http://llvm.org/bugs/show_bug.cgi?id=9517
Summary: Input constraints for inline assembler are not used
for CSE
Product: new-bugs
Version: trunk
Platform: PC
OS/Version: Linux
Status: NEW
Severity: normal
Priority: P
Component: new bugs
AssignedTo: [email protected]
ReportedBy: [email protected]
CC: [email protected]
Consider the following program:
static inline void
outb(unsigned short port, unsigned char data)
{
__asm volatile("outb %%al,%w1" : : "a" (data), "d" (port));
}
extern unsigned short base;
void test(void)
{
outb(base + 16, 0x0);
outb(base + 16, 0x1);
outb(base + 16, 0x2);
outb(base + 16, 0x3);
outb(base + 16, 0x4);
outb(base + 16, 0x5);
outb(base + 16, 0x6);
outb(base + 16, 0x7);
outb(base + 16, 0x8);
}
This should be compiled into something like:
movzwl base, %edx
add 0x10, %edx
xorl %eax, %eax
outb %al,%dx
movb $1, %al
outb %al,%dx
movb $2, %al
outb %al,%dx
...
or even better in terms of code size:
movzwl base, %edx
add 0x10, %edx
xorl %eax, %eax
outb %al,%dx
incl %eax
outb %al,%dx
incl %eax
outb %al,%dx
...
But currently it gets compiled to:
movzwl base, %edx
add 0x10, %edx
xorl %eax, %eax
outb %al,%dx
movzwl base, %edx
add 0x10, %edx
movb $1, %al
outb %al,%dx
movzwl base, %edx
add 0x10, %edx
movb $2, %al
outb %al,%dx
...
Input constraints are not clobbered and therefore, they don't have to be
reloaded. The reload overhead in this case is significant.
--
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.
_______________________________________________
LLVMbugs mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/llvmbugs
