Re: [Freedreno] [PATCH] drm/msm/a6xx: Correct the highestbank configuration

[Rob Clark]

On Fri, Jan 31, 2020 at 12:00 AM Akhil P Oommen <akhilpo@xxxxxxxxxxxxxx> wrote:
>
> On 1/24/2020 11:56 PM, Jordan Crouse wrote:
> > On Fri, Jan 24, 2020 at 05:50:11PM +0530, Akhil P Oommen wrote:
> >> Highest bank bit configuration is different for a618 gpu. Update
> >> it with the correct configuration which is the reset value incidentally.
> >>
> >> Signed-off-by: Akhil P Oommen <akhilpo@xxxxxxxxxxxxxx>
> >> Signed-off-by: Sharat Masetty <smasetty@xxxxxxxxxxxxxx>
> >> ---
> >>   drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 10 ++++++----
> >>   1 file changed, 6 insertions(+), 4 deletions(-)
> >>
> >> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> >> index daf0780..536d196 100644
> >> --- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> >> +++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> >> @@ -470,10 +470,12 @@ static int a6xx_hw_init(struct msm_gpu *gpu)
> >>      /* Select CP0 to always count cycles */
> >>      gpu_write(gpu, REG_A6XX_CP_PERFCTR_CP_SEL_0, PERF_CP_ALWAYS_COUNT);
> >>
> >> -    gpu_write(gpu, REG_A6XX_RB_NC_MODE_CNTL, 2 << 1);
> >> -    gpu_write(gpu, REG_A6XX_TPL1_NC_MODE_CNTL, 2 << 1);
> >> -    gpu_write(gpu, REG_A6XX_SP_NC_MODE_CNTL, 2 << 1);
> >> -    gpu_write(gpu, REG_A6XX_UCHE_MODE_CNTL, 2 << 21);
> >> +    if (adreno_is_a630(adreno_gpu)) {
> >> +            gpu_write(gpu, REG_A6XX_RB_NC_MODE_CNTL, 2 << 1);
> >> +            gpu_write(gpu, REG_A6XX_TPL1_NC_MODE_CNTL, 2 << 1);
> >> +            gpu_write(gpu, REG_A6XX_SP_NC_MODE_CNTL, 2 << 1);
> >> +            gpu_write(gpu, REG_A6XX_UCHE_MODE_CNTL, 2 << 21);
> >> +    }
> > it shouldn't come as a surprise that everything in the a6xx family is going to
> > have a highest bank bit setting. Even though the a618 uses the reset value, I
> > think it would be less confusing to future folks if we explicitly program it:
> >
> > if (adreno_is_a630(adreno_dev))
> >    hbb = 2;
> > else
> >    hbb = 0;
>
> I think it would be better if we keep this in the adreno_info. Yes, this
> would waste a tiny bit of space for other gpu
> entries in the gpulist. It is also possible to move this to a separate
> struct and keep a pointer to it in the adreno_info.
> But that is something we should try when there are more a6xx specific
> configurations in future.
>
> I have a new patch, but testing it is taking longer that I expected. I
> will share it as soon as possible.
>

I'm going to pull this in as-is for msm-fixes.  Please rebase the
change that you haven't posted yet on top of this patch, and send it
as a cleanup for the next cycle.  Thanks

BR,
-R