[Mesa-dev] [PATCH 1/3] nvc0: fix compute state initialization on GK110+

[Samuel Pitoiset]

On 02/14/2016 12:41 AM, Ilia Mirkin wrote:
> On Sat, Feb 13, 2016 at 6:31 PM, Samuel Pitoiset
> <samuel.pitoiset at gmail.com> wrote:
>> For some unknown reasons, the command 0x518, which is FIRMWARE[0x6]
>
> Please make a note that the blob calls it, along with the args passed
> in (usually via scratch regs iirc). It's probably one of the context
> register modifiers. Our firmware doesn't implement any of those
> commands.
>
>> from rnndb, totally hangs the GPU, and a reboot is needed. With
>> that removed, I can succesfully launch compute shaders and so the
>> compute support seems to work as expected.
>>
>> Tested on GK208 and GM107.
>>
>> Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
>> ---
>>   src/gallium/drivers/nouveau/nvc0/nve4_compute.c | 9 +++------
>>   1 file changed, 3 insertions(+), 6 deletions(-)
>>
>> diff --git a/src/gallium/drivers/nouveau/nvc0/nve4_compute.c b/src/gallium/drivers/nouveau/nvc0/nve4_compute.c
>> index aaf4282..31b571d 100644
>> --- a/src/gallium/drivers/nouveau/nvc0/nve4_compute.c
>> +++ b/src/gallium/drivers/nouveau/nvc0/nve4_compute.c
>> @@ -39,7 +39,7 @@ nve4_screen_compute_setup(struct nvc0_screen *screen,
>>   {
>>      struct nouveau_device *dev = screen->base.device;
>>      struct nouveau_object *chan = screen->base.channel;
>> -   unsigned i;
>> +   int i;
>>      int ret;
>>      uint32_t obj_class;
>>
>> @@ -115,13 +115,10 @@ nve4_screen_compute_setup(struct nvc0_screen *screen,
>>      PUSH_DATA (push, NVC0_TSC_MAX_ENTRIES - 1);
>>
>>      if (obj_class >= NVF0_COMPUTE_CLASS) {
>> -      BEGIN_NVC0(push, SUBC_COMPUTE(0x0248), 1);
>> -      PUSH_DATA (push, 0x100);
>> -      BEGIN_NIC0(push, SUBC_COMPUTE(0x0248), 63);
>> -      for (i = 63; i >= 1; --i)
>> +      BEGIN_NIC0(push, SUBC_COMPUTE(0x0248), 64);
>
> You verified that this is also needed on GM107, yes?

Yes, it is.
>
> With the comment re firmware added in, this patch is
>
> Reviewed-by: Ilia Mirkin <imirkin at alum.mit.edu>

Okay, I'll add a comment, thanks.

>
>> +      for (i = 63; i >= 0; i--)
>>            PUSH_DATA(push, 0x38000 | i);
>>         IMMED_NVC0(push, SUBC_COMPUTE(NV50_GRAPH_SERIALIZE), 0);
>> -      IMMED_NVC0(push, SUBC_COMPUTE(0x518), 0);
>>      }
>>
>>      BEGIN_NVC0(push, NVE4_COMPUTE(TEX_CB_INDEX), 1);
>> --
>> 2.6.4
>>
>> _______________________________________________
>> mesa-dev mailing list
>> mesa-dev at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev