On Mon, Aug 10, 2015 at 03:17:26PM +0300, Martin Peres wrote:
> >+static int
> >+intel_detect_pipelined_oacontrol(struct intel_screen *screen,
> >+ struct detect_pipelined_register *detect)
> > {
> > if (screen->devinfo->gen < 6 || screen->devinfo->gen >= 8)
> >- return false;
> >+ return 0;
> > /* Set "Select Context ID" to a particular address (which is likely not a
> > * context), but leave all counting disabled. This should be harmless.
> > */
> >- return intel_detect_pipelined_register(screen,
> >- OACONTROL,
> >- 0x31337000,
> >- true);
> >+ detect->reg = OACONTROL;
> >+ detect->expected_value = 0x31337000;
> >+ detect->result = HW_HAS_PIPELINED_OACONTROL;
> >+ detect->reset = true;
> >+ return 1;
> >+}
> >+
> >+static void
> >+intel_detect_pipelined_register_access(struct intel_screen *screen)
> >+{
> >+ struct detect_pipelined_register regs[2], *r =regs;
> >+
> >+ /* Combine the multiple register access validation into a single
> >+ * round trip through the kernel + GPU.
> >+ */
> >+ r += intel_detect_pipelined_so(screen, r);
> >+ r += intel_detect_pipelined_oacontrol(screen, r);
>
> Not a fan of this construct. How about changing the return types of
> the detect functions to int?.
Do you mean
if (intel_detect_pipelined_so(screen, r))
r++;
or
int index = 0;
index += intel_detect_pipelined_so(screen, &r[index]);
?
-Chris
--
Chris Wilson, Intel Open Source Technology Centre