About using AMMCLIB

578294688 — Sun, 27 Oct 2019 12:14:42 GMT

Hello,engineers!

I recently had a problem with AMMCLIB math library. The version of AMMCLIB math library is MPC577xK_AMMCLIB_v1.1.9 - s32ds_ppc, MCU is S32R274.

I tested the sine function, and I found that GFLIB_Sin_FLT is OK, but GFLIB_Sin_F32 and GFLIB_Sin_F16 are not OK when I use MPC577xK_AMMCLIB.a in e200z7.If I want to use the GFLIB_Sin_F32 function, what do I need to do?

Another, when I use MPC577xK_AMMCLIB.a in e200z4, I found that the three functions are all OK, but GFLIB_Sin_F32 and GFLIB_Sin_F16 running time is more than GFLIB_Sin_FLT, I can't understand that because I think integral types should be less computation-intensive than floating point types.

What is the answer to the above question?

Or are there other ways to improve software performance?

I am looking forward to your reply!

Best wishes

Re: About using AMMCLIB

stanish — Tue, 12 Nov 2019 16:39:25 GMT

Hello,

I assume your goal is to improve speed performance.

as a general suggestion in order I'd recommend to enable

acceleration features such as branch target buffer, Cache, memory WaitStates, HW floating point, SPT...
enable compiler optimizations, use intrinsics for critical code, place critical code into RAM
select better performance compiler (such as GHS)

Regarding AMMCLIB issues I'd first recommend you to update the library which is v1.1.18

Automotive Math and Motor Control Library Set | NXP

and check the performance.

Hope it helps.

Stan

S32 Design StudioのトピックAbout using AMMCLIB

About using AMMCLIB

Re: About using AMMCLIB