Rework intel 2D driver glamor support
cover Rework intel 2D driver glamor support
I spent the day just cleaning up this patch series and testing. I think it's ready for others to use and review. I've been running it on two machines for a couple of days now and it's been solid.
I ran three different desktop environments (current Debian unstable versions):
XFCE 4.10
Gnome 3.12
KDE 4.13.3
I have not merged DRI2 support when running with Glamor; I've got that working locally, but if you accidentally try indirect rendering, you'll crash the X server with an assert failure. So, this is DRI3 only, at least for now.
This patch series also adds a "none" acceleration mode. It's different from other unaccelerated drivers in offering DRI2 and DRI3 support so that you can run direct rendering.
A brief synopsis of the series
1-4 cleanup patches
5 Identify and isolate UXA code; UXA-specific functions are
renamed and moved into uxa-specific files.
6 Remove UXA-based Glamor support. This pulls out all of the
Glamor calls from UXA rendering paths. This patch only builds with glamor disabled; I didn't worry about the existing glamor code or support for glamor within the rest of the driver.
7-9 Prepare for glamor support. Creates a couple of abstract
functions for accel-dependent functionality needed by the initialization and modesetting code. Gets rid of the glamor stubs in intel_glamor.h
10 Add glamor support back in, using the regular glamor
API.
11 Add an unaccelerated option ("none"). This offers fb-only
support and is always compiled in to the driver.
12 Delay initial mode set operation until the root window
is painted and the server is ready to go. This includes potentially copying an existing fbcon frame buffer to the root window in background none mode, providing support for this in all three acceleration modes. The driver used to support this by copying the fbcon buffer to the screen buffer during early server initialization; this change allows the driver to use regular GC-based CopyArea instead of needing custom rendering code.
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
PATCH 1/12 Stop trying to out-guess mesa for BO allocation
--- src/uxa/intel_uxa.c | 2 ++ 1 file changed, 2 insertions(+)
src/uxa/intel_uxa.c +2 -0
@@ -758,6 +758,7 @@ free_priv: goto free_priv; } +#if 0 if (tiling !=I915_TILING_NONE ) {int height;
@@ -780,6 +781,7 @@ free_priv: bo =NULL ; goto free_priv; } +#endif } BAIL:
Signature · 1 line
2.0.1
A message that is not in the archive (sent off-list or not yet received): 1406243908-1123-2-git-send-email-keithp-aN4HjG94KOLQT0dZR+AlfA@public.gmane.org
Re: [PATCH 01/12] Stop trying to out-guess mesa for BO allocation
Keith Packard <[email protected]> writes:
I don't see anything indicating that this code path is only used by glamor.
Quoted 31 lines
--- src/uxa/intel_uxa.c | 2 ++ 1 file changed, 2 insertions(+)src/uxa/intel_uxa.c +2
-0@@ -758,6 +758,7 @@ free_priv: goto free_priv; } +#if 0 if (tiling !=I915_TILING_NONE ) {int height;@@ -780,6 +781,7 @@ free_priv: bo =NULL ; goto free_priv; } +#endif } BAIL:Signature · 6 lines
2.0.1 _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
Re: [PATCH 01/12] Stop trying to out-guess mesa for BO allocation
Quoted 4 lines · Eric Anholt
Keith Packard <[email protected]> writes:
I don't see anything indicating that this code path is only used by glamor.
True. It's a fix for DRI3 for either UXA or "none". Mesa allocates a single page for a 1x1 texture, but this code thinks that should take two pages causing a texture-to-pixmap operation to fail.
In particular, gnome-shell fails without this patch.
It's useful independent of this series, which is why I stuck it first.
Signature · 1 line
[email protected]
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 01/12] Stop trying to out-guess mesa for BO allocation
Quoted 10 lines · Keith Packard
Quoted 4 lines · Eric Anholt
Keith Packard <[email protected]> writes:
I don't see anything indicating that this code path is only used by glamor.
True. It's a fix for DRI3 for either UXA or "none". Mesa allocates a single page for a 1x1 texture, but this code thinks that should take two pages causing a texture-to-pixmap operation to fail.
OK, but isn't the problem with the code you're #if 0ing (which, really? #if 0 in a patch being submitted for review?) that it's aligning to 2*height instead of height?
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 01/12] Stop trying to out-guess mesa for BO allocation
Quoted 3 lines · Eric Anholt
OK, but isn't the problem with the code you're #if 0ing (which, really? #if 0 in a patch being submitted for review?) that it's aligning to 2*height instead of height?
I went and did a bit of archaeology to figure out why it's using 2*height instead of just height. The initial change to pixmap allocation, which was then copied to the BO validation logic, was here:
commit 736b89504a32239a0c7dfb5961c1b8292dd744bd Author: Chris Wilson <[email protected]> Date: Sun Dec 30 10:32:18 2012 +0000 uxa: Align surface allocations to even tile rows Align surface sizes to an even number of tile rows to cater for sampler prefetch. If we read beyond the last page we may catch the PTE in a state of flux and trigger a GPU hang. Also detected by enabling invalid PTE access checking. References: https://bugs.freedesktop.org/show_bug.cgi?id=56916 References: https://bugs.freedesktop.org/show_bug.cgi?id=55984 Signed-off-by: Chris Wilson <[email protected]> diff --git a/src/intel_uxa.c b/src/intel_uxa.c index f5ac0a6..2f14173 100644 --- a/src/intel_uxa.c +++ b/src/intel_uxa.c @@ -209,7 +209,7 @@ intel_uxa_pixmap_compute_size(PixmapPtr pixmap, tile_height = 8; else tile_height = 32; - aligned_h = ALIGN(h, tile_height); + aligned_h = ALIGN(h, 2*tile_height); *stride = intel_get_fence_pitch(intel, ALIGN(pitch, 512),
Look at the referenced bugs, what I found was a long list of random GPU hangs on ILK and SNB hardware that appear to have been caused by a kernel change. Daniel and Chris created a number of scatter shot fixes across the kernel, and this patch to the 2D driver. However, this patch doesn't appear to have actually solved anything; Norbert Preining was still crashing with this patch applied:
https://bugs.freedesktop.org/show_bug.cgi?id=55984#c129
Furthermore, SNA has some similar code, but it applies it conditionally for hardware which doesn't have 'relaxed fencing', which is only hardware older than i965 when running on a older kernel that doesn't recognize the HAS_RELAXED_FENCING parameter (the current kernel always returns 'true').
So, as near as I can tell, this fix should never be necessary as the reported bug wasn't fixed by it, and SNA does not apply this rule to any hardware on which either bug was reproduced.
If this fix is actually useful, wouldn't we want it in Mesa as well?
Signature · 1 line
[email protected]
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
A message that is not in the archive (sent off-list or not yet received): 1406243908-1123-1-git-send-email-keithp-aN4HjG94KOLQT0dZR+AlfA@public.gmane.org
PATCH 2/12 Fix present debug output
Was using 'low_msc', which isn't defined anymore.
--- src/uxa/intel_present.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
src/uxa/intel_present.c +2 -1
@@ -193,7 +193,8 @@ intel_present_queue_vblank(RRCrtcPtr crtc,if (errno !=EBUSY || !intel_present_flush_drm_events (screen))return BadAlloc; } -DebugPresent (("\t\tiq %lld seq %u msc %u (hw msc %u)\n" , event_id, seq, low_msc, vbl.request.sequence)); +DebugPresent (("\t\tiq %lld seq %u msc %llu (hw msc %u)\n" , + (long long ) event_id, seq, (long long ) msc, vbl.request.sequence));return Success; }
Signature · 6 lines
2.0.1 _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
PATCH 7/12 Add intel_flush to abstract flushing pending acceleration operations
In reply to a message not in the archiveintel_flush flushes any pending acceleration operations to the hardware, just like intel_uxa_batch_submit does today except that it is not uxa-specific.
--- src/uxa/intel.h | 3 +++ src/uxa/intel_display.c | 11 +++-------- src/uxa/intel_driver.c | 14 +++++++++----- 3 files changed, 15 insertions(+), 13 deletions(-)
src/uxa/intel.h +3 -0
@@ -535,4 +535,7 @@ intel_get_pixmap_bo(PixmapPtr pixmap);void intel_set_pixmap_bo (PixmapPtr pixmap,dri_bo *bo); +void +intel_flush (intel_screen_private *intel); +#endif /* _I830_H_ */
src/uxa/intel_display.c +3 -8
@@ -57,8 +57,6 @@#include "fb.h" #include "uxa.h" -#include "intel_glamor.h" -#define KNOWN_MODE_FLAGS ((1<<14)-1)struct intel_drm_queue {
@@ -437,8 +435,7 @@ intel_crtc_set_mode_major(xf86CrtcPtr crtc, DisplayModePtr mode, crtc->y = y; crtc->rotation = rotation; -intel_glamor_flush (intel); -intel_batch_submit (crtc->scrn); +intel_flush (intel);mode_to_kmode (crtc->scrn, &intel_crtc->kmode, mode); ret =intel_crtc_apply (crtc);
@@ -1437,8 +1434,7 @@ intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height)if (scrn->virtualX == width && scrn->virtualY == height)return TRUE; -intel_glamor_flush (intel); -intel_batch_submit (scrn); +intel_flush (intel); old_width = scrn->virtualX; old_height = scrn->virtualY;
@@ -1552,8 +1548,7 @@ intel_do_pageflip(intel_screen_private *intel, goto error_out;drm_intel_bo_disable_reuse (new_front); -intel_glamor_flush (intel); -intel_batch_submit (scrn); +intel_flush (intel); mode->pageflip_data = pageflip_data; mode->pageflip_handler = pageflip_handler;
src/uxa/intel_driver.c +9 -5
@@ -181,6 +181,12 @@ intel_get_pixmap_bo(PixmapPtr pixmap)return intel_uxa_get_pixmap_bo (pixmap); } +void +intel_flush (intel_screen_private *intel) +{ +intel_batch_submit (intel->scrn); +} +static void PreInitCleanup (ScrnInfoPtr scrn) {if (!scrn || !scrn->driverPrivate)
@@ -647,7 +653,7 @@ redisplay_dirty(ScreenPtr screen, PixmapDirtyUpdatePtr dirty)PixmapSyncDirtyHelper (dirty, &pixregion);RegionUninit (&pixregion); -intel_batch_submit (scrn); +intel_flush (intel);if (!intel->has_prime_vmap_flush) {drm_intel_bo *bo =intel_uxa_get_pixmap_bo (dirty->slave_dst->master_pixmap); was_blocked =xf86BlockSIGIO ();
@@ -732,10 +738,8 @@ intel_flush_callback(CallbackListPtr *list, pointer user_data, pointer call_data) {ScrnInfoPtr scrn = user_data; -if (scrn->vtSema) { -intel_batch_submit (scrn); -intel_glamor_flush (intel_get_screen_private (scrn)); - } +if (scrn->vtSema) +intel_flush (intel_get_screen_private (scrn)); }#if HAVE_UDEV
Signature · 6 lines
2.0.1 _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
A message that is not in the archive (sent off-list or not yet received): 1406243908-1123-8-git-send-email-keithp-aN4HjG94KOLQT0dZR+AlfA@public.gmane.org
Eric Anholt
Reviewed-byRe: [PATCH 07/12] Add intel_flush to abstract flushing pending acceleration operations
Quoted 3 lines · Keith Packard
intel_flushflushes any pending acceleration operations to the hardware, just likeintel_uxa_batch_submitdoes today except that it is not uxa-specific.
This seems like an obvious cleanup (and in one case a fix), and we ought to even apply it right away as long as it still calls intel_glamor_flush(). If it did, it gets:
even before we sort out the rest of the series.
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
PATCH 8/12 Get rid of glamor stubs in intel_glamor.h
In reply to a message not in the archiveThis eliminates the stubs in intel_glamor.h and replaces them with ifdefs instead.
