topic Re: sqrt function with hard FPU instrustion in Kinetis Microcontrollers

sqrt function with hard FPU instrustion

MK60FX512 — Fri, 16 Jan 2015 06:49:55 GMT

Dear Sir

K60 microcontroller has inbuit FPU, So it support floating point instuction by hard ware itself rather than softcode

I am using KDS for K60 and i using sqrtf() function in .list file it shows software root function. as below

****	b_out = sqrtf(a);
82	.loc 1 82 0 discriminator 1
83 0036 40F20003	movw	r3, #:lower16:a
84 003a C0F20003	movt	r3, #:upper16:a
85 003e D3ED007A	flds	s15, [r3]
86 0042 B0EE670A	fcpys	s0, s15
87 0046 FFF7FEFF	bl	sqrtf
88 004a F0EE407A	fcpys	s15, s0
89 004e 40F20003	movw	r3, #:lower16:b_out
90 0052 C0F20003	movt	r3, #:upper16:b_out
91 0056 C3ED007A	fsts	s15, [r3]
83:../Sources/main.c ****	timecount = timecount+1.2;

1.can somebody say how to invoke floting instuction ?

2. is possible to do square root by excecuting sqrt hardware instruction? if yes then how.

Re: sqrt function with hard FPU instrustion

BlackNight — Fri, 16 Jan 2015 08:31:10 GMT

In the compiler settings, make sure you have these options set:

Erich

Re: Re: sqrt function with hard FPU instrustion

MK60FX512 — Fri, 16 Jan 2015 11:16:04 GMT

There are option of setting FPU type whether hard or soft in KDS but if i set "Toolchaing default" then it is taking as software and generating branch and software code without any FPU instrucation and not giving any error

but I set as

mfloat-abi=hard

mfpu=fpv4-sp-d16

from project setting then while running it is giving Hard fault and haging at

PE_ISR(Cpu_Interrupt)

{

/* This code can be changed using the CPU component property "Build Options / Unhandled int code" */

PE_DEBUGHALT();

}

but same thing i tried in CW10/6 it is working and it generate FSQRTS instruction.

Please help to sort out issue of KDS.

Thanks and regards

Bipin

Re: sqrt function with hard FPU instrustion

BlackNight — Sat, 17 Jan 2015 11:33:54 GMT

It looks like you have not enabled the floating point unit?

Add this code to your application

void vPortEnableVFP(void) {

/* The FPU enable bits are in the CPACR. */

__asm volatile (

" ldr.w r0, =0xE000ED88 \n" /* CAPCR, 0xE000ED88 */

" ldr r1, [r0] \n" /* read CAPR */

" orr r1, r1, #(0xf<<20) \n" /* enable CP10 and CP11 coprocessors */

" str r1, [r0] \n" /* store to new value back */

: /* no output */

: /* no input */

: "r0","r1" /* clobber */

);

}

and call it in main() as the first thing.

Erich

Re: sqrt function with hard FPU instrustion

MK60FX512 — Mon, 16 Mar 2015 05:45:11 GMT

As i was using KDS but because all this problem again i started to use CW10.6 with GCC compiler instead of freescale compiler

In this case float multiplication is generting Instruction like only with "f" only with GCC compiler.

.loc 1 134 0

flds s15, [r5, #0]

fmuls s14, s15, s17

fcpys s0, s16

fmacs s0, s15, s14

movw r2, #:lower16:.LANCHOR1

movt r2, #:upper16:.LANCHOR1

fsts s0, [r2, #0]

but in freescale compiler , it was generating with starting of "V" like vmov.F32

so the above instruction instruction with starting of "f" like fcpys is correct ????

but still it is generating instruction both fsqrt and also branch of sqrt like follow

138:../Sources/main.c ****			dummy_float = sqrtf(dummy_float);
181		.loc 1 138 0
182 0078 F1EEC00A		fsqrts	s1, s0
183 007c F4EE600A		fcmps	s1, s1
184 0080 F1EE10FA		fmstat
185 0084 03D0		beq	.L4
186 0086 FFF7FEFF		bl	sqrtf
187	.LVL14:
188 008a F0EE400A		fcpys	s1, s0
189	.L4:
190 008e 40F20003		movw	r3, #:lower16:.LANCHOR1
191 0092 C0F20003		movt	r3, #:upper16:.LANCHOR1
192 0096 C3ED000A		fsts	s1, [r3, #0]

so in above instruction sqrt is executing with FPU or it is software execution.

Please Help