Re: [PATCH v2 2/3] drm/i915/perf: enable OAR context save/restore of performance counters

[Lionel Landwerlin <lionel.g.landwerlin@intel.com>]

On 19/10/2019 02:22, Chris Wilson wrote:
> Quoting Umesh Nerlige Ramappa (2019-10-18 01:50:27)
>> From: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
>>
>> We want this so we can preempt performance queries and keep the system
>> responsive even when long running queries are ongoing. We avoid doing
>> it for all contexts.
>>
>> v2: use LRI to modify context control (Chris)
>> v3: use MASKED_FIELD to program just the masked bits (Chris)
>> v4: reuse request created during emit_oa_config (Lionel)
>>
>> Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
>> Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
>> ---
>>   drivers/gpu/drm/i915/gt/intel_lrc.h |  1 +
>>   drivers/gpu/drm/i915/i915_perf.c    | 35 ++++++++++++++++++++++++++++-
>>   2 files changed, 35 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h b/drivers/gpu/drm/i915/gt/intel_lrc.h
>> index 99dc576a4e25..b6daac712c9e 100644
>> --- a/drivers/gpu/drm/i915/gt/intel_lrc.h
>> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.h
>> @@ -43,6 +43,7 @@ struct intel_engine_cs;
>>   #define          CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT   (1 << 0)
>>   #define   CTX_CTRL_RS_CTX_ENABLE               (1 << 1)
>>   #define          CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT      (1 << 2)
>> +#define          GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE     (1 << 8)
>>   #define RING_CONTEXT_STATUS_PTR(base)          _MMIO((base) + 0x3a0)
>>   #define RING_EXECLIST_SQ_CONTENTS(base)                _MMIO((base) + 0x510)
>>   #define RING_EXECLIST_CONTROL(base)            _MMIO((base) + 0x550)
>> diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
>> index 91707558a0f5..ce97af484a32 100644
>> --- a/drivers/gpu/drm/i915/i915_perf.c
>> +++ b/drivers/gpu/drm/i915/i915_perf.c
>> @@ -1862,14 +1862,36 @@ get_oa_vma(struct i915_perf_stream *stream, struct i915_oa_config *oa_config)
>>          return i915_vma_get(oa_bo->vma);
>>   }
>>   
>> +static int gen12_emit_oar_config(struct i915_request *rq,
>> +                                struct intel_context *ce,
>> +                                bool enable)
>> +{
>> +       u32 *cs;
>> +
>> +       cs = intel_ring_begin(rq, 4);
>> +       if (IS_ERR(cs))
>> +               return PTR_ERR(cs);
>> +
>> +       *cs++ = MI_LOAD_REGISTER_IMM(1);
>> +       *cs++ = i915_mmio_reg_offset(RING_CONTEXT_CONTROL(ce->engine->mmio_base));
>> +       *cs++ = _MASKED_FIELD(GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE,
>> +                             enable ? GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE : 0);
>> +       *cs++ = MI_NOOP;
>> +
>> +       intel_ring_advance(rq, cs);
>> +
>> +       return 0;
>> +}
>> +
>>   static int emit_oa_config(struct i915_perf_stream *stream,
>>                            struct intel_context *ce)
>>   {
>> +       struct i915_oa_config *oa_config = stream->oa_config;
>>          struct i915_request *rq;
>>          struct i915_vma *vma;
>>          int err;
>>   
>> -       vma = get_oa_vma(stream, stream->oa_config);
>> +       vma = get_oa_vma(stream, oa_config);
>>          if (IS_ERR(vma))
>>                  return PTR_ERR(vma);
>>   
>> @@ -1891,6 +1913,17 @@ static int emit_oa_config(struct i915_perf_stream *stream,
>>          if (err)
>>                  goto err_add_request;
>>   
>> +       /*
>> +        * For Gen12, performance counters are context saved/restored.
>> +        * Only enable it for the context that requested this.
>> +        */
>> +       if (ce == stream->pinned_ctx && IS_GEN(stream->perf->i915, 12)) {
>> +               err = gen12_emit_oar_config(rq, ce, oa_config != NULL);
>> +               if (err)
>> +                       goto err_add_request;
>> +
>> +       }
> Do we need to do emit_oa_config(stream, NULL) in
> gen8_disable_metric_set()?
> -Chris
>
Oh crap... I must have pointed to the wrong function then...

I thought this was called from enable/disable_metric_set().


I guess the best thing would be to call this from 
gen12_enable/disable_metric_set().

But then we would need to create a request for the pinned_ctx.

Sorry :(


-Lionel

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx