lkml.org 
[lkml]   [2022]   [Mar]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH 1/2] drm/amd/display: detach fpu operations from dcn10_validate_bandwidth in calcs
From
Am 28.03.22 um 19:17 schrieb Melissa Wen:
> On 03/28, Christian König wrote:
>> Am 26.03.22 um 21:24 schrieb Melissa Wen:
>>> dcn10_validate_bandwidth is only used on dcn10 files, but is declared in
>>> dcn_calcs files. Rename dcn10_* to dcn_* in calcs, remove DC_FP_* wrapper
>>> inside DML folder and create an specific dcn10_validate_bandwidth in
>>> dcn10_resources that calls dcn_validate_bandwidth and properly wraps that
>>> FPU function with DC_FP_* macro.
>>>
>>> Signed-off-by: Melissa Wen <mwen@igalia.com>
>>> ---
>>> .../gpu/drm/amd/display/dc/dcn10/dcn10_resource.c | 14 ++++++++++++++
>>> .../gpu/drm/amd/display/dc/dml/calcs/dcn_calcs.c | 5 +----
>>> drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h | 2 +-
>>> 3 files changed, 16 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c
>>> index 4048908dd265..1587a060b55a 100644
>>> --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c
>>> +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c
>>> @@ -1141,6 +1141,20 @@ static void dcn10_destroy_resource_pool(struct resource_pool **pool)
>>> *pool = NULL;
>>> }
>>> +static bool dcn10_validate_bandwidth(
>>> + struct dc *dc,
>>> + struct dc_state *context,
>>> + bool fast_validate)
>>> +{
>>> + bool voltage_supported;
>>> +
>>> + DC_FP_START();
>>> + voltage_supported = dcn_validate_bandwidth(dc, context, fast_validate);
>>> + DC_FP_END();
>>> +
>>> + return voltage_supported;
>>> +}
>>> +
>>> static enum dc_status dcn10_validate_plane(const struct dc_plane_state *plane_state, struct dc_caps *caps)
>>> {
>>> if (plane_state->format >= SURFACE_PIXEL_FORMAT_VIDEO_BEGIN
>>> diff --git a/drivers/gpu/drm/amd/display/dc/dml/calcs/dcn_calcs.c b/drivers/gpu/drm/amd/display/dc/dml/calcs/dcn_calcs.c
>>> index e447c74be713..c25023f7d604 100644
>>> --- a/drivers/gpu/drm/amd/display/dc/dml/calcs/dcn_calcs.c
>>> +++ b/drivers/gpu/drm/amd/display/dc/dml/calcs/dcn_calcs.c
>>> @@ -764,7 +764,7 @@ static unsigned int get_highest_allowed_voltage_level(uint32_t chip_family,
>>> return 4;
>>> }
>>> -bool dcn10_validate_bandwidth(
>>> +bool dcn_validate_bandwidth(
>>> struct dc *dc,
>>> struct dc_state *context,
>>> bool fast_validate)
>>> @@ -790,7 +790,6 @@ bool dcn10_validate_bandwidth(
>>> dcn_bw_sync_calcs_and_dml(dc);
>>> memset(v, 0, sizeof(*v));
>>> - DC_FP_START();
>>> v->sr_exit_time = dc->dcn_soc->sr_exit_time;
>>> v->sr_enter_plus_exit_time = dc->dcn_soc->sr_enter_plus_exit_time;
>>> @@ -1323,8 +1322,6 @@ bool dcn10_validate_bandwidth(
>>> bw_limit = dc->dcn_soc->percent_disp_bw_limit * v->fabric_and_dram_bandwidth_vmax0p9;
>>> bw_limit_pass = (v->total_data_read_bandwidth / 1000.0) < bw_limit;
>>> - DC_FP_END();
>>> -
>>> PERFORMANCE_TRACE_END();
>>> BW_VAL_TRACE_FINISH();
>>> diff --git a/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h b/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h
>>> index 337c0161e72d..806f3041db14 100644
>>> --- a/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h
>>> +++ b/drivers/gpu/drm/amd/display/dc/inc/dcn_calcs.h
>>> @@ -619,7 +619,7 @@ struct dcn_ip_params {
>>> };
>>> extern const struct dcn_ip_params dcn10_ip_defaults;
>>> -bool dcn10_validate_bandwidth(
>>> +bool dcn_validate_bandwidth(
>>> struct dc *dc,
>>> struct dc_state *context,
>>> bool fast_validate);
>> Just for the record: That's not really usual kernel coding style, but that's
>> not topic of this patch set.
> Yeah. I didn't change the code style to ease any version conflict management.
>> The series is Acked-by: Christian König <christian.koenig@amd.com>
> Thanks!
>> And it would be really nice if we could make the DC_FP_* macros somehow fail
>> in the dml folder.
> And if we include a kind of dc_assert_fp_disabled() in the dc_fpu_begin()
> (DC_FP_START) - more or less the reverse of dc_assert_fp_enabled(). Does
> it meet the `make the DC_FP_* macros somehow fail in the dml folder` ?
> It is not restricted to the dml folder, but I think it would work
> similarly... Does it make sense?

No, IIRC our display team even mentioned to me that those macros could
potentially be used recursively.

What I mean here is that we somehow raise a compiler warning if somebody
tries to use those defines inside the folder.

Maybe check of gcc supports hardware floating point or something like this.

Christian.

>
> Melissa
>
>> Thanks,
>> Christian.
>>
>>

\
 
 \ /
  Last update: 2022-03-28 19:47    [W:0.072 / U:0.960 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site