FBC frontbuffer tracking conversion, v3
Subsystems (2)
cover FBC frontbuffer tracking conversion, v3
From: Paulo Zanoni <[email protected]>
Hi
Here are the patches with Daniel's latest suggestions applied. I also reordered them a little bit so the stuff already reviewed by Rodrigo can come first. I also included a patch that was part of another series in the end.
Thanks,
Paulo
Paulo Zanoni (7):
drm/i915: extract intel_fbc_find_crtc()
drm/i915: HSW+ FBC is tied to pipe A
drm/i915: gen5+ can have FBC with multiple pipes drm/i915: pass which operation triggered the frontbuffer tracking drm/i915: also do frontbuffer tracking on pwrites drm/i915: add frontbuffer tracking to FBC
drm/i915: don't reallocate the compressed FB at every frame
drivers/gpu/drm/i915/i915_drv.h | 17 ++-- drivers/gpu/drm/i915/i915_gem.c | 27 ++++-- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/i915_gem_stolen.c | 2 +- drivers/gpu/drm/i915/intel_drv.h | 9 +- drivers/gpu/drm/i915/intel_fbc.c | 148 ++++++++++++++++++----------- drivers/gpu/drm/i915/intel_frontbuffer.c | 18 +--- drivers/gpu/drm/i915/intel_ringbuffer.c | 41 +------- drivers/gpu/drm/i915/intel_ringbuffer.h | 1 - 9 files changed, 139 insertions(+), 126 deletions(-)
Signature · 6 lines
2.1.4 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 1/7 drm/i915: extract intel_fbc_find_crtc()
From: Paulo Zanoni <[email protected]>
I want to make this code a little more complicated, so let's extract the function first.
--- drivers/gpu/drm/i915/intel_fbc.c | 46 +++++++++++++++++++++++++--------------- 1 file changed, 29 insertions(+), 17 deletions(-)
drivers/gpu/drm/i915/intel_fbc.c +29 -17
@@ -473,6 +473,32 @@ static bool set_no_fbc_reason(struct drm_i915_private *dev_priv,return true; } +static struct drm_crtc *intel_fbc_find_crtc (struct drm_i915_private *dev_priv) +{ +struct drm_device *dev = dev_priv->dev; +struct drm_crtc *crtc =NULL , *tmp_crtc; + +for_each_crtc (dev, tmp_crtc) { +if (intel_crtc_active (tmp_crtc) && +to_intel_crtc (tmp_crtc)->primary_enabled) { +if (crtc) { +if (set_no_fbc_reason (dev_priv,FBC_MULTIPLE_PIPES )) +DRM_DEBUG_KMS ("more than one pipe active, disabling compression\n" ); +return NULL ; + } + crtc = tmp_crtc; + } + } + +if (!crtc || crtc->primary->fb ==NULL ) { +if (set_no_fbc_reason (dev_priv,FBC_NO_OUTPUT )) +DRM_DEBUG_KMS ("no output, disabling\n" ); +return NULL ; + } + +return crtc; +} +/** * intel_fbc_update - enable/disable FBC as needed * @dev: the drm_device
@@ -495,7 +521,7 @@ static bool set_no_fbc_reason(struct drm_i915_private *dev_priv,void intel_fbc_update (struct drm_device *dev) {struct drm_i915_private *dev_priv = dev->dev_private; -struct drm_crtc *crtc =NULL , *tmp_crtc; +struct drm_crtc *crtc =NULL ;struct intel_crtc *intel_crtc;struct drm_framebuffer *fb;struct drm_i915_gem_object *obj;
@@ -530,23 +556,9 @@ void intel_fbc_update(struct drm_device *dev)* - new fb is too large to fit in compressed buffer * - going to an unsupported config (interlace, pixel multiply, etc.) */ -for_each_crtc (dev, tmp_crtc) { -if (intel_crtc_active (tmp_crtc) && -to_intel_crtc (tmp_crtc)->primary_enabled) { -if (crtc) { -if (set_no_fbc_reason (dev_priv,FBC_MULTIPLE_PIPES )) -DRM_DEBUG_KMS ("more than one pipe active, disabling compression\n" ); - goto out_disable; - } - crtc = tmp_crtc; - } - } - -if (!crtc || crtc->primary->fb ==NULL ) { -if (set_no_fbc_reason (dev_priv,FBC_NO_OUTPUT )) -DRM_DEBUG_KMS ("no output, disabling\n" ); + crtc =intel_fbc_find_crtc (dev_priv); +if (!crtc) goto out_disable; - } intel_crtc =to_intel_crtc (crtc); fb = crtc->primary->fb;
Signature · 6 lines
2.1.4 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 2/7 drm/i915: HSW+ FBC is tied to pipe A
In reply to Paulo ZanoniFrom: Paulo Zanoni <[email protected]>
So add code to consider this case.
v2: Reorder the series, so drop the possible_framebuffer_bits chunk.
--- drivers/gpu/drm/i915/intel_fbc.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_fbc.c +11 -2
@@ -475,10 +475,16 @@ static bool set_no_fbc_reason(struct drm_i915_private *dev_priv,static struct drm_crtc *intel_fbc_find_crtc (struct drm_i915_private *dev_priv) { -struct drm_device *dev = dev_priv->dev;struct drm_crtc *crtc =NULL , *tmp_crtc; +enum pipe pipe; +bool pipe_a_only = false; + +if (IS_HASWELL (dev_priv) ||INTEL_INFO (dev_priv)->gen >=8 ) + pipe_a_only = true; + +for_each_pipe (dev_priv, pipe) { + tmp_crtc = dev_priv->pipe_to_crtc_mapping[pipe]; -for_each_crtc (dev, tmp_crtc) {if (intel_crtc_active (tmp_crtc) &&to_intel_crtc (tmp_crtc)->primary_enabled) {if (crtc) {
@@ -488,6 +494,9 @@ static struct drm_crtc *intel_fbc_find_crtc(struct drm_i915_private *dev_priv) } crtc = tmp_crtc; } + +if (pipe_a_only) +break ; }if (!crtc || crtc->primary->fb ==NULL ) {
Signature · 6 lines
2.1.4 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 3/7 drm/i915: gen5+ can have FBC with multiple pipes
In reply to Paulo ZanoniFrom: Paulo Zanoni <[email protected]>
So allow it.
--- drivers/gpu/drm/i915/intel_fbc.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_fbc.c +4 -2
@@ -477,17 +477,19 @@ static struct drm_crtc *intel_fbc_find_crtc(struct drm_i915_private *dev_priv) {struct drm_crtc *crtc =NULL , *tmp_crtc;enum pipe pipe; -bool pipe_a_only = false; +bool pipe_a_only = false, one_pipe_only = false;if (IS_HASWELL (dev_priv) ||INTEL_INFO (dev_priv)->gen >=8 ) pipe_a_only = true; +else if (INTEL_INFO (dev_priv)->gen <=4 ) + one_pipe_only = true;for_each_pipe (dev_priv, pipe) { tmp_crtc = dev_priv->pipe_to_crtc_mapping[pipe];if (intel_crtc_active (tmp_crtc) &&to_intel_crtc (tmp_crtc)->primary_enabled) { -if (crtc) { +if (one_pipe_only && crtc) {if (set_no_fbc_reason (dev_priv,FBC_MULTIPLE_PIPES ))DRM_DEBUG_KMS ("more than one pipe active, disabling compression\n" );return NULL ;
Signature · 6 lines
2.1.4 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 3/7] drm/i915: gen5+ can have FBC with multiple pipes
Quoted 6 lines · Paulo Zanoni
From: Paulo Zanoni <[email protected]>
So allow it.
Reviewed-by Rodrigo Vivi <[email protected]>Signed-off-by Paulo Zanoni <[email protected]>
Merged up to this one to dinq. The subsequent patches look good, but as per our irc discussion I think it would be useful if Rodrigo could take another look and update his review.
Thanks, Daniel
Quoted 37 lines
--- drivers/gpu/drm/i915/intel_fbc.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)drivers/gpu/drm/i915/intel_fbc.c +4
-2@@ -477,17 +477,19 @@ static struct drm_crtc *intel_fbc_find_crtc(struct drm_i915_private *dev_priv) {struct drm_crtc *crtc =NULL , *tmp_crtc;enum pipe pipe; -bool pipe_a_only = false; +bool pipe_a_only = false, one_pipe_only = false;if (IS_HASWELL (dev_priv) ||INTEL_INFO (dev_priv)->gen >=8 ) pipe_a_only = true; +else if (INTEL_INFO (dev_priv)->gen <=4 ) + one_pipe_only = true;for_each_pipe (dev_priv, pipe) { tmp_crtc = dev_priv->pipe_to_crtc_mapping[pipe];if (intel_crtc_active (tmp_crtc) &&to_intel_crtc (tmp_crtc)->primary_enabled) { -if (crtc) { +if (one_pipe_only && crtc) {if (set_no_fbc_reason (dev_priv,FBC_MULTIPLE_PIPES ))DRM_DEBUG_KMS ("more than one pipe active, disabling compression\n" );return NULL ;Signature · 6 lines
2.1.4 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 4/7 drm/i915: pass which operation triggered the frontbuffer tracking
In reply to Paulo ZanoniFrom: Paulo Zanoni <[email protected]>
We want to port FBC to the frontbuffer tracking infrastructure, but for that we need to know what caused the object invalidation so we can react accordingly: CPU mmaps need manual, GTT mmaps and flips don't need handling and ring rendering needs nukes.
v2: - s/ORIGIN_RENDER/ORIGIN_CS/ (Daniel, Rodrigo)
- Fix copy/pasted wrong documentation
- Rebase
v3: - Rebase
v4: - Don't pass the operation to flushes (Daniel).
