https://bugs.kde.org/show_bug.cgi?id=523626

Mark Wielaard <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REPORTED                    |CONFIRMED
                 CC|                            |[email protected]
     Ever confirmed|0                           |1

--- Comment #9 from Mark Wielaard <[email protected]> ---
(In reply to mcermak from comment #8)
> So the problem appears to be that movdqa expects aligned double quadword.
> That is 128-bit (16-byte) value whose address is a multiple of 16 bytes. 
> Which in this case isn't true ^^ and that seems to cause the segfault.

Nice analysis. I have been staring at the code. But I cannot see whether we are
passing around non-aligned or wrongly typed pointers or if this is a wrong
optimization (apparently the compiler does think it is producing correctly
aligned data that can be efficiently moved around with movdqa). But I am not
sure what it is.

I would push the workaround with __attribute__((optnone)) for sad_8x4 and then
try to find a clang/llvm hacker to see if they understand why the optimization
uses movdqa in this case.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to