Select vector shift with non-immediate i32 shift amount operand by first moving the operand into the right register.

llvm-svn: 50619

using dyn_cast<> on already freed memory.

llvm-svn: 50618

llvm-svn: 50616

llvm-svn: 50615

llvm-svn: 50614

llvm-svn: 50607

llvm-svn: 50604

llvm-svn: 50603

Add separate intrinsics for MMX / SSE shifts with i32 integer operands. This allow us to simplify the horribly complicated matching code.

llvm-svn: 50601

llvm-svn: 50600

llvm-svn: 50596

llvm-svn: 50594

llvm-svn: 50593

llvm-svn: 50591

llvm-svn: 50590

llvm-svn: 50586

llvm-svn: 50582

llvm-svn: 50578

llvm-svn: 50576

llvm-svn: 50575

Not safe folding a load + FsXORPSrr into FsXORPSrm. It's loading a FR64 value but the load folding variant expects a 16-byte aligned address.

llvm-svn: 50574

llvm-svn: 50564

llvm-svn: 50562

llvm-svn: 50561

llvm-svn: 50560

llvm-svn: 50558

llvm-svn: 50548

from x86-64 calling convention work.

llvm-svn: 50545

fixes the 254.gap regression on x86 and the 403.gcc regression on x86-64.

llvm-svn: 50537

llvm-svn: 50536

llvm-svn: 50527

ffastmath mode.  This fixes rdar://5902801, a miscompilation
of gcc.dg/builtins-8.c.

Bill, please pull this into Tak.

llvm-svn: 50523

2) Return NULL instead of false in several places for tidiness.
3) fix a bug optimizing  sprintf(p, "%c", x)

llvm-svn: 50521

a FunctionPass.  This makes it simpler, fixes dozens of bugs, adds
a couple of minor features, and shrinks is considerably: from
2214 to 1437 lines.

llvm-svn: 50520

llvm-svn: 50519

llvm-svn: 50518

llvm-svn: 50517

llvm-svn: 50516

llvm-svn: 50515

a function call that takes two Value*'s as arguments.

llvm-svn: 50514