--- src/uxa/intel_dri.c | 20 ++++++++++++++++++-- src/uxa/intel_driver.c | 6 ++++++ src/uxa/intel_glamor.h | 21 --------------------- src/uxa/intel_present.c | 2 ++ src/uxa/intel_video.c | 4 ++++ 5 files changed, 30 insertions(+), 23 deletions(-)
src/uxa/intel_dri.c +18 -2
@@ -66,7 +66,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.#include "dri2.h" +#if USE_GLAMOR #include "intel_glamor.h" +#endif #include "uxa.h" typedef struct {
@@ -107,6 +109,7 @@ static PixmapPtr get_front_buffer(DrawablePtr drawable)return pixmap; } +#if USE_GLAMOR static PixmapPtr fixup_glamor (DrawablePtr drawable,PixmapPtr pixmap) {ScreenPtr screen = drawable->pScreen;
@@ -158,6 +161,7 @@ static PixmapPtr fixup_glamor(DrawablePtr drawable, PixmapPtr pixmap)intel_get_screen_private (xf86ScreenToScrn (screen))->needs_flush = TRUE;return old; } +#endif #if DRI2INFOREC_VERSION <2 static DRI2BufferPtr
@@ -170,7 +174,9 @@ I830DRI2CreateBuffers(DrawablePtr drawable, unsigned int *attachments,DRI2BufferPtr buffers;I830DRI2BufferPrivatePtr privates;PixmapPtr pixmap, pDepthPixmap; +#if USE_GLAMOR Bool is_glamor_pixmap = FALSE; +#endif int i; buffers =calloc (count,sizeof *buffers);
@@ -187,11 +193,12 @@ I830DRI2CreateBuffers(DrawablePtr drawable, unsigned int *attachments, pixmap =NULL ;if (attachments[i] == DRI2BufferFrontLeft) { pixmap =get_front_buffer (drawable); - +#if USE_GLAMOR if (pixmap ==NULL ) { drawable = &(get_drawable_pixmap (drawable)->drawable); is_glamor_pixmap = TRUE; } +#endif }else if (attachments[i] == DRI2BufferStencil && pDepthPixmap) { pixmap = pDepthPixmap; pixmap->refcnt++;
@@ -230,8 +237,10 @@ I830DRI2CreateBuffers(DrawablePtr drawable, unsigned int *attachments, goto unwind; } +#if USE_GLAMOR if (is_glamor_pixmap) pixmap =fixup_glamor (drawable, pixmap); +#endif }if (attachments[i] == DRI2BufferDepth)
@@ -292,7 +301,9 @@ I830DRI2CreateBuffer(DrawablePtr drawable, unsigned int attachment,DRI2Buffer2Ptr buffer;I830DRI2BufferPrivatePtr privates;PixmapPtr pixmap; +#if USE_GLAMOR Bool is_glamor_pixmap = FALSE; +#endif buffer =calloc (1 ,sizeof *buffer);if (buffer ==NULL )
@@ -306,11 +317,12 @@ I830DRI2CreateBuffer(DrawablePtr drawable, unsigned int attachment, pixmap =NULL ;if (attachment == DRI2BufferFrontLeft) { pixmap =get_front_buffer (drawable); - +#if USE_GLAMOR if (pixmap ==NULL ) { drawable = &(get_drawable_pixmap (drawable)->drawable); is_glamor_pixmap = TRUE; } +#endif }if (pixmap ==NULL ) {
@@ -383,8 +395,10 @@ I830DRI2CreateBuffer(DrawablePtr drawable, unsigned int attachment,free (buffer);return NULL ; } +#if USE_GLAMOR if (is_glamor_pixmap) pixmap =fixup_glamor (drawable, pixmap); +#endif } buffer->attachment = attachment;
@@ -762,7 +776,9 @@ intel_exchange_pixmap_buffers(struct intel_screen_private *intel, PixmapPtr fron new_front->busy =1 ; new_back->busy =-1 ; +#if USE_GLAMOR intel_glamor_exchange_buffers (intel, front, back); +#endif DamageRegionProcessPending (&front->drawable);
src/uxa/intel_driver.c +6 -0
@@ -72,7 +72,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.#include "i915_drm.h" +#if USE_GLAMOR #include "intel_glamor.h" +#endif #include "intel_options.h" static void i830AdjustFrame (ADJUST_FRAME_ARGS_DECL );
@@ -610,12 +612,14 @@ static Bool I830PreInit(ScrnInfoPtr scrn, int flags)return FALSE; } +#if USE_GLAMOR if (!intel_glamor_pre_init (scrn)) {PreInitCleanup (scrn);xf86DrvMsg (scrn->scrnIndex,X_ERROR ,"Failed to pre init glamor display.\n" );return FALSE; } +#endif /* Load the dri modules if requested. */ #if HAVE_DRI2
@@ -1125,7 +1129,9 @@ static Bool I830CloseScreen(CLOSE_SCREEN_ARGS_DECL)DeleteCallback (&FlushCallback, intel_flush_callback, scrn); +#if USE_GLAMOR intel_glamor_close_screen (screen); +#endif TimerFree (intel->cache_expire); intel->cache_expire =NULL ;
src/uxa/intel_glamor.h +0 -21
@@ -32,8 +32,6 @@#include <xf86xv.h> -#ifdef USE_GLAMOR -Bool intel_glamor_pre_init (ScrnInfoPtr scrn);Bool intel_glamor_init (ScreenPtr screen);Bool intel_glamor_create_screen_resources (ScreenPtr screen);
@@ -48,24 +46,5 @@ PixmapPtr intel_glamor_create_pixmap(ScreenPtr screen, int w, int h, int depth, unsigned int usage);void intel_glamor_exchange_buffers (struct intel_screen_private *intel,PixmapPtr src,PixmapPtr dst);XF86VideoAdaptorPtr intel_glamor_xv_init (ScreenPtr screen,int num_ports); -#else - -static inline Bool intel_glamor_pre_init (ScrnInfoPtr scrn) {return TRUE; } -static inline Bool intel_glamor_init (ScreenPtr screen) {return TRUE; } -static inline Bool intel_glamor_create_screen_resources (ScreenPtr screen) {return TRUE; } -static inline Bool intel_glamor_close_screen (ScreenPtr screen) {return TRUE; } -static inline void intel_glamor_free_screen (int scrnIndex,int flags) { } - -static inline void intel_glamor_flush (intel_screen_private * intel) { } - -static inline Bool intel_glamor_create_textured_pixmap (PixmapPtr pixmap) {return TRUE; } -static inline void intel_glamor_destroy_pixmap (PixmapPtr pixmap) { } - -static inline PixmapPtr intel_glamor_create_pixmap (ScreenPtr screen,int w,int h, -int depth,unsigned int usage) {return NULL ; } - -static inline void intel_glamor_exchange_buffers (struct intel_screen_private *intel,PixmapPtr src,PixmapPtr dst) {} -static inline XF86VideoAdaptorPtr intel_glamor_xv_init (ScreenPtr screen,int num_ports) {return NULL ; } -#endif #endif /* INTEL_GLAMOR_H */
src/uxa/intel_present.c +2 -0
@@ -54,7 +54,9 @@#include "present.h" +#if USE_GLAMOR #include "intel_glamor.h" +#endif #include "uxa.h" struct intel_present_vblank_event {
src/uxa/intel_video.c +4 -0
@@ -78,7 +78,9 @@#define _INTEL_XVMC_SERVER_#include "intel_xvmc.h" #endif +#if USE_GLAMOR #include "intel_glamor.h" +#endif #include "intel_uxa.h" #include "intel_video_overlay.h"
@@ -213,10 +215,12 @@ void intel_video_init(ScreenPtr screen) } } +#if USE_GLAMOR glamorAdaptor =intel_glamor_xv_init (screen,16 );if (glamorAdaptor !=NULL )xf86DrvMsg (scrn->scrnIndex,X_INFO ,"Set up textured video using glamor\n" ); +#endif if (overlayAdaptor && intel->XvPreferOverlay)
Signature · 6 lines
2.0.1 _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
A message that is not in the archive (sent off-list or not yet received): 1406243908-1123-9-git-send-email-keithp-aN4HjG94KOLQT0dZR+AlfA@public.gmane.org
Eric Anholt
Acked-byRe: [PATCH 08/12] Get rid of glamor stubs in intel_glamor.h
Quoted 2 lines · Keith Packard
This eliminates the stubs in
intel_glamor.hand replaces them with ifdefs instead.
I don't feel strongly about this either way -- ifdefs are more traditional userland style, while stubs are more kernel style.
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
Re: [PATCH 08/12] Get rid of glamor stubs in intel_glamor.h
Quoted 7 lines · Eric Anholt
Quoted 2 lines · Keith Packard
This eliminates the stubs in
intel_glamor.hand replaces them with ifdefs instead.I don't feel strongly about this either way -- ifdefs are more traditional userland style, while stubs are more kernel style.
Right, just trying to keep things all the same and #ifdef USE_GLAMOR seemed more common.
Quoted 1 line
Acked-by Eric Anholt <[email protected]>
Signature · 1 line
[email protected]
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 10/12 Add glamor back into the driver
In reply to a message not in the archiveThis adds glamor support back into the driver, but instad of going through UXA, this uses it directly instead.
