https://bugs.llvm.org/show_bug.cgi?id=43835
Bug ID: 43835
Summary: Differences between `base+width*i` and `base+=width`
Product: libraries
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: enhancement
Priority: P
Component: Loop Optimizer
Assignee: [email protected]
Reporter: [email protected]
CC: [email protected]
https://godbolt.org/z/TR-s8A
#include <array>
#include <algorithm>
void sink(int* row);
void v0(int* data, int len, int width) {
for(int i = 0; i < len; i++) {
int *row = data + width*i;
sink(row);
}
}
void v1(int* row, int len, int width) {
for(int i = 0; i < len; i++, row+=width) {
sink(row);
}
}
In first case we get
9: ; preds = %9, %5
%10 = phi i64 [ 0, %5 ], [ %13, %9 ]
%11 = mul nsw i64 %10, %6
%12 = getelementptr inbounds i32, i32* %0, i64 %11
tail call void @_Z4sinkPi(i32* %12)
%13 = add nuw nsw i64 %10, 1
but in second:
%9 = phi i32 [ 0, %5 ], [ %11, %8 ]
%10 = phi i32* [ %0, %5 ], [ %12, %8 ]
tail call void @_Z4sinkPi(i32* %10)
%11 = add nuw nsw i32 %9, 1
%12 = getelementptr inbounds i32, i32* %10, i64 %6
I'd guess the second variant is better?
The difference is interesting.
--
You are receiving this mail because:
You are on the CC list for the bug._______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs