Add writeback scaling
Subsystems (3)
v3 cover Add writeback scaling
Add support for scaling on writeback. To do this add writeback_dest_w and writeback_dest_h as writeback connector properties to specify the desired output dimensions.
Then implement downscaling on writeback for Malidp-550 and Malidp-650 (upscaling on writeback is not supported on these devices).
v2: Use 0 as default for writeback_w,h and so update range to use 1 as
minimum.
v3: Rename properties to specify they are destination width/height.
Make sure the values from the properties are passed to enable_memwrite rather than the framebuffer dimensions
Ben Davis (2):
drm: Add writeback_dest_w,h properties
drm/malidp: Enable writeback scaling
drivers/gpu/drm/arm/malidp_crtc.c | 49 +++++++++++--------- drivers/gpu/drm/arm/malidp_crtc.h | 12 +++++ drivers/gpu/drm/arm/malidp_drv.c | 10 +++-- drivers/gpu/drm/arm/malidp_hw.c | 45 +++++++++++++------ drivers/gpu/drm/arm/malidp_hw.h | 19 ++++++-- drivers/gpu/drm/arm/malidp_mw.c | 94 ++++++++++++++++++++++++++++++--------- drivers/gpu/drm/arm/malidp_regs.h | 1 + drivers/gpu/drm/drm_atomic_uapi.c | 8 ++++ drivers/gpu/drm/drm_writeback.c | 30 +++++++++++++ include/drm/drm_connector.h | 4 ++ include/drm/drm_mode_config.h | 10 +++++ 11 files changed, 220 insertions(+), 62 deletions(-) create mode 100644 drivers/gpu/drm/arm/malidp_crtc.h
Signature · 1 line
2.7.4
PATCH v3 1/2 drm: Add writeback_dest_w,h properties
Add new properties to specify width and height for writeback.
--- drivers/gpu/drm/drm_atomic_uapi.c | 8 ++++++++ drivers/gpu/drm/drm_writeback.c | 30 ++++++++++++++++++++++++++++++ include/drm/drm_connector.h | 4 ++++ include/drm/drm_mode_config.h | 10 ++++++++++ 4 files changed, 52 insertions(+)
drivers/gpu/drm/drm_atomic_uapi.c +8 -0
@@ -765,6 +765,10 @@ static int drm_atomic_connector_set_property(struct drm_connector *connector,return -EINVAL ; } state->content_protection = val; + }else if (property == config->prop_writeback_dest_w) { + state->writeback_dest_w = val; + }else if (property == config->prop_writeback_dest_h) { + state->writeback_dest_h = val; }else if (property == config->writeback_fb_id_property) {struct drm_framebuffer *fb =drm_framebuffer_lookup (dev,NULL , val);int ret =drm_atomic_set_writeback_fb_for_connector (state, fb);
@@ -837,6 +841,10 @@ drm_atomic_connector_get_property(struct drm_connector *connector, *val = state->scaling_mode; }else if (property == connector->content_protection_property) { *val = state->content_protection; + }else if (property == config->prop_writeback_dest_w) { + *val = state->writeback_dest_w; + }else if (property == config->prop_writeback_dest_h) { + *val = state->writeback_dest_h; }else if (property == config->writeback_fb_id_property) {/* Writeback framebuffer is one-shot, write and forget */ *val =0 ;
drivers/gpu/drm/drm_writeback.c +30 -0
