[Mesa-dev] [PATCH 2/2] radeonsi: Enable loading into CE RAM.

[Bas Nieuwenhuizen]

Hi Marek,

The hang was most likely related to something else I tried to fix the
loads and that started doing something once the loads were enabled.
With the kernel patch I had the hang too until I cleaned up the
leftovers of all my previous attempts at fixes in mesa.

- Bas



On Thu, Apr 21, 2016 at 11:44 AM, Marek Olšák <maraeo at gmail.com> wrote:
> On Thu, Apr 21, 2016 at 1:49 AM, Bas Nieuwenhuizen
> <bas at basnieuwenhuizen.nl> wrote:
>> We need to enable a bit in the CONTEXT_CONTROL packet for the
>> loads to work.
>>
>> Signed-off-by: Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
>> ---
>>  src/gallium/drivers/radeonsi/si_descriptors.c | 6 ++++++
>>  src/gallium/drivers/radeonsi/si_hw_context.c  | 5 +++++
>>  src/gallium/drivers/radeonsi/si_state.h       | 1 +
>>  3 files changed, 12 insertions(+)
>>
>> diff --git a/src/gallium/drivers/radeonsi/si_descriptors.c b/src/gallium/drivers/radeonsi/si_descriptors.c
>> index 1580e61..30e65a9 100644
>> --- a/src/gallium/drivers/radeonsi/si_descriptors.c
>> +++ b/src/gallium/drivers/radeonsi/si_descriptors.c
>> @@ -185,6 +185,12 @@ static void si_reinitialize_ce_ram(struct si_context *sctx,
>>         desc->ce_ram_dirty = false;
>>  }
>>
>> +void si_ce_enable_loads(struct radeon_winsys_cs *ib) {
>> +       radeon_emit(ib, PKT3(PKT3_CONTEXT_CONTROL, 1, 0));
>> +       radeon_emit(ib, LOAD_CONTROL_UPDATE(1) | LOAD_CONTROL_CE_RAM_EN(1));
>> +       radeon_emit(ib, SHADOW_ENABLE_UPDATE(1));
>> +}
>> +
>>  static bool si_upload_descriptors(struct si_context *sctx,
>>                                   struct si_descriptors *desc,
>>                                   struct r600_atom * atom)
>> diff --git a/src/gallium/drivers/radeonsi/si_hw_context.c b/src/gallium/drivers/radeonsi/si_hw_context.c
>> index e3abb7f..e6018f3 100644
>> --- a/src/gallium/drivers/radeonsi/si_hw_context.c
>> +++ b/src/gallium/drivers/radeonsi/si_hw_context.c
>> @@ -202,6 +202,11 @@ void si_begin_new_cs(struct si_context *ctx)
>>         if (ctx->init_config_gs_rings)
>>                 si_pm4_emit(ctx, ctx->init_config_gs_rings);
>>
>> +       if (ctx->ce_preamble_ib)
>> +               si_ce_enable_loads(ctx->ce_preamble_ib);
>> +       else if (ctx->ce_ib)
>> +               si_ce_enable_loads(ctx->ce_ib);
>
> So what was the cause of hangs you were seeing? Does it not hang when
> there is CONTEXT_CONTROL in the CE IB?
>
> Marek