--- drivers/gpu/drm/i915/i915_drv.h | 7 +++++++ drivers/gpu/drm/i915/i915_gem.c | 4 ++-- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/intel_drv.h | 3 ++- drivers/gpu/drm/i915/intel_frontbuffer.c | 4 +++- 5 files changed, 15 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +7 -0
@@ -772,6 +772,13 @@ struct intel_context {struct list_head link; }; +enum fb_op_origin { +ORIGIN_GTT , +ORIGIN_CPU , +ORIGIN_CS , +ORIGIN_FLIP , +}; +struct i915_fbc {unsigned long uncompressed_size;unsigned threshold;
drivers/gpu/drm/i915/i915_gem.c +2 -2
@@ -3769,7 +3769,7 @@ i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write) }if (write) -intel_fb_obj_invalidate (obj,NULL ); +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_GTT ); trace_i915_gem_object_change_domain(obj, old_read_domains,
@@ -4084,7 +4084,7 @@ i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write) }if (write) -intel_fb_obj_invalidate (obj,NULL ); +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_CPU ); trace_i915_gem_object_change_domain(obj, old_read_domains,
drivers/gpu/drm/i915/i915_gem_execbuffer.c +1 -1
@@ -971,7 +971,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas, obj->dirty =1 ;i915_gem_request_assign (&obj->last_write_req, req); -intel_fb_obj_invalidate (obj, ring); +intel_fb_obj_invalidate (obj, ring,ORIGIN_CS );/* update for the implicit flush after a batch */ obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS ;
drivers/gpu/drm/i915/intel_drv.h +2 -1
@@ -853,7 +853,8 @@ void intel_ddi_set_vc_payload_alloc(struct drm_crtc *crtc, bool state);/* intel_frontbuffer.c */ void intel_fb_obj_invalidate (struct drm_i915_gem_object *obj, -struct intel_engine_cs *ring); +struct intel_engine_cs *ring, +enum fb_op_origin origin);void intel_frontbuffer_flip_prepare (struct drm_device *dev,unsigned frontbuffer_bits); void intel_frontbuffer_flip_complete(struct drm_device *dev,
drivers/gpu/drm/i915/intel_frontbuffer.c +3 -1
@@ -127,6 +127,7 @@ static void intel_mark_fb_busy(struct drm_device *dev,* intel_fb_obj_invalidate - invalidate frontbuffer object * @obj: GEM object to invalidate * @ring: set for asynchronous rendering +* @origin: which operation caused the invalidation * * This function gets called every time rendering on the given object starts and * frontbuffer caching (fbc, low refresh rate for DRRS, panel self refresh) must
@@ -135,7 +136,8 @@ static void intel_mark_fb_busy(struct drm_device *dev,* scheduled. */ void intel_fb_obj_invalidate (struct drm_i915_gem_object *obj, -struct intel_engine_cs *ring) +struct intel_engine_cs *ring, +enum fb_op_origin origin) {struct drm_device *dev = obj->base.dev;struct drm_i915_private *dev_priv = dev->dev_private;
Signature · 6 lines
2.1.4 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 4/7] drm/i915: pass which operation triggered the frontbuffer tracking
Sorry for the long delay here.
Quoted 118 lines · Paulo Zanoni
From: Paulo Zanoni <[email protected]>
We want to port FBC to the frontbuffer tracking infrastructure, but for that we need to know what caused the object invalidation so we can react accordingly: CPU mmaps need manual, GTT mmaps and flips don't need handling and ring rendering needs nukes.
v2: - s/ORIGIN_RENDER/ORIGIN_CS/ (Daniel, Rodrigo)
- Fix copy/pasted wrong documentation
- Rebase
v3: - Rebase
v4: - Don't pass the operation to flushes (Daniel).Signed-off-by Paulo Zanoni <[email protected]>--- drivers/gpu/drm/i915/i915_drv.h | 7 +++++++ drivers/gpu/drm/i915/i915_gem.c | 4 ++-- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/intel_drv.h | 3 ++- drivers/gpu/drm/i915/intel_frontbuffer.c | 4 +++- 5 files changed, 15 insertions(+), 5 deletions(-)drivers/gpu/drm/i915/i915_drv.h +7
-0@@ -772,6 +772,13 @@ struct intel_context {struct list_head link; }; +enum fb_op_origin { +ORIGIN_GTT , +ORIGIN_CPU , +ORIGIN_CS , +ORIGIN_FLIP , +}; +struct i915_fbc {unsigned long uncompressed_size;unsigned threshold;drivers/gpu/drm/i915/i915_gem.c +2
-2@@ -3769,7 +3769,7 @@ i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write) }if (write) -intel_fb_obj_invalidate (obj,NULL ); +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_GTT ); trace_i915_gem_object_change_domain(obj, old_read_domains,@@ -4084,7 +4084,7 @@ i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write) }if (write) -intel_fb_obj_invalidate (obj,NULL ); +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_CPU ); trace_i915_gem_object_change_domain(obj, old_read_domains,drivers/gpu/drm/i915/i915_gem_execbuffer.c +1
-1@@ -971,7 +971,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas, obj->dirty =1 ;i915_gem_request_assign (&obj->last_write_req, req); -intel_fb_obj_invalidate (obj, ring); +intel_fb_obj_invalidate (obj, ring,ORIGIN_CS );/* update for the implicit flush after a batch */ obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS ;drivers/gpu/drm/i915/intel_drv.h +2
-1@@ -853,7 +853,8 @@ void intel_ddi_set_vc_payload_alloc(struct drm_crtc *crtc, bool state);/* intel_frontbuffer.c */ void intel_fb_obj_invalidate (struct drm_i915_gem_object *obj, -struct intel_engine_cs *ring); +struct intel_engine_cs *ring, +enum fb_op_origin origin);void intel_frontbuffer_flip_prepare (struct drm_device *dev,unsigned frontbuffer_bits); void intel_frontbuffer_flip_complete(struct drm_device *dev,drivers/gpu/drm/i915/intel_frontbuffer.c +3
-2@@ -127,6 +127,7 @@ static void intel_mark_fb_busy(struct drm_device *dev,* intel_fb_obj_invalidate - invalidate frontbuffer object * @obj: GEM object to invalidate * @ring: set for asynchronous rendering +* @origin: which operation caused the invalidation * * This function gets called every time rendering on the given object starts and * frontbuffer caching (fbc, low refresh rate for DRRS, panel self refresh) must @@ -135,7 +136,8 @@ static void intel_mark_fb_busy(struct drm_device *dev,* scheduled. */ void intel_fb_obj_invalidate (struct drm_i915_gem_object *obj, -struct intel_engine_cs *ring) +struct intel_engine_cs *ring, +enum fb_op_origin origin) {struct drm_device *dev = obj->base.dev;struct drm_i915_private *dev_priv = dev->dev_private; --2.1.4
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 6 lines
Rodrigo Vivi Blog: http://blog.vivi.eng.br _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 5/7 drm/i915: also do frontbuffer tracking on pwrites
In reply to Paulo ZanoniFrom: Paulo Zanoni <[email protected]>
We need this for FBC, and possibly for PSR too.
v2: Don't only flush: invalidate too (Daniel).
--- drivers/gpu/drm/i915/i915_gem.c | 23 ++++++++++++++++++----- 1 file changed, 18 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/i915_gem.c +18 -5
@@ -351,7 +351,7 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,struct drm_device *dev = obj->base.dev;void *vaddr = obj->phys_handle->vaddr + args->offset;char __user *user_data =to_user_ptr (args->data_ptr); -int ret; +int ret =0 ;/* We manually control the domain here and pretend that it * remains coherent i.e. in the GTT domain, like shmem_pwrite.
@@ -360,6 +360,7 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,if (ret)return ret; +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_CPU );if (__copy_from_user_inatomic_nocache (vaddr, user_data, args->size)) {unsigned long unwritten;
@@ -370,13 +371,18 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,mutex_unlock (&dev->struct_mutex); unwritten =copy_from_user (vaddr, user_data, args->size);mutex_lock (&dev->struct_mutex); -if (unwritten) -return -EFAULT ; +if (unwritten) { + ret = -EFAULT ; + goto out; + } }drm_clflush_virt_range (vaddr, args->size);i915_gem_chipset_flush (dev); -return 0 ; + +out: +intel_fb_obj_flush (obj, false); +return ret; }void *i915_gem_object_alloc (struct drm_device *dev)
@@ -810,6 +816,8 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev, offset =i915_gem_obj_ggtt_offset (obj) + args->offset; +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_GTT ); +while (remain >0 ) {/* Operation in this page *
@@ -830,7 +838,7 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev,if (fast_user_write (dev_priv->gtt.mappable, page_base, page_offset, user_data, page_length)) { ret = -EFAULT ; - goto out_unpin; + goto out_flush; } remain -= page_length;
@@ -838,6 +846,8 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev, offset += page_length; } +out_flush: +intel_fb_obj_flush (obj, false); out_unpin:i915_gem_object_ggtt_unpin (obj); out:
@@ -952,6 +962,8 @@ i915_gem_shmem_pwrite(struct drm_device *dev,if (ret)return ret; +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_CPU ); +i915_gem_object_pin_pages (obj); offset = args->offset;
@@ -1030,6 +1042,7 @@ out:if (needs_clflush_after)i915_gem_chipset_flush (dev); +intel_fb_obj_flush (obj, false);return ret; }
Signature · 6 lines
2.1.4 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 5/7] drm/i915: also do frontbuffer tracking on pwrites
Quoted 105 lines · Paulo Zanoni
From: Paulo Zanoni <[email protected]>
We need this for FBC, and possibly for PSR too.
v2: Don't only flush: invalidate too (Daniel).
Signed-off-by Paulo Zanoni <[email protected]>--- drivers/gpu/drm/i915/i915_gem.c | 23 ++++++++++++++++++----- 1 file changed, 18 insertions(+), 5 deletions(-)drivers/gpu/drm/i915/i915_gem.c +18
-6@@ -351,7 +351,7 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,struct drm_device *dev = obj->base.dev;void *vaddr = obj->phys_handle->vaddr + args->offset;char __user *user_data =to_user_ptr (args->data_ptr); -int ret; +int ret =0 ;/* We manually control the domain here and pretend that it * remains coherent i.e. in the GTT domain, like shmem_pwrite. @@ -360,6 +360,7 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,if (ret)return ret; +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_CPU );if (__copy_from_user_inatomic_nocache (vaddr, user_data, args->size)) {unsigned long unwritten;@@ -370,13 +371,18 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,mutex_unlock (&dev->struct_mutex); unwritten =copy_from_user (vaddr, user_data, args->size);mutex_lock (&dev->struct_mutex); -if (unwritten) -return -EFAULT ; +if (unwritten) { + ret = -EFAULT ; + goto out; + } }drm_clflush_virt_range (vaddr, args->size);i915_gem_chipset_flush (dev); -return 0 ; + +out: +intel_fb_obj_flush (obj, false); +return ret; }void *i915_gem_object_alloc (struct drm_device *dev)@@ -810,6 +816,8 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev, offset =i915_gem_obj_ggtt_offset (obj) + args->offset; +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_GTT ); +while (remain >0 ) {/* Operation in this page * @@ -830,7 +838,7 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev,if (fast_user_write (dev_priv->gtt.mappable, page_base, page_offset, user_data, page_length)) { ret = -EFAULT ; - goto out_unpin; + goto out_flush; } remain -= page_length;@@ -838,6 +846,8 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev, offset += page_length; } +out_flush: +intel_fb_obj_flush (obj, false); out_unpin:i915_gem_object_ggtt_unpin (obj); out:@@ -952,6 +962,8 @@ i915_gem_shmem_pwrite(struct drm_device *dev,if (ret)return ret; +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_CPU ); +i915_gem_object_pin_pages (obj); offset = args->offset;@@ -1030,6 +1042,7 @@ out:if (needs_clflush_after)i915_gem_chipset_flush (dev); +intel_fb_obj_flush (obj, false);return ret; } --2.1.4
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 6 lines
Rodrigo Vivi Blog: http://blog.vivi.eng.br _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 5/7] drm/i915: also do frontbuffer tracking on pwrites
Quoted 10 lines · Rodrigo Vivi
Reviewed-by Rodrigo Vivi <[email protected]>Quoted 7 lines · Paulo Zanoni
From: Paulo Zanoni <[email protected]>
We need this for FBC, and possibly for PSR too.
v2: Don't only flush: invalidate too (Daniel).
Signed-off-by Paulo Zanoni <[email protected]>
Merged up to this patch, thanks.
-Daniel
Quoted 108 lines
Quoted 98 lines
--- drivers/gpu/drm/i915/i915_gem.c | 23 ++++++++++++++++++----- 1 file changed, 18 insertions(+), 5 deletions(-)drivers/gpu/drm/i915/i915_gem.c +18
-6@@ -351,7 +351,7 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,struct drm_device *dev = obj->base.dev;void *vaddr = obj->phys_handle->vaddr + args->offset;char __user *user_data =to_user_ptr (args->data_ptr); -int ret; +int ret =0 ;/* We manually control the domain here and pretend that it * remains coherent i.e. in the GTT domain, like shmem_pwrite. @@ -360,6 +360,7 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,if (ret)return ret; +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_CPU );if (__copy_from_user_inatomic_nocache (vaddr, user_data, args->size)) {unsigned long unwritten;@@ -370,13 +371,18 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,mutex_unlock (&dev->struct_mutex); unwritten =copy_from_user (vaddr, user_data, args->size);mutex_lock (&dev->struct_mutex); -if (unwritten) -return -EFAULT ; +if (unwritten) { + ret = -EFAULT ; + goto out; + } }drm_clflush_virt_range (vaddr, args->size);i915_gem_chipset_flush (dev); -return 0 ; + +out: +intel_fb_obj_flush (obj, false); +return ret; }void *i915_gem_object_alloc (struct drm_device *dev)@@ -810,6 +816,8 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev, offset =i915_gem_obj_ggtt_offset (obj) + args->offset; +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_GTT ); +while (remain >0 ) {/* Operation in this page * @@ -830,7 +838,7 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev,if (fast_user_write (dev_priv->gtt.mappable, page_base, page_offset, user_data, page_length)) { ret = -EFAULT ; - goto out_unpin; + goto out_flush; } remain -= page_length;@@ -838,6 +846,8 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev, offset += page_length; } +out_flush: +intel_fb_obj_flush (obj, false); out_unpin:i915_gem_object_ggtt_unpin (obj); out:@@ -952,6 +962,8 @@ i915_gem_shmem_pwrite(struct drm_device *dev,if (ret)return ret; +intel_fb_obj_invalidate (obj,NULL ,ORIGIN_CPU ); +i915_gem_object_pin_pages (obj); offset = args->offset;@@ -1030,6 +1042,7 @@ out:if (needs_clflush_after)i915_gem_chipset_flush (dev); +intel_fb_obj_flush (obj, false);return ret; } --2.1.4
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfxSignature · 6 lines
Rodrigo Vivi Blog: http://blog.vivi.eng.br _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 6/7 drm/i915: add frontbuffer tracking to FBC
In reply to Paulo ZanoniFrom: Paulo Zanoni <[email protected]>
Kill the blt/render tracking we currently have and use the frontbuffer tracking infrastructure.
Don't enable things by default yet.
v2: (Rodrigo) Fix small conflict on rebase and typo at subject.
v3: (Paulo) Rebase on RENDER_CS change.
v4: (Paulo) Rebase.
v5: (Paulo) Simplify: flushes don't have origin (Daniel).
Also rebase due to patch order changes.
--- drivers/gpu/drm/i915/i915_drv.h | 10 +--- drivers/gpu/drm/i915/intel_drv.h | 6 ++- drivers/gpu/drm/i915/intel_fbc.c | 91 +++++++++++++++++++------------- drivers/gpu/drm/i915/intel_frontbuffer.c | 14 +---- drivers/gpu/drm/i915/intel_ringbuffer.c | 41 +------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 1 - 6 files changed, 65 insertions(+), 98 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +2 -8
@@ -783,6 +783,8 @@ struct i915_fbc {unsigned long uncompressed_size;unsigned threshold;unsigned int fb_id; +unsigned int possible_framebuffer_bits; +unsigned int busy_bits;struct intel_crtc *crtc;int y;
@@ -795,14 +797,6 @@ struct i915_fbc {* possible. */ bool enabled; -/* On gen8 some rings cannont perform fbc clean operation so for now -* we are doing this on SW with mmio. -* This variable works in the opposite information direction -* of ring->fbc_dirty telling software on frontbuffer tracking -* to perform the cache clean on sw side. -*/ -bool need_sw_cache_clean; -struct intel_fbc_work {struct delayed_work work;struct drm_crtc *crtc;
drivers/gpu/drm/i915/intel_drv.h +5 -1
@@ -1091,7 +1091,11 @@ bool intel_fbc_enabled(struct drm_device *dev);void intel_fbc_update (struct drm_device *dev);void intel_fbc_init (struct drm_i915_private *dev_priv);void intel_fbc_disable (struct drm_device *dev); -void bdw_fbc_sw_flush (struct drm_device *dev,u32 value); +void intel_fbc_invalidate (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits, +enum fb_op_origin origin); +void intel_fbc_flush (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits);/* intel_hdmi.c */ void intel_hdmi_init (struct drm_device *dev,int hdmi_reg,enum port port);
drivers/gpu/drm/i915/intel_fbc.c +54 -37
136 lines not on this page · raw message
drivers/gpu/drm/i915/intel_frontbuffer.c +2 -12
@@ -118,8 +118,6 @@ static void intel_mark_fb_busy(struct drm_device *dev,continue ;intel_increase_pllclock (dev, pipe); -if (ring &&intel_fbc_enabled (dev)) - ring->fbc_dirty = true; } }
@@ -160,6 +158,7 @@ void intel_fb_obj_invalidate(struct drm_i915_gem_object *obj,intel_psr_invalidate (dev, obj->frontbuffer_bits);intel_edp_drrs_invalidate (dev, obj->frontbuffer_bits); +intel_fbc_invalidate (dev_priv, obj->frontbuffer_bits, origin); }/**
@@ -187,16 +186,7 @@ void intel_frontbuffer_flush(struct drm_device *dev,intel_edp_drrs_flush (dev, frontbuffer_bits);intel_psr_flush (dev, frontbuffer_bits); - -/* -* FIXME: Unconditional fbc flushing here is a rather gross hack and -* needs to be reworked into a proper frontbuffer tracking scheme like -* psr employs. -*/ -if (dev_priv->fbc.need_sw_cache_clean) { - dev_priv->fbc.need_sw_cache_clean = false; -bdw_fbc_sw_flush (dev,FBC_REND_CACHE_CLEAN ); - } +intel_fbc_flush (dev_priv, frontbuffer_bits); }/**
drivers/gpu/drm/i915/intel_ringbuffer.c +2 -39
83 lines not on this page · raw message
drivers/gpu/drm/i915/intel_ringbuffer.h +0 -1
@@ -267,7 +267,6 @@ struct intel_engine_cs {*/ struct drm_i915_gem_request *outstanding_lazy_request;bool gpu_caches_dirty; -bool fbc_dirty;wait_queue_head_t irq_queue;
Signature · 6 lines
2.1.4 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 6/7] drm/i915: add frontbuffer tracking to FBC
Quoted 29 lines · Paulo Zanoni
From: Paulo Zanoni <[email protected]>
Kill the blt/render tracking we currently have and use the frontbuffer tracking infrastructure.
Don't enable things by default yet.
v2: (Rodrigo) Fix small conflict on rebase and typo at subject.
v3: (Paulo) Rebase onRENDER_CSchange.
v4: (Paulo) Rebase.
v5: (Paulo) Simplify: flushes don't have origin (Daniel).
Also rebase due to patch order changes.Signed-off-by Rodrigo Vivi <[email protected]>Signed-off-by Paulo Zanoni <[email protected]>--- drivers/gpu/drm/i915/i915_drv.h | 10 +--- drivers/gpu/drm/i915/intel_drv.h | 6 ++- drivers/gpu/drm/i915/intel_fbc.c | 91 +++++++++++++++++++------------- drivers/gpu/drm/i915/intel_frontbuffer.c | 14 +---- drivers/gpu/drm/i915/intel_ringbuffer.c | 41 +------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 1 - 6 files changed, 65 insertions(+), 98 deletions(-)drivers/gpu/drm/i915/i915_drv.h +0
-0
drivers/gpu/drm/i915/i915_drv.h struct i915_fbc { · line 787
Quoted 5 lines
drivers/gpu/drm/i915/i915_drv.h +2
-0@@ -783,6 +783,8 @@ struct i915_fbc {unsigned long uncompressed_size;unsigned threshold;unsigned int fb_id; +unsigned int possible_framebuffer_bits; +unsigned int busy_bits;
I'm not sure if I understood why you keep 2 variables here?
Is it because we keep enabling/disabling fbc with updates all over the code? In this case it is ok we just need to kill it when we kill update_fbc...
