comparing a long against zero got us this:

        sub %ESP, 8
        mov DWORD PTR [%ESP + 4], %ESI
        mov DWORD PTR [%ESP], %EDI
        mov %EAX, DWORD PTR [%ESP + 12]
        mov %EDX, DWORD PTR [%ESP + 16]
        mov %ECX, 0
        mov %ESI, 0
        mov %EDI, %EAX
        xor %EDI, %ECX
        mov %ECX, %EDX
        xor %ECX, %ESI
        or %EDI, %ECX
        sete %CL
        test %CL, %CL
        je .LBB2 # PC rel: F

Now it gets us this:

        mov %EAX, DWORD PTR [%ESP + 4]
        mov %EDX, DWORD PTR [%ESP + 8]
        mov %ECX, %EAX
        or %ECX, %EDX
        sete %CL
        test %CL, %CL
        je .LBB2 # PC rel: F

llvm-svn: 12696

example, multiplying X*(1 + (1LL << 32)) now produces:

test:
        mov %ECX, DWORD PTR [%ESP + 4]
        mov %EDX, DWORD PTR [%ESP + 8]
        mov %EAX, %ECX
        add %EDX, %ECX
        ret

[[[Note to Alkis: why isn't linear scan generating this code??  This might be a
 problem with your intervals being too conservative:

test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %EDX, DWORD PTR [%ESP + 8]
        add %EDX, %EAX
        ret

end note]]]

Whereas GCC produces this:

T:
        sub     %esp, 12
        mov     %edx, DWORD PTR [%esp+16]
        mov     DWORD PTR [%esp+8], %edi
        mov     %ecx, DWORD PTR [%esp+20]
        xor     %edi, %edi
        mov     DWORD PTR [%esp], %ebx
        mov     %ebx, %edi
        mov     %eax, %edx
        mov     DWORD PTR [%esp+4], %esi
        add     %ebx, %edx
        mov     %edi, DWORD PTR [%esp+8]
        lea     %edx, [%ecx+%ebx]
        mov     %esi, DWORD PTR [%esp+4]
        mov     %ebx, DWORD PTR [%esp]
        add     %esp, 12
        ret

I'm not sure example what GCC is smoking here, but it looks like it has just
confused itself with a bunch of stack slots or something.  The intel compiler
is better, but still not good:

T:
        movl      4(%esp), %edx                                 #2.11
        movl      8(%esp), %eax                                 #2.11
        lea       (%eax,%edx), %ecx                             #3.12
        movl      $1, %eax                                      #3.12
        mull      %edx                                          #3.12
        addl      %ecx, %edx                                    #3.12
        ret                                                     #3.12

llvm-svn: 12693

long %test(long %X) {
        %Y = mul long %X, 123
        ret long %Y
}

we used to generate:

test:
        sub %ESP, 12
        mov DWORD PTR [%ESP + 8], %ESI
        mov DWORD PTR [%ESP + 4], %EDI
        mov DWORD PTR [%ESP], %EBX
        mov %ECX, DWORD PTR [%ESP + 16]
        mov %ESI, DWORD PTR [%ESP + 20]
        mov %EDI, 123
        mov %EBX, 0
        mov %EAX, %ECX
        mul %EDI
        imul %ESI, %EDI
        add %ESI, %EDX
        imul %ECX, %EBX
        add %ESI, %ECX
        mov %EDX, %ESI
        mov %EBX, DWORD PTR [%ESP]
        mov %EDI, DWORD PTR [%ESP + 4]
        mov %ESI, DWORD PTR [%ESP + 8]
        add %ESP, 12
        ret

Now we emit:
test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
        mov %EDX, 123
        mul %EDX
        imul %ECX, %ECX, 123
        add %ECX, %EDX
        mov %EDX, %ECX
        ret