--- src/uxa/intel.h | 17 +++ src/uxa/intel_display.c | 30 ++++- src/uxa/intel_dri.c | 20 +++ src/uxa/intel_driver.c | 247 ++++++++++++++++++++++++++++--------- src/uxa/intel_glamor.c | 316 +++++++++++++++++++++++++----------------------- src/uxa/intel_glamor.h | 18 ++- src/uxa/intel_present.c | 5 - src/uxa/intel_uxa.c | 5 + src/uxa/intel_video.c | 75 +++++++----- 9 files changed, 480 insertions(+), 253 deletions(-)
src/uxa/intel.h +17 -0
@@ -111,10 +111,21 @@ enum dri_type {DRI_ACTIVE }; +enum accel_type { +#if USE_GLAMOR +ACCEL_GLAMOR , +#endif +#if USE_UXA +ACCEL_UXA , +#endif +}; + typedef struct intel_screen_private { ScrnInfoPtr scrn; int cpp; + enum accel_type accel; + #define RENDER_BATCH I915_EXEC_RENDER #define BLT_BATCH I915_EXEC_BLT unsigned int current_batch;
@@ -127,6 +138,7 @@ typedef struct intel_screen_private {dri_bufmgr *bufmgr; +#if USE_UXA uint32_t batch_ptr[4096 ];/** Byte offset in batch_ptr for the next dword to be emitted. */ unsigned int batch_used;
@@ -142,6 +154,7 @@ typedef struct intel_screen_private {struct list batch_pixmaps;drm_intel_bo *wa_scratch_bo;OsTimerPtr cache_expire; +#endif /* For Xvideo */ Bool use_overlay;
@@ -178,8 +191,10 @@ typedef struct intel_screen_private {void (*batch_flush) (struct intel_screen_private *intel);void (*batch_commit_notify) (struct intel_screen_private *intel); +#if USE_UXA struct _UxaDriver *uxa_driver;int uxa_flags; +#endif Bool need_sync;int accel_pixmap_offset_alignment;int accel_max_x;
@@ -212,6 +227,7 @@ typedef struct intel_screen_private {drm_intel_bo *gen6_depth_stencil_bo; } video; +#if USE_UXA /* Render accel state */ float scale_units[2 ][2 ];/** Transform pointers for src/mask, or NULL if identity */
@@ -269,6 +285,7 @@ typedef struct intel_screen_private {/* 965 render acceleration state */ struct gen4_render_state *gen4_render_state; +#endif /* DRI enabled this generation. */ enum dri_type dri2, dri3;
src/uxa/intel_display.c +24 -6
@@ -56,6 +56,9 @@#include "xf86DDC.h" #include "fb.h" #include "uxa.h" +#if USE_GLAMOR +#include "intel_glamor.h" +#endif #define KNOWN_MODE_FLAGS ((1<<14)-1)
@@ -182,7 +185,7 @@ intel_output_backlight_init(xf86OutputPtr output) {struct intel_output *intel_output = output->driver_private;intel_screen_private *intel =intel_get_screen_private (output->scrn); -char *str; +const char *str;#if !USE_BACKLIGHT return ;
@@ -683,7 +686,10 @@ intel_set_scanout_pixmap(xf86CrtcPtr crtc, PixmapPtr ppix)return TRUE; } - bo =intel_get_pixmap_bo (ppix); + bo =intel_get_pixmap_bo (ppix); +if (!bo) +return FALSE; +if (intel->front_buffer) {ErrorF ("have front buffer\n" ); }
@@ -1472,8 +1478,20 @@ intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height) scrn->virtualX = width; scrn->virtualY = height; -if (!intel_uxa_create_screen_resources (scrn->pScreen)) - goto fail; +switch (intel->accel) { +#if USE_GLAMOR +case ACCEL_GLAMOR : +if (!intel_glamor_create_screen_resources (scrn->pScreen)) + goto fail; +break ; +#endif +#if USE_UXA +case ACCEL_UXA : +if (!intel_uxa_create_screen_resources (scrn->pScreen)) + goto fail; +break ; +#endif + }for (i =0 ; i < xf86_config->num_crtc; i++) {xf86CrtcPtr crtc = xf86_config->crtc[i];
@@ -1532,7 +1550,7 @@ intel_do_pageflip(intel_screen_private *intel,xf86CrtcConfigPtr config =XF86_CRTC_CONFIG_PTR (scrn);struct intel_crtc *crtc = config->crtc[0 ]->driver_private;struct intel_mode *mode = crtc->mode; -unsigned int pitch = scrn->displayWidth * intel->cpp; +unsigned int pitch = scrn->displayWidth * intel->cpp;struct intel_pageflip *flip;uint32_t new_fb_id;uint32_t flags;
@@ -2235,7 +2253,7 @@ intel_create_pixmap_for_bo(ScreenPtr pScreen, dri_bo *bo,return NullPixmap; } -intel_set_pixmap_bo (pixmap, bo); +intel_set_pixmap_bo (pixmap, bo);return pixmap; }
src/uxa/intel_dri.c +20 -0
@@ -883,6 +883,26 @@ I830DRI2ScheduleFlip(struct intel_screen_private *intel,drm_intel_bo_disable_reuse (new_back);dri_bo_flink (new_back, &intel->back_name); +#if USE_GLAMOR +if (intel->accel ==ACCEL_GLAMOR ) { +I830DRI2BufferPrivatePtr drvpriv; +PixmapPtr front_pixmap, back_pixmap; +ScreenPtr screen; + + screen = draw->pScreen; + drvpriv = info->front->driverPrivate; + front_pixmap = drvpriv->pixmap; + + back_pixmap =intel_glamor_create_back_pixmap (screen, + front_pixmap, + new_back); +if (back_pixmap ==NULL ) { +drm_intel_bo_unreference (new_back); +return FALSE; + } + intel->back_pixmap = back_pixmap; + } +#endif }else { new_back = intel->back_buffer; intel->back_buffer =NULL ;
src/uxa/intel_driver.c +189 -58
346 lines not on this page · raw message
src/uxa/intel_glamor.c +166 -150
396 lines not on this page · raw message
src/uxa/intel_glamor.h +14 -4
@@ -40,11 +40,21 @@ void intel_glamor_free_screen(int scrnIndex, int flags);void intel_glamor_flush (intel_screen_private * intel); -Bool intel_glamor_create_textured_pixmap (PixmapPtr pixmap); -void intel_glamor_destroy_pixmap (PixmapPtr pixmap); -PixmapPtr intel_glamor_create_pixmap (ScreenPtr screen,int w,int h, -int depth,unsigned int usage); +#if HAVE_DRI2 +PixmapPtr +intel_glamor_create_back_pixmap (ScreenPtr screen, +PixmapPtr front_pixmap, +drm_intel_bo *back_bo); +void intel_glamor_exchange_buffers (struct intel_screen_private *intel,PixmapPtr src,PixmapPtr dst); +#endif +XF86VideoAdaptorPtr intel_glamor_xv_init (ScreenPtr screen,int num_ports); +dri_bo * +intel_glamor_get_pixmap_bo (PixmapPtr pixmap); + +Bool +intel_glamor_set_pixmap_bo (PixmapPtr pixmap,dri_bo *bo); +#endif /* INTEL_GLAMOR_H */
src/uxa/intel_present.c +0 -5
@@ -54,11 +54,6 @@#include "present.h" -#if USE_GLAMOR -#include "intel_glamor.h" -#endif -#include "uxa.h" -struct intel_present_vblank_event {uint64_t event_id; };
src/uxa/intel_uxa.c +5 -0
@@ -1272,6 +1272,11 @@ Bool intel_uxa_init(ScreenPtr screen)ScrnInfoPtr scrn =xf86ScreenToScrn (screen);intel_screen_private *intel =intel_get_screen_private (scrn); +intel_batch_init (scrn); + +if (INTEL_INFO (intel)->gen >=040 &&INTEL_INFO (intel)->gen <0100 ) +gen4_render_state_init (scrn); +#if HAS_DIXREGISTERPRIVATEKEY if (!dixRegisterPrivateKey (&uxa_pixmap_index,PRIVATE_PIXMAP ,0 )) #else
src/uxa/intel_video.c +45 -30
119 lines not on this page · raw message
Signature · 6 lines
2.0.1 _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
A message that is not in the archive (sent off-list or not yet received): 1406243908-1123-11-git-send-email-keithp-aN4HjG94KOLQT0dZR+AlfA@public.gmane.org
Re: [PATCH 10/12] Add glamor back into the driver
Quoted 2 lines · Keith Packard
This adds glamor support back into the driver, but instad of going through UXA, this uses it directly instead.
This is hard to read with the conditionalizing all of the UXA code in the same commit as adding the glamor code. Then there are a bunch of unrelated whitespace changes, or flattening of a bunch of nested conditionals.
The only substantive problem I see is intel_glamor_set_pixmap_bo().
Quoted 23 lines
src/uxa/intel_driver.c +23
-0+static void intel_check_accel_option (ScrnInfoPtr scrn) +{ +intel_screen_private *intel =intel_get_screen_private (scrn); +enum {NONE ,SNA ,UXA ,GLAMOR } accel_method =DEFAULT_ACCEL_METHOD ; +const char *s; + + s =xf86GetOptValString (intel->Options,OPTION_ACCEL_METHOD ); +if (s !=NULL ) { +#if USE_GLAMOR +if (strcasecmp (s,"glamor" ) ==0 ) + accel_method =GLAMOR ; + else +#endif +#if USE_UXA +if (strcasecmp (s,"uxa" ) ==0 ) + accel_method =UXA ; + else +#endif + accel_method =DEFAULT_ACCEL_METHOD ; + } +switch (accel_method) { +default : +#if USE_GLAMOR
src/uxa/intel_driver.c static void intel_check_chipset_option(ScrnInfoPtr scrn) · line 293
Quoted 10 lines
src/uxa/intel_driver.c +10
-0+case GLAMOR : + intel->accel =ACCEL_GLAMOR ; +break ; +#endif +#if USE_UXA +case UXA : + intel->accel =ACCEL_UXA ; +break ; +#endif + }
The extra enum and temporary variable introduced here seems pretty pointless (even if that pattern had happened before).
Quoted 7 lines
src/uxa/intel_glamor.c +2
-0+Bool +intel_glamor_set_pixmap_bo (PixmapPtr pixmap,drm_intel_bo *bo)
src/uxa/intel_glamor.c line 117
Quoted 10 lines
src/uxa/intel_glamor.c +10
-0+{ +if (bo ==NULL ||glamor_egl_create_textured_pixmap (pixmap, + bo->handle, +intel_pixmap_pitch (pixmap))) + { +intel_glamor_reference_pixmap_bo (pixmap, bo); +return TRUE; + } +return FALSE; +}
I don't think this will work -- glamor_egl uses a different fd from intel->bufmgr, so you're attaching some unrelated BO, if you're even that lucky.
src/uxa/intel_glamor.c intel_glamor_pre_init(ScrnInfoPtr scrn) · line 215
Quoted 11 lines
Bool
intel_glamor_init(ScreenPtr screen)
{src/uxa/intel_glamor.c +3
-3ScrnInfoPtr scrn =xf86ScreenToScrn (screen); -intel_screen_private *intel =intel_get_screen_private (scrn); -if ((intel->uxa_flags &UXA_GLAMOR_EGL_INITIALIZED ) ==0 ) - goto fail; +if (!dixPrivateKeyRegistered (&intel_glamor_pixmap_key)) +if (!dixRegisterPrivateKey (&intel_glamor_pixmap_key,PRIVATE_PIXMAP ,sizeof (struct intel_glamor_pixmap ))) +return FALSE;
No need to check for initiailization -- double-calling it is safe (as long as the size is consistent).
src/uxa/intel_glamor.c intel_glamor_flush(intel_screen_private * intel) · line 244
Quoted 9 lines
void
src/uxa/intel_glamor.c +1
-2@@ -258,18 +240,52 @@ intel_glamor_flush(intel_screen_private * intel)ScreenPtr screen; screen =xf86ScrnToScreen (intel->scrn); -if (intel->uxa_flags &UXA_USE_GLAMOR ) -glamor_block_handler (screen); +glamor_block_handler (screen); }
glamor_block_handler() is pretty awfully named. We should do something about that.
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
A message that is not in the archive (sent off-list or not yet received): 8761iedzxh.fsf-omZaPlIz5HhaEpDpdNBo/[email protected]
Re: [PATCH 10/12] Add glamor back into the driver
Quoted 9 lines · Eric Anholt
Quoted 2 lines · Keith Packard
This adds glamor support back into the driver, but instad of going through UXA, this uses it directly instead.
This is hard to read with the conditionalizing all of the UXA code in the same commit as adding the glamor code. Then there are a bunch of unrelated whitespace changes, or flattening of a bunch of nested conditionals.
I'm not sure how to make the patch easier to review; I guess I could make UXA conditional first? That would be 'crazy' in that the driver would fail to ever work if you didn't ask for UXA, but might make the patch easier to read?
Quoted 4 lines
The only substantive problem I see is
intel_glamor_set_pixmap_bo().The extra enum and temporary variable introduced here seems pretty pointless (even if that pattern had happened before).
Agreed. The problem is that 'DEFAULT_ACCEL_METHOD' is defined as 'GLAMOR', 'UXA' or 'NONE' by configure.ac. This seemed like the cleanest solution in some ways. I also liked having the accel_type enum *not* define acceleration types which weren't compiled into the driver; that caught a few missing #ifdefs
Quoted 3 lines
I don't think this will work --
glamor_egluses a different fd fromintel->bufmgr, so you're attaching some unrelated BO, if you're even that lucky.
This API uses the same FD as the intel bufmgr.
From intel_glamor.c:
if (!glamor_egl_init(scrn, intel->drmSubFD)) {From glamor_egl.c:
Bool
glamor_egl_init(ScrnInfoPtr scrn, int fd)
...
glamor_egl->fd = fd;
...
glamor_egl->has_gem = glamor_egl_check_has_gem(fd);
...
Bool
glamor_egl_create_textured_pixmap(PixmapPtr pixmap, int handle, int stride)
...
if (glamor_egl->has_gem) {
if (!glamor_get_flink_name(glamor_egl->fd, handle, &name)) {So, you pass in a GEM handle for the intel driver bufmgr and glamor does the flink to get a global name.
We should be using an FD passing mechanism here instead, to avoid creating more global names...
Quoted 2 lines
No need to check for initiailization -- double-calling it is safe (as long as the size is consistent).
Thanks. Will fix.
Quoted 4 lines
Quoted 4 lines
void
src/uxa/intel_glamor.c +0
-0@@ -258,18 +240,52 @@ intel_glamor_flush(intel_screen_private * intel)ScreenPtr screen;
Quoted 8 lines
src/uxa/intel_glamor.c intel_glamor_flush(intel_screen_private * intel) · line 244
Quoted 5 lines
src/uxa/intel_glamor.c +1
-2screen =xf86ScrnToScreen (intel->scrn); -if (intel->uxa_flags &UXA_USE_GLAMOR ) -glamor_block_handler (screen); +glamor_block_handler (screen); }
glamor_block_handler()is pretty awfully named. We should do something about that.
Suggestions welcome :-)
Thanks for the careful review.
Signature · 1 line
[email protected]
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
PATCH 11/12 Add "none" acceleration option
In reply to a message not in the archiveThis just uses fb directly
--- configure.ac | 24 ++++- src/Makefile.am | 9 +- src/intel_module.c | 12 ++- src/uxa/Makefile.am | 46 ++++++--- src/uxa/intel.h | 5 + src/uxa/intel_display.c | 6 ++ src/uxa/intel_dri3.c | 49 ++++++---- src/uxa/intel_driver.c | 245 ++++++++++++++++++++++++++++++------------------ src/uxa/intel_glamor.c | 10 ++ src/uxa/intel_none.c | 189 +++++++++++++++++++++++++++++++++++++ src/uxa/intel_none.h | 38 ++++++++ src/uxa/intel_uxa.c | 9 ++ 12 files changed, 504 insertions(+), 138 deletions(-) create mode 100644 src/uxa/intel_none.c create mode 100644 src/uxa/intel_none.h
configure.ac +20 -4
@@ -396,7 +396,6 @@ if test "x$UXA" = "xauto"; then fi if test "x$UXA" != "xno"; then AC_DEFINE(USE_UXA, 1, [Enable UXA support]) - PKG_CHECK_MODULES(DRMINTEL, [libdrm_intel >= $uxa_requires_libdrm]) required_pixman_version=0.24 UXA=yes fi
@@ -407,9 +406,6 @@ AC_ARG_ENABLE(glamor, [GLAMOR="$enableval"], [GLAMOR="no"]) if test "x$GLAMOR" != "xno"; then - if test "x$UXA" != "xyes"; then - AC_MSG_ERROR([Glamor acceleration requested but UXA is not enabled]) - fi if pkg-config --exists "xorg-server >= 1.15.99.901"; then GLAMOR="yes (using Xorg glamor module)" else
@@ -417,9 +413,12 @@ if test "x$GLAMOR" != "xno"; then PKG_CHECK_MODULES(LIBGLAMOR_EGL, [glamor-egl]) GLAMOR="yes (using libglamor)" fi + GLAMOR=yes AC_DEFINE(USE_GLAMOR, 1, [Enable glamor acceleration]) fi +PKG_CHECK_MODULES(DRMINTEL, [libdrm_intel >= $uxa_requires_libdrm]) + PKG_CHECK_MODULES(XORG, [xorg-server >= $required_xorg_server_version xproto fontsproto pixman-1 >= $required_pixman_version $REQUIRED_MODULES]) ABI_VERSION=`$PKG_CONFIG --variable=abi_videodrv xorg-server`
@@ -661,6 +660,12 @@ if test "x$accel" = "xauto"; then else if test "x$UXA" != "xno"; then accel="uxa" + else + if test "x$GLAMOR" != "xno"; then + accel="glamor" + else + accel="none" + fi fi fi if test "x$accel" = "xauto" -a "x$KMS" = "xyes"; then
@@ -696,6 +701,11 @@ if test "x$accel" = "xglamor"; then fi fi +if test "x$accel" = "xnone"; then + AC_DEFINE(DEFAULT_ACCEL_METHOD, NONE, [Default acceleration method]) + have_accel="yes" +fi + if test "x$have_accel" = "xnone"; then if test "x$KMS" = "xyes"; then AC_MSG_ERROR([Invalid default acceleration option])
@@ -886,6 +896,12 @@ if test "x$GLAMOR" != "xno"; then fi fi +if test "$accel" = "none"; then + accel_msg="$accel_msg *none" +else + accel_msg="$accel_msg none" +fi + if test "x$dri_msg" = "x"; then dri_msg=" none" fi
src/Makefile.am +2 -7
@@ -18,7 +18,7 @@ # IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN # CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. -SUBDIRS = render_program legacy +SUBDIRS = render_program legacy uxa # this is obnoxious: # -module lets us name the module exactly how we want
@@ -41,18 +41,13 @@ endif intel_drv_la_LTLIBRARIES = intel_drv.la intel_drv_la_LDFLAGS = -module -avoid-version intel_drv_ladir = $(moduledir)/drivers -intel_drv_la_LIBADD = legacy/liblegacy.la $(PCIACCESS_LIBS) $(XORG_LIBS) +intel_drv_la_LIBADD = legacy/liblegacy.la uxa/libuxa.la $(PCIACCESS_LIBS) $(XORG_LIBS) if SNA SUBDIRS += sna intel_drv_la_LIBADD += sna/libsna.la endif -if UXA -SUBDIRS += uxa -intel_drv_la_LIBADD += uxa/libuxa.la -endif - NULL:=# intel_drv_la_SOURCES = \
src/intel_module.c +8 -4
@@ -489,7 +489,7 @@ _xf86findDriver(const char *ident, XF86ConfDevicePtr p)return NULL ; } -static enum accel_method {SNA ,UXA ,GLAMOR }get_accel_method (void ) +static enum accel_method {NONE ,SNA ,UXA ,GLAMOR }get_accel_method (void ) {enum accel_method accel_method =DEFAULT_ACCEL_METHOD ;XF86ConfDevicePtr dev;
@@ -509,6 +509,8 @@ static enum accel_method { SNA, UXA, GLAMOR } get_accel_method(void) accel_method =UXA ; else if (strcasecmp(s, "glamor") == 0) accel_method =GLAMOR ; + else if (strcasecmp(s, "none") == 0) + accel_method =NONE ; } }
@@ -564,12 +566,14 @@ intel_scrn_create(DriverPtr driver,#if USE_SNA case SNA :return sna_init_scrn (scrn, entity_num);#endif -#if USE_UXA +#if USE_GLAMOR case GLAMOR : +#endif +#if USE_UXA case UXA : -return intel_init_scrn (scrn);#endif - +case NONE : +return intel_init_scrn (scrn); default: break; }#endif
src/uxa/Makefile.am +33 -13
78 lines not on this page · raw message
src/uxa/intel.h +5 -0
@@ -112,6 +112,7 @@ enum dri_type { };enum accel_type { +ACCEL_NONE ,#if USE_GLAMOR ACCEL_GLAMOR ,#endif
@@ -126,9 +127,11 @@ typedef struct intel_screen_private {enum accel_type accel; +#if USE_UXA #define RENDER_BATCH I915_EXEC_RENDER#define BLT_BATCH I915_EXEC_BLTunsigned int current_batch; +#endif void *modes;drm_intel_bo *front_buffer, *back_buffer;
@@ -544,7 +547,9 @@ Bool intel_present_screen_init(ScreenPtr screen);static inline Bool intel_present_screen_init (ScreenPtr screen) {return 0 ; }#endif +#if USE_UXA #include "intel_uxa.h" +#endif dri_bo *intel_get_pixmap_bo (PixmapPtr pixmap);
src/uxa/intel_display.c +6 -0
@@ -56,9 +56,11 @@#include "xf86DDC.h" #include "fb.h" #include "uxa.h" +#if USE_GLAMOR #include "intel_glamor.h" #endif +#include "intel_none.h" #define KNOWN_MODE_FLAGS ((1<<14)-1)
@@ -1491,6 +1493,10 @@ intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height) goto fail;break ;#endif +case ACCEL_NONE : +if (!intel_none_create_screen_resources (scrn->pScreen)) + goto fail; +break ; }for (i =0 ; i < xf86_config->num_crtc; i++) {
src/uxa/intel_dri3.c +30 -19
74 lines not on this page · raw message
src/uxa/intel_driver.c +154 -91
404 lines not on this page · raw message
src/uxa/intel_glamor.c +10 -0
@@ -36,6 +36,7 @@#define GLAMOR_FOR_XORG 1#include <glamor.h> #include <unistd.h> +#include <mipict.h> #include "intel.h" #include "i915_drm.h"
@@ -210,6 +211,15 @@ intel_glamor_init(ScreenPtr screen) {ScrnInfoPtr scrn =xf86ScreenToScrn (screen); +/* Must be first, before anything else installs screen callbacks. */ +if (!fbScreenInit (screen,NULL , + scrn->virtualX, scrn->virtualY, + scrn->xDpi, scrn->yDpi, + scrn->displayWidth, scrn->bitsPerPixel)) +return FALSE; + +fbPictureInit (screen,NULL ,0 ); +if (!dixPrivateKeyRegistered (&intel_glamor_pixmap_key))if (!dixRegisterPrivateKey (&intel_glamor_pixmap_key,PRIVATE_PIXMAP ,sizeof (struct intel_glamor_pixmap )))return FALSE;
src/uxa/intel_none.c new file +189 -0
189 lines not on this page · raw message
src/uxa/intel_none.h new file +38 -0
@@ -0,0 +1,38 @@ +/* +* Copyright © 2014 Keith Packard +* +* Permission to use, copy, modify, distribute, and sell this software and its +* documentation for any purpose is hereby granted without fee, provided that +* the above copyright notice appear in all copies and that both that copyright +* notice and this permission notice appear in supporting documentation, and +* that the name of the copyright holders not be used in advertising or +* publicity pertaining to distribution of the software without specific, +* written prior permission. The copyright holders make no representations +* about the suitability of this software for any purpose. It is provided "as +* is" without express or implied warranty. +* +* THE COPYRIGHT HOLDERS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, +* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO +* EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY SPECIAL, INDIRECT OR +* CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, +* DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER +* TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE +* OF THIS SOFTWARE. +*/ + +#ifndef _INTEL_NONE_H_ +#define _INTEL_NONE_H_ + +Bool +intel_none_init (ScreenPtr screen); + +Bool +intel_none_create_screen_resources (ScreenPtr screen); + +dri_bo * +intel_none_get_pixmap_bo (PixmapPtr pixmap); + +Bool +intel_none_set_pixmap_bo (PixmapPtr pixmap,dri_bo *dri_bo); + +#endif /* _INTEL_NONE_H_ */
src/uxa/intel_uxa.c +9 -0
@@ -1272,6 +1272,15 @@ Bool intel_uxa_init(ScreenPtr screen)ScrnInfoPtr scrn =xf86ScreenToScrn (screen);intel_screen_private *intel =intel_get_screen_private (scrn); +/* Must be first, before anything else installs screen callbacks. */ +if (!fbScreenInit (screen,NULL , + scrn->virtualX, scrn->virtualY, + scrn->xDpi, scrn->yDpi, + scrn->displayWidth, scrn->bitsPerPixel)) +return FALSE; + +fbPictureInit (screen,NULL ,0 ); +intel_batch_init (scrn);if (INTEL_INFO (intel)->gen >=040 &&INTEL_INFO (intel)->gen <0100 )
Signature · 6 lines
2.0.1 _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
PATCH 12/12 Delay initial modeset until root window contents are prepared
In reply to a message not in the archiveWait until the root window has been painted for the first time before doing the modeset. This avoids flashing black while the root window gets set up.