Quoted 57 lines
drivers/gpu/drm/i915/i915_drv.h +0
-8struct intel_crtc *crtc;int y;@@ -795,14 +797,6 @@ struct i915_fbc {* possible. */ bool enabled; -/* On gen8 some rings cannont perform fbc clean operation so for now -* we are doing this on SW with mmio. -* This variable works in the opposite information direction -* of ring->fbc_dirty telling software on frontbuffer tracking -* to perform the cache clean on sw side. -*/ -bool need_sw_cache_clean; -struct intel_fbc_work {struct delayed_work work;struct drm_crtc *crtc;drivers/gpu/drm/i915/intel_drv.h +5
-1@@ -1091,7 +1091,11 @@ bool intel_fbc_enabled(struct drm_device *dev);void intel_fbc_update (struct drm_device *dev);void intel_fbc_init (struct drm_i915_private *dev_priv);void intel_fbc_disable (struct drm_device *dev); -void bdw_fbc_sw_flush (struct drm_device *dev,u32 value); +void intel_fbc_invalidate (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits, +enum fb_op_origin origin); +void intel_fbc_flush (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits);/* intel_hdmi.c */ void intel_hdmi_init (struct drm_device *dev,int hdmi_reg,enum port port);drivers/gpu/drm/i915/intel_fbc.c +1
-12@@ -174,29 +174,10 @@ static bool g4x_fbc_enabled(struct drm_device *dev)return I915_READ (DPFC_CONTROL ) &DPFC_CTL_EN ; } -static void snb_fbc_blit_update (struct drm_device *dev) +static void intel_fbc_nuke (struct drm_i915_private *dev_priv) { -struct drm_i915_private *dev_priv = dev->dev_private; -u32 blt_ecoskpd; - -/* Make sure blitter notifies FBC of writes */ - -/* Blitter is part of Media powerwell on VLV. No impact of -* his param in other platforms for now */ -intel_uncore_forcewake_get (dev_priv,FORCEWAKE_MEDIA ); - - blt_ecoskpd =I915_READ (GEN6_BLITTER_ECOSKPD ); - blt_ecoskpd |= GEN6_BLITTER_FBC_NOTIFY <<
drivers/gpu/drm/i915/intel_fbc.c static bool g4x_fbc_enabled(struct drm_device *dev) · line 199
Quoted 10 lines
drivers/gpu/drm/i915/intel_fbc.c +0
-10-GEN6_BLITTER_LOCK_SHIFT ; -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); - blt_ecoskpd |=GEN6_BLITTER_FBC_NOTIFY ; -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); - blt_ecoskpd &= ~(GEN6_BLITTER_FBC_NOTIFY << -GEN6_BLITTER_LOCK_SHIFT ); -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); -POSTING_READ (GEN6_BLITTER_ECOSKPD ); - -intel_uncore_forcewake_put (dev_priv,FORCEWAKE_MEDIA );
Are you sure this continue working on snb? or should we fully remove fbc support for snb and older?
Anyway I believe this could be in a different patch.
Quoted 262 lines
drivers/gpu/drm/i915/intel_fbc.c +53
-15+I915_WRITE (MSG_FBC_REND_STATE ,FBC_REND_NUKE ); +POSTING_READ (MSG_FBC_REND_STATE ); }static void ilk_fbc_enable (struct drm_crtc *crtc)@@ -239,9 +220,10 @@ static void ilk_fbc_enable(struct drm_crtc *crtc)I915_WRITE (SNB_DPFC_CTL_SA ,SNB_CPU_FENCE_ENABLE | obj->fence_reg);I915_WRITE (DPFC_CPU_FENCE_OFFSET , crtc->y); -snb_fbc_blit_update (dev); } +intel_fbc_nuke (dev_priv); +DRM_DEBUG_KMS ("enabled fbc on plane %c\n" ,plane_name (intel_crtc->plane)); }@@ -320,7 +302,7 @@ static void gen7_fbc_enable(struct drm_crtc *crtc) SNB_CPU_FENCE_ENABLE | obj->fence_reg);I915_WRITE (DPFC_CPU_FENCE_OFFSET , crtc->y); -snb_fbc_blit_update (dev); +intel_fbc_nuke (dev_priv);DRM_DEBUG_KMS ("enabled fbc on plane %c\n" ,plane_name (intel_crtc->plane)); }@@ -340,19 +322,6 @@ bool intel_fbc_enabled(struct drm_device *dev)return dev_priv->fbc.enabled; } -void bdw_fbc_sw_flush (struct drm_device *dev,u32 value) -{ -struct drm_i915_private *dev_priv = dev->dev_private; - -if (!IS_GEN8 (dev)) -return ; - -if (!intel_fbc_enabled (dev)) -return ; - -I915_WRITE (MSG_FBC_REND_STATE , value); -} -static void intel_fbc_work_fn (struct work_struct *__work) { struct intel_fbc_work *work =@@ -685,6 +654,44 @@ out_disable:i915_gem_stolen_cleanup_compression (dev); } +void intel_fbc_invalidate (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits, +enum fb_op_origin origin) +{ +struct drm_device *dev = dev_priv->dev; +unsigned int fbc_bits; + +if (origin ==ORIGIN_GTT ) +return ; + +if (dev_priv->fbc.enabled) + fbc_bits =INTEL_FRONTBUFFER_PRIMARY (dev_priv->fbc.crtc->pipe); +else if (dev_priv->fbc.fbc_work) + fbc_bits =INTEL_FRONTBUFFER_PRIMARY ( +to_intel_crtc (dev_priv->fbc.fbc_work->crtc)->pipe); +else + fbc_bits = dev_priv->fbc.possible_framebuffer_bits; + + dev_priv->fbc.busy_bits |= (fbc_bits & frontbuffer_bits); + +if (dev_priv->fbc.busy_bits) +intel_fbc_disable (dev); +} + +void intel_fbc_flush (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits) +{ +struct drm_device *dev = dev_priv->dev; + +if (!dev_priv->fbc.busy_bits) +return ; + + dev_priv->fbc.busy_bits &= ~frontbuffer_bits; + +if (!dev_priv->fbc.busy_bits) +intel_fbc_update (dev); +} +/** * intel_fbc_init - Initialize FBC * @dev_priv: the i915 device @@ -693,12 +700,22 @@ out_disable:*/ void intel_fbc_init (struct drm_i915_private *dev_priv) { +enum pipe pipe; +if (!HAS_FBC (dev_priv)) { dev_priv->fbc.enabled = false; dev_priv->fbc.no_fbc_reason =FBC_UNSUPPORTED ;return ; } +for_each_pipe (dev_priv, pipe) { + dev_priv->fbc.possible_framebuffer_bits |= +INTEL_FRONTBUFFER_PRIMARY (pipe); + +if (IS_HASWELL (dev_priv) ||INTEL_INFO (dev_priv)->gen >=8 ) +break ; + } +if (INTEL_INFO (dev_priv)->gen >=7 ) { dev_priv->display.fbc_enabled = ilk_fbc_enabled; dev_priv->display.enable_fbc = gen7_fbc_enable;drivers/gpu/drm/i915/intel_frontbuffer.c +2
-12@@ -118,8 +118,6 @@ static void intel_mark_fb_busy(struct drm_device *dev,continue ;intel_increase_pllclock (dev, pipe); -if (ring &&intel_fbc_enabled (dev)) - ring->fbc_dirty = true; } }@@ -160,6 +158,7 @@ void intel_fb_obj_invalidate(struct drm_i915_gem_object *obj,intel_psr_invalidate (dev, obj->frontbuffer_bits);intel_edp_drrs_invalidate (dev, obj->frontbuffer_bits); +intel_fbc_invalidate (dev_priv, obj->frontbuffer_bits, origin); }/** @@ -187,16 +186,7 @@ void intel_frontbuffer_flush(struct drm_device *dev,intel_edp_drrs_flush (dev, frontbuffer_bits);intel_psr_flush (dev, frontbuffer_bits); - -/* -* FIXME: Unconditional fbc flushing here is a rather gross hack and -* needs to be reworked into a proper frontbuffer tracking scheme like -* psr employs. -*/ -if (dev_priv->fbc.need_sw_cache_clean) { - dev_priv->fbc.need_sw_cache_clean = false; -bdw_fbc_sw_flush (dev,FBC_REND_CACHE_CLEAN ); - } +intel_fbc_flush (dev_priv, frontbuffer_bits); }/** drivers/gpu/drm/i915/intel_ringbuffer.c +2
-39@@ -317,29 +317,6 @@ gen7_render_ring_cs_stall_wa(struct intel_engine_cs *ring)return 0 ; } -static int gen7_ring_fbc_flush (struct intel_engine_cs *ring,u32 value) -{ -int ret; - -if (!ring->fbc_dirty) -return 0 ; - - ret =intel_ring_begin (ring,6 ); -if (ret) -return ret; -/* WaFbcNukeOn3DBlt:ivb/hsw */ -intel_ring_emit (ring,MI_LOAD_REGISTER_IMM (1 )); -intel_ring_emit (ring,MSG_FBC_REND_STATE ); -intel_ring_emit (ring, value); -intel_ring_emit (ring,MI_STORE_REGISTER_MEM (1 ) |MI_SRM_LRM_GLOBAL_GTT ); -intel_ring_emit (ring,MSG_FBC_REND_STATE ); -intel_ring_emit (ring, ring->scratch.gtt_offset +256 ); -intel_ring_advance (ring); - - ring->fbc_dirty = false; -return 0 ; -} -static int gen7_render_ring_flush (struct intel_engine_cs *ring,u32 invalidate_domains,u32 flush_domains)@@ -398,9 +375,6 @@ gen7_render_ring_flush(struct intel_engine_cs *ring,intel_ring_emit (ring,0 );intel_ring_advance (ring); -if (!invalidate_domains && flush_domains) -return gen7_ring_fbc_flush (ring,FBC_REND_NUKE ); -return 0 ; }@@ -462,9 +436,6 @@ gen8_render_ring_flush(struct intel_engine_cs *ring,if (ret)return ret; -if (!invalidate_domains && flush_domains) -return gen7_ring_fbc_flush (ring,FBC_REND_NUKE ); -return 0 ; }@@ -2420,7 +2391,6 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, u32 invalidate, u32 flush) {struct drm_device *dev = ring->dev; -struct drm_i915_private *dev_priv = dev->dev_private;uint32_t cmd;int ret;@@ -2429,7 +2399,7 @@ static int gen6_ring_flush(struct intel_engine_cs *ring,return ret; cmd =MI_FLUSH_DW ; -if (INTEL_INFO (ring->dev)->gen >=8 ) +if (INTEL_INFO (dev)->gen >=8 ) cmd +=1 ;/* We always require a command barrier so that subsequent @@ -2449,7 +2419,7 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, cmd |=MI_INVALIDATE_TLB ;intel_ring_emit (ring, cmd);intel_ring_emit (ring,I915_GEM_HWS_SCRATCH_ADDR |MI_FLUSH_DW_USE_GTT ); -if (INTEL_INFO (ring->dev)->gen >=8 ) { +if (INTEL_INFO (dev)->gen >=8 ) {intel_ring_emit (ring,0 );/* upper addr */ intel_ring_emit (ring,0 );/* value */ }else {@@ -2458,13 +2428,6 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, }intel_ring_advance (ring); -if (!invalidate && flush) { -if (IS_GEN7 (dev)) -return gen7_ring_fbc_flush (ring,FBC_REND_CACHE_CLEAN ); -else if (IS_BROADWELL (dev)) - dev_priv->fbc.need_sw_cache_clean = true; - } -return 0 ; }drivers/gpu/drm/i915/intel_ringbuffer.h +0
-2@@ -267,7 +267,6 @@ struct intel_engine_cs {*/ struct drm_i915_gem_request *outstanding_lazy_request;bool gpu_caches_dirty; -bool fbc_dirty;wait_queue_head_t irq_queue; --2.1.4
Quoted 4 lines
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Thanks,
Signature · 6 lines
Rodrigo Vivi Blog: http://blog.vivi.eng.br _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 6/7] drm/i915: add frontbuffer tracking to FBC
2015-03-03 21:57 GMT-03:00 Rodrigo Vivi <[email protected]>:
Quoted 25 lines · Paulo Zanoni
Quoted 24 lines
From: Paulo Zanoni <[email protected]>
Kill the blt/render tracking we currently have and use the frontbuffer tracking infrastructure.
