AW: optimizer discards sign information

[Xi Ruoyao]

On Wed, 2024-04-10 at 17:57 +0800, LIU Hao via Gcc-help wrote:
> 在 2024-04-10 17:49, stefan@franke.ms 写道:
> > But I keep considering this as a bug. And clang behaves correctly!
> 
> Yes there have been many reports [1]. It's a missed optimization.

Note that for this specific case:

   typedef unsigned long long int u64;
   typedef unsigned int u32;
   typedef unsigned short u16;
   
   u64 foo(u16 *a, u16 *b) {
       u32 x = *a * *b;
       u64 r = x;
       return r >> 31;
   }
   
   gcc yields
   
   foo:
           xor     eax, eax
           ret
   
   clang yields
   
   foo:                                    # @foo
           movzx   ecx, word ptr [rdi]
           movzx   eax, word ptr [rsi]
           imul    eax, ecx
           shr     eax, 31
           ret
   
It's actually a missed-optimization of **clang**.  Optimizing this
function to always return 0 **is** correct.

But for the general case:

u64 foo(u16 a, u16 b) {
    u32 x = a * b;
    u64 r = x;
    return r;
}

there is a missed-optimization of GCC (redundant sign extension).

> You may work around it by using 32-bit parameters, or casting either
> operand to u32; casting the result will not help.

Indeed.


-- 
Xi Ruoyao <xry111@xry111.site>
School of Aerospace Science and Technology, Xidian University