I have a question about floating point performance relative with fpsr register.
When i initialize hardware, there is floating point exception(inexactly floating-point exception).
I did not set fpcr.IXE=0, so fpsr.IXC is set. not occur exception.
After that, my floating point instruction(fmadd) is faster than before.
Why fmadd instruction is faster(fpsr.IXC=1) than before(fpsr.IXC=0)?
Can you help me?
Just a guess: Checking needs time.
From the manual (ARM DDI 0406C.d, A2.69):FPSCR.IXC Inexact. The bit is set to 1 if the result of an operation is not equivalent to the value that would be produced if the operation were performed with unbounded precision and exponent range.
Is there more information about why FPSCR.IXC=1 is faster than FPSCR.IXC=0 on floating point calculation?
Additionally, floating point calculation is not occur inexact exception eventhough FPSCR.IXC=0.
View all questions in Cortex-A / A-Profile forum