Don't enable things by default yet.
v2: (Rodrigo) Fix small conflict on rebase and typo at subject.
v3: (Paulo) Rebase onRENDER_CSchange.
v4: (Paulo) Rebase.
v5: (Paulo) Simplify: flushes don't have origin (Daniel).
Also rebase due to patch order changes.Signed-off-by Rodrigo Vivi <[email protected]>Signed-off-by Paulo Zanoni <[email protected]>--- drivers/gpu/drm/i915/i915_drv.h | 10 +--- drivers/gpu/drm/i915/intel_drv.h | 6 ++- drivers/gpu/drm/i915/intel_fbc.c | 91 +++++++++++++++++++------------- drivers/gpu/drm/i915/intel_frontbuffer.c | 14 +---- drivers/gpu/drm/i915/intel_ringbuffer.c | 41 +------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 1 - 6 files changed, 65 insertions(+), 98 deletions(-)
Quoted 12 lines
drivers/gpu/drm/i915/i915_drv.h struct i915_fbc { · line 787
Quoted 10 lines
drivers/gpu/drm/i915/i915_drv.h +2
-0@@ -783,6 +783,8 @@ struct i915_fbc {unsigned long uncompressed_size;unsigned threshold;unsigned int fb_id; +unsigned int possible_framebuffer_bits; +unsigned int busy_bits;I'm not sure if I understood why you keep 2 variables here?
After Daniel's last suggestion, the possible_framebuffer_bits could probably be removed and left as a local variable at intel_fbc_validate
now: it's just a matter of coding style. Some platforms support FBC on more than just pipe A, so this variable is used to simplify checks related to this.
The busy_bits was also part of Daniel's last request: we need to store which bits triggered intel_fbc_invalidate calls, and then check them when intel_fbc_flush is called.
Quoted 2 lines
Is it because we keep enabling/disabling fbc with updates all over the code? In this case it is ok we just need to kill it when we kill
update_fbc...
I don't understand what you mean here. Please clarify.
Quoted 72 lines
Quoted 58 lines
drivers/gpu/drm/i915/i915_drv.h +0
-8struct intel_crtc *crtc;int y;@@ -795,14 +797,6 @@ struct i915_fbc {* possible. */ bool enabled; -/* On gen8 some rings cannont perform fbc clean operation so for now -* we are doing this on SW with mmio. -* This variable works in the opposite information direction -* of ring->fbc_dirty telling software on frontbuffer tracking -* to perform the cache clean on sw side. -*/ -bool need_sw_cache_clean; -struct intel_fbc_work {struct delayed_work work;struct drm_crtc *crtc;drivers/gpu/drm/i915/intel_drv.h +5
-1@@ -1091,7 +1091,11 @@ bool intel_fbc_enabled(struct drm_device *dev);void intel_fbc_update (struct drm_device *dev);void intel_fbc_init (struct drm_i915_private *dev_priv);void intel_fbc_disable (struct drm_device *dev); -void bdw_fbc_sw_flush (struct drm_device *dev,u32 value); +void intel_fbc_invalidate (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits, +enum fb_op_origin origin); +void intel_fbc_flush (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits);/* intel_hdmi.c */ void intel_hdmi_init (struct drm_device *dev,int hdmi_reg,enum port port);drivers/gpu/drm/i915/intel_fbc.c +1
-12@@ -174,29 +174,10 @@ static bool g4x_fbc_enabled(struct drm_device *dev)return I915_READ (DPFC_CONTROL ) &DPFC_CTL_EN ; } -static void snb_fbc_blit_update (struct drm_device *dev) +static void intel_fbc_nuke (struct drm_i915_private *dev_priv) { -struct drm_i915_private *dev_priv = dev->dev_private; -u32 blt_ecoskpd; - -/* Make sure blitter notifies FBC of writes */ - -/* Blitter is part of Media powerwell on VLV. No impact of -* his param in other platforms for now */ -intel_uncore_forcewake_get (dev_priv,FORCEWAKE_MEDIA ); - - blt_ecoskpd =I915_READ (GEN6_BLITTER_ECOSKPD ); - blt_ecoskpd |= GEN6_BLITTER_FBC_NOTIFY <<drivers/gpu/drm/i915/intel_fbc.c static bool g4x_fbc_enabled(struct drm_device *dev) · line 199
Quoted 10 lines
drivers/gpu/drm/i915/intel_fbc.c +0
-10-GEN6_BLITTER_LOCK_SHIFT ; -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); - blt_ecoskpd |=GEN6_BLITTER_FBC_NOTIFY ; -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); - blt_ecoskpd &= ~(GEN6_BLITTER_FBC_NOTIFY << -GEN6_BLITTER_LOCK_SHIFT ); -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); -POSTING_READ (GEN6_BLITTER_ECOSKPD ); - -intel_uncore_forcewake_put (dev_priv,FORCEWAKE_MEDIA );Are you sure this continue working on snb? or should we fully remove fbc support for snb and older?
Quoted 1 line
Anyway I believe this could be in a different patch.
Well, the whole idea of using the sofware-based frontbuffer tracking mechanism is to get rid of all the hardware-based stuff.
Quoted 273 lines
Quoted 263 lines
drivers/gpu/drm/i915/intel_fbc.c +53
-15+I915_WRITE (MSG_FBC_REND_STATE ,FBC_REND_NUKE ); +POSTING_READ (MSG_FBC_REND_STATE ); }static void ilk_fbc_enable (struct drm_crtc *crtc)@@ -239,9 +220,10 @@ static void ilk_fbc_enable(struct drm_crtc *crtc)I915_WRITE (SNB_DPFC_CTL_SA ,SNB_CPU_FENCE_ENABLE | obj->fence_reg);I915_WRITE (DPFC_CPU_FENCE_OFFSET , crtc->y); -snb_fbc_blit_update (dev); } +intel_fbc_nuke (dev_priv); +DRM_DEBUG_KMS ("enabled fbc on plane %c\n" ,plane_name (intel_crtc->plane)); }@@ -320,7 +302,7 @@ static void gen7_fbc_enable(struct drm_crtc *crtc) SNB_CPU_FENCE_ENABLE | obj->fence_reg);I915_WRITE (DPFC_CPU_FENCE_OFFSET , crtc->y); -snb_fbc_blit_update (dev); +intel_fbc_nuke (dev_priv);DRM_DEBUG_KMS ("enabled fbc on plane %c\n" ,plane_name (intel_crtc->plane)); }@@ -340,19 +322,6 @@ bool intel_fbc_enabled(struct drm_device *dev)return dev_priv->fbc.enabled; } -void bdw_fbc_sw_flush (struct drm_device *dev,u32 value) -{ -struct drm_i915_private *dev_priv = dev->dev_private; - -if (!IS_GEN8 (dev)) -return ; - -if (!intel_fbc_enabled (dev)) -return ; - -I915_WRITE (MSG_FBC_REND_STATE , value); -} -static void intel_fbc_work_fn (struct work_struct *__work) { struct intel_fbc_work *work =@@ -685,6 +654,44 @@ out_disable:i915_gem_stolen_cleanup_compression (dev); } +void intel_fbc_invalidate (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits, +enum fb_op_origin origin) +{ +struct drm_device *dev = dev_priv->dev; +unsigned int fbc_bits; + +if (origin ==ORIGIN_GTT ) +return ; + +if (dev_priv->fbc.enabled) + fbc_bits =INTEL_FRONTBUFFER_PRIMARY (dev_priv->fbc.crtc->pipe); +else if (dev_priv->fbc.fbc_work) + fbc_bits =INTEL_FRONTBUFFER_PRIMARY ( +to_intel_crtc (dev_priv->fbc.fbc_work->crtc)->pipe); +else + fbc_bits = dev_priv->fbc.possible_framebuffer_bits; + + dev_priv->fbc.busy_bits |= (fbc_bits & frontbuffer_bits); + +if (dev_priv->fbc.busy_bits) +intel_fbc_disable (dev); +} + +void intel_fbc_flush (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits) +{ +struct drm_device *dev = dev_priv->dev; + +if (!dev_priv->fbc.busy_bits) +return ; + + dev_priv->fbc.busy_bits &= ~frontbuffer_bits; + +if (!dev_priv->fbc.busy_bits) +intel_fbc_update (dev); +} +/** * intel_fbc_init - Initialize FBC * @dev_priv: the i915 device @@ -693,12 +700,22 @@ out_disable:*/ void intel_fbc_init (struct drm_i915_private *dev_priv) { +enum pipe pipe; +if (!HAS_FBC (dev_priv)) { dev_priv->fbc.enabled = false; dev_priv->fbc.no_fbc_reason =FBC_UNSUPPORTED ;return ; } +for_each_pipe (dev_priv, pipe) { + dev_priv->fbc.possible_framebuffer_bits |= +INTEL_FRONTBUFFER_PRIMARY (pipe); + +if (IS_HASWELL (dev_priv) ||INTEL_INFO (dev_priv)->gen >=8 ) +break ; + } +if (INTEL_INFO (dev_priv)->gen >=7 ) { dev_priv->display.fbc_enabled = ilk_fbc_enabled; dev_priv->display.enable_fbc = gen7_fbc_enable;drivers/gpu/drm/i915/intel_frontbuffer.c +2
-12@@ -118,8 +118,6 @@ static void intel_mark_fb_busy(struct drm_device *dev,continue ;intel_increase_pllclock (dev, pipe); -if (ring &&intel_fbc_enabled (dev)) - ring->fbc_dirty = true; } }@@ -160,6 +158,7 @@ void intel_fb_obj_invalidate(struct drm_i915_gem_object *obj,intel_psr_invalidate (dev, obj->frontbuffer_bits);intel_edp_drrs_invalidate (dev, obj->frontbuffer_bits); +intel_fbc_invalidate (dev_priv, obj->frontbuffer_bits, origin); }/** @@ -187,16 +186,7 @@ void intel_frontbuffer_flush(struct drm_device *dev,intel_edp_drrs_flush (dev, frontbuffer_bits);intel_psr_flush (dev, frontbuffer_bits); - -/* -* FIXME: Unconditional fbc flushing here is a rather gross hack and -* needs to be reworked into a proper frontbuffer tracking scheme like -* psr employs. -*/ -if (dev_priv->fbc.need_sw_cache_clean) { - dev_priv->fbc.need_sw_cache_clean = false; -bdw_fbc_sw_flush (dev,FBC_REND_CACHE_CLEAN ); - } +intel_fbc_flush (dev_priv, frontbuffer_bits); }/** drivers/gpu/drm/i915/intel_ringbuffer.c +2
-39@@ -317,29 +317,6 @@ gen7_render_ring_cs_stall_wa(struct intel_engine_cs *ring)return 0 ; } -static int gen7_ring_fbc_flush (struct intel_engine_cs *ring,u32 value) -{ -int ret; - -if (!ring->fbc_dirty) -return 0 ; - - ret =intel_ring_begin (ring,6 ); -if (ret) -return ret; -/* WaFbcNukeOn3DBlt:ivb/hsw */ -intel_ring_emit (ring,MI_LOAD_REGISTER_IMM (1 )); -intel_ring_emit (ring,MSG_FBC_REND_STATE ); -intel_ring_emit (ring, value); -intel_ring_emit (ring,MI_STORE_REGISTER_MEM (1 ) |MI_SRM_LRM_GLOBAL_GTT ); -intel_ring_emit (ring,MSG_FBC_REND_STATE ); -intel_ring_emit (ring, ring->scratch.gtt_offset +256 ); -intel_ring_advance (ring); - - ring->fbc_dirty = false; -return 0 ; -} -static int gen7_render_ring_flush (struct intel_engine_cs *ring,u32 invalidate_domains,u32 flush_domains)@@ -398,9 +375,6 @@ gen7_render_ring_flush(struct intel_engine_cs *ring,intel_ring_emit (ring,0 );intel_ring_advance (ring); -if (!invalidate_domains && flush_domains) -return gen7_ring_fbc_flush (ring,FBC_REND_NUKE ); -return 0 ; }@@ -462,9 +436,6 @@ gen8_render_ring_flush(struct intel_engine_cs *ring,if (ret)return ret; -if (!invalidate_domains && flush_domains) -return gen7_ring_fbc_flush (ring,FBC_REND_NUKE ); -return 0 ; }@@ -2420,7 +2391,6 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, u32 invalidate, u32 flush) {struct drm_device *dev = ring->dev; -struct drm_i915_private *dev_priv = dev->dev_private;uint32_t cmd;int ret;@@ -2429,7 +2399,7 @@ static int gen6_ring_flush(struct intel_engine_cs *ring,return ret; cmd =MI_FLUSH_DW ; -if (INTEL_INFO (ring->dev)->gen >=8 ) +if (INTEL_INFO (dev)->gen >=8 ) cmd +=1 ;/* We always require a command barrier so that subsequent @@ -2449,7 +2419,7 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, cmd |=MI_INVALIDATE_TLB ;intel_ring_emit (ring, cmd);intel_ring_emit (ring,I915_GEM_HWS_SCRATCH_ADDR |MI_FLUSH_DW_USE_GTT ); -if (INTEL_INFO (ring->dev)->gen >=8 ) { +if (INTEL_INFO (dev)->gen >=8 ) {intel_ring_emit (ring,0 );/* upper addr */ intel_ring_emit (ring,0 );/* value */ }else {@@ -2458,13 +2428,6 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, }intel_ring_advance (ring); -if (!invalidate && flush) { -if (IS_GEN7 (dev)) -return gen7_ring_fbc_flush (ring,FBC_REND_CACHE_CLEAN ); -else if (IS_BROADWELL (dev)) - dev_priv->fbc.need_sw_cache_clean = true; - } -return 0 ; }drivers/gpu/drm/i915/intel_ringbuffer.h +0
-2@@ -267,7 +267,6 @@ struct intel_engine_cs {*/ struct drm_i915_gem_request *outstanding_lazy_request;bool gpu_caches_dirty; -bool fbc_dirty;wait_queue_head_t irq_queue; --2.1.4
Quoted 4 lines
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfxThanks,
--
Rodrigo Vivi
Blog: http://blog.vivi.eng.br
Signature · 5 lines
Paulo Zanoni _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 6/7] drm/i915: add frontbuffer tracking to FBC
Quoted 45 lines · Paulo Zanoni
2015-03-03 21:57 GMT-03:00 Rodrigo Vivi <[email protected]>:
Quoted 25 lines · Paulo Zanoni
Quoted 24 lines
From: Paulo Zanoni <[email protected]>
Kill the blt/render tracking we currently have and use the frontbuffer tracking infrastructure.
Don't enable things by default yet.
v2: (Rodrigo) Fix small conflict on rebase and typo at subject.
v3: (Paulo) Rebase onRENDER_CSchange.
v4: (Paulo) Rebase.
v5: (Paulo) Simplify: flushes don't have origin (Daniel).
Also rebase due to patch order changes.Signed-off-by Rodrigo Vivi <[email protected]>Signed-off-by Paulo Zanoni <[email protected]>--- drivers/gpu/drm/i915/i915_drv.h | 10 +--- drivers/gpu/drm/i915/intel_drv.h | 6 ++- drivers/gpu/drm/i915/intel_fbc.c | 91 +++++++++++++++++++------------- drivers/gpu/drm/i915/intel_frontbuffer.c | 14 +---- drivers/gpu/drm/i915/intel_ringbuffer.c | 41 +------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 1 - 6 files changed, 65 insertions(+), 98 deletions(-)Quoted 12 lines
drivers/gpu/drm/i915/i915_drv.h struct i915_fbc { · line 787
Quoted 10 lines
drivers/gpu/drm/i915/i915_drv.h +2
-0@@ -783,6 +783,8 @@ struct i915_fbc {unsigned long uncompressed_size;unsigned threshold;unsigned int fb_id; +unsigned int possible_framebuffer_bits; +unsigned int busy_bits;I'm not sure if I understood why you keep 2 variables here?
After Daniel's last suggestion, the
possible_framebuffer_bitscould probably be removed and left as a local variable atintel_fbc_validate
now: it's just a matter of coding style. Some platforms support FBC on more than just pipe A, so this variable is used to simplify checks related to this.
Quoted 3 lines
The
busy_bitswas also part of Daniel's last request: we need to store which bits triggeredintel_fbc_invalidatecalls, and then check them whenintel_fbc_flushis called.
Got it.
Thanks for explaining it.
Quoted 5 lines
Quoted 3 lines
Is it because we keep enabling/disabling fbc with updates all over the code? In this case it is ok we just need to kill it when we kill
update_fbc...I don't understand what you mean here. Please clarify.
I had missunderstood the reason of possible_frontbuffer bits, sorry. But about killing update_fbc I believe that after frontbuffer rework is in place we don't need to use that update_fbc function that disable and enable fbc on the cases we knew it could fail. With frontbuffer bits we could let it always enabled and kicking with frontbuffer bits.
Quoted 73 lines
Quoted 73 lines
Quoted 59 lines
drivers/gpu/drm/i915/i915_drv.h +0
-8struct intel_crtc *crtc;int y;@@ -795,14 +797,6 @@ struct i915_fbc {* possible. */ bool enabled; -/* On gen8 some rings cannont perform fbc clean operation so for now -* we are doing this on SW with mmio. -* This variable works in the opposite information direction -* of ring->fbc_dirty telling software on frontbuffer tracking -* to perform the cache clean on sw side. -*/ -bool need_sw_cache_clean; -struct intel_fbc_work {struct delayed_work work;struct drm_crtc *crtc;drivers/gpu/drm/i915/intel_drv.h +5
-1@@ -1091,7 +1091,11 @@ bool intel_fbc_enabled(struct drm_device *dev);void intel_fbc_update (struct drm_device *dev);void intel_fbc_init (struct drm_i915_private *dev_priv);void intel_fbc_disable (struct drm_device *dev); -void bdw_fbc_sw_flush (struct drm_device *dev,u32 value); +void intel_fbc_invalidate (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits, +enum fb_op_origin origin); +void intel_fbc_flush (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits);/* intel_hdmi.c */ void intel_hdmi_init (struct drm_device *dev,int hdmi_reg,enum port port);drivers/gpu/drm/i915/intel_fbc.c +1
-12@@ -174,29 +174,10 @@ static bool g4x_fbc_enabled(struct drm_device *dev)return I915_READ (DPFC_CONTROL ) &DPFC_CTL_EN ; } -static void snb_fbc_blit_update (struct drm_device *dev) +static void intel_fbc_nuke (struct drm_i915_private *dev_priv) { -struct drm_i915_private *dev_priv = dev->dev_private; -u32 blt_ecoskpd; - -/* Make sure blitter notifies FBC of writes */ - -/* Blitter is part of Media powerwell on VLV. No impact of -* his param in other platforms for now */ -intel_uncore_forcewake_get (dev_priv,FORCEWAKE_MEDIA ); - - blt_ecoskpd =I915_READ (GEN6_BLITTER_ECOSKPD ); - blt_ecoskpd |= GEN6_BLITTER_FBC_NOTIFY <<drivers/gpu/drm/i915/intel_fbc.c static bool g4x_fbc_enabled(struct drm_device *dev) · line 199
Quoted 10 lines
drivers/gpu/drm/i915/intel_fbc.c +0
-10-GEN6_BLITTER_LOCK_SHIFT ; -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); - blt_ecoskpd |=GEN6_BLITTER_FBC_NOTIFY ; -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); - blt_ecoskpd &= ~(GEN6_BLITTER_FBC_NOTIFY << -GEN6_BLITTER_LOCK_SHIFT ); -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); -POSTING_READ (GEN6_BLITTER_ECOSKPD ); - -intel_uncore_forcewake_put (dev_priv,FORCEWAKE_MEDIA );Are you sure this continue working on snb? or should we fully remove fbc support for snb and older?
Quoted 4 lines
Quoted 1 line
Anyway I believe this could be in a different patch.
Well, the whole idea of using the sofware-based frontbuffer tracking mechanism is to get rid of all the hardware-based stuff.
Hm, I still have a concern here. We would need to test on those platforms to be certain. The reason for that is that HW tracking still works so it can let the buffer in a stage that compression is unable to start... same reason that we still need both lines below on BDW.
Anyway this wouldn't break anything and it this rework was a great improvement so feel free to use
and we continue any other necessary improvement or fix on top.
Quoted 271 lines
Quoted 271 lines
Quoted 264 lines
drivers/gpu/drm/i915/intel_fbc.c +53
-15+I915_WRITE (MSG_FBC_REND_STATE ,FBC_REND_NUKE ); +POSTING_READ (MSG_FBC_REND_STATE ); }static void ilk_fbc_enable (struct drm_crtc *crtc)@@ -239,9 +220,10 @@ static void ilk_fbc_enable(struct drm_crtc *crtc)I915_WRITE (SNB_DPFC_CTL_SA ,SNB_CPU_FENCE_ENABLE | obj->fence_reg);I915_WRITE (DPFC_CPU_FENCE_OFFSET , crtc->y); -snb_fbc_blit_update (dev); } +intel_fbc_nuke (dev_priv); +DRM_DEBUG_KMS ("enabled fbc on plane %c\n" ,plane_name (intel_crtc->plane)); }@@ -320,7 +302,7 @@ static void gen7_fbc_enable(struct drm_crtc *crtc) SNB_CPU_FENCE_ENABLE | obj->fence_reg);I915_WRITE (DPFC_CPU_FENCE_OFFSET , crtc->y); -snb_fbc_blit_update (dev); +intel_fbc_nuke (dev_priv);DRM_DEBUG_KMS ("enabled fbc on plane %c\n" ,plane_name (intel_crtc->plane)); }@@ -340,19 +322,6 @@ bool intel_fbc_enabled(struct drm_device *dev)return dev_priv->fbc.enabled; } -void bdw_fbc_sw_flush (struct drm_device *dev,u32 value) -{ -struct drm_i915_private *dev_priv = dev->dev_private; - -if (!IS_GEN8 (dev)) -return ; - -if (!intel_fbc_enabled (dev)) -return ; - -I915_WRITE (MSG_FBC_REND_STATE , value); -} -static void intel_fbc_work_fn (struct work_struct *__work) { struct intel_fbc_work *work =@@ -685,6 +654,44 @@ out_disable:i915_gem_stolen_cleanup_compression (dev); } +void intel_fbc_invalidate (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits, +enum fb_op_origin origin) +{ +struct drm_device *dev = dev_priv->dev; +unsigned int fbc_bits; + +if (origin ==ORIGIN_GTT ) +return ; + +if (dev_priv->fbc.enabled) + fbc_bits =INTEL_FRONTBUFFER_PRIMARY (dev_priv->fbc.crtc->pipe); +else if (dev_priv->fbc.fbc_work) + fbc_bits =INTEL_FRONTBUFFER_PRIMARY ( +to_intel_crtc (dev_priv->fbc.fbc_work->crtc)->pipe); +else + fbc_bits = dev_priv->fbc.possible_framebuffer_bits; + + dev_priv->fbc.busy_bits |= (fbc_bits & frontbuffer_bits); + +if (dev_priv->fbc.busy_bits) +intel_fbc_disable (dev); +} + +void intel_fbc_flush (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits) +{ +struct drm_device *dev = dev_priv->dev; + +if (!dev_priv->fbc.busy_bits) +return ; + + dev_priv->fbc.busy_bits &= ~frontbuffer_bits; + +if (!dev_priv->fbc.busy_bits) +intel_fbc_update (dev); +} +/** * intel_fbc_init - Initialize FBC * @dev_priv: the i915 device @@ -693,12 +700,22 @@ out_disable:*/ void intel_fbc_init (struct drm_i915_private *dev_priv) { +enum pipe pipe; +if (!HAS_FBC (dev_priv)) { dev_priv->fbc.enabled = false; dev_priv->fbc.no_fbc_reason =FBC_UNSUPPORTED ;return ; } +for_each_pipe (dev_priv, pipe) { + dev_priv->fbc.possible_framebuffer_bits |= +INTEL_FRONTBUFFER_PRIMARY (pipe); + +if (IS_HASWELL (dev_priv) ||INTEL_INFO (dev_priv)->gen >=8 ) +break ; + } +if (INTEL_INFO (dev_priv)->gen >=7 ) { dev_priv->display.fbc_enabled = ilk_fbc_enabled; dev_priv->display.enable_fbc = gen7_fbc_enable;drivers/gpu/drm/i915/intel_frontbuffer.c +2
-12@@ -118,8 +118,6 @@ static void intel_mark_fb_busy(struct drm_device *dev,continue ;intel_increase_pllclock (dev, pipe); -if (ring &&intel_fbc_enabled (dev)) - ring->fbc_dirty = true; } }@@ -160,6 +158,7 @@ void intel_fb_obj_invalidate(struct drm_i915_gem_object *obj,intel_psr_invalidate (dev, obj->frontbuffer_bits);intel_edp_drrs_invalidate (dev, obj->frontbuffer_bits); +intel_fbc_invalidate (dev_priv, obj->frontbuffer_bits, origin); }/** @@ -187,16 +186,7 @@ void intel_frontbuffer_flush(struct drm_device *dev,intel_edp_drrs_flush (dev, frontbuffer_bits);intel_psr_flush (dev, frontbuffer_bits); - -/* -* FIXME: Unconditional fbc flushing here is a rather gross hack and -* needs to be reworked into a proper frontbuffer tracking scheme like -* psr employs. -*/ -if (dev_priv->fbc.need_sw_cache_clean) { - dev_priv->fbc.need_sw_cache_clean = false; -bdw_fbc_sw_flush (dev,FBC_REND_CACHE_CLEAN ); - } +intel_fbc_flush (dev_priv, frontbuffer_bits); }/** drivers/gpu/drm/i915/intel_ringbuffer.c +2
-39@@ -317,29 +317,6 @@ gen7_render_ring_cs_stall_wa(struct intel_engine_cs *ring)return 0 ; } -static int gen7_ring_fbc_flush (struct intel_engine_cs *ring,u32 value) -{ -int ret; - -if (!ring->fbc_dirty) -return 0 ; - - ret =intel_ring_begin (ring,6 ); -if (ret) -return ret; -/* WaFbcNukeOn3DBlt:ivb/hsw */ -intel_ring_emit (ring,MI_LOAD_REGISTER_IMM (1 )); -intel_ring_emit (ring,MSG_FBC_REND_STATE ); -intel_ring_emit (ring, value); -intel_ring_emit (ring,MI_STORE_REGISTER_MEM (1 ) |MI_SRM_LRM_GLOBAL_GTT ); -intel_ring_emit (ring,MSG_FBC_REND_STATE ); -intel_ring_emit (ring, ring->scratch.gtt_offset +256 ); -intel_ring_advance (ring); - - ring->fbc_dirty = false; -return 0 ; -} -static int gen7_render_ring_flush (struct intel_engine_cs *ring,u32 invalidate_domains,u32 flush_domains)@@ -398,9 +375,6 @@ gen7_render_ring_flush(struct intel_engine_cs *ring,intel_ring_emit (ring,0 );intel_ring_advance (ring); -if (!invalidate_domains && flush_domains) -return gen7_ring_fbc_flush (ring,FBC_REND_NUKE ); -return 0 ; }@@ -462,9 +436,6 @@ gen8_render_ring_flush(struct intel_engine_cs *ring,if (ret)return ret; -if (!invalidate_domains && flush_domains) -return gen7_ring_fbc_flush (ring,FBC_REND_NUKE ); -return 0 ; }@@ -2420,7 +2391,6 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, u32 invalidate, u32 flush) {struct drm_device *dev = ring->dev; -struct drm_i915_private *dev_priv = dev->dev_private;uint32_t cmd;int ret;@@ -2429,7 +2399,7 @@ static int gen6_ring_flush(struct intel_engine_cs *ring,return ret; cmd =MI_FLUSH_DW ; -if (INTEL_INFO (ring->dev)->gen >=8 ) +if (INTEL_INFO (dev)->gen >=8 ) cmd +=1 ;/* We always require a command barrier so that subsequent @@ -2449,7 +2419,7 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, cmd |=MI_INVALIDATE_TLB ;intel_ring_emit (ring, cmd);intel_ring_emit (ring,I915_GEM_HWS_SCRATCH_ADDR |MI_FLUSH_DW_USE_GTT ); -if (INTEL_INFO (ring->dev)->gen >=8 ) { +if (INTEL_INFO (dev)->gen >=8 ) {intel_ring_emit (ring,0 );/* upper addr */ intel_ring_emit (ring,0 );/* value */ }else {@@ -2458,13 +2428,6 @@ static int gen6_ring_flush(struct intel_engine_cs *ring, }intel_ring_advance (ring); -if (!invalidate && flush) { -if (IS_GEN7 (dev)) -return gen7_ring_fbc_flush (ring,FBC_REND_CACHE_CLEAN ); -else if (IS_BROADWELL (dev)) - dev_priv->fbc.need_sw_cache_clean = true; - } -return 0 ; }drivers/gpu/drm/i915/intel_ringbuffer.h +0
-2@@ -267,7 +267,6 @@ struct intel_engine_cs {*/ struct drm_i915_gem_request *outstanding_lazy_request;bool gpu_caches_dirty; -bool fbc_dirty;wait_queue_head_t irq_queue; --2.1.4
Quoted 4 lines
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfxThanks,
Quoted 6 lines
Quoted 3 lines
--
Rodrigo Vivi
Blog: http://blog.vivi.eng.br
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Daniel Vetter
AppliedRe: [PATCH 6/7] drm/i915: add frontbuffer tracking to FBC
Quoted 140 lines · Rodrigo
Quoted 45 lines · Paulo Zanoni
2015-03-03 21:57 GMT-03:00 Rodrigo Vivi <[email protected]>:
Quoted 25 lines · Paulo Zanoni
Quoted 24 lines
From: Paulo Zanoni <[email protected]>
Kill the blt/render tracking we currently have and use the frontbuffer tracking infrastructure.
