On Fri, 5 Apr 2002, José Fonseca wrote:
> On 2002.04.04 18:47 Tony Rogvall wrote:
> > "José Fonseca" wrote:
> >
> > > On 2002.04.04 11:46 Tony Rogvall wrote:
> > > > ...
> > > >
> > > > Found it, I assumed it was a printk and removed DRM_INFO voila it
> > started
> > > > without
> > > > a crash. In mach64_dma there is code for printing AGP_BASE !! Just
> > put
> > > > some
> > > > __REALLY_HAVE_AGP around it.
> > > >
> > >
> > > __REALLY_HAVE_AGP is a constant macro. It must be some variable that
> > > indicates wether agpgart is or not loaded.
> >
>
> > well nearly constant it is defined by:
> >
> > #define __REALLY_HAVE_AGP (__HAVE_AGP && (defined(CONFIG_AGP) || \
> >
> > defined(CONFIG_AGP_MODULE)))
> >
> > And for my system (running redhat) agpgart is compiled into the kernel,
> > but
> > that
> > does not say anything about if I have AGP or not!
> >
>
> I just commited a patch to get over this in the meanwhile:
>
> --- mach64_dma.c 26 Feb 2002 22:02:37 -0000 1.1.6.1
> +++ mach64_dma.c 5 Apr 2002 00:59:47 -0000
> @@ -93,8 +93,13 @@
> static void dump_engine_info( drm_mach64_private_t *dev_priv )
> {
> DRM_INFO( "\n" );
> - DRM_INFO( " AGP_BASE = 0x%08x\n", MACH64_READ(
> MACH64_AGP_BASE ) );
> - DRM_INFO( " AGP_CNTL = 0x%08x\n", MACH64_READ(
> MACH64_AGP_CNTL ) );
> +#if 0 /* disable while is_pci is not implemented */
> + if ( !dev_priv->is_pci)
> + {
> + DRM_INFO( " AGP_BASE = 0x%08x\n", MACH64_READ(
> MACH64_AGP_BASE ) );
> + DRM_INFO( " AGP_CNTL = 0x%08x\n", MACH64_READ(
> MACH64_AGP_CNTL ) );
> + }
> +#endif
> DRM_INFO( " ALPHA_TST_CNTL = 0x%08x\n", MACH64_READ(
> MACH64_ALPHA_TST_CNTL ) );
> DRM_INFO( "\n" );
> DRM_INFO( " BM_COMMAND = 0x%08x\n", MACH64_READ(
> MACH64_BM_COMMAND ) );
We could really remove the call to dump_engine_info altogether in the
mach64-0-0-3 branch, it's just debugging output leftover from the earlier
dma testing. The busmaster test is also not necessary in this branch.
BTW, I did some testing on ReadPixels for very simple cases, and it works
fine, as does the readPixSanity glean test. I haven't determined the
cause of the readback errors yet, but I did find something strange. The
maskedClear glean test fails, always on the second test (everything but
green masked), for the GL_FRONT buffer. However, if I set a breakpoint
before the glClear, step that call, then continue, and repeat for each
iteration, the test passes! I haven't figured this one out yet. At first
I thought maybe the ReadPixels was called while the draw engine was still
busy, and stepping allowed drawing to complete, but ReadPixels appears to
always flush vertices and wait for draw engine idle in the HW_LOCK macro.
So I'm still looking...
--
Leif Delgass
http://www.retinalburn.net
_______________________________________________
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel
