Cordic

[Tugdual]

Does anybdoy knows if the Prime still relies on Cordic or uses a math coprocessor (which itself may as well rely on cordic...)?

[Tim Wessman]

The numeric math library is essentially the same BCD math library from the 48 series. So yes.

The CAS however uses binary and I don't know exactly what Bernard does there.

[debrouxl]

The ancient ARM9 processor series doesn't have floating-point acceleration like, say, the Cortex-A series does.

[parisse]

Floating point arithmetic in the CAS is using the double format representation (truncated to 48 bits mantissa) and the standard libm/libstdc++ functions (except for a few special functions).

[Tugdual]

Thank you all for quick and detailed answers!

[cyrille de brébisson]

Hello

(01-28-2014 05:39 PM)debrouxl Wrote:  The ancient ARM9 processor series doesn't have floating-point acceleration like, say, the Cortex-A series does.

Incorrect, some ARM9 do, and some Cortex-A do not...
but in our case, we do not...

However, the ARM FPU does not integrate sin/cos anyway... so the conversation is moot :-) ARM FPU only have +, -, *, / and sqrt...

Cyrille