--- src/uxa/intel.h | 2 ++ src/uxa/intel_display.c | 39 ++++++++++++++++++--------------------- src/uxa/intel_driver.c | 40 +++++++++++++++++++++++++++++++++++----- 3 files changed, 55 insertions(+), 26 deletions(-)
src/uxa/intel.h +2 -0
@@ -194,6 +194,8 @@ typedef struct intel_screen_private {void (*batch_flush) (struct intel_screen_private *intel);void (*batch_commit_notify) (struct intel_screen_private *intel); +Bool need_entervt; +#if USE_UXA struct _UxaDriver *uxa_driver;int uxa_flags;
src/uxa/intel_display.c +18 -21
@@ -2319,7 +2319,7 @@ void intel_copy_fb(ScrnInfoPtr scrn)ScreenPtr pScreen =xf86ScrnToScreen (scrn);intel_screen_private *intel =intel_get_screen_private (scrn);PixmapPtr src, dst; -unsigned int pitch = scrn->displayWidth * intel->cpp; +GCPtr gc;struct intel_crtc *intel_crtc;int i, fbcon_id;
@@ -2339,30 +2339,27 @@ void intel_copy_fb(ScrnInfoPtr scrn)if (src ==NULL )return ; -/* We dont have a screen Pixmap yet */ - dst =intel_create_pixmap_for_bo (pScreen, intel->front_buffer, - scrn->virtualX, scrn->virtualY, - scrn->depth, scrn->bitsPerPixel, - pitch); + dst = (*pScreen->GetScreenPixmap)(pScreen); +if (dst == NullPixmap) goto cleanup_src; -if (!intel->uxa_driver->prepare_copy (src, dst, --1 ,-1 , - GXcopy,FB_ALLONES )) - goto cleanup_dst; - - intel->uxa_driver->copy (dst, -0 ,0 , -0 ,0 , - scrn->virtualX, scrn->virtualY); - intel->uxa_driver->done_copy (dst); -#if ABI_VIDEODRV_VERSION >=SET_ABI_VERSION (10 ,0 ) - pScreen->canDoBGNoneRoot = TRUE; -#endif + gc =GetScratchGC (scrn->depth, pScreen); + +if (!gc) + goto cleanup_src; + +ValidateGC (&dst->drawable, gc); + + (*gc->ops->CopyArea)(&src->drawable, + &dst->drawable, + gc, +0 ,0 , + scrn->virtualX, scrn->virtualY, +0 ,0 ); + +FreeScratchGC (gc); -cleanup_dst: - (*pScreen->DestroyPixmap)(dst); cleanup_src: (*pScreen->DestroyPixmap)(src); }
src/uxa/intel_driver.c +35 -5
@@ -170,13 +170,19 @@ static Bool i830CreateScreenResources(ScreenPtr screen)case ACCEL_GLAMOR :if (!intel_glamor_create_screen_resources (screen))return FALSE; +#if ABI_VIDEODRV_VERSION >=SET_ABI_VERSION (10 ,0 ) + screen->canDoBGNoneRoot = TRUE; +#endif break ;#endif #if USE_UXA case ACCEL_UXA :if (!intel_uxa_create_screen_resources (screen))return FALSE; -intel_copy_fb (scrn); +#if ABI_VIDEODRV_VERSION >=SET_ABI_VERSION (10 ,0 ) + screen->canDoBGNoneRoot = TRUE; +#endif +break ;#endif case ACCEL_NONE :if (!intel_none_create_screen_resources (screen))
@@ -830,6 +836,29 @@ I830BlockHandler(BLOCKHANDLER_ARGS_DECL) screen->BlockHandler = intel->BlockHandler; +/* At server init time, get the root window bits from fbcon if +* available and then update the protocol-level information +* for both RandR and XINERAMA +*/ +if (intel->need_entervt) { + intel->need_entervt = FALSE; + +if (screen->root->backgroundState == None) +intel_copy_fb (scrn); + +/* Must force it before EnterVT, so we are in control of VT and +* later memory should be bound when allocating, e.g rotate_mem */ + scrn->vtSema = TRUE; + +if (!I830EnterVT (VT_FUNC_ARGS (0 ))) { +xf86DrvMsg (scrn->scrnIndex,X_ERROR , +"Failed to perform initial modeset.\n" ); +FatalError ("Mode set failed\n" ); + } + +xf86RandR12CreateScreenResources (screen); + } + (*screen->BlockHandler) (BLOCKHANDLER_ARGS ); intel->BlockHandler = screen->BlockHandler;
@@ -1211,11 +1240,12 @@ I830ScreenInit(SCREEN_INIT_ARGS_DECL)I830UeventInit (scrn);#endif -/* Must force it before EnterVT, so we are in control of VT and -* later memory should be bound when allocating, e.g rotate_mem */ - scrn->vtSema = TRUE; +/* Delay the call to I830EnterVT until +* the server is all ready to accept protocol requests. +*/ + intel->need_entervt = TRUE; -return I830EnterVT (VT_FUNC_ARGS (0 )); +return TRUE; }static void i830AdjustFrame (ADJUST_FRAME_ARGS_DECL )
Signature · 6 lines
2.0.1 _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
Quoted 3 lines · Keith Packard
I spent the day just cleaning up this patch series and testing. I think it's ready for others to use and review. I've been running it on two machines for a couple of days now and it's been solid.
Patches 2, 4 are:
Patch 5 is:
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
Quoted 6 lines · Eric Anholt
Quoted 3 lines · Keith Packard
I spent the day just cleaning up this patch series and testing. I think it's ready for others to use and review. I've been running it on two machines for a couple of days now and it's been solid.
Quoted 7 lines
Patches 2, 4 are:
Reviewed-by Eric Anholt <[email protected]>Patch 5 is:
Acked-by Eric Anholt <[email protected]>
Thanks.
Signature · 1 line
[email protected]
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 6 lines · Eric Anholt
Quoted 3 lines · Keith Packard
I spent the day just cleaning up this patch series and testing. I think it's ready for others to use and review. I've been running it on two machines for a couple of days now and it's been solid.
Quoted 7 lines
Patches 2, 4 are:
Reviewed-by Eric Anholt <[email protected]>Patch 5 is:
Acked-by Eric Anholt <[email protected]>
I've pushed a new version of the tree with your review marked and a few changes:
* Split the GetScratchPixmapHeader change into two pieces; the first just removes the redundant calls to drm_intel_bo_disable_reuse and the second contains the scratch pixmap changes and the other misc stuff. Yes, I could split the misc changes out into another cleanup patch if you want.
* Removed the tiling check from intel_present when flipping. The kernel doesn't appear to ever require matching tiling.
* Removed call to dixPrivateKeyRegistered in the 'Add glamor back' patch.
What I didn't do is clean up the 'remove glamor' patch so that it would compile even with --enable-glamor. Fixing that seems like noise to me; all of the changes needed to make it work would be immediately un-done when adding glamor back.
Signature · 1 line
[email protected]
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 3/12 Don't use GetScratchPixmapHeader for shadow pixmaps
In reply to Keith PackardGetScratchPixmapHeader should only be used for local memory pixmaps, as used by PutImage and friends. That's because when you free the scratch pixmap header, it doesn't actually free the pixmap; instead, it gets stuffed in pScreen->pScratchPixmap and any private data stored on it will be left hanging around forever.
In the case of glamor, that private data includes all of the GL state. Using that scratch pixmap later results in glamor getting mightily confused as the pixmap and underlying objects do not match.
Avoid this by allocating pixmap headers explicitly for this purpose.
--- src/uxa/intel_display.c | 44 ++++++++++++++++++++++++++++++-------------- 1 file changed, 30 insertions(+), 14 deletions(-)
src/uxa/intel_display.c +30 -14
@@ -545,13 +545,31 @@ intel_crtc_shadow_allocate(xf86CrtcPtr crtc, int width, int height)return NULL ; } -drm_intel_bo_disable_reuse (intel_crtc->rotate_bo); - intel_crtc->rotate_pitch = rotate_pitch;return intel_crtc->rotate_bo; }static PixmapPtr +intel_create_pixmap_header (ScreenPtr pScreen,int width,int height,int depth, +int bitsPerPixel,int devKind,void *pPixData) +{ +PixmapPtr pixmap; + +/* width and height of 0 means don't allocate any pixmap data */ + pixmap = (*pScreen->CreatePixmap) (pScreen,0 ,0 , depth,0 ); + +if (pixmap) { +if ((*pScreen->ModifyPixmapHeader) (pixmap, width, height, depth, + bitsPerPixel, devKind, pPixData)) + { +return pixmap; + } + (*pScreen->DestroyPixmap) (pixmap); + } +return NullPixmap; +} + +static PixmapPtr intel_crtc_shadow_create (xf86CrtcPtr crtc,void *data,int width,int height) {ScrnInfoPtr scrn = crtc->scrn;
@@ -573,12 +591,12 @@ intel_crtc_shadow_create(xf86CrtcPtr crtc, void *data, int width, int height)return NULL ; } - rotate_pixmap =GetScratchPixmapHeader (scrn->pScreen, - width, height, - scrn->depth, - scrn->bitsPerPixel, - intel_crtc->rotate_pitch, -NULL ); + rotate_pixmap =intel_create_pixmap_header (scrn->pScreen, + width, height, + scrn->depth, + scrn->bitsPerPixel, + intel_crtc->rotate_pitch, +NULL );if (rotate_pixmap ==NULL ) { xf86DrvMsg(scrn->scrnIndex, X_ERROR,
@@ -602,8 +620,8 @@ intel_crtc_shadow_destroy(xf86CrtcPtr crtc, PixmapPtr rotate_pixmap, void *data)struct intel_mode *mode = intel_crtc->mode;if (rotate_pixmap) { -intel_set_pixmap_bo (rotate_pixmap,NULL ); -FreeScratchPixmapHeader (rotate_pixmap); +intel_set_pixmap_bo (rotate_pixmap,NULL ); + rotate_pixmap->drawable.pScreen->DestroyPixmap (rotate_pixmap); }if (data) {
@@ -1415,7 +1433,6 @@ intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height)int i, old_width, old_height, old_pitch;int pitch;uint32_t tiling; -ScreenPtr screen;if (scrn->virtualX == width && scrn->virtualY == height)return TRUE;
@@ -1430,8 +1447,7 @@ intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height) old_front = intel->front_buffer;if (intel->back_pixmap) { - screen = intel->back_pixmap->drawable.pScreen; - screen->DestroyPixmap (intel->back_pixmap); + scrn->pScreen->DestroyPixmap (intel->back_pixmap); intel->back_pixmap =NULL ; }
@@ -1454,7 +1470,6 @@ intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height)if (ret) goto fail; -drm_intel_bo_disable_reuse (intel->front_buffer); intel->front_pitch = pitch; intel->front_tiling = tiling;
@@ -2204,6 +2219,7 @@ Bool intel_crtc_on(xf86CrtcPtr crtc)return ret; } + static PixmapPtr intel_create_pixmap_for_bo(ScreenPtr pScreen, dri_bo *bo, int width, int height,
Signature · 1 line
2.0.1
A message that is not in the archive (sent off-list or not yet received): 1406243908-1123-4-git-send-email-keithp-aN4HjG94KOLQT0dZR+AlfA@public.gmane.org
Re: [PATCH 03/12] Don't use GetScratchPixmapHeader for shadow pixmaps
Quoted 22 lines · Keith Packard
GetScratchPixmapHeader should only be used for local memory pixmaps, as used by PutImage and friends. That's because when you free the scratch pixmap header, it doesn't actually free the pixmap; instead, it gets stuffed in
pScreen->pScratchPixmapand any private data stored on it will be left hanging around forever.In the case of glamor, that private data includes all of the GL state. Using that scratch pixmap later results in glamor getting mightily confused as the pixmap and underlying objects do not match.
Avoid this by allocating pixmap headers explicitly for this purpose.
Signed-off-by Keith Packard <[email protected]>--- src/uxa/intel_display.c | 44 ++++++++++++++++++++++++++++++-------------- 1 file changed, 30 insertions(+), 14 deletions(-)src/uxa/intel_display.c +0
-0
src/uxa/intel_display.c intel_crtc_shadow_allocate(xf86CrtcPtr crtc, int width, int height) · line 550
Quoted 8 lines
src/uxa/intel_display.c +0
-2@@ -545,13 +545,31 @@ intel_crtc_shadow_allocate(xf86CrtcPtr crtc, int width, int height)return NULL ; } -drm_intel_bo_disable_reuse (intel_crtc->rotate_bo); - intel_crtc->rotate_pitch = rotate_pitch;return intel_crtc->rotate_bo; }
See comment below...
Quoted 20 lines
src/uxa/intel_display.c +2
-3@@ -602,8 +620,8 @@ intel_crtc_shadow_destroy(xf86CrtcPtr crtc, PixmapPtr rotate_pixmap, void *data)struct intel_mode *mode = intel_crtc->mode;if (rotate_pixmap) { -intel_set_pixmap_bo (rotate_pixmap,NULL ); -FreeScratchPixmapHeader (rotate_pixmap); +intel_set_pixmap_bo (rotate_pixmap,NULL ); + rotate_pixmap->drawable.pScreen->DestroyPixmap (rotate_pixmap); }if (data) {@@ -1415,7 +1433,6 @@ intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height)int i, old_width, old_height, old_pitch;int pitch;uint32_t tiling; -ScreenPtr screen;if (scrn->virtualX == width && scrn->virtualY == height)return TRUE;
src/uxa/intel_display.c intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height) · line 1453
Quoted 9 lines
src/uxa/intel_display.c +1
-2@@ -1430,8 +1447,7 @@ intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height) old_front = intel->front_buffer;if (intel->back_pixmap) { - screen = intel->back_pixmap->drawable.pScreen; - screen->DestroyPixmap (intel->back_pixmap); + scrn->pScreen->DestroyPixmap (intel->back_pixmap); intel->back_pixmap =NULL ; }
Grumble, unrelated noise in the patch.
src/uxa/intel_display.c intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height) · line 1475
Quoted 8 lines
src/uxa/intel_display.c +0
-1@@ -1454,7 +1470,6 @@ intel_xf86crtc_resize(ScrnInfoPtr scrn, int width, int height)if (ret) goto fail; -drm_intel_bo_disable_reuse (intel->front_buffer); intel->front_pitch = pitch; intel->front_tiling = tiling;
This change appears to be unrelated, and possibly harmful (if X has dropped the last ref to the BO, but it's still the scanout buffer, a new allocation would now reuse the BO and scribble on scanout until the next modeset happens).
src/uxa/intel_display.c Bool intel_crtc_on(xf86CrtcPtr crtc) · line 2225
Quoted 8 lines
src/uxa/intel_display.c +1
-0@@ -2204,6 +2219,7 @@ Bool intel_crtc_on(xf86CrtcPtr crtc)return ret; } + static PixmapPtr intel_create_pixmap_for_bo(ScreenPtr pScreen, dri_bo *bo, int width, int height,
Unrelated whitespace.
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
Re: [PATCH 03/12] Don't use GetScratchPixmapHeader for shadow pixmaps
Quoted 4 lines · Eric Anholt
This change appears to be unrelated, and possibly harmful (if X has dropped the last ref to the BO, but it's still the scanout buffer, a new allocation would now reuse the BO and scribble on scanout until the next modeset happens).
Yeah, it's unrelated. intel_allocate_framebuffer calls disable_reuse, so there's no need to call it from these two other places. I'll split that change out into a separate patch with separate comment.
Quoted 1 line
Unrelated whitespace.
There are a bunch of whitespace fixups; should I pull those into a separate patch or just leave them scattered in the first patch to change a file?
Signature · 1 line
[email protected]
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 03/12] Don't use GetScratchPixmapHeader for shadow pixmaps
Quoted 11 lines · Keith Packard
Quoted 4 lines · Eric Anholt
This change appears to be unrelated, and possibly harmful (if X has dropped the last ref to the BO, but it's still the scanout buffer, a new allocation would now reuse the BO and scribble on scanout until the next modeset happens).
Yeah, it's unrelated.
intel_allocate_framebuffercallsdisable_reuse, so there's no need to call it from these two other places. I'll split that change out into a separate patch with separate comment.
Quoted 5 lines
Quoted 1 line
Unrelated whitespace.
There are a bunch of whitespace fixups; should I pull those into a separate patch or just leave them scattered in the first patch to change a file?
One patch at the front is fine with me.
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 4/12 Move intel_alloc_framebuffer to intel_memory.c
In reply to Keith PackardUXA and Glamor both share this function, so move it out of the UXA file.
--- src/uxa/intel.h | 5 ++ src/uxa/intel_memory.c | 138 ++++++++++++++++++++++++++++++++++++++++++ src/uxa/intel_uxa.c | 158 ++++--------------------------------------------- 3 files changed, 154 insertions(+), 147 deletions(-)
src/uxa/intel.h +5 -0
@@ -539,6 +539,11 @@ unsigned long intel_get_fence_pitch(intel_screen_private *intel, unsigned long pBool intel_check_display_stride (ScrnInfoPtr scrn,int stride,Bool tiling);void intel_set_gem_max_sizes (ScrnInfoPtr scrn); +unsigned int +intel_compute_size (struct intel_screen_private *intel, +int w,int h,int bpp,unsigned usage, +uint32_t *tiling,int *stride); + drm_intel_bo *intel_allocate_framebuffer(ScrnInfoPtr scrn, int width, int height, int cpp, int *out_stride,
src/uxa/intel_memory.c +138 -0
141 lines not on this page · raw message
src/uxa/intel_uxa.c +11 -147
182 lines not on this page · raw message
Signature · 1 line
2.0.1
PATCH 5/12 Rename uxa-specific functions and structs
In reply to Keith PackardThis starts to separate uxa from the rest of the driver by renaming uxa-specific functions and structs to make it clear which portions of the driver are uxa-specific and which are common.
--- src/uxa/Makefile.am | 4 + src/uxa/i830_render.c | 82 +-- src/uxa/i915_render.c | 100 ++-- src/uxa/i915_video.c | 10 +- src/uxa/i965_render.c | 98 +-- src/uxa/i965_video.c | 48 +- src/uxa/intel.h | 267 +-------- src/uxa/intel_batchbuffer.c | 8 +- src/uxa/intel_batchbuffer.h | 4 +- src/uxa/intel_display.c | 78 +++ src/uxa/intel_dri.c | 26 +- src/uxa/intel_dri3.c | 10 +- src/uxa/intel_driver.c | 39 +- src/uxa/intel_glamor.c | 4 +- src/uxa/intel_uxa.c | 136 +++-- src/uxa/intel_uxa.h | 300 ++++++++++ src/uxa/intel_uxa_video.c | 389 ++++++++++++ src/uxa/intel_video.c | 1312 +++++++---------------------------------- src/uxa/intel_video.h | 156 +++-- src/uxa/intel_video_overlay.c | 554 +++++++++++++++++ src/uxa/intel_video_overlay.h | 51 ++ 21 files changed, 2004 insertions(+), 1672 deletions(-) create mode 100644 src/uxa/intel_uxa.h create mode 100644 src/uxa/intel_uxa_video.c create mode 100644 src/uxa/intel_video_overlay.c create mode 100644 src/uxa/intel_video_overlay.h
src/uxa/Makefile.am +4 -0
@@ -25,6 +25,7 @@ AM_CFLAGS += -I$(top_srcdir)/xvmc -I$(top_srcdir)/src -I$(top_srcdir)/src/render noinst_LTLIBRARIES = libuxa.la libuxa_la_LIBADD = $(UDEV_LIBS) $(DRMINTEL_LIBS) $(DRM_LIBS) libuxa_la_SOURCES = \ + intel_uxa.h \ brw_defines.h \ brw_structs.h \ common.h \
@@ -38,6 +39,9 @@ libuxa_la_SOURCES = \ intel_uxa.c \ intel_video.c \ intel_video.h \ + intel_video_overlay.c \ + intel_video_overlay.h \ + intel_uxa_video.c \ i830_3d.c \ i830_render.c \ i830_reg.h \
src/uxa/i830_render.c +41 -41
286 lines not on this page · raw message
src/uxa/i915_render.c +50 -50
350 lines not on this page · raw message
src/uxa/i915_video.c +5 -5
@@ -67,7 +67,7 @@ I915DisplayVideoTextured(ScrnInfoPtr scrn, dyo = dstRegion->extents.y1;if (pixmap->drawable.width >2048 || pixmap->drawable.height >2048 || - !intel_check_pitch_3d (pixmap)) { + !intel_uxa_check_pitch_3d (pixmap)) {ScreenPtr screen = pixmap->drawable.pScreen; target = screen->CreatePixmap(screen,
@@ -78,7 +78,7 @@ I915DisplayVideoTextured(ScrnInfoPtr scrn,if (target ==NULL )return ; -if (intel_get_pixmap_bo (target) ==NULL ) { +if (intel_uxa_get_pixmap_bo (target) ==NULL ) { screen->DestroyPixmap (target);return ; }
@@ -159,9 +159,9 @@ I915DisplayVideoTextured(ScrnInfoPtr scrn, DSTORG_VERT_BIAS(0x8) | format);/* front buffer, pitch, offset */ -if (intel_pixmap_tiled (target)) { +if (intel_uxa_pixmap_tiled (target)) { tiling =BUF_3D_TILED_SURFACE ; -if (intel_get_pixmap_private (target)->tiling ==I915_TILING_Y ) +if (intel_uxa_get_pixmap_private (target)->tiling ==I915_TILING_Y ) tiling |=BUF_3D_TILE_WALK_Y ; }else tiling =0 ;
@@ -488,5 +488,5 @@ I915DisplayVideoTextured(ScrnInfoPtr scrn, target->drawable.pScreen->DestroyPixmap (target); } -intel_debug_flush (scrn); +intel_uxa_debug_flush (scrn); }
src/uxa/i965_render.c +49 -49
377 lines not on this page · raw message
src/uxa/i965_video.c +24 -24
189 lines not on this page · raw message
src/uxa/intel.h +18 -249
302 lines not on this page · raw message
src/uxa/intel_batchbuffer.c +4 -4
@@ -142,7 +142,7 @@ void intel_batch_teardown(ScrnInfoPtr scrn)static void intel_batch_do_flush (ScrnInfoPtr scrn) {intel_screen_private *intel =intel_get_screen_private (scrn); -struct intel_pixmap *priv; +struct intel_uxa_pixmap *priv;list_for_each_entry (priv, &intel->batch_pixmaps, batch) priv->dirty =0 ;
@@ -289,10 +289,10 @@ void intel_batch_submit(ScrnInfoPtr scrn) }while (!list_is_empty (&intel->batch_pixmaps)) { -struct intel_pixmap *entry; +struct intel_uxa_pixmap *entry; entry =list_first_entry (&intel->batch_pixmaps, - struct intel_pixmap, + struct intel_uxa_pixmap, batch); entry->busy =-1 ;
@@ -311,7 +311,7 @@ void intel_batch_submit(ScrnInfoPtr scrn) intel->current_batch =0 ; } -void intel_debug_flush (ScrnInfoPtr scrn) +void intel_uxa_debug_flush (ScrnInfoPtr scrn) {intel_screen_private *intel =intel_get_screen_private (scrn);
src/uxa/intel_batchbuffer.h +2 -2
@@ -129,7 +129,7 @@ intel_batch_emit_reloc(intel_screen_private *intel,static inline void intel_batch_mark_pixmap_domains (intel_screen_private *intel, -struct intel_pixmap *priv, +struct intel_uxa_pixmap *priv,uint32_t read_domains,uint32_t write_domain) {assert (read_domains);
@@ -149,7 +149,7 @@ intel_batch_emit_reloc_pixmap(intel_screen_private *intel, PixmapPtr pixmap, uint32_t read_domains, uint32_t write_domain, uint32_t delta, int needs_fence) { - struct intel_pixmap *priv = intel_get_pixmap_private(pixmap); + struct intel_uxa_pixmap *priv = intel_uxa_get_pixmap_private(pixmap);intel_batch_mark_pixmap_domains (intel, priv, read_domains, write_domain);
src/uxa/intel_display.c +78 -0
81 lines not on this page · raw message
src/uxa/intel_dri.c +13 -13
@@ -82,7 +82,7 @@ static int i830_client_key;static uint32_t pixmap_flink (PixmapPtr pixmap) { -struct intel_pixmap *priv =intel_get_pixmap_private (pixmap); +struct intel_uxa_pixmap *priv =intel_uxa_get_pixmap_private (pixmap);uint32_t name;if (priv ==NULL || priv->bo ==NULL )
@@ -113,7 +113,7 @@ static PixmapPtr fixup_glamor(DrawablePtr drawable, PixmapPtr pixmap)ScrnInfoPtr scrn =xf86ScreenToScrn (screen);intel_screen_private *intel =intel_get_screen_private (scrn);PixmapPtr old =get_drawable_pixmap (drawable); -struct intel_pixmap *priv =intel_get_pixmap_private (pixmap); +struct intel_uxa_pixmap *priv =intel_uxa_get_pixmap_private (pixmap);GCPtr gc;/* With a glamor pixmap, 2D pixmaps are created in texture
@@ -140,12 +140,12 @@ static PixmapPtr fixup_glamor(DrawablePtr drawable, PixmapPtr pixmap)FreeScratchGC (gc); } -intel_set_pixmap_private (pixmap,NULL ); +intel_uxa_set_pixmap_private (pixmap,NULL );/* Exchange the underlying texture/image. */ intel_glamor_exchange_buffers (intel, old, pixmap);/* And redirect the pixmap to the new bo (for 3D). */ -intel_set_pixmap_private (old, priv); +intel_uxa_set_pixmap_private (old, priv); old->refcnt++; screen->ModifyPixmapHeader(old,
@@ -739,10 +739,10 @@ i830_dri2_del_frame_event(DRI2FrameEventPtr info)free (info); } -static struct intel_pixmap * +static struct intel_uxa_pixmap *intel_exchange_pixmap_buffers (struct intel_screen_private *intel,PixmapPtr front,PixmapPtr back) { -struct intel_pixmap *new_front, *new_back; +struct intel_uxa_pixmap *new_front =NULL , *new_back;RegionRec region;/* Post damage on the front buffer so that listeners, such
@@ -755,10 +755,10 @@ intel_exchange_pixmap_buffers(struct intel_screen_private *intel, PixmapPtr fron region.data =NULL ;DamageRegionAppend (&front->drawable, ®ion); - new_front =intel_get_pixmap_private (back); - new_back =intel_get_pixmap_private (front); -intel_set_pixmap_private (front, new_front); -intel_set_pixmap_private (back, new_back); + new_front =intel_uxa_get_pixmap_private (back); + new_back =intel_uxa_get_pixmap_private (front); +intel_uxa_set_pixmap_private (front, new_front); +intel_uxa_set_pixmap_private (back, new_back); new_front->busy =1 ; new_back->busy =-1 ;
@@ -774,7 +774,7 @@ I830DRI2ExchangeBuffers(struct intel_screen_private *intel, DRI2BufferPtr front, {I830DRI2BufferPrivatePtr front_priv, back_priv;int tmp; -struct intel_pixmap *new_front; +struct intel_uxa_pixmap *new_front; front_priv = front->driverPrivate; back_priv = back->driverPrivate;
