| Issue |
164749
|
| Summary |
ARM64EC Performance Issue
|
| Labels |
new issue
|
| Assignees |
|
| Reporter |
XuSai1998
|
When building the following code for ARM64EC, I noticed that the call to Point in method2 results in the insertion of __os_arm64x_check_icall, which negatively impacts performance compared to method1.
Is there an attribute or compiler directive I can use to disable this check, If I can guarantee that pfun1 and pfun2 will never point to x64 code?
<img width="872" height="654" alt="Image" src="" />
`
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <windows.h>
int fun1(int a, int b) {
return a + b;
}
int fun2(int a, int b) {
return a * b;
}
typedef int (*AddFunc)(int, int);
int main() {
HANDLE hProcess = GetCurrentProcess();
DWORD_PTR mask = 1 << 10;
SetProcessAffinityMask(hProcess, mask);
AddFunc pfun1 = fun1; //Point the function pointer to fun1
AddFunc pfun2 = fun2; //Point the function pointer to fun1
//# Method 1
int sum = 0;
Sleep(1000);
for (int i = 0; i < 10000; i++) {
for(int j = 0; j < 200000; j++) {
sum = sum + fun1(i, j); //Function call
sum = sum + fun2(i, j);
}
}
printf("sum=%d\n", sum);
//# Method 2
sum = 0;
Sleep(1000);
for (int i = 0; i < 10000; i++) {
for (int j = 0; j < 200000; j++) {
sum = sum + pfun1(i, j); //Pointer call
sum = sum + pfun2(i, j);
}
}
printf("sum=%d\n", sum);
}
`
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