@@ -74,6 +74,14 @@* applications making use of writeback connectors *always* retrieve an * out-fence for the commit and use it appropriately. * From userspace, this property will always read as zero. +* +* "WRITEBACK_DEST_W": +* The width of the composition to write back. 0 acts as default. If +* non-zero the composition will be scaled to match the given width. +* +* "WRITEBACK_DEST_H": +* The height of the composition to write back. 0 acts as default. If +* non-zero the composition will be scaled to match the given height. */ #define fence_to_wb_connector (x) container_of(x->lock, \
@@ -141,6 +149,22 @@ static int create_writeback_properties(struct drm_device *dev) dev->mode_config.writeback_out_fence_ptr_property = prop; } +if (!dev->mode_config.prop_writeback_dest_w) { + prop =drm_property_create_range (dev,DRM_MODE_PROP_ATOMIC , +"WRITEBACK_DEST_W" ,1 ,UINT_MAX ); +if (!prop) +return -ENOMEM ; + dev->mode_config.prop_writeback_dest_w = prop; + } + +if (!dev->mode_config.prop_writeback_dest_h) { + prop =drm_property_create_range (dev,DRM_MODE_PROP_ATOMIC , +"WRITEBACK_DEST_H" ,1 ,UINT_MAX ); +if (!prop) +return -ENOMEM ; + dev->mode_config.prop_writeback_dest_h = prop; + } +return 0 ; }
@@ -225,6 +249,12 @@ int drm_writeback_connector_init(struct drm_device *dev,drm_object_attach_property (&connector->base, config->writeback_pixel_formats_property, blob->base.id); + +drm_object_attach_property (&connector->base, + config->prop_writeback_dest_w,0 ); +drm_object_attach_property (&connector->base, + config->prop_writeback_dest_h,0 ); + wb_connector->pixel_formats_blob_ptr = blob;return 0 ;
include/drm/drm_connector.h +4 -0
@@ -515,6 +515,10 @@ struct drm_connector_state {*/ struct drm_writeback_job *writeback_job; +/** @writeback_dest_w: width of plane to write to wb buffer */ +/** @writeback_dest_h: height of plane to write to wb buffer */ +uint32_t writeback_dest_w, writeback_dest_h; +/** * @max_requested_bpc: Connector property to limit the maximum bit * depth of the pixels.
include/drm/drm_mode_config.h +10 -0
@@ -622,6 +622,16 @@ struct drm_mode_config {*/ struct drm_property *prop_crtc_h;/** +* @prop_writeback_dest_w: Writeback connector property for the plane +* destination position in the writeback buffer. +*/ +struct drm_property *prop_writeback_dest_w; +/** +* @prop_writeback_dest_h: Writeback connector property for the plane +* destination position in the writeback buffer. +*/ +struct drm_property *prop_writeback_dest_h; +/** * @prop_fb_id: Default atomic plane property to specify the * &drm_framebuffer. */
Signature · 6 lines
2.7.4 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v3 2/2 drm/malidp: Enable writeback scaling
In reply to Ben DavisThe phase setting part of malidp_crtc_atomic_check_scaling is refactored to allow use in writeback scaling.
Also the enable_memwrite function prototype is simplified by directly passing mw_state.
--- drivers/gpu/drm/arm/malidp_crtc.c | 49 +++++++++++--------- drivers/gpu/drm/arm/malidp_crtc.h | 12 +++++ drivers/gpu/drm/arm/malidp_drv.c | 10 +++-- drivers/gpu/drm/arm/malidp_hw.c | 45 +++++++++++++------ drivers/gpu/drm/arm/malidp_hw.h | 19 ++++++-- drivers/gpu/drm/arm/malidp_mw.c | 94 ++++++++++++++++++++++++++++++--------- drivers/gpu/drm/arm/malidp_regs.h | 1 + 7 files changed, 168 insertions(+), 62 deletions(-) create mode 100644 drivers/gpu/drm/arm/malidp_crtc.h
