Implement support for using modeling implicit-zero-extension on x86-64
with SUBREG_TO_REG, teach SimpleRegisterCoalescing to coalesce
SUBREG_TO_REG instructions (which are similar to INSERT_SUBREG
instructions), and teach the DAGCombiner to take advantage of this on
targets which support it. This eliminates many redundant
zero-extension operations on x86-64.
This adds a new TargetLowering hook, isZExtFree. It's similar to
isTruncateFree, except it only applies to actual definitions, and not
no-op truncates which may not zero the high bits.
Also, this adds a new optimization to SimplifyDemandedBits: transform
operations like x+y into (zext (add (trunc x), (trunc y))) on targets
where all the casts are no-ops. In contexts where the high part of the
add is explicitly masked off, this allows the mask operation to be
eliminated. Fix the DAGCombiner to avoid undoing these transformations
to eliminate casts on targets where the casts are no-ops.
Also, this adds a new two-address lowering heuristic. Since
two-address lowering runs before coalescing, it helps to be able to
look through copies when deciding whether commuting and/or
three-address conversion are profitable.
Also, fix a bug in LiveInterval::MergeInClobberRanges. It didn't handle
the case that a clobber range extended both before and beyond an
existing live range. In that case, multiple live ranges need to be
added. This was exposed by the new subreg coalescing code.
Remove 2008-05-06-SpillerBug.ll. It was bugpoint-reduced, and the
spiller behavior it was looking for no longer occurrs with the new
instruction selection.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@68576 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-08 00:15:30 +00:00
|
|
|
; RUN: llvm-as < %s | llc -march=x86 | grep mov | count 5
|
Teach 2addr pass to be do more commuting. If both uses of a two-address instruction are killed, but the first operand has a use before and after the def, commute if the second operand does not suffer from the same issue.
%reg1028<def> = EXTRACT_SUBREG %reg1027<kill>, 1
%reg1029<def> = MOV8rr %reg1028
%reg1029<def> = SHR8ri %reg1029, 7, %EFLAGS<imp-def,dead>
insert => %reg1030<def> = MOV8rr %reg1028
%reg1030<def> = ADD8rr %reg1028<kill>, %reg1029<kill>, %EFLAGS<imp-def,dead>
In this case, it might not be possible to coalesce the second MOV8rr
instruction if the first one is coalesced. So it would be profitable to
commute it:
%reg1028<def> = EXTRACT_SUBREG %reg1027<kill>, 1
%reg1029<def> = MOV8rr %reg1028
%reg1029<def> = SHR8ri %reg1029, 7, %EFLAGS<imp-def,dead>
insert => %reg1030<def> = MOV8rr %reg1029
%reg1030<def> = ADD8rr %reg1029<kill>, %reg1028<kill>, %EFLAGS<imp-def,dead>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@62954 91177308-0d34-0410-b5e6-96231b3b80d8
2009-01-25 03:53:59 +00:00
|
|
|
; rdar://6523745
|
|
|
|
|
|
|
|
|
|
@"\01LC" = internal constant [4 x i8] c"%d\0A\00" ; <[4 x i8]*> [#uses=1]
|
|
|
|
|
|
|
|
|
|
define i32 @main() nounwind {
|
|
|
|
|
bb1.thread:
|
|
|
|
|
br label %bb1
|
|
|
|
|
|
|
|
|
|
bb1: ; preds = %bb1, %bb1.thread
|
|
|
|
|
%i.0.reg2mem.0 = phi i32 [ 0, %bb1.thread ], [ %indvar.next, %bb1 ] ; <i32> [#uses=2]
|
|
|
|
|
%0 = trunc i32 %i.0.reg2mem.0 to i8 ; <i8> [#uses=1]
|
|
|
|
|
%1 = sdiv i8 %0, 2 ; <i8> [#uses=1]
|
|
|
|
|
%2 = sext i8 %1 to i32 ; <i32> [#uses=1]
|
|
|
|
|
%3 = tail call i32 (i8*, ...)* @printf(i8* getelementptr ([4 x i8]* @"\01LC", i32 0, i32 0), i32 %2) nounwind ; <i32> [#uses=0]
|
|
|
|
|
%indvar.next = add i32 %i.0.reg2mem.0, 1 ; <i32> [#uses=2]
|
|
|
|
|
%exitcond = icmp eq i32 %indvar.next, 258 ; <i1> [#uses=1]
|
|
|
|
|
br i1 %exitcond, label %bb2, label %bb1
|
|
|
|
|
|
|
|
|
|
bb2: ; preds = %bb1
|
|
|
|
|
ret i32 0
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
declare i32 @printf(i8*, ...) nounwind
|
