Preston Faiks commented: 
https://gitlab.rtems.org/rtems/rtos/rtems/-/issues/5214#note_119882


- FPSR upon entering a function is effectively random
  * NO, it should not be random
- may have any value on entry to a subroutine
  * YES, due to the cumulative effects of the FP instructions prior to the 
function call

The issue here is that the AArch64 PCS only applies to a single thread of 
execution. The fact that RTEMS is doing a context switch and allowing other 
threads of execution to affect the FPSR is outside the scope of that document.

What you are effectively saying is that the RTEMS threads are not truly 
separate threads of execution, but in fact subroutines called by the 
Context_Switch() function. The only reason it needs to be saved is because its 
invoking another thread, not because its a callee-saved register.

-- 
View it on GitLab: 
https://gitlab.rtems.org/rtems/rtos/rtems/-/issues/5214#note_119882
You're receiving this email because of your account on gitlab.rtems.org.


_______________________________________________
bugs mailing list
[email protected]
http://lists.rtems.org/mailman/listinfo/bugs

Reply via email to