drivers/gpu/drm/arm/malidp_crtc.c +27 -22
@@ -265,6 +265,31 @@ static int malidp_crtc_atomic_check_ctm(struct drm_crtc *crtc,return 0 ; } +void malidp_set_se_config_phase (struct malidp_se_config *s) +{ +#define SE_N_PHASE 4 +#define SE_SHIFT_N_PHASE 12 +u32 phase; + +/* Calculate initial_phase and delta_phase for horizontal. */ + phase = s->input_w; + s->h_init_phase = ((phase <<SE_N_PHASE ) / s->output_w +1 ) /2 ; + + phase = s->input_w; + phase <<= (SE_SHIFT_N_PHASE +SE_N_PHASE ); + s->h_delta_phase = phase / s->output_w; + +/* Same for vertical. */ + phase = s->input_h; + s->v_init_phase = ((phase <<SE_N_PHASE ) / s->output_h +1 ) /2 ; + + phase = s->input_h; + phase <<= (SE_SHIFT_N_PHASE +SE_N_PHASE ); + s->v_delta_phase = phase / s->output_h; +#undef SE_N_PHASE +#undef SE_SHIFT_N_PHASE +} +static int malidp_crtc_atomic_check_scaling (struct drm_crtc *crtc,struct drm_crtc_state *state) {
@@ -291,7 +316,6 @@ static int malidp_crtc_atomic_check_scaling(struct drm_crtc *crtc,drm_atomic_crtc_state_for_each_plane_state (plane, pstate, state) {struct malidp_plane *mp =to_malidp_plane (plane); -u32 phase;if (!(mp->layer->id & scaling))continue ;
@@ -319,27 +343,8 @@ static int malidp_crtc_atomic_check_scaling(struct drm_crtc *crtc, s->output_w = pstate->crtc_w; s->output_h = pstate->crtc_h; -#define SE_N_PHASE 4 -#define SE_SHIFT_N_PHASE 12 -/* Calculate initial_phase and delta_phase for horizontal. */ - phase = s->input_w; - s->h_init_phase = - ((phase <<SE_N_PHASE ) / s->output_w +1 ) /2 ; - - phase = s->input_w; - phase <<= (SE_SHIFT_N_PHASE +SE_N_PHASE ); - s->h_delta_phase = phase / s->output_w; - -/* Same for vertical. */ - phase = s->input_h; - s->v_init_phase = - ((phase <<SE_N_PHASE ) / s->output_h +1 ) /2 ; - - phase = s->input_h; - phase <<= (SE_SHIFT_N_PHASE +SE_N_PHASE ); - s->v_delta_phase = phase / s->output_h; -#undef SE_N_PHASE -#undef SE_SHIFT_N_PHASE + +malidp_set_se_config_phase (s); s->plane_src_id = mp->layer->id; }
drivers/gpu/drm/arm/malidp_crtc.h new file +12 -0
@@ -0,0 +1,12 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* +* (C) COPYRIGHT 2019 ARM Limited. All rights reserved. +* +*/ + +#include "malidp_hw.h" + +#ifndef __MALIDP_CRTC_H__ +#define __MALIDP_CRTC_H__ +void malidp_set_se_config_phase (struct malidp_se_config *s); +#endif
drivers/gpu/drm/arm/malidp_drv.c +7 -3
@@ -138,7 +138,7 @@ static void malidp_atomic_commit_se_config(struct drm_crtc *crtc,u32 val;/* Set SE_CONTROL */ -if (!s->scale_enable) { +if (!s->scale_enable && !s->wb_scale_enable) { val =malidp_hw_read (hwdev, se_control); val &= ~MALIDP_SE_SCALING_EN ;malidp_hw_write (hwdev, val, se_control);
@@ -147,12 +147,16 @@ static void malidp_atomic_commit_se_config(struct drm_crtc *crtc, hwdev->hw->se_set_scaling_coeffs (hwdev, s, old_s); val =malidp_hw_read (hwdev, se_control); - val |=MALIDP_SE_SCALING_EN |MALIDP_SE_ALPHA_EN ; + val |=MALIDP_SE_SCALING_EN ; val &= ~MALIDP_SE_ENH (MALIDP_SE_ENH_MASK ); val |= s->enhancer_enable ?MALIDP_SE_ENH (3 ) :0 ; - val |=MALIDP_SE_RGBO_IF_EN ; +if (s->scale_enable) + val |= (MALIDP_SE_RGBO_IF_EN |MALIDP_SE_ALPHA_EN ); +else + val &= ~((MALIDP_SE_RGBO_IF_EN |MALIDP_SE_ALPHA_EN )); +malidp_hw_write (hwdev, val, se_control);/* Set IN_SIZE & OUT_SIZE. */