Don't enable things by default yet.
v2: (Rodrigo) Fix small conflict on rebase and typo at subject.
v3: (Paulo) Rebase onRENDER_CSchange.
v4: (Paulo) Rebase.
v5: (Paulo) Simplify: flushes don't have origin (Daniel).
Also rebase due to patch order changes.Signed-off-by Rodrigo Vivi <[email protected]>Signed-off-by Paulo Zanoni <[email protected]>--- drivers/gpu/drm/i915/i915_drv.h | 10 +--- drivers/gpu/drm/i915/intel_drv.h | 6 ++- drivers/gpu/drm/i915/intel_fbc.c | 91 +++++++++++++++++++------------- drivers/gpu/drm/i915/intel_frontbuffer.c | 14 +---- drivers/gpu/drm/i915/intel_ringbuffer.c | 41 +------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 1 - 6 files changed, 65 insertions(+), 98 deletions(-)Quoted 12 lines
drivers/gpu/drm/i915/i915_drv.h struct i915_fbc { · line 787
Quoted 10 lines
drivers/gpu/drm/i915/i915_drv.h +2
-0@@ -783,6 +783,8 @@ struct i915_fbc {unsigned long uncompressed_size;unsigned threshold;unsigned int fb_id; +unsigned int possible_framebuffer_bits; +unsigned int busy_bits;I'm not sure if I understood why you keep 2 variables here?
After Daniel's last suggestion, the
possible_framebuffer_bitscould probably be removed and left as a local variable atintel_fbc_validate
now: it's just a matter of coding style. Some platforms support FBC on more than just pipe A, so this variable is used to simplify checks related to this.Quoted 3 lines
The
busy_bitswas also part of Daniel's last request: we need to store which bits triggeredintel_fbc_invalidatecalls, and then check them whenintel_fbc_flushis called.Got it.
Thanks for explaining it.Quoted 5 lines
Quoted 3 lines
Is it because we keep enabling/disabling fbc with updates all over the code? In this case it is ok we just need to kill it when we kill
update_fbc...I don't understand what you mean here. Please clarify.
I had missunderstood the reason of
possible_frontbufferbits, sorry. But about killingupdate_fbcI believe that after frontbuffer rework is in place we don't need to use thatupdate_fbcfunction that disable and enable fbc on the cases we knew it could fail. With frontbuffer bits we could let it always enabled and kicking with frontbuffer bits.Quoted 75 lines
Quoted 74 lines
Quoted 59 lines
drivers/gpu/drm/i915/i915_drv.h +0
-8struct intel_crtc *crtc;int y;@@ -795,14 +797,6 @@ struct i915_fbc {* possible. */ bool enabled; -/* On gen8 some rings cannont perform fbc clean operation so for now -* we are doing this on SW with mmio. -* This variable works in the opposite information direction -* of ring->fbc_dirty telling software on frontbuffer tracking -* to perform the cache clean on sw side. -*/ -bool need_sw_cache_clean; -struct intel_fbc_work {struct delayed_work work;struct drm_crtc *crtc;drivers/gpu/drm/i915/intel_drv.h +5
-1@@ -1091,7 +1091,11 @@ bool intel_fbc_enabled(struct drm_device *dev);void intel_fbc_update (struct drm_device *dev);void intel_fbc_init (struct drm_i915_private *dev_priv);void intel_fbc_disable (struct drm_device *dev); -void bdw_fbc_sw_flush (struct drm_device *dev,u32 value); +void intel_fbc_invalidate (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits, +enum fb_op_origin origin); +void intel_fbc_flush (struct drm_i915_private *dev_priv, +unsigned int frontbuffer_bits);/* intel_hdmi.c */ void intel_hdmi_init (struct drm_device *dev,int hdmi_reg,enum port port);drivers/gpu/drm/i915/intel_fbc.c +1
-12@@ -174,29 +174,10 @@ static bool g4x_fbc_enabled(struct drm_device *dev)return I915_READ (DPFC_CONTROL ) &DPFC_CTL_EN ; } -static void snb_fbc_blit_update (struct drm_device *dev) +static void intel_fbc_nuke (struct drm_i915_private *dev_priv) { -struct drm_i915_private *dev_priv = dev->dev_private; -u32 blt_ecoskpd; - -/* Make sure blitter notifies FBC of writes */ - -/* Blitter is part of Media powerwell on VLV. No impact of -* his param in other platforms for now */ -intel_uncore_forcewake_get (dev_priv,FORCEWAKE_MEDIA ); - - blt_ecoskpd =I915_READ (GEN6_BLITTER_ECOSKPD ); - blt_ecoskpd |= GEN6_BLITTER_FBC_NOTIFY <<drivers/gpu/drm/i915/intel_fbc.c static bool g4x_fbc_enabled(struct drm_device *dev) · line 199
Quoted 10 lines
drivers/gpu/drm/i915/intel_fbc.c +0
-10-GEN6_BLITTER_LOCK_SHIFT ; -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); - blt_ecoskpd |=GEN6_BLITTER_FBC_NOTIFY ; -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); - blt_ecoskpd &= ~(GEN6_BLITTER_FBC_NOTIFY << -GEN6_BLITTER_LOCK_SHIFT ); -I915_WRITE (GEN6_BLITTER_ECOSKPD , blt_ecoskpd); -POSTING_READ (GEN6_BLITTER_ECOSKPD ); - -intel_uncore_forcewake_put (dev_priv,FORCEWAKE_MEDIA );Are you sure this continue working on snb? or should we fully remove fbc support for snb and older?
Anyway I believe this could be in a different patch.
Quoted 7 lines
Quoted 2 lines
Well, the whole idea of using the sofware-based frontbuffer tracking mechanism is to get rid of all the hardware-based stuff.
Hm, I still have a concern here. We would need to test on those platforms to be certain. The reason for that is that HW tracking still works so it can let the buffer in a stage that compression is unable to start... same reason that we still need both lines below on BDW.
Afaik the fence based tracking only invalidates specific lines. Everything else shouldn't be required any more with this patch.
Quoted 4 lines
Anyway this wouldn't break anything and it this rework was a great improvement so feel free to use
Reviewed-by Rodrigo Vivi <[email protected]>
Queued for -next, thanks for the patch.
-Daniel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 7/7 drm/i915: don't reallocate the compressed FB at every frame
In reply to Paulo ZanoniFrom: Paulo Zanoni <[email protected]>
With the current code we just reallocate the compressed FB at every FBC update: we have X in one frame, then in the other frame we need X again, but we check "needed < have" instead of "needed <= have".
v2: Rebase after Jani addressed the other problems described in v1.
--- drivers/gpu/drm/i915/i915_gem_stolen.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_gem_stolen.c +1 -1
@@ -253,7 +253,7 @@ int i915_gem_stolen_setup_compression(struct drm_device *dev, int size, int fb_cif (!drm_mm_initialized (&dev_priv->mm.stolen))return -ENODEV ; -if (size < dev_priv->fbc.uncompressed_size) +if (size <= dev_priv->fbc.uncompressed_size)return 0 ;/* Release any current block */
Signature · 6 lines
2.1.4 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 7/7] drm/i915: don't reallocate the compressed FB at every frame
Task id: 5775 -------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied
PNV -3 281/281 278/281
ILK 313/313 313/313 SNB 309/309 309/309 IVB 382/382 382/382 BYT 296/296 296/296 HSW 426/426 426/426 BDW -1 318/318 317/318 -------------------------------------Detailed------------------------------------- Platform Test drm-intel-nightly Series Applied *PNV igt_gen3_render_mixed_blits PASS(1) NRUN(1) *PNV igt_gen3_render_tiledx_blits PASS(1) NRUN(1) *PNV igt_gen3_render_tiledy_blits PASS(1) NRUN(1) *BDW igt_gem_gtt_hog PASS(1) DMESG_WARN(1)
Note: You need to pay more attention to line start with '*'
_______________________________________________
Intel-gfx mailing list [email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 7/7] drm/i915: don't reallocate the compressed FB at every frame
In reply to Paulo ZanoniQuoted 10 lines · Paulo Zanoni
From: Paulo Zanoni <[email protected]>
With the current code we just reallocate the compressed FB at every FBC update: we have X in one frame, then in the other frame we need X again, but we check "needed < have" instead of "needed <= have".
v2: Rebase after Jani addressed the other problems described in v1.
Cc Jani Nikula <[email protected]>Signed-off-by Paulo Zanoni <[email protected]>
Quoted 20 lines
--- drivers/gpu/drm/i915/i915_gem_stolen.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)drivers/gpu/drm/i915/i915_gem_stolen.c +1
-1@@ -253,7 +253,7 @@ int i915_gem_stolen_setup_compression(struct drm_device *dev, int size, int fb_cif (!drm_mm_initialized (&dev_priv->mm.stolen))return -ENODEV ; -if (size < dev_priv->fbc.uncompressed_size) +if (size <= dev_priv->fbc.uncompressed_size)return 0 ;/* Release any current block */ Signature · 1 line
2.1.4
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Daniel Vetter
AppliedRe: [PATCH 7/7] drm/i915: don't reallocate the compressed FB at every frame
Quoted 7 lines · Jani Nikula
Quoted 6 lines · Paulo Zanoni
From: Paulo Zanoni <[email protected]>
With the current code we just reallocate the compressed FB at every FBC update: we have X in one frame, then in the other frame we need X again, but we check "needed < have" instead of "needed <= have".
Quoted 6 lines
Quoted 4 lines
v2: Rebase after Jani addressed the other problems described in v1.
Cc Jani Nikula <[email protected]>Signed-off-by Paulo Zanoni <[email protected]>Reviewed-by Jani Nikula <[email protected]>
Queued for -next, thanks for the patch.
-Daniel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx