mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-09-02 11:30:10 +00:00
This commit teaches the X86 backend to create the same X86 instructions when it lowers an sadd/ssub with overflow intrinsic and a conditional branch that uses that overflow result. This allows SelectionDAG to recognize and remove one of the redundant operations. This fixes <rdar://problem/15874016> and <rdar://problem/15661073>. Reviewed by Nadav git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@199976 91177308-0d34-0410-b5e6-96231b3b80d8
35 lines
805 B
LLVM
35 lines
805 B
LLVM
; RUN: llc < %s -mtriple=x86_64-apple-darwin | FileCheck %s
|
|
|
|
define void @redundant_add(i64 %n) {
|
|
; Check that we don't create two additions for the sadd.with.overflow.
|
|
; CHECK-LABEL: redundant_add
|
|
; CHECK-NOT: leaq
|
|
; CHECK-NOT: addq
|
|
; CHECK: incq
|
|
; CHECK-NEXT: jno
|
|
entry:
|
|
br label %exit_check
|
|
|
|
exit_check:
|
|
%i = phi i64 [ 0, %entry ], [ %i.next, %loop ]
|
|
%c = icmp slt i64 %i, %n
|
|
br i1 %c, label %loop, label %exit
|
|
|
|
loop:
|
|
%i.o = tail call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 %i, i64 1)
|
|
%i.next = extractvalue { i64, i1 } %i.o, 0
|
|
%o = extractvalue { i64, i1 } %i.o, 1
|
|
br i1 %o, label %overflow, label %exit_check
|
|
|
|
exit:
|
|
ret void
|
|
|
|
overflow:
|
|
tail call void @llvm.trap()
|
|
unreachable
|
|
}
|
|
|
|
declare { i64, i1 } @llvm.sadd.with.overflow.i64(i64, i64)
|
|
declare void @llvm.trap()
|
|
|