[U-Boot] [PATCH 4/9] openrisc: Add library functions

[Marek Vasut]

> On Mon, Nov 21, 2011 at 11:52:59PM +0100, Marek Vasut wrote:
> > >  create mode 100644 arch/openrisc/lib/timer.c
> > 
> > Timer support isn't a library function but a CPU function, so move it to
> > 3/9.
> 
> I never quite worked out where the timer functions belongs,
> some have them in interupts.c and some in their own file.
> Some have them cpu/ and some in lib/
> 
> > > +	asm("l.nop 0x1"); /* Kill any simulation */
> > 
> > Simulation? Oh, it's an FPGA based CPU or what?
> 
> Well, yes, FPGAs are probably the most common case, but also ASIC
> implementations exists.
> The extra argument to the nop instruction is ignored by hardware,
> but have special meanings when ran in simulation.

Hmm ... I'm not quite sure this is right.

> 
> > > +int timer_init(void)
> > > +{
> > > +	/* Install timer exception handler */
> > > +	exception_install_handler(EXC_TIMER, timer_isr);
> > > +
> > > +	/* Set up the timer for the first expiration. */
> > > +	timestamp = 0;
> > > +
> > > +	mtspr(SPR_TTMR, SPR_TTMR_IE | SPR_TTMR_RT |
> > > +		(TIMER_COUNTER_CYCLES & SPR_TTMR_TP));
> > > +
> > > +	/* Enable tick timer exception in supervisor register */
> > > +	mtspr(SPR_SR, mfspr(SPR_SR) | SPR_SR_TEE);
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +void reset_timer(void)
> > > +{
> > > +	timestamp = 0;
> > > +
> > > +	mtspr(SPR_TTMR, SPR_TTMR_IE | SPR_TTMR_RT |
> > > +		(TIMER_COUNTER_CYCLES & SPR_TTMR_TP));
> > > +}
> > > +
> > > +/*
> > > + * The timer value in ms is calculated by taking the
> > > + * value accumulated by full timer revolutions plus the value
> > > + * accumulated in this period
> > > + */
> > > +ulong get_timer(ulong base)
> > > +{
> > > +	return timestamp + mfspr(SPR_TTCR)/TIMER_CYCLES_MS - base;
> > > +}
> > > +
> > > +void set_timer(ulong t)
> > > +{
> > > +	reset_timer();
> > > +	timestamp = t;
> > > +}
> > > +
> > > +void __udelay(ulong usec)
> > > +{
> > > +	ulong elapsed = 0;
> > > +	ulong tick;
> > > +	ulong last_tick;
> > > +
> > > +	last_tick = mfspr(SPR_TTCR);
> > > +	while ((elapsed / TIMER_CYCLES_US) < usec) {
> > > +		tick = mfspr(SPR_TTCR);
> > > +		if (tick >= last_tick)
> > > +			elapsed += (tick - last_tick);
> > > +		else
> > > +			elapsed += TIMER_COUNTER_CYCLES - (last_tick - tick);
> > > +		last_tick = tick;
> > > +	}
> > > +}
> > 
> > I'm not sure if this conforms with current timer api, can you cross-check
> > with arch/arm/arm926ejs/mx28/timer.c ? That's the latest addition and
> > should conform.
> 
> In my opinion it seems to do the same thing as that, what exactly did you
> find non-conforming?

I was just curious if it's ok. If it is, so be it.
> 
> Stefan