Commit Graph

21598 Commits

Author SHA1 Message Date
Chris Lattner 726075fdf8 Teach the addressing mode stuff to fold "%lo" into 'ri' addressing modes,
allowing us to compile this:

to this:

%G1 = external global int
%G2 = external global int
void %test() {
        %X = load int* %G1
        store int %X, int* %G2
        ret void
}

test:
        save -96, %sp, %sp
        sethi %hi(G1), %l0
        ld [%l0+%lo(G1)], %l0
        sethi %hi(G2), %l1
        st %l0, [%l1+%lo(G2)]
        restore %g0, %g0, %g0
        retl
        nop

instead of this:

test:
        save -96, %sp, %sp
        sethi %hi(G1), %l0
        or %g0, %lo(G1), %l1
        ld [%l1+%l0], %l0
        sethi %hi(G2), %l1
        or %g0, %lo(G2), %l2
        st %l0, [%l2+%l1]
        restore %g0, %g0, %g0
        retl
        nop

llvm-svn: 24812
2005-12-18 02:27:00 +00:00
Chris Lattner a983c3df1c Add initial support for global variables, and fix a bug in addr mode selection
where we didn't select the operands.

llvm-svn: 24811
2005-12-18 02:10:39 +00:00
Chris Lattner a49b652a62 Claiming that branch targets are registers is not very wholesome. Change them
to be basic blocks.  Also, add uncond branches.

llvm-svn: 24810
2005-12-18 01:46:58 +00:00
Chris Lattner b29957500e Add unordered comparisons
llvm-svn: 24809
2005-12-18 01:41:39 +00:00
Chris Lattner e58481be36 Add patterns to the rest of the int condbranches and some of the fp branches
llvm-svn: 24808
2005-12-18 01:38:19 +00:00
Chris Lattner 9cf4bb2867 Add initial conditional branch support. This doesn't actually work yet due
to a bug in the scheduler.

llvm-svn: 24807
2005-12-18 01:20:35 +00:00
Chris Lattner cf12118965 Simplify code
llvm-svn: 24806
2005-12-18 01:03:46 +00:00
Chris Lattner 00759eac78 Eliminate CMPri, which is a synonym for SUBCCri
llvm-svn: 24805
2005-12-17 23:52:08 +00:00
Chris Lattner bf0bd99e03 allow custom expansion of BR_CC
llvm-svn: 24804
2005-12-17 23:46:46 +00:00
Chris Lattner 5a6b03c1b8 add fneg,fabs,fsqrt instructions
llvm-svn: 24803
2005-12-17 23:20:27 +00:00
Chris Lattner 32f19262d5 Add patterns for fround/fextend and the funny fsmuld instruction
llvm-svn: 24802
2005-12-17 23:14:30 +00:00
Chris Lattner 06952dfced Add FP +,-,*,/
llvm-svn: 24801
2005-12-17 23:10:46 +00:00
Chris Lattner 89078880f2 Give patterns to F3_3 instructions
llvm-svn: 24800
2005-12-17 23:05:35 +00:00
Chris Lattner 829572cdca Implement 64-bit add/sub, make sure to receive and return 64-bit args with
the right halves in the right regs

llvm-svn: 24799
2005-12-17 22:55:57 +00:00
Chris Lattner ebfa06a2de implement div and rem
llvm-svn: 24798
2005-12-17 22:39:19 +00:00
Chris Lattner 8eaf9f4cb3 implement MULHU/MULHS for 64-bit multiplies
llvm-svn: 24797
2005-12-17 22:30:00 +00:00
Chris Lattner 4abe9528f9 Add patterns for multiply, simplify Y register handling stuff, add RDY instruction
llvm-svn: 24796
2005-12-17 22:22:53 +00:00
Chris Lattner 2616a0b56f Make the addressing modes smarter
llvm-svn: 24795
2005-12-17 21:25:27 +00:00
Chris Lattner 19ff62dc67 remove some unused instructions
llvm-svn: 24794
2005-12-17 21:13:50 +00:00
Chris Lattner e39ab718c0 add andn/orn/xorn patterns. This allows us to compile this:
long %test(ubyte, short, long %X, long %Y) {
  %A = xor long %X, -1
  %B = and long %Y, %A
  ret long %B
}

to this:

test:
        save -96, %sp, %sp
        andn %i4, %i2, %i0
        andn %i5, %i3, %i1
        restore %g0, %g0, %g0
        retl
        nop

instead of this:

test:
        save -96, %sp, %sp
        xor %i2, -1, %l0
        xor %i3, -1, %l1
        and %i4, %l0, %i0
        and %i5, %l1, %i1
        restore %g0, %g0, %g0
        retl
        nop

The simpleisel emits:  :(

test:
        save -96, %sp, %sp
        or %g0, -1, %l0
        or %g0, -1, %l0
        or %g0, -1, %l0
        or %g0, -1, %l1
        xor %i2, %l0, %l0
        xor %i3, %l1, %l1
        and %i4, %l0, %i0
        and %i5, %l1, %i1
        restore %g0, %g0, %g0
        retl
        nop

llvm-svn: 24793
2005-12-17 21:05:49 +00:00
Chris Lattner ea752cc50a Add support for 64-bit arguments
llvm-svn: 24792
2005-12-17 20:59:06 +00:00
Chris Lattner 6be83f9959 Sparc doesn't have sext_inreg
llvm-svn: 24791
2005-12-17 20:50:42 +00:00
Chris Lattner d10995cb26 add patterns for FP stores
llvm-svn: 24790
2005-12-17 20:47:16 +00:00
Chris Lattner 490a6edf52 Add [reg+reg] integer stores
llvm-svn: 24789
2005-12-17 20:44:36 +00:00
Chris Lattner b57bb13253 Add store patterns
llvm-svn: 24788
2005-12-17 20:42:55 +00:00
Chris Lattner 233e044738 add truncstore
llvm-svn: 24787
2005-12-17 20:42:29 +00:00
Chris Lattner 7e7c355154 add fp load patterns, switch rest of loads and stores to use addrmodes
llvm-svn: 24786
2005-12-17 20:32:47 +00:00
Chris Lattner c4f3a7adea Add integer load[r+r] forms.
llvm-svn: 24785
2005-12-17 20:26:45 +00:00
Chris Lattner 1c02c45f18 Rename load/store instructions to include an RI suffix
llvm-svn: 24784
2005-12-17 20:18:49 +00:00
Chris Lattner 4fa86e1d55 Add patterns for the rest of the loads. Add 'ri' suffixes to the load and store insts
llvm-svn: 24783
2005-12-17 20:18:24 +00:00
Chris Lattner 5d15f9ed60 Add basic addressing mode support and one load.
llvm-svn: 24782
2005-12-17 20:04:49 +00:00
Chris Lattner db8e888fb5 eliminate some redundancy
llvm-svn: 24781
2005-12-17 19:47:05 +00:00
Chris Lattner 5e68639009 Use a combination of sethi and or to build arbitrary immediates.
llvm-svn: 24780
2005-12-17 19:41:43 +00:00
Chris Lattner 8546257435 Use sethi to build large immediates with zeros at the bottom
llvm-svn: 24779
2005-12-17 19:37:00 +00:00
Chris Lattner 7b6f2e879d Add shift and small immediate support
llvm-svn: 24778
2005-12-17 19:07:57 +00:00
Chris Lattner 30f924e3ca Add some basic reg-reg instructions
llvm-svn: 24777
2005-12-17 18:53:33 +00:00
Chris Lattner 2edb4b7f99 Add empty patterns to all F3_1 instructions
llvm-svn: 24776
2005-12-17 18:49:14 +00:00
Evan Cheng 1d71248392 Darwin API issue: indirect load of external and weak symbols.
llvm-svn: 24775
2005-12-17 09:13:43 +00:00
Chris Lattner 866cef563b Add some simple integer patterns. This allows us to compile this:
int %test(int %A) {
  %B = add int %A, 1
  %C = xor int %B, 123
  ret int %C
}

into this:

test:
        save -96, %sp, %sp
        add %i0, 1, %l0
        xor %l0, 123, %i0
        restore %g0, %g0, %g0
        retl
        nop

for example.  I guess it would make sense to add reg/reg versions too.

llvm-svn: 24774
2005-12-17 08:26:38 +00:00
Chris Lattner 80a3875bc1 Implement ret with operand, giving us this:
int %test(int %A) {
  ret int %A
}

llvm-svn: 24773
2005-12-17 08:15:09 +00:00
Chris Lattner 1136b7a2e0 Add a pattern for 'ret'. This now compiles:
void %test() { ret void }

:)

llvm-svn: 24772
2005-12-17 08:08:42 +00:00
Chris Lattner 1549e4d590 Add empty patterns for F3_2 instructions
llvm-svn: 24771
2005-12-17 08:06:43 +00:00
Chris Lattner 9f1c860e1e Implement LowerArguments, at least for the first 6 integer args
llvm-svn: 24770
2005-12-17 08:03:24 +00:00
Chris Lattner 4f34e9f7ff Add the framework for a dag-dag isel
llvm-svn: 24769
2005-12-17 07:47:01 +00:00
Evan Cheng f3b16bc5a0 Remove a few lines of dead code.
llvm-svn: 24768
2005-12-17 07:18:44 +00:00
Chris Lattner 69b5d17f92 asmprinter done, added crucial missing step
llvm-svn: 24767
2005-12-17 07:17:59 +00:00
Chris Lattner 55f9dbe1ea Use the AsmPrinter for global variable init printing. This eliminates a
bunch of code and causes V8 to start using the fancy .asciz directive that
the sun assembler supports.

llvm-svn: 24766
2005-12-17 07:17:08 +00:00
Chris Lattner fb7fd98cd4 Switch constant pool printing over to use the Shared AsmPrinter version
llvm-svn: 24765
2005-12-17 07:11:43 +00:00
Chris Lattner b808c8e2e4 Use the shared AsmPrinter code for some basic stuff. No functionality
change except for fewer .section directives emitted

llvm-svn: 24764
2005-12-17 07:04:29 +00:00
Evan Cheng 7087cd275b Added an idea about any_extend for performance tuning.
llvm-svn: 24763
2005-12-17 06:54:43 +00:00