[PATCH v2] drm/amdgpu/OLAND: clip the ref divider max value

Christian König christian.koenig at amd.com
Fri Aug 20 06:38:13 UTC 2021


Sounds like a good idea to me, but I would limit this generally or at 
least for the whole generation and not just one particular chipset.

Regards,
Christian.

Am 20.08.21 um 08:05 schrieb Sharma, Shashank:
> From 4841e5ba60e33ff798bde6cb69fbd7e137b6db9c Mon Sep 17 00:00:00 2001
> From: Shashank Sharma <shashank.sharma at amd.com>
> Date: Fri, 20 Aug 2021 10:20:02 +0530
> Subject: [PATCH v2] drm/amdgpu/OLAND: clip the ref divider max value
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
>
> This patch limits the ref_div_max value to 100, during the
> calculation of PLL feedback reference divider. With current
> value (128), the produced fb_ref_div value generates unstable
> output at particular frequencies. Radeon driver limits this
> value at 100.
>
> On Oland, when we try to setup mode 2048x1280 at 60 (a bit weird,
> I know), it demands a clock of 221270 Khz. It's been observed
> that the PLL calculations using values 128 and 100 are vastly
> different, and look like this:
>
> +------------------------------------------+
> |Parameter    |AMDGPU        |Radeon       |
> |             |              |             |
> +-------------+----------------------------+
> |Clock feedback              |             |
> |divider max  |  128         |   100       |
> |cap value    |              |             |
> |             |              |             |
> |             |              |             |
> +------------------------------------------+
> |ref_div_max  |              |             |
> |             |  42          |  20         |
> |             |              |             |
> |             |              |             |
> +------------------------------------------+
> |ref_div      |  42          |  20         |
> |             |              |             |
> +------------------------------------------+
> |fb_div       |  10326       |  8195       |
> +------------------------------------------+
> |fb_div       |  1024        |  163        |
> +------------------------------------------+
> |fb_dev_p     |  4           |  9          |
> |frac fb_de^_p|              |             |
> +----------------------------+-------------+
>
> With ref_div_max value clipped at 100, AMDGPU driver can also
> drive videmode 2048x1280 at 60 (221Mhz) and produce proper output
> without any blanking and distortion on the screen.
>
> PS: This value was changed from 128 to 100 in Radeon driver also, here:
> https://github.com/freedesktop/drm-tip/commit/4b21ce1b4b5d262e7d4656b8ececc891fc3cb806 
>
>
> V1:
> Got acks from:
> Acked-by: Alex Deucher <alexander.deucher at amd.com>
> Acked-by: Christian König <christian.koenig at amd.com>
>
> V2:
> - Restricting the changes only for OLAND, just to avoid any regression
>   for other cards.
> - Changed unsigned -> unsigned int to make checkpatch quiet.
>
> Cc: Alex Deucher <Alexander.Deucher at amd.com>
> Cc: Christian König <christian.koenig at amd.com>
> Cc: Eddy Qin <Eddy.Qin at amd.com>
> Signed-off-by: Shashank Sharma <shashank.sharma at amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c    | 20 +++++++++++++-------
>  drivers/gpu/drm/amd/amdgpu/amdgpu_pll.h    |  3 ++-
>  drivers/gpu/drm/amd/amdgpu/atombios_crtc.c |  2 +-
>  3 files changed, 16 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c 
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c
> index f2e20666c9c1..6d04c1d25bfb 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c
> @@ -80,12 +80,17 @@ static void amdgpu_pll_reduce_ratio(unsigned *nom, 
> unsigned *den,
>   * Calculate feedback and reference divider for a given post divider. 
> Makes
>   * sure we stay within the limits.