@@ -967,8 +967,8 @@ can_exchange(DrawablePtr drawable, DRI2BufferPtr front, DRI2BufferPtr back)I830DRI2BufferPrivatePtr back_priv = back->driverPrivate;PixmapPtr front_pixmap = front_priv->pixmap;PixmapPtr back_pixmap = back_priv->pixmap; -struct intel_pixmap *front_intel =intel_get_pixmap_private (front_pixmap); -struct intel_pixmap *back_intel =intel_get_pixmap_private (back_pixmap); +struct intel_uxa_pixmap *front_intel =intel_uxa_get_pixmap_private (front_pixmap); +struct intel_uxa_pixmap *back_intel =intel_uxa_get_pixmap_private (back_pixmap);if (!pScrn->vtSema)return FALSE;
src/uxa/intel_dri3.c +5 -5
@@ -56,7 +56,7 @@ static PixmapPtr intel_dri3_pixmap_from_fd(ScreenPtr screen, {ScrnInfoPtr scrn =xf86ScreenToScrn (screen);intel_screen_private *intel =intel_get_screen_private (scrn); -struct intel_pixmap *priv; +struct intel_uxa_pixmap *priv;PixmapPtr pixmap;dri_bo *bo;
@@ -84,10 +84,10 @@ static PixmapPtr intel_dri3_pixmap_from_fd(ScreenPtr screen,if (bo ==NULL ) goto free_pixmap; -intel_set_pixmap_bo (pixmap, bo); +intel_uxa_set_pixmap_bo (pixmap, bo);dri_bo_unreference (bo); - priv =intel_get_pixmap_private (pixmap); + priv =intel_uxa_get_pixmap_private (pixmap);if (priv ==NULL ) goto free_pixmap;
@@ -105,10 +105,10 @@ static int intel_dri3_fd_from_pixmap(ScreenPtr screen, CARD16 *stride, CARD32 *size) { -struct intel_pixmap *priv; +struct intel_uxa_pixmap *priv;int fd; - priv =intel_get_pixmap_private (pixmap); + priv =intel_uxa_get_pixmap_private (pixmap);if (!priv)return -1 ;
src/uxa/intel_driver.c +14 -25
@@ -169,6 +169,18 @@ static Bool i830CreateScreenResources(ScreenPtr screen)return TRUE; } +void +intel_set_pixmap_bo (PixmapPtr pixmap,dri_bo *bo) +{ +intel_uxa_set_pixmap_bo (pixmap, bo); +} + +dri_bo * +intel_get_pixmap_bo (PixmapPtr pixmap) +{ +return intel_uxa_get_pixmap_bo (pixmap); +} +static void PreInitCleanup (ScrnInfoPtr scrn) {if (!scrn || !scrn->driverPrivate)
@@ -612,29 +624,6 @@ static Bool I830PreInit(ScrnInfoPtr scrn, int flags)return TRUE; } -/** -* Intialiazes the hardware for the 3D pipeline use in the 2D driver. -* -* Some state caching is performed to avoid redundant state emits. This -* function is also responsible for marking the state as clobbered for DRI -* clients. -*/ -void IntelEmitInvarientState (ScrnInfoPtr scrn) -{ -intel_screen_private *intel =intel_get_screen_private (scrn); - -/* If we've emitted our state since the last clobber by another client, -* skip it. -*/ -if (intel->last_3d !=LAST_3D_OTHER ) -return ; - -if (IS_GEN2 (intel)) -I830EmitInvarientState (scrn); - else if IS_GEN3(intel) -I915EmitInvarientState (scrn); -} -#ifdef INTEL_PIXMAP_SHARING static void redisplay_dirty (ScreenPtr screen,PixmapDirtyUpdatePtr dirty)
@@ -660,7 +649,7 @@ redisplay_dirty(ScreenPtr screen, PixmapDirtyUpdatePtr dirty)intel_batch_submit (scrn);if (!intel->has_prime_vmap_flush) { -drm_intel_bo *bo =intel_get_pixmap_bo (dirty->slave_dst->master_pixmap); +drm_intel_bo *bo =intel_uxa_get_pixmap_bo (dirty->slave_dst->master_pixmap); was_blocked =xf86BlockSIGIO ();drm_intel_bo_map (bo, FALSE);drm_intel_bo_unmap (bo);
@@ -1000,7 +989,7 @@ I830ScreenInit(SCREEN_INIT_ARGS_DECL)#endif /* Init video */ if (intel->XvEnabled) -I830InitVideo (screen); +intel_video_init (screen);#if HAVE_DRI2 switch (intel->dri2) {
src/uxa/intel_glamor.c +2 -2
@@ -156,12 +156,12 @@ intel_glamor_create_textured_pixmap(PixmapPtr pixmap) {ScrnInfoPtr scrn =xf86ScreenToScrn (pixmap->drawable.pScreen);intel_screen_private *intel =intel_get_screen_private (scrn); -struct intel_pixmap *priv; +struct intel_uxa_pixmap *priv;if ((intel->uxa_flags &UXA_USE_GLAMOR ) ==0 )return TRUE; - priv =intel_get_pixmap_private (pixmap); + priv =intel_uxa_get_pixmap_private (pixmap);if (glamor_egl_create_textured_pixmap (pixmap, priv->bo->handle, priv->stride)) {drm_intel_bo_disable_reuse (priv->bo);
src/uxa/intel_uxa.c +77 -59
385 lines not on this page · raw message
src/uxa/intel_uxa.h new file +300 -0
300 lines not on this page · raw message
src/uxa/intel_uxa_video.c new file +389 -0
389 lines not on this page · raw message
src/uxa/intel_video.c +212 -1100
65 lines not on this page · raw message
src/uxa/intel_video.h +112 -44
156 lines not on this page · raw message
src/uxa/intel_video_overlay.c new file +554 -0
554 lines not on this page · raw message
src/uxa/intel_video_overlay.h new file +51 -0
51 lines not on this page · raw message
The rest of this diff is not shown. View the raw message.
Signature · 7 lines
2.0.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Part of the diff is not shown here. View the raw message.
PATCH 6/12 Remove glamor support from UXA acceleration
In reply to Keith PackardMakes UXA no longer include calls to glamor acceleration functions.
--- src/uxa/Makefile.am | 9 -- src/uxa/intel_dri.c | 60 +------------ src/uxa/intel_driver.c | 1 - src/uxa/intel_uxa.c | 49 ----------- src/uxa/intel_uxa.h | 1 - src/uxa/uxa-accel.c | 230 ------------------------------------------------- src/uxa/uxa-glamor.h | 65 -------------- src/uxa/uxa-glyphs.c | 28 ------ src/uxa/uxa-render.c | 79 ----------------- src/uxa/uxa.c | 7 -- src/uxa/uxa.h | 16 ---- 11 files changed, 2 insertions(+), 543 deletions(-) delete mode 100644 src/uxa/uxa-glamor.h
src/uxa/Makefile.am +0 -9
@@ -34,7 +34,6 @@ libuxa_la_SOURCES = \ intel_batchbuffer.h \ intel_display.c \ intel_driver.c \ - intel_glamor.h \ intel_memory.c \ intel_uxa.c \ intel_video.c \
@@ -65,14 +64,6 @@ libuxa_la_SOURCES = \ uxa-unaccel.c $(NULL) -if GLAMOR -AM_CFLAGS += $(LIBGLAMOR_CFLAGS) -libuxa_la_LIBADD += $(LIBGLAMOR_LIBS) -libuxa_la_SOURCES += \ - intel_glamor.c \ - $(NULL) -endif - if DRI2 AM_CFLAGS += $(DRI2_CFLAGS) libuxa_la_SOURCES += \
src/uxa/intel_dri.c +2 -58
84 lines not on this page · raw message
src/uxa/intel_driver.c +0 -1
@@ -963,7 +963,6 @@ I830ScreenInit(SCREEN_INIT_ARGS_DECL) intel->CreateScreenResources = screen->CreateScreenResources; screen->CreateScreenResources = i830CreateScreenResources; -intel_glamor_init (screen);if (!xf86CrtcScreenInit (screen))return FALSE;
src/uxa/intel_uxa.c +0 -49
101 lines not on this page · raw message
src/uxa/intel_uxa.h +0 -1
@@ -40,7 +40,6 @@ struct intel_uxa_pixmap {#define PIN_DRI2 0x2#define PIN_DRI3 0x4#define PIN_PRIME 0x8 -#define PIN_GLAMOR 0x10 };#if HAS_DEVPRIVATEKEYREC
src/uxa/uxa-accel.c +0 -230
323 lines not on this page · raw message
src/uxa/uxa-glamor.h deleted +0 -65
65 lines not on this page · raw message
src/uxa/uxa-glyphs.c +0 -28
@@ -65,7 +65,6 @@#include <stdlib.h> #include "uxa-priv.h" -#include "uxa-glamor.h" #include "common.h" /* Width of the pixmaps we use for the caches; this should be less than
@@ -210,10 +209,6 @@ bail:Bool uxa_glyphs_init (ScreenPtr pScreen) { -uxa_screen_t *uxa_screen =uxa_get_screen (pScreen); - -if (uxa_screen->info->flags &UXA_USE_GLAMOR ) -return TRUE;#if HAS_DIXREGISTERPRIVATEKEY if (!dixRegisterPrivateKey (&uxa_glyph_key,PRIVATE_GLYPH ,0 ))return FALSE;
@@ -308,12 +303,6 @@ uxa_glyph_unrealize(ScreenPtr screen, GlyphPtr glyph) {struct uxa_glyph *priv; -uxa_screen_t *uxa_screen =uxa_get_screen (screen); - -if (uxa_screen->info->flags &UXA_USE_GLAMOR ) { -glamor_glyph_unrealize (screen, glyph); -return ; - }/* Use Lookup in case we have not attached to this glyph. */ priv =dixLookupPrivate (&glyph->devPrivates, &uxa_glyph_key);
@@ -975,23 +964,6 @@ uxa_glyphs(CARD8 op,ScreenPtr screen = pDst->pDrawable->pScreen;uxa_screen_t *uxa_screen =uxa_get_screen (screen); -if (uxa_screen->info->flags &UXA_USE_GLAMOR ) { -int ok; - -uxa_picture_prepare_access (pDst,UXA_GLAMOR_ACCESS_RW ); -uxa_picture_prepare_access (pSrc,UXA_GLAMOR_ACCESS_RO ); - ok =glamor_glyphs_nf (op, - pSrc, pDst, maskFormat, - xSrc, ySrc, nlist, list, glyphs); -uxa_picture_finish_access (pSrc,UXA_GLAMOR_ACCESS_RO ); -uxa_picture_finish_access (pDst,UXA_GLAMOR_ACCESS_RW ); - -if (!ok) - goto fallback; - -return ; - } - if (!uxa_screen->info->prepare_composite || uxa_screen->force_fallback || !uxa_drawable_is_offscreen(pDst->pDrawable) ||
src/uxa/uxa-render.c +0 -79
110 lines not on this page · raw message
src/uxa/uxa.c +0 -7
@@ -38,7 +38,6 @@#include <X11/fonts/fontstruct.h> #include "dixfontstr.h" #include "uxa.h" -#include "uxa-glamor.h" #if HAS_DEVPRIVATEKEYREC DevPrivateKeyRec uxa_screen_index;
@@ -184,7 +183,6 @@ void uxa_finish_access(DrawablePtr pDrawable, uxa_access_t access)static void uxa_validate_gc (GCPtr pGC,unsigned long changes,DrawablePtr pDrawable) { -uxa_screen_t *uxa_screen =uxa_get_screen (pGC->pScreen);/* fbValidateGC will do direct access to pixmaps if the tiling has * changed. * Preempt fbValidateGC by doing its work and masking the change out, so
@@ -195,10 +193,6 @@ uxa_validate_gc(GCPtr pGC, unsigned long changes, DrawablePtr pDrawable)* may be pure GLAMOR pixmap, then we should let the glamor * to do the validation. */ -if (uxa_screen->info->flags &UXA_USE_GLAMOR ) { -glamor_validate_gc (pGC, changes, pDrawable); - goto set_ops; - }#ifdef FB_24_32BIT if ((changes & GCTile) &&fbGetRotatedPixmap (pGC)) { (*pGC->pScreen->DestroyPixmap) (fbGetRotatedPixmap (pGC));
@@ -267,7 +261,6 @@ uxa_validate_gc(GCPtr pGC, unsigned long changes, DrawablePtr pDrawable)fbValidateGC (pGC, changes, pDrawable); } -set_ops: pGC->ops = (GCOps *) & uxa_ops; }
src/uxa/uxa.h +0 -16
@@ -46,8 +46,6 @@typedef enum {UXA_ACCESS_RO ,UXA_ACCESS_RW , -UXA_GLAMOR_ACCESS_RO , -UXA_GLAMOR_ACCESS_RW }uxa_access_t ;/**
@@ -545,20 +543,6 @@ typedef struct _UxaDriver {*/ #define UXA_TWO_BITBLT_DIRECTIONS (1 << 2) -/** -* UXA_USE_GLAMOR indicates to use glamor acceleration to perform rendering. -* And if glamor fail to accelerate the rendering, then goto fallback to -* use CPU to do the rendering. This flag will be set only when glamor get -* initialized successfully. -* Note, in ddx close screen, this bit need to be cleared. -*/ -#define UXA_USE_GLAMOR (1 << 3) - -/* UXA_GLAMOR_EGL_INITIALIZED indicates glamor egl layer get initialized -* successfully. UXA layer does not use this flag, before call to -* glamor_init, ddx need to check this flag. */ -#define UXA_GLAMOR_EGL_INITIALIZED (1 << 4) -/** @} */ /** @name UXA CreatePixmap hint flags * @{
Signature · 6 lines
2.0.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
A message that is not in the archive (sent off-list or not yet received): 1406243908-1123-7-git-send-email-keithp-aN4HjG94KOLQT0dZR+AlfA@public.gmane.org
Re: [PATCH 06/12] Remove glamor support from UXA acceleration
Quoted 32 lines · Keith Packard
Makes UXA no longer include calls to glamor acceleration functions.
Signed-off-by Keith Packard <[email protected]>--- src/uxa/Makefile.am | 9 -- src/uxa/intel_dri.c | 60 +------------ src/uxa/intel_driver.c | 1 - src/uxa/intel_uxa.c | 49 ----------- src/uxa/intel_uxa.h | 1 - src/uxa/uxa-accel.c | 230 ------------------------------------------------- src/uxa/uxa-glamor.h | 65 -------------- src/uxa/uxa-glyphs.c | 28 ------ src/uxa/uxa-render.c | 79 ----------------- src/uxa/uxa.c | 7 -- src/uxa/uxa.h | 16 ---- 11 files changed, 2 insertions(+), 543 deletions(-) delete mode 100644 src/uxa/uxa-glamor.hsrc/uxa/Makefile.am +0
-1@@ -34,7 +34,6 @@ libuxa_la_SOURCES = \ intel_batchbuffer.h \ intel_display.c \ intel_driver.c \ - intel_glamor.h \ intel_memory.c \ intel_uxa.c \ intel_video.c \@@ -65,14 +64,6 @@ libuxa_la_SOURCES = \ uxa-unaccel.c
src/uxa/Makefile.am libuxa_la_SOURCES = \ · line 75
Quoted 10 lines
src/uxa/Makefile.am +0
-8$(NULL) -if GLAMOR -AM_CFLAGS += $(LIBGLAMOR_CFLAGS) -libuxa_la_LIBADD += $(LIBGLAMOR_LIBS) -libuxa_la_SOURCES += \ - intel_glamor.c \ - $(NULL) -endif -
Something weird happened in this commit.
uxa-glamor.h is removed from the repo, but the makefile only drops intel_glamor.h, which keeps getting used. I suspect make distcheck is broken.
intel_glamor.c is no longer built, but USE_GLAMOR might still be defined I think, and if so intel_glamor_flush() from it is still getting called.
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
Re: [PATCH 06/12] Remove glamor support from UXA acceleration
Quoted 5 lines · Eric Anholt
Something weird happened in this commit.
uxa-glamor.his removed from the repo, but the makefile only dropsintel_glamor.h, which keeps getting used. I suspect make distcheck is broken.
Yeah, almost certainly. In fact, after this patch, the driver does *not* build if you ask for glamor.
Quoted 3 lines
intel_glamor.cis no longer built, butUSE_GLAMORmight still be defined I think, and if sointel_glamor_flush()from it is still getting called.
Right, the driver explicitly doesn't build with --enable-glamor after this patch. Cleaning it up so that it would build makes more of a mess though, so I just didn't worry about it as I was re-adding glamor when all of the USE_GLAMOR bits would be needed again.
I can make it work if you like, but this patch and the re-enable glamor patch would both get larger.
Signature · 1 line
[email protected]
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 9/12 Do more checks for proposed flip pixmaps
In reply to Keith PackardMake sure the pitch and tiling are correct.
Make sure there's a BO we can get at.
--- src/uxa/intel_present.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)
src/uxa/intel_present.c +18 -0
@@ -248,6 +248,8 @@ intel_present_check_flip(RRCrtcPtr crtc,ScreenPtr screen = window->drawable.pScreen;ScrnInfoPtr scrn =xf86ScreenToScrn (screen);intel_screen_private *intel =intel_get_screen_private (scrn); +dri_bo *bo; +uint32_t tiling, swizzle;if (!scrn->vtSema)return FALSE;
@@ -261,6 +263,22 @@ intel_present_check_flip(RRCrtcPtr crtc,if (crtc && !intel_crtc_on (crtc->devPrivate))return FALSE; +/* Check stride, can't change that on flip */ +if (pixmap->devKind != intel->front_pitch) +return FALSE; + +/* Make sure there's a bo we can get to */ + bo =intel_get_pixmap_bo (pixmap); +if (!bo) +return FALSE; + +/* Check tiling, can't change that on flip */ +if (drm_intel_bo_get_tiling ((drm_intel_bo *) bo, &tiling, &swizzle) !=0 ) +return FALSE; + +if (tiling != intel->front_tiling) +return FALSE; +return TRUE; }
Signature · 1 line
2.0.1
A message that is not in the archive (sent off-list or not yet received): 1406243908-1123-10-git-send-email-keithp-aN4HjG94KOLQT0dZR+AlfA@public.gmane.org
Re: [PATCH 09/12] Do more checks for proposed flip pixmaps
Quoted 2 lines · Keith Packard
Make sure the pitch and tiling are correct.
Make sure there's a BO we can get at.
I thought we couldn't change these parameters, but now I can't find what prevents them from changing. Can you cite sources?
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel
Re: [PATCH 09/12] Do more checks for proposed flip pixmaps
Quoted 7 lines · Eric Anholt
Quoted 2 lines · Keith Packard
Make sure the pitch and tiling are correct.
Make sure there's a BO we can get at.I thought we couldn't change these parameters, but now I can't find what prevents them from changing. Can you cite sources?
Looks like we *can* change tiling format. That actually makes me kinda happy as that explains why we were able to allocate a linear frame buffer for the X front buffer (due to a bug) and page flip to DRI3 buffers which are always tiled.
However, we can't change the pitch. From the kernel driver:
/* * TILEOFF/LINOFF registers can't be changed via MI display flips. * Note that pitch changes could also affect these register. */ if (INTEL_INFO(dev)->gen > 3 && (fb->offsets[0] != crtc->primary->fb->offsets[0] || fb->pitches[0] != crtc->primary->fb->pitches[0])) return -EINVAL;
I'll remove the tiling check.
Signature · 1 line
[email protected]
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 09/12] Do more checks for proposed flip pixmaps
Quoted 26 lines · Keith Packard
Quoted 7 lines · Eric Anholt
Quoted 2 lines · Keith Packard
Make sure the pitch and tiling are correct.
Make sure there's a BO we can get at.I thought we couldn't change these parameters, but now I can't find what prevents them from changing. Can you cite sources?
Looks like we *can* change tiling format. That actually makes me kinda happy as that explains why we were able to allocate a linear frame buffer for the X front buffer (due to a bug) and page flip to DRI3 buffers which are always tiled.
However, we can't change the pitch. From the kernel driver:
/* * TILEOFF/LINOFF registers can't be changed via MI display flips. * Note that pitch changes could also affect these register. */ if (INTEL_INFO(dev)->gen > 3 && (fb->offsets[0] != crtc->primary->fb->offsets[0] || fb->pitches[0] != crtc->primary->fb->pitches[0])) return -EINVAL;
Quoted 1 line
I'll remove the tiling check.
Now that we have mmio flips in the kernel we can start to relax that restriction. That still needs a bit more work in the mmio flip code but I believe some people working on just that.
We could even change the pixel format, except a check was added to drm_mode_page_flip_ioctl() to prevent that, so I guess it was deemed that the API isn't meant to allow that.
Signature · 2 lines
Ville Syrjälä Intel OTC
Re: [PATCH 09/12] Do more checks for proposed flip pixmaps
Quoted 3 lines · Ville Syrjälä
Now that we have mmio flips in the kernel we can start to relax that restriction. That still needs a bit more work in the mmio flip code but I believe some people working on just that.
I couldn't find any tiling restrictions in the current (ring-based) flip code; did I just miss them?
Quoted 3 lines
We could even change the pixel format, except a check was added to
drm_mode_page_flip_ioctl()to prevent that, so I guess it was deemed that the API isn't meant to allow that.
Yeah, not sure I care about this; 32bpp is pretty much the only format I want.
Signature · 1 line
[email protected]
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
A message that is not in the archive (sent off-list or not yet received): 86a97pd07f.fsf-6d7jPg3SX/[email protected]
Re: [Intel-gfx] [PATCH 09/12] Do more checks for proposed flip pixmaps
Quoted 8 lines · Keith Packard
Quoted 3 lines · Ville Syrjälä
Now that we have mmio flips in the kernel we can start to relax that restriction. That still needs a bit more work in the mmio flip code but I believe some people working on just that.
I couldn't find any tiling restrictions in the current (ring-based) flip code; did I just miss them?
No, changing tiling is supposed to work via cs flips. Except it doesn't actually work on VLV for some reason. We now fall back to mmio flip on VLV for that, so given a recent enough kernel it should just work (tm) on all platforms.
I was referring to the relaxing the restrictions on stride changes.
Quoted 7 lines
Quoted 4 lines
We could even change the pixel format, except a check was added to
drm_mode_page_flip_ioctl()to prevent that, so I guess it was deemed that the API isn't meant to allow that.Yeah, not sure I care about this; 32bpp is pretty much the only format I want.
You just need to be careful with the X vs. A because the kernel considers those distinct pixel formats. I now regret adding that distinciton to the pixel formats, but sadly I don't own a time machine so I can't undo it.
Signature · 6 lines
Ville Syrjälä Intel OTC _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel