unsigned long a, unsigned long b) { return -(a < b); }
compiles to
  _test:                              ## @test
    cmpq  %rsi, %rdi                  ## encoding: [0x48,0x39,0xf7]
    sbbl  %eax, %eax                  ## encoding: [0x19,0xc0]
    ret                               ## encoding: [0xc3]
instead of
  _test:                              ## @test
    xorl  %ecx, %ecx                  ## encoding: [0x31,0xc9]
    cmpq  %rsi, %rdi                  ## encoding: [0x48,0x39,0xf7]
    movl  $-1, %eax                   ## encoding: [0xb8,0xff,0xff,0xff,0xff]
    cmovael %ecx, %eax                ## encoding: [0x0f,0x43,0xc1]
    ret                               ## encoding: [0xc3]

llvm-svn: 122451
 2010-12-22 23:09:28 +00:00
Benjamin Kramer
369872edfc Add some x86 specific dagcombines for conditional increments. 
(add Y, (sete  X, 0)) -> cmp X, 1; adc  0, Y
(add Y, (setne X, 0)) -> cmp X, 1; sbb -1, Y
(sub (sete  X, 0), Y) -> cmp X, 1; sbb  0, Y
(sub (setne X, 0), Y) -> cmp X, 1; adc -1, Y

for
  unsigned foo(unsigned a, unsigned b) {
    if (a == 0) b++;
    return b;
  }
we now get:
  foo:
    cmpl  $1, %edi
    movl  %esi, %eax
    adcl  $0, %eax
    ret
instead of:
  foo:
    testl %edi, %edi
    sete  %al
    movzbl  %al, %eax
    addl  %esi, %eax
    ret

llvm-svn: 122364
 2010-12-21 21:41:44 +00:00
Chris Lattner
65c5243bd6 rename MVT::Flag to MVT::Glue. "Flag" is a terrible name for 
something that just glues two nodes together, even if it is
sometimes used for flags.

llvm-svn: 122310
 2010-12-21 02:38:05 +00:00
Nate Begeman
c7dfecb10e Implement feedback from Bruno on making pblendvb an x86-specific ISD node in addition to being an intrinsic, and convert 
lowering to use it.  Hopefully the pattern fragment is doing the right thing with XMM0, looks correct in testing.

llvm-svn: 122277
 2010-12-20 22:04:24 +00:00
Chris Lattner
bee7320c3c now that addc/adde are gone, "ADDC" in the X86 backend uses EFLAGS results, 
the same as setcc.  Optimize ADDC(0,0,FLAGS) -> SET_CARRY(FLAGS).  This is
a step towards finishing off PR5443.  In the testcase in that bug we now  get:

	movq	%rdi, %rax
	addq	%rsi, %rax
	sbbq	%rcx, %rcx
	testb	$1, %cl
	setne	%dl
	ret

instead of:

	movq	%rdi, %rax
	addq	%rsi, %rax
	movl	$0, %ecx
	adcq	$0, %rcx
	testq	%rcx, %rcx
	setne	%dl
	ret

llvm-svn: 122219
 2010-12-20 01:37:09 +00:00
Chris Lattner
16ea7f257f use for loop over types. 
llvm-svn: 122214
 2010-12-20 01:03:27 +00:00
Chris Lattner
8b1f76cad6 Change the X86 backend to stop using the evil ADDC/ADDE/SUBC/SUBE nodes (which 
their carry depenedencies with MVT::Flag operands) and use clean and beautiful
EFLAGS dependences instead.

We do this by changing the modelling of SBB/ADC to have EFLAGS input and outputs
(which is what requires the previous scheduler change) and change X86 ISelLowering
to custom lower ADDC and friends down to X86ISD::ADD/ADC/SUB/SBB nodes.

With the previous series of changes, this causes no changes in the testsuite, woo.

llvm-svn: 122213
 2010-12-20 00:59:46 +00:00
Mon P Wang
d3adab7a64 Prevents PerformShuffleCombine from creating a node with an illegal type after legalize types 
has run, e.g., prevent creating an i64 node from a v2i64 when i64 is not a legal type.

llvm-svn: 122206
 2010-12-19 23:55:53 +00:00
Chris Lattner
297259f6f1 improve the setcc -> setcc_carry optimization to happen more 
consistently by moving it out of lowering into dag combine.

Add some missing patterns for matching away extended versions of setcc_c.

llvm-svn: 122201
 2010-12-19 22:08:31 +00:00
Chris Lattner
1f31c7fa15 simplify some code to just reuse a setcc if we can instead of 
going through the CSE maps to get it.

llvm-svn: 122196
 2010-12-19 21:23:48 +00:00
Chris Lattner
2d59eef5fd now that generic vector types aren't selected onto MMX operations, 
we don't need -disable-mmx anymore.

llvm-svn: 122189
 2010-12-19 20:19:20 +00:00
Chris Lattner
30438e63c8 reduce copy/paste programming with the power of for loops. 
llvm-svn: 122187
 2010-12-19 20:07:10 +00:00
Chris Lattner
29475c23d0 X86 supports i8/i16 overflow ops (except i8 multiplies), we should 
generate them.  

Now we compile:

define zeroext i8 @X(i8 signext %a, i8 signext %b) nounwind ssp {
entry:
  %0 = tail call %0 @llvm.sadd.with.overflow.i8(i8 %a, i8 %b)
  %cmp = extractvalue %0 %0, 1
  br i1 %cmp, label %if.then, label %if.end

into:

_X:                                     ## @X
## BB#0:                                ## %entry
	subl	$12, %esp
	movb	16(%esp), %al
	addb	20(%esp), %al
	jo	LBB0_2

Before we were generating:

_X:                                     ## @X
## BB#0:                                ## %entry
	pushl	%ebp
	movl	%esp, %ebp
	subl	$8, %esp
	movb	12(%ebp), %al
	testb	%al, %al
	setge	%cl
	movb	8(%ebp), %dl
	testb	%dl, %dl
	setge	%ah
	cmpb	%cl, %ah
	sete	%cl
	addb	%al, %dl
	testb	%dl, %dl
	setge	%al
	cmpb	%al, %ah
	setne	%al
	andb	%cl, %al
	testb	%al, %al
	jne	LBB0_2

llvm-svn: 122186
 2010-12-19 20:03:11 +00:00
Nate Begeman
ef5f3c0fa7 Add support for matching psign & plendvb to the x86 target 
Remove unnecessary pandn patterns, 'vnot' patfrag looks through bitcasts

llvm-svn: 122098
 2010-12-17 22:55:37 +00:00
Nate Begeman
cb6d1c8193 Formalize the notion that AVX and SSE are non-overlapping extensions from the compiler's point of view. Per email discussion, we either want to always use VEX-prefixed instructions or never use them, and are taking "HasAVX" to mean "Always use VEX". Passing -mattr=-avx,+sse42 should serve to restore legacy SSE support when desirable. 
llvm-svn: 121439
 2010-12-10 00:26:57 +00:00
Eric Christopher
0e40452eb0 Rewrite the darwin tlv support to use a chain and return to copying 
the output to the correct register. Fixes a hidden problem uncovered
by the last patch where we'd try to DAG combine our MVT::Other node
oddly.

llvm-svn: 121358
 2010-12-09 06:25:53 +00:00
Eric Christopher
64e662fce9 Stop confusing people, it's not really a chain, or a tumor. 
llvm-svn: 121340
 2010-12-09 00:57:19 +00:00
Eric Christopher
0100a8fda4 Remove extraneous copy from DAG conversion for darwin tls. This was 
popping up at O0 when it wasn't folded and the fast allocator would
complain.

llvm-svn: 121330
 2010-12-09 00:27:58 +00:00
Chris Lattner
e30adfb732 Teach X86ISelLowering that the second result of X86ISD::UMUL is a flags 
result.  This allows us to compile:

void *test12(long count) {
      return new int[count];
}

into:

test12:
	movl	$4, %ecx
	movq	%rdi, %rax
	mulq	%rcx
	movq	$-1, %rdi
	cmovnoq	%rax, %rdi
	jmp	__Znam                  ## TAILCALL

instead of:

test12:
	movl	$4, %ecx
	movq	%rdi, %rax
	mulq	%rcx
	seto	%cl
	testb	%cl, %cl
	movq	$-1, %rdi
	cmoveq	%rax, %rdi
	jmp	__Znam

Of course it would be even better if the regalloc inverted the cmov to 'cmovoq',
which would eliminate the need for the 'movq %rdi, %rax'.

llvm-svn: 120936
 2010-12-05 07:49:54 +00:00
Chris Lattner
76601e7a99 it turns out that when ".with.overflow" intrinsics were added to the X86 
backend that they were all implemented except umul.  This one fell back
to the default implementation that did a hi/lo multiply and compared the
top.  Fix this to check the overflow flag that the 'mul' instruction
sets, so we can avoid an explicit test.  Now we compile:

void *func(long count) {
      return new int[count];
}

into:

__Z4funcl:                              ## @_Z4funcl
	movl	$4, %ecx                ## encoding: [0xb9,0x04,0x00,0x00,0x00]
	movq	%rdi, %rax              ## encoding: [0x48,0x89,0xf8]
	mulq	%rcx                    ## encoding: [0x48,0xf7,0xe1]
	seto	%cl                     ## encoding: [0x0f,0x90,0xc1]
	testb	%cl, %cl                ## encoding: [0x84,0xc9]
	movq	$-1, %rdi               ## encoding: [0x48,0xc7,0xc7,0xff,0xff,0xff,0xff]
	cmoveq	%rax, %rdi              ## encoding: [0x48,0x0f,0x44,0xf8]
	jmp	__Znam                  ## TAILCALL

instead of:

__Z4funcl:                              ## @_Z4funcl
	movl	$4, %ecx                ## encoding: [0xb9,0x04,0x00,0x00,0x00]
	movq	%rdi, %rax              ## encoding: [0x48,0x89,0xf8]
	mulq	%rcx                    ## encoding: [0x48,0xf7,0xe1]
	testq	%rdx, %rdx              ## encoding: [0x48,0x85,0xd2]
	movq	$-1, %rdi               ## encoding: [0x48,0xc7,0xc7,0xff,0xff,0xff,0xff]
	cmoveq	%rax, %rdi              ## encoding: [0x48,0x0f,0x44,0xf8]
	jmp	__Znam                  ## TAILCALL

Other than the silly seto+test, this is using the o bit directly, so it's going in the right
direction.

llvm-svn: 120935
 2010-12-05 07:30:36 +00:00
Chris Lattner
16bafb2414 generalize the previous check to handle -1 on either side of the 
select, inserting a not to compensate.  Add a missing isZero check
that I lost somehow.

This improves codegen of:

void *func(long count) {
      return new int[count];
}

from:

__Z4funcl:                              ## @_Z4funcl
	movl	$4, %ecx                ## encoding: [0xb9,0x04,0x00,0x00,0x00]
	movq	%rdi, %rax              ## encoding: [0x48,0x89,0xf8]
	mulq	%rcx                    ## encoding: [0x48,0xf7,0xe1]
	testq	%rdx, %rdx              ## encoding: [0x48,0x85,0xd2]
	movq	$-1, %rdi               ## encoding: [0x48,0xc7,0xc7,0xff,0xff,0xff,0xff]
	cmoveq	%rax, %rdi              ## encoding: [0x48,0x0f,0x44,0xf8]
	jmp	__Znam                  ## TAILCALL
                                        ## encoding: [0xeb,A]

to:

__Z4funcl:                              ## @_Z4funcl
	movl	$4, %ecx                ## encoding: [0xb9,0x04,0x00,0x00,0x00]
	movq	%rdi, %rax              ## encoding: [0x48,0x89,0xf8]
	mulq	%rcx                    ## encoding: [0x48,0xf7,0xe1]
	cmpq	$1, %rdx                ## encoding: [0x48,0x83,0xfa,0x01]
	sbbq	%rdi, %rdi              ## encoding: [0x48,0x19,0xff]
	notq	%rdi                    ## encoding: [0x48,0xf7,0xd7]
	orq	%rax, %rdi              ## encoding: [0x48,0x09,0xc7]
	jmp	__Znam                  ## TAILCALL
                                        ## encoding: [0xeb,A]

llvm-svn: 120932
 2010-12-05 02:00:51 +00:00
Chris Lattner
474ed0aa9b Improve an integer select optimization in two ways: 
1. generalize 
    (select (x == 0), -1, 0) -> (sign_bit (x - 1))
to:
    (select (x == 0), -1, y) -> (sign_bit (x - 1)) | y

2. Handle the identical pattern that happens with !=:
   (select (x != 0), y, -1) -> (sign_bit (x - 1)) | y

cmov is often high latency and can't fold immediates or
memory operands.  For example for (x == 0) ? -1 : 1, before 
we got:

< 	testb	%sil, %sil
< 	movl	$-1, %ecx
< 	movl	$1, %eax
< 	cmovel	%ecx, %eax

now we get:

> 	cmpb	$1, %sil
> 	sbbl	%eax, %eax
> 	orl	$1, %eax

llvm-svn: 120929
 2010-12-05 01:23:24 +00:00
Benjamin Kramer
851691ddb2 Add patterns for the x86 popcnt instruction. 
- Also adds a new POPCNT subtarget feature that is currently enabled if the target
  supports SSE4.2 (nehalem) or SSE4A (barcelona).

llvm-svn: 120917
 2010-12-04 20:32:23 +00:00
Benjamin Kramer
77faee6ba1 Simplify code. No functionality change. 
llvm-svn: 120907
 2010-12-04 14:22:24 +00:00
Evan Cheng
4118b24aca Fix and re-enable tail call optimization of expanded libcalls. 
llvm-svn: 120622
 2010-12-01 22:59:46 +00:00
Duncan Sands
cd4f56b8e2 I don't think it makes any sense to assert that the target supports SSE3 here. 
The user (i.e. whoever generated a call to the intrinsic in the first place) is
essentially asking for a particular instruction to be placed in the assembler.
If that instruction won't execute on the target machine, that's their problem
not ours.  Two buildbots with processors that don't support SSE3 were barfing
on the apm.ll test in CodeGen/X86 because of this assertion.

llvm-svn: 120574
 2010-12-01 12:58:13 +00:00
Evan Cheng
84162760b7 Speculatively disable x86 portion of r120501 to appease the x86_64 buildbot. 
llvm-svn: 120549
 2010-12-01 03:27:20 +00:00
Evan Cheng
f7e586d749 Enable sibling call optimization of libcalls which are expanded during 
legalization time. Since at legalization time there is no mapping from
SDNode back to the corresponding LLVM instruction and the return
SDNode is target specific, this requires a target hook to check for
eligibility. Only x86 and ARM support this form of sibcall optimization
right now.
rdar://8707777

llvm-svn: 120501
 2010-11-30 23:55:39 +00:00
Eric Christopher
73365ae8b6 Fix insertion point in pcmp expander. 
While I'm there, clean up too many \n even for me.

llvm-svn: 120411
 2010-11-30 08:20:21 +00:00
Eric Christopher
2170738538 Fix some cleanups from my last patch. 
llvm-svn: 120410
 2010-11-30 08:10:28 +00:00
Eric Christopher
f27f0b5234 Rewrite mwait and monitor support and custom lower arguments. 
Fixes PR8573.

llvm-svn: 120404
 2010-11-30 07:20:12 +00:00
Rafael Espindola
9287c4b38f Move lowering of TLS_addr32 and TLS_addr64 to X86MCInstLower. 
llvm-svn: 120263
 2010-11-28 21:16:39 +00:00
Rafael Espindola
45cd9713f2 Lower TLS_addr32 and TLS_addr64. 
llvm-svn: 120225
 2010-11-27 20:43:02 +00:00
Wesley Peck
d589353ad0 Renaming ISD::BIT_CONVERT to ISD::BITCAST to better reflect the LLVM IR concept. 
llvm-svn: 119990
 2010-11-23 03:31:01 +00:00
Anton Korobeynikov
269e7d3be1 Move hasFP() and few related hooks to TargetFrameInfo. 
llvm-svn: 119740
 2010-11-18 21:19:35 +00:00
Chris Lattner
9a0a840839 add targetoperand flags for jump tables, constant pool and block address 
nodes to indicate when ha16/lo16 modifiers should be used.  This lets
us pass PowerPC/indirectbr.ll.

The one annoying thing about this patch is that the MCSymbolExpr isn't
expressive enough to represent ha16(label1-label2) which we need on
PowerPC.  I have a terrible hack in the meantime, but this will have
to be revisited at some point.

Last major conversion item left is global variable references.

llvm-svn: 119105
 2010-11-15 02:46:57 +00:00
Chris Lattner
51168d6510 move the pic base symbol stuff up to MachineFunction 
since it is trivial and will be shared between ppc and x86.
This substantially simplifies the X86 backend also.

llvm-svn: 119089
 2010-11-14 22:48:15 +00:00
Chris Lattner
09ea2ce78f simplify getPICBaseSymbol a bit. 
llvm-svn: 119088
 2010-11-14 22:37:11 +00:00
Peter Collingbourne
4ec6b2dbe7 Recognise 32-bit ror-based bswap implementation used by uclibc 
llvm-svn: 119007
 2010-11-13 19:54:30 +00:00
Peter Collingbourne
6c3094234e Support ; as asm separator 
llvm-svn: 119006
 2010-11-13 19:54:23 +00:00
Dale Johannesen
659061fcd3 Remove possibly useful info from comment, per Chris. 
llvm-svn: 118865
 2010-11-12 00:43:18 +00:00
Duncan Sands
41edf30895 Simplify uses of MVT and EVT. An MVT can be compared directly 
with a SimpleValueType, while an EVT supports equality and
inequality comparisons with SimpleValueType.

llvm-svn: 118169
 2010-11-03 12:17:33 +00:00
Duncan Sands
92f33ea784 Factorize the duplicated logic for choosing the right argument 
calling convention out of the fast and normal ISel files, and
into the calling convention TD file.

llvm-svn: 117856
 2010-10-31 13:21:44 +00:00
John Thompson
6115a7f1d4 Inline asm multiple alternative constraints development phase 2 - improved basic logic, added initial platform support. 
llvm-svn: 117667
 2010-10-29 17:29:13 +00:00
Michael J. Spencer
5518dda87e x86-Win32: Switch ftol2 calling convention from stdcall to C. 
llvm-svn: 117474
 2010-10-27 18:52:38 +00:00
Dale Johannesen
2566d39d07 An stdcall function calling a non-stdcall function 
cannot use tailcall.  PR 8461.

llvm-svn: 117322
 2010-10-25 22:17:05 +00:00
Duncan Sands
8c5f243aa0 Add parentheses to pacify gcc, which warns otherwise. 
llvm-svn: 117020
 2010-10-21 16:02:12 +00:00
Michael J. Spencer
5a68d7ce94 X86: Add alloca probing to dynamic alloca on Windows. Fixes PR8424. 
llvm-svn: 116984
 2010-10-21 01:41:01 +00:00
Dale Johannesen
58fe3193a6 Remove Synthesizable from the Type system; as MMX vector 
types are no longer Legal on X86, we don't need it.
No functional change.  8499854.

llvm-svn: 116947
 2010-10-20 21:32:10 +00:00
Michael J. Spencer
e528f2588f X86: Add MS-CRT libcalls. 
llvm-svn: 116801
 2010-10-19 07:32:52 +00:00
Michael J. Spencer
cd6be63d05 Fix Whitespace. 
llvm-svn: 116800
 2010-10-19 07:32:42 +00:00
Eric Christopher
303692631e Combine these together - should probably have some text associated 
that says what why what we just asserted is wrong.

llvm-svn: 116333
 2010-10-12 19:44:17 +00:00
Nick Lewycky
a6815ae877 Mark variable 'NoImplicitFloatOps' used only in an assert as used. 
llvm-svn: 116323
 2010-10-12 18:18:03 +00:00
Dan Gohman
d904add908 Initial va_arg support for x86-64. Patch by David Meyer! 
llvm-svn: 116319
 2010-10-12 18:00:49 +00:00
Andrew Trick
5704a15e36 Fixes bug 8297: i386 cmpxchg8b, missing MachineMemOperand 
llvm-svn: 116214
 2010-10-11 19:02:04 +00:00
Michael J. Spencer
08b96a9209 X86: Call ulldiv and ftol2 on Windows instead of their libgcc eqivilents. 
llvm-svn: 116188
 2010-10-11 05:29:15 +00:00
Michael J. Spencer
8f7251a0e9 X86: MinGW should always use libgcc on Windows. 
llvm-svn: 116177
 2010-10-10 23:11:06 +00:00
Michael J. Spencer
b13cb4dd72 X86: Call _alldiv instead of __divdi3 on Windows (excluding cygwin). 
llvm-svn: 116174
 2010-10-10 22:04:34 +00:00
Michael J. Spencer
ae56350c76 Fix Whitespace. 
llvm-svn: 116173
 2010-10-10 22:04:20 +00:00
Cameron Esfahani
664317d6cd Recommit 116056, now with the missing file... 
llvm-svn: 116083
 2010-10-08 19:24:18 +00:00
Andrew Trick
0d8a3d67e0 reverting 116056: win64_params.ll may need to be conditionalized? 
llvm-svn: 116063
 2010-10-08 17:22:42 +00:00
Cameron Esfahani
a9f8bb1356 Small patch to restore home register stack space allocation for the Win64 case. Add test case. This code eventually needs to be tighter, since it's always allocating it, even in leaf routines. 
llvm-svn: 116056
 2010-10-08 10:31:30 +00:00
Evan Cheng
7c89d70f27 Canonicalize X86ISD::MOVDDUP nodes to v2f64 to make sure all cases match. Also eliminate unneeded isel patterns. rdar://8520311 
llvm-svn: 115977
 2010-10-07 20:50:20 +00:00
Anton Korobeynikov
f1acea8615 va_args support for Win64. 
Patch by Cameron!

llvm-svn: 115480
 2010-10-03 22:52:07 +00:00
Dale Johannesen
c14a1eda84 Massive rewrite of MMX: 
The x86_mmx type is used for MMX intrinsics, parameters and
return values where these use MMX registers, and is also
supported in load, store, and bitcast.

Only the above operations generate MMX instructions, and optimizations
do not operate on or produce MMX intrinsics. 

MMX-sized vectors <2 x i32> etc. are lowered to XMM or split into
smaller pieces.  Optimizations may occur on these forms and the
result casted back to x86_mmx, provided the result feeds into a
previous existing x86_mmx operation.

The point of all this is prevent optimizations from introducing
MMX operations, which is unsafe due to the EMMS problem.

llvm-svn: 115243
 2010-09-30 23:57:10 +00:00
Chris Lattner
2e3f9253fd improve indentation 
llvm-svn: 114815
 2010-09-27 06:34:01 +00:00
Eric Christopher
467683e4ab This code should never fire on non-darwin subtargets. 
llvm-svn: 114811
 2010-09-27 06:01:51 +00:00
Dale Johannesen
d75242fa84 We can't return SSE/MMX vectors if SSE is disabled. 
llvm-svn: 114745
 2010-09-24 19:05:48 +00:00
Bob Wilson
3ff9f2d102 Attempt to fix llvm-gcc build. It was crashing when building gcov.o for an 
ARM cross-compiler on x86, because the MMO size did not match the type size.
This fixes the MMO size and also the size of the stack object to match the
type size.

llvm-svn: 114554
 2010-09-22 17:35:14 +00:00
Chris Lattner
26d11d7501 reimplement elf TLS support in terms of addressing modes, eliminating SegmentBaseAddress. 
llvm-svn: 114529
 2010-09-22 04:39:11 +00:00
Chris Lattner
c81dcbec9e convert the last 4 X86ISD nodes that should have memoperands to have them. 
llvm-svn: 114523
 2010-09-22 01:28:21 +00:00
Chris Lattner
29754fc406 give X86ISD::FNSTCW16m a memoperand, since it touches memory. It only 
can access the stack due to how it is generated though.

llvm-svn: 114522
 2010-09-22 01:11:26 +00:00
Chris Lattner
fee1ac61bd give FP_TO_INT16_IN_MEM and friends a memoperand. They are only 
used with stack slots, but hey, lets be safe.

llvm-svn: 114521
 2010-09-22 01:05:16 +00:00
Chris Lattner
e52da86fab give VZEXT_LOAD a memory operand, it now works with segment registers. 
llvm-svn: 114515
 2010-09-22 00:34:38 +00:00
Chris Lattner
f9861312cb give LCMPXCHG_DAG[8] a memory operand, allowing it to work with addrspace 256/257 
llvm-svn: 114508
 2010-09-21 23:59:42 +00:00
Owen Anderson
f6dd8e7f5c Reimplement r114460 in target-independent DAGCombine rather than target-dependent, by using 
the predicate to discover the number of sign bits.  Enhance X86's target lowering to provide
a useful response to this query.

llvm-svn: 114473
 2010-09-21 20:42:50 +00:00
Chris Lattner
3dde58c15a convert a couple more places to use the new getStore() 
llvm-svn: 114463
 2010-09-21 18:51:21 +00:00
Owen Anderson
97a8fdc19c When adding the carry bit to another value on X86, exploit the fact that the carry-materialization 
(sbbl x, x) sets the registers to 0 or ~0.  Combined with two's complement arithmetic, we can fold
the intermediate AND and the ADD into a single SUB.

This fixes <rdar://problem/8449754>.

llvm-svn: 114460
 2010-09-21 18:41:19 +00:00
Chris Lattner
5f584efd31 eliminate some uses of the getStore overload. 
llvm-svn: 114453
 2010-09-21 17:50:43 +00:00
Chris Lattner
4320dda4fb convert the targets off the non-MachinePointerInfo of getLoad. 
llvm-svn: 114410
 2010-09-21 06:44:06 +00:00
Chris Lattner
112cf9bc89 it's more elegant to put the "getConstantPool" and 
"getFixedStack" on the MachinePointerInfo class.  While
this isn't the problem I'm setting out to solve, it is the
right way to eliminate PseudoSourceValue, so lets go with it.

llvm-svn: 114406
 2010-09-21 06:22:23 +00:00
Chris Lattner
810a630851 update the X86 backend to use the MachinePointerInfo version of one 
of the getLoad methods.  This fixes at least one bug where an incorrect
svoffset is passed in (a potential combiner-aa miscompile).

llvm-svn: 114404
 2010-09-21 06:02:19 +00:00
Chris Lattner
f94de5bf46 reimplement memcpy/memmove/memset lowering to use MachinePointerInfo 
instead of srcvalue/offset pairs.  This corrects SV info for mem 
operations whose size is > 32-bits.

llvm-svn: 114401
 2010-09-21 05:40:29 +00:00
Chris Lattner
2edbad8a3d convert targets to the new MF.getMachineMemOperand interface. 
llvm-svn: 114391
 2010-09-21 04:39:43 +00:00
John Thompson
ae3a86d6de Added skeleton for inline asm multiple alternative constraint support. 
llvm-svn: 113766
 2010-09-13 18:15:37 +00:00
Bruno Cardoso Lopes
a74c787dc8 Minor change. Fix comments and remove unused and redundant code 
llvm-svn: 113378
 2010-09-08 18:12:31 +00:00
Bruno Cardoso Lopes
892c337123 x86 vector shuffle lowering now relies only on target specific 
nodes to emit shuffles and don't do isel mask matching anymore.
- Add the selection of the remaining shuffle opcode (movddup)
- Introduce two new functions to "recognize" where we may get
potential folds and add several comments to them explaining why
they are not yet in the desidered shape.
- Add more patterns to fallback the case where we select
a specific shuffle opcode as if it could fold a load, but it
can't, so remap to a valid instruction.
- Add a couple of FIXMEs to address in the following days once
there's a good solution to the current folding problem.

llvm-svn: 113369
 2010-09-08 17:43:25 +00:00
Bruno Cardoso Lopes
a4ca8c3ac5 Factor out some x86 vector shuffle rewriting and add comments about the direction the shuffle lowering is heading to 
llvm-svn: 113286
 2010-09-07 21:03:14 +00:00
Bruno Cardoso Lopes
e33983dba9 Move code around to prepare for moving some of the logic together to another function 
llvm-svn: 113267
 2010-09-07 20:20:27 +00:00
Bill Wendling
9bb7ac566f Add an MVT::x86mmx type. It will take the place of all current MMX vector types. 
llvm-svn: 113261
 2010-09-07 20:03:56 +00:00
Bruno Cardoso Lopes
21e1fc67c3 decouple MMX check from regular splat checks. Some refactoring is coming, and MMX should be left alone to be easily removed after moving to intrinsics 
llvm-svn: 113247
 2010-09-07 18:41:45 +00:00
Bruno Cardoso Lopes
dcc8690051 Remove now useless check, because the code can be matched below, no need to leave it for isel 
llvm-svn: 113242
 2010-09-07 18:29:03 +00:00
Bruno Cardoso Lopes
e6f7e4684d Minor change. Since the checks are equivalent, use isMMX 
llvm-svn: 113239
 2010-09-07 18:24:00 +00:00
Bruno Cardoso Lopes
8d76bbe31c Remove the last bit of isShuffleMaskLegal checks and improve the comment regarding mmx shuffles 
llvm-svn: 113059
 2010-09-04 02:58:56 +00:00
Bruno Cardoso Lopes
fccf00be8c make explicit that we not handle several mmx shuffles 
llvm-svn: 113058
 2010-09-04 02:50:13 +00:00
Bruno Cardoso Lopes
0562140ec3 Emit target specific nodes to handle palignr. Do not touch it for MMX versions yet. 
llvm-svn: 113056
 2010-09-04 02:36:07 +00:00
Bruno Cardoso Lopes
73651844b2 Emit target specific nodes to handle splats starting at zero indicies 
llvm-svn: 113055
 2010-09-04 02:02:14 +00:00