mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-22 18:54:02 +01:00
48dc5bb962
Don't spend the entire iteration space in the scalar loop prologue if computing the trip count overflows. This change also gets rid of the backedge check in the prologue loop and the extra check for overflowing trip-count. Differential Revision: http://reviews.llvm.org/D7715 llvm-svn: 229731
38 lines
1.2 KiB
LLVM
38 lines
1.2 KiB
LLVM
; RUN: opt < %s -S -unroll-runtime -unroll-count=2 -loop-unroll | FileCheck %s
|
|
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
|
|
|
|
; This test case documents how runtime loop unrolling handles the case
|
|
; when the backedge-count is -1.
|
|
|
|
; If %N, the backedge-taken count, is -1 then %0 unsigned-overflows
|
|
; and is 0. %xtraiter too is 0, signifying that the total trip-count
|
|
; is divisible by 2. The prologue then branches to the unrolled loop
|
|
; and executes the 2^32 iterations there, in groups of 2.
|
|
|
|
|
|
; CHECK: entry:
|
|
; CHECK-NEXT: %0 = add i32 %N, 1
|
|
; CHECK-NEXT: %xtraiter = and i32 %0, 1
|
|
; CHECK-NEXT: %lcmp.mod = icmp ne i32 %xtraiter, 0
|
|
; CHECK-NEXT: br i1 %lcmp.mod, label %while.body.prol, label %entry.split
|
|
|
|
; CHECK: while.body.prol:
|
|
; CHECK: br label %entry.split
|
|
|
|
; CHECK: entry.split:
|
|
|
|
; Function Attrs: nounwind readnone ssp uwtable
|
|
define i32 @foo(i32 %N) {
|
|
entry:
|
|
br label %while.body
|
|
|
|
while.body: ; preds = %while.body, %entry
|
|
%i = phi i32 [ 0, %entry ], [ %inc, %while.body ]
|
|
%cmp = icmp eq i32 %i, %N
|
|
%inc = add i32 %i, 1
|
|
br i1 %cmp, label %while.end, label %while.body
|
|
|
|
while.end: ; preds = %while.body
|
|
ret i32 %i
|
|
}
|