Which, incidently, is substantially nicer than what GCC manages:
T:
        sub     %esp, 8
        mov     %eax, 123
        mov     DWORD PTR [%esp], %ebx
        mov     %ebx, DWORD PTR [%esp+16]
        mov     DWORD PTR [%esp+4], %esi
        mov     %esi, DWORD PTR [%esp+12]
        imul    %ecx, %ebx, 123
        mov     %ebx, DWORD PTR [%esp]
        mul     %esi
        mov     %esi, DWORD PTR [%esp+4]
        add     %esp, 8
        lea     %edx, [%ecx+%edx]
        ret

llvm-svn: 12692

On this testcase:

long %test(long %X) {
        %Y = shr long %X, ubyte 32
        ret long %Y
}

instead of:
t:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %EAX, DWORD PTR [%ESP + 8]
        sar %EAX, 0
        mov %EDX, 0
        ret


we now emit:
test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %EAX, DWORD PTR [%ESP + 8]
        mov %EDX, 0
        ret

llvm-svn: 12688

llvm-svn: 12687

For example, on this instruction:

        call void %test(long 1234)

Instead of this:
        mov %EAX, 1234
        mov %ECX, 0
        mov DWORD PTR [%ESP], %EAX
        mov DWORD PTR [%ESP + 4], %ECX
        call test

We now emit this:
        mov DWORD PTR [%ESP], 1234
        mov DWORD PTR [%ESP + 4], 0
        call test

llvm-svn: 12686

of the words of the constant is zeros.  For example:
  Y = and long X, 1234

now generates:
  Yl = and Xl, 1234
  Yh = 0

instead of:
  Yl = and Xl, 1234
  Yh = and Xh, 0

llvm-svn: 12685

llvm-svn: 12684

This allows us to handle code like 'add long %X, 123456789012' more efficiently.

llvm-svn: 12683

long %test(long %X) {
        %Y = sub long 0, %X
        ret long %Y
}

We used to generate:

test:
        sub %ESP, 4
        mov DWORD PTR [%ESP], %ESI
        mov %ECX, DWORD PTR [%ESP + 8]
        mov %ESI, DWORD PTR [%ESP + 12]
        mov %EAX, 0
        mov %EDX, 0
        sub %EAX, %ECX
        sbb %EDX, %ESI
        mov %ESI, DWORD PTR [%ESP]
        add %ESP, 4
        ret

Now we generate:

test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %EDX, DWORD PTR [%ESP + 8]
        neg %EAX
        adc %EDX, 0
        neg %EDX
        ret

llvm-svn: 12681

llvm-svn: 12680

  * In promote32, if we can just promote a constant value, do so instead of
    promoting a constant dynamically.
  * In visitReturn inst, actually USE the promote32 argument that takes a
    Value*

The end result of this is that we now generate this:

test:
        mov %EAX, 0
        ret

instead of...

test:
        mov %AX, 0
        movzx %EAX, %AX
        ret

for:

ushort %test() {
        ret ushort 0
}

llvm-svn: 12679

types and can have arbitrary 32- and 64-bit integer types indexing into
sequential types.

llvm-svn: 12653

llvm-svn: 12615

llvm-svn: 12610

llvm-svn: 12579

the X86 does not support a full set of fp cmove instructions, so we can't always
fold the condition into the select.  :(  Yuck.

llvm-svn: 12577

using our broad selection of movcc instructions.  :)

llvm-svn: 12560

folding compares into the select yet.

llvm-svn: 12553

an incoming value from a block, the selector would evaluate the constant
at the TOP of the block instead of at the end of the block.  This made the
live range for the constant span the entire block, increasing register
pressure needlessly.

llvm-svn: 12542

llvm-svn: 12500

folding load instructions into other instructions across free instruction
boundaries.  Perhaps this will also fix the other strange failures?

llvm-svn: 12494

llvm-svn: 12357

Intrinsic::va*.  This avoid conflicting with macros in the stdlib.h file.

llvm-svn: 12356

testcase like this:

int %test(int* %P, int %A) {
        %Pv = load int* %P
        %B = add int %A, %Pv
        ret int %B
}

We now generate:
test:
        mov %ECX, DWORD PTR [%ESP + 4]
        mov %EAX, DWORD PTR [%ESP + 8]
        add %EAX, DWORD PTR [%ECX]
        ret

Instead of:
test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
        mov %EAX, DWORD PTR [%EAX]
        add %EAX, %ECX
        ret

... saving one instruction, and often a register.  Note that there are a lot
of other instructions that could use this, but they aren't handled.  I'm not
really interested in adding them, but mul/div and all of the FP instructions
could be supported as well if someone wanted to add them.

llvm-svn: 12204

llvm-svn: 12203

llvm-svn: 12064

of generating this code:

        mov %EAX, 4
        mov DWORD PTR [%ESP], %EAX
        mov %AX, 123
        movsx %EAX, %AX
        mov DWORD PTR [%ESP + 4], %EAX
        call Y

we now generate:
        mov DWORD PTR [%ESP], 4
        mov DWORD PTR [%ESP + 4], 123
        call Y

Which hurts the eyes less.  :)

Considering that register pressure around call sites is already high (with all
of the callee clobber registers n stuff), this may help a lot.

llvm-svn: 12028

to function calls, we would emit dead code, like this:

int Y(int, short, double);
int X() {
  Y(4, 123, 4);
}

--- Old
X:
        sub %ESP, 20
        mov %EAX, 4
        mov DWORD PTR [%ESP], %EAX
***     mov %AX, 123
        mov %AX, 123
        movsx %EAX, %AX
        mov DWORD PTR [%ESP + 4], %EAX
        fld QWORD PTR [.CPIX_0]
        fstp QWORD PTR [%ESP + 8]
        call Y
        mov %EAX, 0
        # IMPLICIT_USE %EAX %ESP
        add %ESP, 20
        ret

Now we emit:
X:
        sub %ESP, 20
        mov %EAX, 4
        mov DWORD PTR [%ESP], %EAX
        mov %AX, 123
        movsx %EAX, %AX
        mov DWORD PTR [%ESP + 4], %EAX
        fld QWORD PTR [.CPIX_0]
        fstp QWORD PTR [%ESP + 8]
        call Y
        mov %EAX, 0
        # IMPLICIT_USE %EAX %ESP
        add %ESP, 20
        ret

Next up, eliminate the mov AX and movsx entirely!

llvm-svn: 12026

their names more decriptive. A name consists of the base name, a
default operand size followed by a character per operand with an
optional special size. For example:

ADD8rr -> add, 8-bit register, 8-bit register

IMUL16rmi -> imul, 16-bit register, 16-bit memory, 16-bit immediate

IMUL16rmi8 -> imul, 16-bit register, 16-bit memory, 8-bit immediate

MOVSX32rm16 -> movsx, 32-bit register, 16-bit memory

llvm-svn: 11995

Replace uses of addZImm with addImm.

llvm-svn: 11992

block loops.

llvm-svn: 11990

llvm-svn: 11984

to denote this fact.

llvm-svn: 11972

denote this fact.

llvm-svn: 11971

consistent with the rest and also pepare for the addition of their
memory operand variants.

llvm-svn: 11902

MRegisterInfo::is{Physical,Virtual}Register. Apply appropriate fixes
to relevant files.

llvm-svn: 11882

bugs.  Thanks Brian!

llvm-svn: 11859

scaled indexes.  This allows us to compile GEP's like this:

int* %test([10 x { int, { int } }]* %X, int %Idx) {
        %Idx = cast int %Idx to long
        %X = getelementptr [10 x { int, { int } }]* %X, long 0, long %Idx, ubyte 1, ubyte 0
        ret int* %X
}

Into a single address computation:

test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
        lea %EAX, DWORD PTR [%EAX + 8*%ECX + 4]
        ret

Before it generated:
test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
        shl %ECX, 3
        add %EAX, %ECX
        lea %EAX, DWORD PTR [%EAX + 4]
        ret

This is useful for things like int/float/double arrays, as the indexing can be folded into
the loads&stores, reducing register pressure and decreasing the pressure on the decode unit.
With these changes, I expect our performance on 256.bzip2 and gzip to improve a lot.  On
bzip2 for example, we go from this:

10665 asm-printer           - Number of machine instrs printed
   40 ra-local              - Number of loads/stores folded into instructions
 1708 ra-local              - Number of loads added
 1532 ra-local              - Number of stores added
 1354 twoaddressinstruction - Number of instructions added
 1354 twoaddressinstruction - Number of two-address instructions
 2794 x86-peephole          - Number of peephole optimization performed

to this:
9873 asm-printer           - Number of machine instrs printed
  41 ra-local              - Number of loads/stores folded into instructions
1710 ra-local              - Number of loads added
1521 ra-local              - Number of stores added
 789 twoaddressinstruction - Number of instructions added
 789 twoaddressinstruction - Number of two-address instructions
2142 x86-peephole          - Number of peephole optimization performed

... and these types of instructions are often in tight loops.

Linear scan is also helped, but not as much.  It goes from:

8787 asm-printer           - Number of machine instrs printed
2389 liveintervals         - Number of identity moves eliminated after coalescing
2288 liveintervals         - Number of interval joins performed
3522 liveintervals         - Number of intervals after coalescing
5810 liveintervals         - Number of original intervals
 700 spiller               - Number of loads added
 487 spiller               - Number of stores added
 303 spiller               - Number of register spills
1354 twoaddressinstruction - Number of instructions added
1354 twoaddressinstruction - Number of two-address instructions
 363 x86-peephole          - Number of peephole optimization performed

to:

7982 asm-printer           - Number of machine instrs printed
1759 liveintervals         - Number of identity moves eliminated after coalescing
1658 liveintervals         - Number of interval joins performed
3282 liveintervals         - Number of intervals after coalescing
4940 liveintervals         - Number of original intervals
 635 spiller               - Number of loads added
 452 spiller               - Number of stores added
 288 spiller               - Number of register spills
 789 twoaddressinstruction - Number of instructions added
 789 twoaddressinstruction - Number of two-address instructions
 258 x86-peephole          - Number of peephole optimization performed

Though I'm not complaining about the drop in the number of intervals.  :)

llvm-svn: 11820

  to do analysis.

*** FOLD getelementptr instructions into loads and stores when possible,
    making use of some of the crazy X86 addressing modes.

For example, the following C++ program fragment:

struct complex {
    double re, im;
    complex(double r, double i) : re(r), im(i) {}
};
inline complex operator+(const complex& a, const complex& b) {
    return complex(a.re+b.re, a.im+b.im);
}
complex addone(const complex& arg) {
    return arg + complex(1,0);
}

Used to be compiled to:
_Z6addoneRK7complex:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
***     mov %EDX, %ECX
        fld QWORD PTR [%EDX]
        fld1
        faddp %ST(1)
***     add %ECX, 8
        fld QWORD PTR [%ECX]
        fldz
        faddp %ST(1)
***     mov %ECX, %EAX
        fxch %ST(1)
        fstp QWORD PTR [%ECX]
***     add %EAX, 8
        fstp QWORD PTR [%EAX]
        ret

Now it is compiled to:
_Z6addoneRK7complex:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
        fld QWORD PTR [%ECX]
        fld1
        faddp %ST(1)
        fld QWORD PTR [%ECX + 8]
        fldz
        faddp %ST(1)
        fxch %ST(1)
        fstp QWORD PTR [%EAX]
        fstp QWORD PTR [%EAX + 8]
        ret

Other programs should see similar improvements, across the board.  Note that
in addition to reducing instruction count, this also reduces register pressure
a lot, always a good thing on X86.  :)

llvm-svn: 11819