>   */
> -static void amdgpu_pll_get_fb_ref_div(unsigned nom, unsigned den, 
> unsigned post_div,
> -                      unsigned fb_div_max, unsigned ref_div_max,
> -                      unsigned *fb_div, unsigned *ref_div)
> +static void amdgpu_pll_get_fb_ref_div(struct amdgpu_device *adev, 
> unsigned int nom,
> +                      unsigned int den, unsigned int post_div,
> +                      unsigned int fb_div_max, unsigned int ref_div_max,
> +                      unsigned int *fb_div, unsigned int *ref_div)
>  {
> +
>      /* limit reference * post divider to a maximum */
> -    ref_div_max = min(128 / post_div, ref_div_max);
> +    if (adev->asic_type == CHIP_OLAND)
> +        ref_div_max = min(100 / post_div, ref_div_max);
> +    else
> +        ref_div_max = min(128 / post_div, ref_div_max);
>
>      /* get matching reference and feedback divider */
>      *ref_div = min(max(DIV_ROUND_CLOSEST(den, post_div), 1u), 
> ref_div_max);
> @@ -112,7 +117,8 @@ static void amdgpu_pll_get_fb_ref_div(unsigned 
> nom, unsigned den, unsigned post_
>   * Try to calculate the PLL parameters to generate the given frequency:
>   * dot_clock = (ref_freq * feedback_div) / (ref_div * post_div)
>   */
> -void amdgpu_pll_compute(struct amdgpu_pll *pll,
> +void amdgpu_pll_compute(struct amdgpu_device *adev,
> +            struct amdgpu_pll *pll,
>              u32 freq,
>              u32 *dot_clock_p,
>              u32 *fb_div_p,
> @@ -199,7 +205,7 @@ void amdgpu_pll_compute(struct amdgpu_pll *pll,
>
>      for (post_div = post_div_min; post_div <= post_div_max; 
> ++post_div) {
>          unsigned diff;
> -        amdgpu_pll_get_fb_ref_div(nom, den, post_div, fb_div_max,
> +        amdgpu_pll_get_fb_ref_div(adev, nom, den, post_div, fb_div_max,
>                        ref_div_max, &fb_div, &ref_div);
>          diff = abs(target_clock - (pll->reference_freq * fb_div) /
>              (ref_div * post_div));
> @@ -214,7 +220,7 @@ void amdgpu_pll_compute(struct amdgpu_pll *pll,
>      post_div = post_div_best;
>
>      /* get the feedback and reference divider for the optimal value */
> -    amdgpu_pll_get_fb_ref_div(nom, den, post_div, fb_div_max, 
> ref_div_max,
> +    amdgpu_pll_get_fb_ref_div(adev, nom, den, post_div, fb_div_max, 
> ref_div_max,
>                    &fb_div, &ref_div);
>
>      /* reduce the numbers to a simpler ratio once more */
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.h 
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.h
> index db6136f68b82..44a583d6c9b4 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.h
> @@ -24,7 +24,8 @@
>  #ifndef __AMDGPU_PLL_H__
>  #define __AMDGPU_PLL_H__
>
> -void amdgpu_pll_compute(struct amdgpu_pll *pll,
> +void amdgpu_pll_compute(struct amdgpu_device *adev,
> +             struct amdgpu_pll *pll,
>               u32 freq,
>               u32 *dot_clock_p,
>               u32 *fb_div_p,
> diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c 
> b/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c
> index 159a2a4385a1..afad094f84c2 100644
> --- a/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c
> +++ b/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c
> @@ -851,7 +851,7 @@ void amdgpu_atombios_crtc_set_pll(struct drm_crtc 
> *crtc, struct drm_display_mode
>      pll->reference_div = amdgpu_crtc->pll_reference_div;
>      pll->post_div = amdgpu_crtc->pll_post_div;
>
> -    amdgpu_pll_compute(pll, amdgpu_crtc->adjusted_clock, &pll_clock,
> +    amdgpu_pll_compute(adev, pll, amdgpu_crtc->adjusted_clock, 
> &pll_clock,
>                  &fb_div, &frac_fb_div, &ref_div, &post_div);
>
>      amdgpu_atombios_crtc_program_ss(adev, ATOM_DISABLE, 
> amdgpu_crtc->pll_id,



More information about the amd-gfx mailing list