llvm-svn: 26740

llvm-svn: 26739

transformation decisions.

llvm-svn: 26738

llvm-svn: 26737

llvm-svn: 26736

llvm-svn: 26735

llvm-svn: 26734

registers, and update it on entry to each function, then restore it on exit.

This compiles:

void func(vfloat *a, vfloat *b, vfloat *c) {
        *a = *b * *c + *c;
}

to this:

_func:
        mfspr r2, 256
        oris r6, r2, 49152
        mtspr 256, r6
        lvx v0, 0, r5
        lvx v1, 0, r4
        vmaddfp v0, v1, v0, v0
        stvx v0, 0, r3
        mtspr 256, r2
        blr

GCC produces this (which has additional stack accesses):

_func:
        mfspr r0,256
        stw r0,-4(r1)
        oris r0,r0,0xc000
        mtspr 256,r0
        lvx v0,0,r5
        lvx v1,0,r4
        lwz r12,-4(r1)
        vmaddfp v0,v0,v1,v0
        stvx v0,0,r3
        mtspr 256,r12
        blr

llvm-svn: 26733

llvm-svn: 26732

llvm-svn: 26731

$(Echo) instead of @echo

llvm-svn: 26730

llvm-svn: 26729

llvm-svn: 26728

Regression/CodeGen/PowerPC/and_add.ll

a case that occurs with dynamic allocas of constant size.

llvm-svn: 26727

llvm-svn: 26726

llvm-svn: 26725

llvm-svn: 26724

a select and FABS/FNEG.

This speeds up a trivial (aka stupid) copysign benchmark I wrote from 6.73s
to 2.64s, woo.

llvm-svn: 26723

llvm-svn: 26722

llvm-svn: 26721

llvm-svn: 26720

1. Use flags on the instructions in the .td file to indicate the PPC970 unit
   type instead of a table in the .cpp file.  Much cleaner.
2. Change the hazard recognizer to build d-groups according to the actual
   algorithm used, not my flawed understanding of it.
3. Model "must be in the first slot" and "must be the only instr in a group"
   accurately.

llvm-svn: 26719

to be emitted.

Don't add one to the latency of a completed instruction if the latency of the
op is 0.

llvm-svn: 26718

predecessor to finish before they can start.

llvm-svn: 26717

operands have all issued, but whose results are not yet available.  This
allows us to compile:

int G;
int test(int A, int B, int* P) {
   return (G+A)*(B+1);
}

to:

_test:
        lis r2, ha16(L_G$non_lazy_ptr)
        addi r4, r4, 1
        lwz r2, lo16(L_G$non_lazy_ptr)(r2)
        lwz r2, 0(r2)
        add r2, r2, r3
        mullw r3, r2, r4
        blr

instead of this, which has a stall between the lis/lwz:

_test:
        lis r2, ha16(L_G$non_lazy_ptr)
        lwz r2, lo16(L_G$non_lazy_ptr)(r2)
        addi r4, r4, 1
        lwz r2, 0(r2)
        add r2, r2, r3
        mullw r3, r2, r4
        blr

llvm-svn: 26716

Regression/Transforms/InstCombine/stacksaverestore.ll

llvm-svn: 26715

which cycle it lands on.

llvm-svn: 26714

llvm-svn: 26713

is together, and direction independent code is together.

llvm-svn: 26712

merge succs/chainsuccs -> succs set

This has no functionality change, simplifies the code, and reduces the size
of sunits.

llvm-svn: 26711

llvm-svn: 26710

llvm-svn: 26709

llvm-svn: 26708

set construction, rather than intersecting various std::sets.  This reduces
the memory usage for the testcase in PR681 from 496 to 26MB of ram on my
darwin system, and reduces the runtime from 32.8 to 0.8 seconds on a
2.5GHz G5.  This also enables future code sharing between Dom and PostDom
now that they share near-identical implementations.

llvm-svn: 26707

llvm-svn: 26706

llvm-svn: 26705

off the result string at the first null terminator.

llvm-svn: 26704

llvm-svn: 26703

llvm-svn: 26702

llvm-svn: 26701