Scalar SSE: a < b ? c : 0.0 -> cmpss, andps
Scalar SSE: float -> i16 needs to be promoted

llvm-svn: 22637

llvm-svn: 22634

llvm-svn: 22619

llvm-svn: 22617

llvm-svn: 22596

llvm-svn: 22595

Patch contributed by Jim Laskey!

llvm-svn: 22594

llvm-svn: 22588

llvm-svn: 22571

llvm-svn: 22569

llvm-svn: 22565

llvm-svn: 22561

1 byte loads and other operations.  This is bad for store-forwarding on
common CPUs.  We now do this:

fnstcw WORD PTR [%ESP]
mov %AX, WORD PTR [%ESP]

instead of:

fnstcw WORD PTR [%ESP]
mov %AL, BYTE PTR [%ESP + 1]

llvm-svn: 22559

FP-to-int-in-memory: this exposes the load from the stored slot to the
selection dag, allowing it to be folded into other operaions.

llvm-svn: 22556

llvm-svn: 22553

that the X86 does not support to the legalizer.  This allows it to be better
optimized, etc, and will help with SSE support.

llvm-svn: 22551

llvm-svn: 22550

long %test4(double %X) {
        %tmp.1 = cast double %X to long         ; <long> [#uses=1]
        ret long %tmp.1
}

to this:

_test4:
        sub %ESP, 12
        fld QWORD PTR [%ESP + 16]
        fistp QWORD PTR [%ESP]
        mov %EDX, DWORD PTR [%ESP + 4]
        mov %EAX, DWORD PTR [%ESP]
        add %ESP, 12
        ret

instead of this:

_test4:
        sub %ESP, 28
        fld QWORD PTR [%ESP + 32]
        fstp QWORD PTR [%ESP]
        call ___fixdfdi
        add %ESP, 28
        ret

llvm-svn: 22549

llvm-svn: 22543

llvm-svn: 22538

have to write arguments to the stack

llvm-svn: 22536

llvm-svn: 22535

For the following code:
double %ext(int %A.0__, long %A.1__) {
        %A_addr = alloca %typedef.DComplex              ; <%typedef.DComplex*> [#uses=2]
        %tmp.1 = cast %typedef.DComplex* %A_addr to int*                ; <int*> [#uses=1]
        store int %A.0__, int* %tmp.1
        %tmp.2 = getelementptr %typedef.DComplex* %A_addr, int 0, uint 1                ; <double*> [#uses=2]
        %tmp.3 = cast double* %tmp.2 to long*           ; <long*> [#uses=1]
        store long %A.1__, long* %tmp.3
        %tmp.5 = load double* %tmp.2            ; <double> [#uses=1]
        ret double %tmp.5
}

We now generate:
_ext:
.LBB_ext_0:     ;
        stw r3, -12(r1)
        stw r4, -8(r1)
        stw r5, -4(r1)
        lfd f1, -8(r1)
        blr

Instead of:
_ext:
.LBB_ext_0:     ;
        stw r3, -12(r1)
        addi r2, r1, -12
        stw r4, 4(r2)
        stw r5, 8(r2)
        lfd f1, 4(r2)
        blr

This also fires hundreds of times on MultiSource.

llvm-svn: 22533

llvm-svn: 22530

llvm-svn: 22523

unnecessary SP manipulation in leaf routines that don't need it.

llvm-svn: 22522

llvm-svn: 22521

llvm-svn: 22520

llvm-svn: 22518

llvm-svn: 22517

llvm-svn: 22516

the need to build PIC.

llvm-svn: 22512

llvm-svn: 22508

llvm-svn: 22507

Handle more imm forms, and load small negative i32 constants without hitting memory (should do the same for arbitrary zero extended small negative constants)

llvm-svn: 22505

llvm-svn: 22502

llvm-svn: 22500

llvm-svn: 22499

llvm-svn: 22498

Remove the LoadHiAddr pseudo-instruction.
Optimization of stores to and loads from statics.
Force JIT to use new non-PIC codepaths.

llvm-svn: 22494