Mailinglist Archive: radeonhd (408 mails)

[Alex Deucher <alexdeucher@xxxxxxxxx>]

2009/6/5 Rafał Miłecki <zajec5@xxxxxxxxx>:
> W dniu 5 czerwca 2009 20:11 użytkownik Alex Deucher
> <alexdeucher@xxxxxxxxx> napisał:
> > 2009/6/5 Rafał Miłecki <zajec5@xxxxxxxxx>:
> > > There is one more command I don't understand:
> > >
> > > SET_REG_BLOCK  0000
> > > SET_REG_BLOCK  0001
> > >
> > > what does it do?
> >
> > There are lots of sets of registers that have multiple instances in
> > register space (e.g., there are two identical sets of CRTC registers,
> > one for each CRTC, DACs and DIG PHYs on newer chips, etc.).  That
> > command is used to select the offset added to the registers in order
> > to access different register blocks.  Take a look at SetCRTC_Timing:
> >
> >  0009: 3da50400          COMP   param[04]  [.X..]  <-  00
> >  000d: 441300            JUMP_Equal  0013
> >  0010: 3a0002            SET_REG_BLOCK  0200
> >
> > if you've selected crtc 1, add an offset of 0x200 dwords (0x800 bytes)
> > to the registers, if it's crtc 0, do not.
> >
> > Alex
>
> Thanks for explaination with nice example. Nice way to handle 2
> pll/crtc/... in one code :)
>
> Still one more (hopefully last?) question. How do work (and what is
> difference) TEST and COMP?

COMP looks to see if src and dst are equal, TEST tests using logical
AND.  You can follow the logic of any of these commands by looking at
AtomBios/CD_Operations.c.  See ProcessCompare() and ProcessTest().

> param[00] == 0x12345678
> X: 4be50004          TEST   param[00]  [X...]  <-  04
> Y: 441700            JUMP_Equal  0017
> 12 != 04, so jump won't happen, right?

This would be equate to:
(0x12 & 0x4) ? NotEqual : Equal
So the jump would happen.

> param[00] == 0x12345678
> 0017: 3d0d000000        COMP   param[00]  [..XX]  <-  0000
> 001c: 448c00            JUMP_Equal  008c
> 5678 != 0000, so jump won't happen, right?

correct.

Alex
--
To unsubscribe, e-mail: radeonhd+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: radeonhd+help@xxxxxxxxxxxx