drivers/gpu/drm/arm/malidp_hw.c +32 -13
75 lines not on this page · raw message
drivers/gpu/drm/arm/malidp_hw.h +16 -3
@@ -85,6 +85,7 @@ enum malidp_scaling_coeff_set {struct malidp_se_config {u8 scale_enable :1 ;u8 enhancer_enable :1 ; +u8 wb_scale_enable :1 ;u8 hcoeff :3 ;u8 vcoeff :3 ;u8 plane_src_id;
@@ -94,6 +95,19 @@ struct malidp_se_config {u32 v_init_phase, v_delta_phase; }; +#define to_mw_state (_state) (struct malidp_mw_connector_state *)(_state) + +struct malidp_mw_connector_state { +struct drm_connector_state base; +dma_addr_t addrs[2 ]; +s32 pitches[2 ]; +u8 format; +u8 n_planes; +bool rgb2yuv_initialized; +const s16 *rgb2yuv_coeffs; +bool wb_scale_enable; +}; +/* regmap features */ #define MALIDP_REGMAP_HAS_CLEARIRQ BIT(0)#define MALIDP_DEVICE_AFBC_SUPPORT_SPLIT BIT(1)
@@ -206,9 +220,8 @@ struct malidp_hw {* @param h - height of the output frame * @param fmt_id - internal format ID of output buffer */ -int (*enable_memwrite)(struct malidp_hw_device *hwdev,dma_addr_t *addrs, -s32 *pitches,int num_planes,u16 w,u16 h,u32 fmt_id, -const s16 *rgb2yuv_coeffs); +int (*enable_memwrite)(struct malidp_hw_device *hwdev,u16 w,u16 h, +struct malidp_mw_connector_state *mw_state);/* * Disable the writing to memory of the next frame's content.
drivers/gpu/drm/arm/malidp_mw.c +73 -21
130 lines not on this page · raw message
drivers/gpu/drm/arm/malidp_regs.h +1 -0
@@ -73,6 +73,7 @@#define MALIDP_DISP_FUNC_CADJ (1 << 4)#define MALIDP_DISP_FUNC_ILACED (1 << 8)#define MALIDP_SCALE_ENGINE_EN (1 << 16) +#define MALIDP_SE_MEMWRITE_SCL_EN (1 << 5)#define MALIDP_SE_MEMWRITE_EN (2 << 5)/* register offsets for IRQ management */
Signature · 6 lines
2.7.4 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Quoted 9 lines · Ben Davis
Add support for scaling on writeback. To do this add
writeback_dest_wandwriteback_dest_has writeback connector properties to specify the desired output dimensions.
Then implement downscaling on writeback for Malidp-550 and Malidp-650 (upscaling on writeback is not supported on these devices).v2: Use 0 as default for writeback_w,h and so update range to use 1 as
minimum.
Hm I missed that, I don't think that's good, since it prevents userspace from blindly writing back the properties it's read. We've tried hard to avoid that, since we're suggesting compositor can take a snapshot of all kms properties (including the ones they don't understand) and restore that on vt switching. Hence stuff like fence fds returning -1, and accepting -1 as NULL to make this work.
tldr; I think range needs to include 0, and we need make that a special thing, maybe enforced with a drm_connector_state_compute_writeback_dst_h/w, which takes crtc_state->mode.v/hdisplay into account if the writeback_dst_h/w is 0.
-Daniel
Quoted 26 lines
v3: Rename properties to specify they are destination width/height.
Make sure the values from the properties are passed toenable_memwriterather than the framebuffer dimensionsBen Davis (2):
drm: Add writeback_dest_w,h properties
drm/malidp: Enable writeback scalingdrivers/gpu/drm/arm/malidp_crtc.c | 49 +++++++++++--------- drivers/gpu/drm/arm/malidp_crtc.h | 12 +++++ drivers/gpu/drm/arm/malidp_drv.c | 10 +++-- drivers/gpu/drm/arm/malidp_hw.c | 45 +++++++++++++------ drivers/gpu/drm/arm/malidp_hw.h | 19 ++++++-- drivers/gpu/drm/arm/malidp_mw.c | 94 ++++++++++++++++++++++++++++++--------- drivers/gpu/drm/arm/malidp_regs.h | 1 + drivers/gpu/drm/drm_atomic_uapi.c | 8 ++++ drivers/gpu/drm/drm_writeback.c | 30 +++++++++++++ include/drm/drm_connector.h | 4 ++ include/drm/drm_mode_config.h | 10 +++++ 11 files changed, 220 insertions(+), 62 deletions(-) create mode 100644 drivers/gpu/drm/arm/malidp_crtc.h--
2.7.4
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Liviu Dudau
maintainerQuoted 11 lines · Daniel Vetter
Quoted 9 lines · Ben Davis
Add support for scaling on writeback. To do this add
writeback_dest_wandwriteback_dest_has writeback connector properties to specify the desired output dimensions.
Then implement downscaling on writeback for Malidp-550 and Malidp-650 (upscaling on writeback is not supported on these devices).v2: Use 0 as default for writeback_w,h and so update range to use 1 as
minimum.
Quoted 6 lines
Hm I missed that, I don't think that's good, since it prevents userspace from blindly writing back the properties it's read. We've tried hard to avoid that, since we're suggesting compositor can take a snapshot of all kms properties (including the ones they don't understand) and restore that on vt switching. Hence stuff like fence fds returning -1, and accepting -1 as NULL to make this work.
Is that they way it should work, though? I remember a few good years back, before KMS, there were lots of issues with X server hanging and not restoring the vt mode correctly. Should it not be that on getting back control of a DRM master, an application sets it last known good mode that it had, before it lost (or gave away) control?
Quoted 6 lines
tldr; I think range needs to include 0, and we need make that a special thing, maybe enforced with a drm_connector_state_compute_writeback_dst_h/w, which takes crtc_state->mode.v/hdisplay into account if the writeback_dst_h/w is 0.
Repeating just to make sure we're on the same page: we allow 0 as valid range value, but writeback_dst_h/w ultimately gets updated before the commit gets passed on to the driver to use crtc_state->mode.v/hdisplay so as not to trigger any scaling. Correct?
Best regards,
Liviu
Quoted 34 lines
-Daniel
Quoted 26 lines
v3: Rename properties to specify they are destination width/height.
Make sure the values from the properties are passed toenable_memwriterather than the framebuffer dimensionsBen Davis (2):
drm: Add writeback_dest_w,h properties
drm/malidp: Enable writeback scalingdrivers/gpu/drm/arm/malidp_crtc.c | 49 +++++++++++--------- drivers/gpu/drm/arm/malidp_crtc.h | 12 +++++ drivers/gpu/drm/arm/malidp_drv.c | 10 +++-- drivers/gpu/drm/arm/malidp_hw.c | 45 +++++++++++++------ drivers/gpu/drm/arm/malidp_hw.h | 19 ++++++-- drivers/gpu/drm/arm/malidp_mw.c | 94 ++++++++++++++++++++++++++++++--------- drivers/gpu/drm/arm/malidp_regs.h | 1 + drivers/gpu/drm/drm_atomic_uapi.c | 8 ++++ drivers/gpu/drm/drm_writeback.c | 30 +++++++++++++ include/drm/drm_connector.h | 4 ++ include/drm/drm_mode_config.h | 10 +++++ 11 files changed, 220 insertions(+), 62 deletions(-) create mode 100644 drivers/gpu/drm/arm/malidp_crtc.h--
2.7.4Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Signature · 12 lines
====================
| I would like to |
| fix the world, |
| but they're not |
| giving me the |
\ source code! /
---------------
¯\_(ツ)_/¯
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-develQuoted 20 lines · Liviu Dudau
Quoted 11 lines · Daniel Vetter
Quoted 9 lines · Ben Davis
Add support for scaling on writeback. To do this add
writeback_dest_wandwriteback_dest_has writeback connector properties to specify the desired output dimensions.
Then implement downscaling on writeback for Malidp-550 and Malidp-650 (upscaling on writeback is not supported on these devices).v2: Use 0 as default for writeback_w,h and so update range to use 1 as
minimum.Quoted 6 lines
Hm I missed that, I don't think that's good, since it prevents userspace from blindly writing back the properties it's read. We've tried hard to avoid that, since we're suggesting compositor can take a snapshot of all kms properties (including the ones they don't understand) and restore that on vt switching. Hence stuff like fence fds returning -1, and accepting -1 as NULL to make this work.
Quoted 5 lines
Is that they way it should work, though? I remember a few good years back, before KMS, there were lots of issues with X server hanging and not restoring the vt mode correctly. Should it not be that on getting back control of a DRM master, an application sets it last known good mode that it had, before it lost (or gave away) control?
Given that we've recommended both approaches I expect there's a mix of both approaches out there in the real world, with people just hacking things up until it works. That's also what we're doing with the in-kernel fbdev emulation. I'd say defacto this means it's all ill-defined, but also that we imo should make a best effort to support either approach.
Quoted 10 lines
Quoted 5 lines
tldr; I think range needs to include 0, and we need make that a special thing, maybe enforced with a drm_connector_state_compute_writeback_dst_h/w, which takes crtc_state->mode.v/hdisplay into account if the writeback_dst_h/w is 0.
Repeating just to make sure we're on the same page: we allow 0 as valid range value, but writeback_dst_h/w ultimately gets updated before the commit gets passed on to the driver to use crtc_state->mode.v/hdisplay so as not to trigger any scaling. Correct?
That would also break stuff, since then you'd scale the next time v/hdisplay are different. Except if we again clear them back to 0 on duplicate_state, which I guess would work since writeback is one-shot. I was thinking more of a helper function, but your approach + resetting always to 0 on duplicate_state + lots of comments/kerneldoc should work too. And I guess would be more robust, since driver authors can't accidentally use the wrong state since it's fixed up. -Daniel
Quoted 49 lines
Best regards,
LiviuQuoted 30 lines
-Daniel
Quoted 23 lines
v3: Rename properties to specify they are destination width/height.
Make sure the values from the properties are passed toenable_memwriterather than the framebuffer dimensionsBen Davis (2):
drm: Add writeback_dest_w,h properties
drm/malidp: Enable writeback scalingdrivers/gpu/drm/arm/malidp_crtc.c | 49 +++++++++++--------- drivers/gpu/drm/arm/malidp_crtc.h | 12 +++++ drivers/gpu/drm/arm/malidp_drv.c | 10 +++-- drivers/gpu/drm/arm/malidp_hw.c | 45 +++++++++++++------ drivers/gpu/drm/arm/malidp_hw.h | 19 ++++++-- drivers/gpu/drm/arm/malidp_mw.c | 94 ++++++++++++++++++++++++++++++--------- drivers/gpu/drm/arm/malidp_regs.h | 1 + drivers/gpu/drm/drm_atomic_uapi.c | 8 ++++ drivers/gpu/drm/drm_writeback.c | 30 +++++++++++++ include/drm/drm_connector.h | 4 ++ include/drm/drm_mode_config.h | 10 +++++ 11 files changed, 220 insertions(+), 62 deletions(-) create mode 100644 drivers/gpu/drm/arm/malidp_crtc.hQuoted 3 lines
--
2.7.4Quoted 4 lines
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch--
====================| I would like to | | fix the world, | | but they're not | | giving me the | \ source code! / --------------- ¯\_(ツ)_/¯
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch