Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Tue, Jan 06, 2015 at 03:15:39PM +0800, Xiang, Haihao wrote: On Mon, 2015-01-05 at 23:03 -0800, Kenneth Graunke wrote: On Tuesday, January 06, 2015 02:39:36 PM Xiang, Haihao wrote: On Mon, 2015-01-05 at 21:54 -0800, Kenneth Graunke wrote: On Tuesday, January 06, 2015 01:11:53 PM Xiang, Haihao wrote: Hi Kenneth, How did you test OSD ? I can't reproduce the issue you mentioned, OSD works well for me when using mplayer-vaapi with the latest libva/libva-intel-driver master branch. I tried your patch, what surprised me is OSD still works well after applying your patch. It seems your patch didn't disable the palette. Thanks Haihao I ran: mplayer -osdlevel 3 -vo vaapi big_buck_bunny_720p_stereo.ogg For me, the OSD text is solid green, with hard edges. The OSD text is white for me when using mplayer -osdlevel 3 -vo vaapi xxx. If possible, could you update your mplayer ? Huh. I'm using the Arch Linux package of mplayer-vaapi 36265-13, which seems to be the most recent subversion commit ID. I've never seen white text on my Haswell system - it seems to be consistently dark green. If you use -vo gl or -vo xv, the OSD is solid white text with a black border around it. I presume that it's supposed to be white with vaapi as well, but I guess I'm not entirely sure. It's possible that the optimization doesn't affect the palette as long as you never use sample_c with the paletted textures. I verified the palette takes effect in the following way: 1. Only support P8A8 format in the driver 2. ran the above command and I saw white OSD text 3. Only support P4A4 format in the driver and don't use 3DSTATE_SAMPLER_PALETTE_LOAD0 to load the value to the texture palette, so the palette keeps unchanged. 4. ran the above command and I saw black OSD text. 5. Load the right value to the texture palette and ran the above command again, I saw white OSD text. Hence I think sample_c with the paletted textures is used in the driver. That sounds like the palette is actually working, then. Great :) I doubt that libva would use sample_c - sampling with a shadow comparison? It looks like it just uses sample and sample+killpix. You are right, libva driver doesn't use sample_c message. I'm pretty sure the sample_c optimization just uses the palette memory as storage for some stuff, so it's quite possible it just works if you're only using sample and sample+killpix. Thanks for the explanation, it makes sense to me. Thanks for digging into this some more, I've added the above discussion in a quote to the commit message. Thanks, Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Mon, 2015-01-05 at 21:54 -0800, Kenneth Graunke wrote: On Tuesday, January 06, 2015 01:11:53 PM Xiang, Haihao wrote: Hi Kenneth, How did you test OSD ? I can't reproduce the issue you mentioned, OSD works well for me when using mplayer-vaapi with the latest libva/libva-intel-driver master branch. I tried your patch, what surprised me is OSD still works well after applying your patch. It seems your patch didn't disable the palette. Thanks Haihao I ran: mplayer -osdlevel 3 -vo vaapi big_buck_bunny_720p_stereo.ogg For me, the OSD text is solid green, with hard edges. The OSD text is white for me when using mplayer -osdlevel 3 -vo vaapi xxx. If possible, could you update your mplayer ? If you use -vo gl or -vo xv, the OSD is solid white text with a black border around it. I presume that it's supposed to be white with vaapi as well, but I guess I'm not entirely sure. It's possible that the optimization doesn't affect the palette as long as you never use sample_c with the paletted textures. I verified the palette takes effect in the following way: 1. Only support P8A8 format in the driver 2. ran the above command and I saw white OSD text 3. Only support P4A4 format in the driver and don't use 3DSTATE_SAMPLER_PALETTE_LOAD0 to load the value to the texture palette, so the palette keeps unchanged. 4. ran the above command and I saw black OSD text. 5. Load the right value to the texture palette and ran the above command again, I saw white OSD text. Hence I think sample_c with the paletted textures is used in the driver. --Ken ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Tue, 2015-01-06 at 14:39 +0800, Xiang, Haihao wrote: On Mon, 2015-01-05 at 21:54 -0800, Kenneth Graunke wrote: On Tuesday, January 06, 2015 01:11:53 PM Xiang, Haihao wrote: Hi Kenneth, How did you test OSD ? I can't reproduce the issue you mentioned, OSD works well for me when using mplayer-vaapi with the latest libva/libva-intel-driver master branch. I tried your patch, what surprised me is OSD still works well after applying your patch. It seems your patch didn't disable the palette. Thanks Haihao I ran: mplayer -osdlevel 3 -vo vaapi big_buck_bunny_720p_stereo.ogg For me, the OSD text is solid green, with hard edges. The OSD text is white for me when using mplayer -osdlevel 3 -vo vaapi xxx. If possible, could you update your mplayer ? If you use -vo gl or -vo xv, the OSD is solid white text with a black border around it. I presume that it's supposed to be white with vaapi as well, but I guess I'm not entirely sure. It's possible that the optimization doesn't affect the palette as long as you never use sample_c with the paletted textures. I verified the palette takes effect in the following way: 1. Only support P8A8 format in the driver 2. ran the above command and I saw white OSD text 3. Only support P4A4 format in the driver and don't use 3DSTATE_SAMPLER_PALETTE_LOAD0 to load the value to the texture palette, so the palette keeps unchanged. 4. ran the above command and I saw black OSD text. 5. Load the right value to the texture palette and ran the above command again, I saw white OSD text. Hence I think sample_c with the paletted textures is used in the driver. Sorry, libva driver doesn't use sample_c message, I mean the paletted texture is used. However corroding to the doc, Palette is disabled for fast mode. --Ken ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Tuesday, January 06, 2015 02:39:36 PM Xiang, Haihao wrote: On Mon, 2015-01-05 at 21:54 -0800, Kenneth Graunke wrote: On Tuesday, January 06, 2015 01:11:53 PM Xiang, Haihao wrote: Hi Kenneth, How did you test OSD ? I can't reproduce the issue you mentioned, OSD works well for me when using mplayer-vaapi with the latest libva/libva-intel-driver master branch. I tried your patch, what surprised me is OSD still works well after applying your patch. It seems your patch didn't disable the palette. Thanks Haihao I ran: mplayer -osdlevel 3 -vo vaapi big_buck_bunny_720p_stereo.ogg For me, the OSD text is solid green, with hard edges. The OSD text is white for me when using mplayer -osdlevel 3 -vo vaapi xxx. If possible, could you update your mplayer ? Huh. I'm using the Arch Linux package of mplayer-vaapi 36265-13, which seems to be the most recent subversion commit ID. I've never seen white text on my Haswell system - it seems to be consistently dark green. If you use -vo gl or -vo xv, the OSD is solid white text with a black border around it. I presume that it's supposed to be white with vaapi as well, but I guess I'm not entirely sure. It's possible that the optimization doesn't affect the palette as long as you never use sample_c with the paletted textures. I verified the palette takes effect in the following way: 1. Only support P8A8 format in the driver 2. ran the above command and I saw white OSD text 3. Only support P4A4 format in the driver and don't use 3DSTATE_SAMPLER_PALETTE_LOAD0 to load the value to the texture palette, so the palette keeps unchanged. 4. ran the above command and I saw black OSD text. 5. Load the right value to the texture palette and ran the above command again, I saw white OSD text. Hence I think sample_c with the paletted textures is used in the driver. That sounds like the palette is actually working, then. Great :) I doubt that libva would use sample_c - sampling with a shadow comparison? It looks like it just uses sample and sample+killpix. I'm pretty sure the sample_c optimization just uses the palette memory as storage for some stuff, so it's quite possible it just works if you're only using sample and sample+killpix. --Ken signature.asc Description: This is a digitally signed message part. ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Mon, 2015-01-05 at 23:03 -0800, Kenneth Graunke wrote: On Tuesday, January 06, 2015 02:39:36 PM Xiang, Haihao wrote: On Mon, 2015-01-05 at 21:54 -0800, Kenneth Graunke wrote: On Tuesday, January 06, 2015 01:11:53 PM Xiang, Haihao wrote: Hi Kenneth, How did you test OSD ? I can't reproduce the issue you mentioned, OSD works well for me when using mplayer-vaapi with the latest libva/libva-intel-driver master branch. I tried your patch, what surprised me is OSD still works well after applying your patch. It seems your patch didn't disable the palette. Thanks Haihao I ran: mplayer -osdlevel 3 -vo vaapi big_buck_bunny_720p_stereo.ogg For me, the OSD text is solid green, with hard edges. The OSD text is white for me when using mplayer -osdlevel 3 -vo vaapi xxx. If possible, could you update your mplayer ? Huh. I'm using the Arch Linux package of mplayer-vaapi 36265-13, which seems to be the most recent subversion commit ID. I've never seen white text on my Haswell system - it seems to be consistently dark green. If you use -vo gl or -vo xv, the OSD is solid white text with a black border around it. I presume that it's supposed to be white with vaapi as well, but I guess I'm not entirely sure. It's possible that the optimization doesn't affect the palette as long as you never use sample_c with the paletted textures. I verified the palette takes effect in the following way: 1. Only support P8A8 format in the driver 2. ran the above command and I saw white OSD text 3. Only support P4A4 format in the driver and don't use 3DSTATE_SAMPLER_PALETTE_LOAD0 to load the value to the texture palette, so the palette keeps unchanged. 4. ran the above command and I saw black OSD text. 5. Load the right value to the texture palette and ran the above command again, I saw white OSD text. Hence I think sample_c with the paletted textures is used in the driver. That sounds like the palette is actually working, then. Great :) I doubt that libva would use sample_c - sampling with a shadow comparison? It looks like it just uses sample and sample+killpix. You are right, libva driver doesn't use sample_c message. I'm pretty sure the sample_c optimization just uses the palette memory as storage for some stuff, so it's quite possible it just works if you're only using sample and sample+killpix. Thanks for the explanation, it makes sense to me. --Ken ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Wed, Dec 31, 2014 at 04:23:00PM -0800, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. This may break sampling of paletted formats (P8/A8P8/P8A8). It's unclear whether it affects sampling of paletted formats in general, or just the sample_c message (which is never used). While libva does have support for using paletted formats (primarily for OSDs), that support appears to have been broken for at least a year, so I couldn't observe a regression from this. Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) Resubmitting the patch to unconditionally enable this. I tried to get libva-intel to use paletted formats, and observe a regression...but the only thing I found that used it was mplayer's OSD (on screen display). Even without my patch, the colors were totally wrong with that, and it's according to a few distro wikis, that's been the case for over a year. If libva's code for paletted formats /is/ broken, they could always add code to disable this bit using the command validator when fixing it. Could we try merging this, and back it out if someone reports a regression? I haven't observed any problems. It's also been quite stable. Yeah makes sense. When resending please incorporated review feedback (Ville dug out the wa name), I've done that. And I've pasted the additional detail about the libva saga, just for reference (since no one will remember that it's mplayer's OSD which uses this 2 months down the road). Also please cc libva mailing lists next time around as an fyi. Done that too. Queued for -next, thanks for the patch. -Daniel diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 40ca873..0f32fd1a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -6167,6 +6167,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN3 0xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS(18) #define GEN8_SAMPLER_POWER_BYPASS_DIS (11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7d99a9c..17e84dc 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5974,6 +5974,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, _MASKED_FIELD(GEN6_WIZ_HASHING_MASK, GEN6_WIZ_HASHING_16x4)); + /* Make sample_c messages faster. */ + I915_WRITE(HALF_SLICE_CHICKEN3, +_MASKED_BIT_ENABLE(HSW_SAMPLE_C_PERFORMANCE)); + /* WaSwitchSolVfFArbitrationPriority:hsw */ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL); -- 2.2.1 ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Monday, January 05, 2015 02:19:15 PM Daniel Vetter wrote: On Wed, Dec 31, 2014 at 04:23:00PM -0800, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. This may break sampling of paletted formats (P8/A8P8/P8A8). It's unclear whether it affects sampling of paletted formats in general, or just the sample_c message (which is never used). While libva does have support for using paletted formats (primarily for OSDs), that support appears to have been broken for at least a year, so I couldn't observe a regression from this. Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) Resubmitting the patch to unconditionally enable this. I tried to get libva-intel to use paletted formats, and observe a regression...but the only thing I found that used it was mplayer's OSD (on screen display). Even without my patch, the colors were totally wrong with that, and it's according to a few distro wikis, that's been the case for over a year. If libva's code for paletted formats /is/ broken, they could always add code to disable this bit using the command validator when fixing it. Could we try merging this, and back it out if someone reports a regression? I haven't observed any problems. It's also been quite stable. Yeah makes sense. When resending please incorporated review feedback (Ville dug out the wa name), I've done that. And I've pasted the additional detail about the libva saga, just for reference (since no one will remember that it's mplayer's OSD which uses this 2 months down the road). Also please cc libva mailing lists next time around as an fyi. Done that too. Queued for -next, thanks for the patch. -Daniel Oh, sorry, I missed that in the review. Thanks, Daniel! --Ken signature.asc Description: This is a digitally signed message part. ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
Hi Kenneth, How did you test OSD ? I can't reproduce the issue you mentioned, OSD works well for me when using mplayer-vaapi with the latest libva/libva-intel-driver master branch. I tried your patch, what surprised me is OSD still works well after applying your patch. It seems your patch didn't disable the palette. Thanks Haihao On Monday, January 05, 2015 02:19:15 PM Daniel Vetter wrote: On Wed, Dec 31, 2014 at 04:23:00PM -0800, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. This may break sampling of paletted formats (P8/A8P8/P8A8). It's unclear whether it affects sampling of paletted formats in general, or just the sample_c message (which is never used). While libva does have support for using paletted formats (primarily for OSDs), that support appears to have been broken for at least a year, so I couldn't observe a regression from this. Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) Resubmitting the patch to unconditionally enable this. I tried to get libva-intel to use paletted formats, and observe a regression...but the only thing I found that used it was mplayer's OSD (on screen display). Even without my patch, the colors were totally wrong with that, and it's according to a few distro wikis, that's been the case for over a year. If libva's code for paletted formats /is/ broken, they could always add code to disable this bit using the command validator when fixing it. Could we try merging this, and back it out if someone reports a regression? I haven't observed any problems. It's also been quite stable. Yeah makes sense. When resending please incorporated review feedback (Ville dug out the wa name), I've done that. And I've pasted the additional detail about the libva saga, just for reference (since no one will remember that it's mplayer's OSD which uses this 2 months down the road). Also please cc libva mailing lists next time around as an fyi. Done that too. Queued for -next, thanks for the patch. -Daniel Oh, sorry, I missed that in the review. Thanks, Daniel! --Ken ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Tuesday, January 06, 2015 01:11:53 PM Xiang, Haihao wrote: Hi Kenneth, How did you test OSD ? I can't reproduce the issue you mentioned, OSD works well for me when using mplayer-vaapi with the latest libva/libva-intel-driver master branch. I tried your patch, what surprised me is OSD still works well after applying your patch. It seems your patch didn't disable the palette. Thanks Haihao I ran: mplayer -osdlevel 3 -vo vaapi big_buck_bunny_720p_stereo.ogg For me, the OSD text is solid green, with hard edges. If you use -vo gl or -vo xv, the OSD is solid white text with a black border around it. I presume that it's supposed to be white with vaapi as well, but I guess I'm not entirely sure. It's possible that the optimization doesn't affect the palette as long as you never use sample_c with the paletted textures. --Ken signature.asc Description: This is a digitally signed message part. ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
Tested-By: PRC QA PRTS (Patch Regression Test System Contact: shuang...@intel.com) -Summary- Platform Delta drm-intel-nightly Series Applied PNV 363/364 363/364 ILK -1 364/366 363/366 SNB +2-1 443/450 444/450 IVB 496/498 496/498 BYT 288/289 288/289 HSW +2-6 542/564 538/564 BDW 415/417 415/417 -Detailed- Platform Testdrm-intel-nightly Series Applied *ILK igt_gem_fenced_exec_thrash_no-spare-fences-busy PASS(2, M37) DMESG_WARN(1, M37) *SNB igt_kms_flip_modeset-vs-vblank-race DMESG_WARN(2, M35) PASS(1, M35) SNB igt_kms_flip_modeset-vs-vblank-race-interruptible DMESG_WARN(2, M35M22)PASS(1, M35) DMESG_WARN(1, M35) SNB igt_kms_plane_plane-position-hole-pipe-B-plane-1 DMESG_WARN(1, M35)PASS(2, M35M22) PASS(1, M35) *HSW igt_kms_flip_flip-vs-expired-vblank PASS(2, M40) DMESG_WARN(1, M40) *HSW igt_kms_flip_flip-vs-expired-vblank-interruptible PASS(2, M40) DMESG_WARN(1, M40) *HSW igt_kms_flip_nonexisting-fb PASS(2, M40) DMESG_WARN(1, M40) *HSW igt_kms_flip_nonexisting-fb-interruptible PASS(2, M40) DMESG_WARN(1, M40) *HSW igt_kms_flip_single-buffer-flip-vs-dpms-off-vs-modeset PASS(3, M40M19) DMESG_WARN(1, M40) *HSW igt_kms_plane_plane-panning-bottom-right-pipe-C-plane-1 TIMEOUT(2, M40)PASS(1, M19) PASS(1, M40) *HSW igt_kms_plane_plane-position-hole-pipe-A-plane-2 PASS(2, M40) DMESG_WARN(1, M40) HSW igt_pm_rpm_modeset-non-lpsp-stress-no-wait NSPT(1, M19)DMESG_WARN(1, M40)PASS(1, M40) PASS(1, M40) Note: You need to pay more attention to line start with '*' ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
[Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. This may break sampling of paletted formats (P8/A8P8/P8A8). It's unclear whether it affects sampling of paletted formats in general, or just the sample_c message (which is never used). While libva does have support for using paletted formats (primarily for OSDs), that support appears to have been broken for at least a year, so I couldn't observe a regression from this. Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) Resubmitting the patch to unconditionally enable this. I tried to get libva-intel to use paletted formats, and observe a regression...but the only thing I found that used it was mplayer's OSD (on screen display). Even without my patch, the colors were totally wrong with that, and it's according to a few distro wikis, that's been the case for over a year. If libva's code for paletted formats /is/ broken, they could always add code to disable this bit using the command validator when fixing it. Could we try merging this, and back it out if someone reports a regression? I haven't observed any problems. It's also been quite stable. diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 40ca873..0f32fd1a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -6167,6 +6167,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN30xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS (18) #define GEN8_SAMPLER_POWER_BYPASS_DIS(11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7d99a9c..17e84dc 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5974,6 +5974,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, _MASKED_FIELD(GEN6_WIZ_HASHING_MASK, GEN6_WIZ_HASHING_16x4)); + /* Make sample_c messages faster. */ + I915_WRITE(HALF_SLICE_CHICKEN3, + _MASKED_BIT_ENABLE(HSW_SAMPLE_C_PERFORMANCE)); + /* WaSwitchSolVfFArbitrationPriority:hsw */ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL); -- 2.2.1 ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Fri, Nov 07, 2014 at 10:46:31AM -0800, Matt Turner wrote: On Wed, Oct 29, 2014 at 2:12 PM, Kenneth Graunke kenn...@whitecape.org wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Discussion seems to have ended, but Mesa really needs this. So, Kernel Team, how are we going to make sure Mesa gets this 18% performance improvement? Someone writes the patches+tests and everything, gets it all reviewed and I'll merge it. I'd prefer if we could do this with the cmd parser, but that thing seems to be eternally stuck. So a new execbuf flag is imo ok too. execbuf already tracks a few other lonesome bits and makes sure userspace actually gets the setting it requests, so should work out. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Wed, Oct 29, 2014 at 2:12 PM, Kenneth Graunke kenn...@whitecape.org wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Discussion seems to have ended, but Mesa really needs this. So, Kernel Team, how are we going to make sure Mesa gets this 18% performance improvement? ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Mon, 2014-11-03 at 13:48 +0100, Daniel Vetter wrote: On Fri, Oct 31, 2014 at 11:27:33AM +0200, Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 12:57:04PM -0700, Kenneth Graunke wrote: Before we get too much further...we should check if libva is actually broken. I don't know if this means the sampler palette completely doesn't work, or if it just means sample_c doesn't work with the palette. If it's the latter, we're probably fine, because I doubt libva uses sample_c. Yeah if we wouldn't break any existing userspace I guess we could just flip the switch in the kernel. If anyone later wants to start doing something that no longer works they'd have to deal with disabling the bit using an LRI. It very much looks like libva uses palettes, since it supports C8 and C4 image formats (well some crazy fourcc nonsense, but meh). And it does so on all generations support by the libva driver, i.e. including hsw afaict Cc'ing people and lists with more clue who should be able to tell whether its not just there but actually works ... Yes, libva driver uses sample (0) and palette on HSW Thanks Haihao -Daniel ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Fri, Oct 31, 2014 at 11:27:33AM +0200, Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 12:57:04PM -0700, Kenneth Graunke wrote: Before we get too much further...we should check if libva is actually broken. I don't know if this means the sampler palette completely doesn't work, or if it just means sample_c doesn't work with the palette. If it's the latter, we're probably fine, because I doubt libva uses sample_c. Yeah if we wouldn't break any existing userspace I guess we could just flip the switch in the kernel. If anyone later wants to start doing something that no longer works they'd have to deal with disabling the bit using an LRI. It very much looks like libva uses palettes, since it supports C8 and C4 image formats (well some crazy fourcc nonsense, but meh). And it does so on all generations support by the libva driver, i.e. including hsw afaict Cc'ing people and lists with more clue who should be able to tell whether its not just there but actually works ... -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On 30/10/14 19:26, Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 10:32:38AM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 01:01:30 PM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 02:32:40AM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 11:00:51 AM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 10:50:03AM +0200, Ville Syrjälä wrote: On Wed, Oct 29, 2014 at 03:12:43PM -0700, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Nice. We need more bits like this ;) Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77fce96..340821a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5952,6 +5952,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN30xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS (18) #define GEN8_SAMPLER_POWER_BYPASS_DIS(11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..50c72a7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5736,6 +5736,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + /* Make sample_c messages faster. */ I found a name for it in the w/a database. WaSampleCChickenBitEnable:hsw Reviewed-by: Ville Syrjälä ville.syrj...@linux.intel.com Oh actually it says palette won't work when this bit is on. I'm assuming that's the texture palette. Do we have any use of that anywhere? That's a good point. 3DSTATE_SAMPLER_PALETTE_LOAD and the A8P8/indexed formats aren't used by Mesa or xf86-video-intel, but it looks like they might be used by libva. Can someone confirm that libva does use the sampler palette? If they do, what do we do about it? I suppose the best option then would be to use an LRI from a batch, which means the register would need to be added to the cmd parser white list. This is one of the context saved registers so doing the LRI just once per context should be enough. I don't like that solution. For one, it's impossible - you can't LRI from userspace batches, even if you add it to the kernel command parser's whitelist, because the hardware scanner is still enabled. Given that I've been waiting two years for this capability, I want to find a more immediate solution. Ah. I've somehow convinced myself the cmd parser might actually be doing something besides just eating CPU cycles these days. But I guess not. Another option is to have some sort of execbuf flag...maybe a 3D/Media usage flag. If set to 3D, write 0x6000200...if media, write 0x600. Or something specific. I do hate adding more junk to the execbuf path, though. Other ideas? Fast vs. slow flag? :) More seriously, one somewhat crappy option would be to initialize that bit to 1 for all explicit contexts, and then have the kernel always turn it off before executing something with the default context. It's not unlike how we imagined the RS stuff would work since old userspace doesn't know to turn RS off when using the default context. But if we would do something like this, I think it would be nice if we could just add it as a temporary hack and potentially drop it if we manage to lose the restore inhibit flag and someone finishes the cmd parser. So maybe Mesa could also try to set it from a batch, which currently would be a nop or rejected by the cmd parser, but when the cmd parser starts to work it might actually do something? But always initializing it to 1 for all explicit context in the kernel obviously requires that libva doesn't use an explicit context itself. In case it does, another idea would be to just add a hack to the cmd parser to emit the LRI on the ring when it enconters this register. Not very pretty, but should work somewhat. It wouldn't allow mid-batch changes to the register value though, but enough for setting it once for each context in Mesa. And of course we'd still need the kernel to turn it off before any default context will see it. This looks like a use for the proposed DRM_I915_GEM_CONTEXT_CREATE2 ioctl, which (so far) allows setting of scheduling priority for batches submitted in the new context, and/or declaring that the context will be used for GPGPU batches. We
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Thu, Oct 30, 2014 at 12:57:04PM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 09:26:01 PM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 10:32:38AM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 01:01:30 PM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 02:32:40AM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 11:00:51 AM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 10:50:03AM +0200, Ville Syrjälä wrote: On Wed, Oct 29, 2014 at 03:12:43PM -0700, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Nice. We need more bits like this ;) Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77fce96..340821a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5952,6 +5952,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN30xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS (18) #define GEN8_SAMPLER_POWER_BYPASS_DIS(11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..50c72a7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5736,6 +5736,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + /* Make sample_c messages faster. */ I found a name for it in the w/a database. WaSampleCChickenBitEnable:hsw Reviewed-by: Ville Syrjälä ville.syrj...@linux.intel.com Oh actually it says palette won't work when this bit is on. I'm assuming that's the texture palette. Do we have any use of that anywhere? That's a good point. 3DSTATE_SAMPLER_PALETTE_LOAD and the A8P8/indexed formats aren't used by Mesa or xf86-video-intel, but it looks like they might be used by libva. Can someone confirm that libva does use the sampler palette? If they do, what do we do about it? I suppose the best option then would be to use an LRI from a batch, which means the register would need to be added to the cmd parser white list. This is one of the context saved registers so doing the LRI just once per context should be enough. I don't like that solution. For one, it's impossible - you can't LRI from userspace batches, even if you add it to the kernel command parser's whitelist, because the hardware scanner is still enabled. Given that I've been waiting two years for this capability, I want to find a more immediate solution. Ah. I've somehow convinced myself the cmd parser might actually be doing something besides just eating CPU cycles these days. But I guess not. Another option is to have some sort of execbuf flag...maybe a 3D/Media usage flag. If set to 3D, write 0x6000200...if media, write 0x600. Or something specific. I do hate adding more junk to the execbuf path, though. Other ideas? Fast vs. slow flag? :) More seriously, one somewhat crappy option would be to initialize that bit to 1 for all explicit contexts, and then have the kernel always turn it off before executing something with the default context. It's not unlike how we imagined the RS stuff would work since old userspace doesn't know to turn RS off when using the default context. Interesting idea - that might work. We don't need mid-batch changes either. I don't think HALF_SLICE_CHICKEN3 is part of the logical context, FWIW. Oh but it is. Lots of chickens like to nest in the context. Before we get too much further...we should check if libva is actually broken. I don't know if this means the sampler palette completely doesn't work, or if it just means sample_c doesn't work with the palette. If it's the latter, we're probably fine, because I doubt libva uses sample_c. Yeah if we wouldn't break any existing userspace I guess we could just flip the switch in the kernel. If anyone
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
[off-topic] On Fri, 31 Oct 2014, Ville Syrjälä ville.syrj...@linux.intel.com wrote: On Thu, Oct 30, 2014 at 12:57:04PM -0700, Kenneth Graunke wrote: I don't think HALF_SLICE_CHICKEN3 is part of the logical context, FWIW. Oh but it is. Lots of chickens like to nest in the context. In the context of lunch, HALF_SLICE_CHICKEN3 is indeed part of the context. But it has certainly ceased to nest. BR, Jani. -- Jani Nikula, Intel Open Source Technology Center ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Wed, Oct 29, 2014 at 03:12:43PM -0700, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Nice. We need more bits like this ;) Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77fce96..340821a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5952,6 +5952,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN3 0xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS(18) #define GEN8_SAMPLER_POWER_BYPASS_DIS (11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..50c72a7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5736,6 +5736,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + /* Make sample_c messages faster. */ I found a name for it in the w/a database. WaSampleCChickenBitEnable:hsw Reviewed-by: Ville Syrjälä ville.syrj...@linux.intel.com + I915_WRITE(HALF_SLICE_CHICKEN3, +_MASKED_BIT_ENABLE(HSW_SAMPLE_C_PERFORMANCE)); + /* WaSwitchSolVfFArbitrationPriority:hsw */ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL); -- 2.1.2 ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx -- Ville Syrjälä Intel OTC ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Thu, Oct 30, 2014 at 10:50:03AM +0200, Ville Syrjälä wrote: On Wed, Oct 29, 2014 at 03:12:43PM -0700, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Nice. We need more bits like this ;) Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77fce96..340821a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5952,6 +5952,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN30xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS (18) #define GEN8_SAMPLER_POWER_BYPASS_DIS(11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..50c72a7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5736,6 +5736,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + /* Make sample_c messages faster. */ I found a name for it in the w/a database. WaSampleCChickenBitEnable:hsw Reviewed-by: Ville Syrjälä ville.syrj...@linux.intel.com Oh actually it says palette won't work when this bit is on. I'm assuming that's the texture palette. Do we have any use of that anywhere? + I915_WRITE(HALF_SLICE_CHICKEN3, + _MASKED_BIT_ENABLE(HSW_SAMPLE_C_PERFORMANCE)); + /* WaSwitchSolVfFArbitrationPriority:hsw */ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL); -- 2.1.2 ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx -- Ville Syrjälä Intel OTC -- Ville Syrjälä Intel OTC ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Thursday, October 30, 2014 11:00:51 AM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 10:50:03AM +0200, Ville Syrjälä wrote: On Wed, Oct 29, 2014 at 03:12:43PM -0700, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Nice. We need more bits like this ;) Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77fce96..340821a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5952,6 +5952,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN3 0xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS(18) #define GEN8_SAMPLER_POWER_BYPASS_DIS (11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..50c72a7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5736,6 +5736,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + /* Make sample_c messages faster. */ I found a name for it in the w/a database. WaSampleCChickenBitEnable:hsw Reviewed-by: Ville Syrjälä ville.syrj...@linux.intel.com Oh actually it says palette won't work when this bit is on. I'm assuming that's the texture palette. Do we have any use of that anywhere? That's a good point. 3DSTATE_SAMPLER_PALETTE_LOAD and the A8P8/indexed formats aren't used by Mesa or xf86-video-intel, but it looks like they might be used by libva. Can someone confirm that libva does use the sampler palette? If they do, what do we do about it? --Ken signature.asc Description: This is a digitally signed message part. ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Thu, Oct 30, 2014 at 02:32:40AM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 11:00:51 AM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 10:50:03AM +0200, Ville Syrjälä wrote: On Wed, Oct 29, 2014 at 03:12:43PM -0700, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Nice. We need more bits like this ;) Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77fce96..340821a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5952,6 +5952,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN30xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS (18) #define GEN8_SAMPLER_POWER_BYPASS_DIS(11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..50c72a7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5736,6 +5736,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + /* Make sample_c messages faster. */ I found a name for it in the w/a database. WaSampleCChickenBitEnable:hsw Reviewed-by: Ville Syrjälä ville.syrj...@linux.intel.com Oh actually it says palette won't work when this bit is on. I'm assuming that's the texture palette. Do we have any use of that anywhere? That's a good point. 3DSTATE_SAMPLER_PALETTE_LOAD and the A8P8/indexed formats aren't used by Mesa or xf86-video-intel, but it looks like they might be used by libva. Can someone confirm that libva does use the sampler palette? If they do, what do we do about it? I suppose the best option then would be to use an LRI from a batch, which means the register would need to be added to the cmd parser white list. This is one of the context saved registers so doing the LRI just once per context should be enough. Well that's assuming libva doesn't use the default context. I'm getting another itch to drop the restore inhibit flag for default contexts. That would actually make it possible to do these sort of things without risking breakage to existing userspace. But I think Chris is going scream unless the patch comes with performance data that shows it doesn't hurt too much. -- Ville Syrjälä Intel OTC ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Thursday, October 30, 2014 01:01:30 PM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 02:32:40AM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 11:00:51 AM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 10:50:03AM +0200, Ville Syrjälä wrote: On Wed, Oct 29, 2014 at 03:12:43PM -0700, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Nice. We need more bits like this ;) Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77fce96..340821a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5952,6 +5952,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN3 0xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS(18) #define GEN8_SAMPLER_POWER_BYPASS_DIS (11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..50c72a7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5736,6 +5736,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + /* Make sample_c messages faster. */ I found a name for it in the w/a database. WaSampleCChickenBitEnable:hsw Reviewed-by: Ville Syrjälä ville.syrj...@linux.intel.com Oh actually it says palette won't work when this bit is on. I'm assuming that's the texture palette. Do we have any use of that anywhere? That's a good point. 3DSTATE_SAMPLER_PALETTE_LOAD and the A8P8/indexed formats aren't used by Mesa or xf86-video-intel, but it looks like they might be used by libva. Can someone confirm that libva does use the sampler palette? If they do, what do we do about it? I suppose the best option then would be to use an LRI from a batch, which means the register would need to be added to the cmd parser white list. This is one of the context saved registers so doing the LRI just once per context should be enough. I don't like that solution. For one, it's impossible - you can't LRI from userspace batches, even if you add it to the kernel command parser's whitelist, because the hardware scanner is still enabled. Given that I've been waiting two years for this capability, I want to find a more immediate solution. Another option is to have some sort of execbuf flag...maybe a 3D/Media usage flag. If set to 3D, write 0x6000200...if media, write 0x600. Or something specific. I do hate adding more junk to the execbuf path, though. Other ideas? Well that's assuming libva doesn't use the default context. I'm getting another itch to drop the restore inhibit flag for default contexts. That would actually make it possible to do these sort of things without risking breakage to existing userspace. But I think Chris is going scream unless the patch comes with performance data that shows it doesn't hurt too much. I suppose it wouldn't affect Mesa much, since we never use the default context on Gen6+. But otherwise I'd probably want to see the data, like Chris... signature.asc Description: This is a digitally signed message part. ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Thu, Oct 30, 2014 at 10:32:38AM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 01:01:30 PM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 02:32:40AM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 11:00:51 AM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 10:50:03AM +0200, Ville Syrjälä wrote: On Wed, Oct 29, 2014 at 03:12:43PM -0700, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Nice. We need more bits like this ;) Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77fce96..340821a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5952,6 +5952,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN30xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS (18) #define GEN8_SAMPLER_POWER_BYPASS_DIS(11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..50c72a7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5736,6 +5736,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + /* Make sample_c messages faster. */ I found a name for it in the w/a database. WaSampleCChickenBitEnable:hsw Reviewed-by: Ville Syrjälä ville.syrj...@linux.intel.com Oh actually it says palette won't work when this bit is on. I'm assuming that's the texture palette. Do we have any use of that anywhere? That's a good point. 3DSTATE_SAMPLER_PALETTE_LOAD and the A8P8/indexed formats aren't used by Mesa or xf86-video-intel, but it looks like they might be used by libva. Can someone confirm that libva does use the sampler palette? If they do, what do we do about it? I suppose the best option then would be to use an LRI from a batch, which means the register would need to be added to the cmd parser white list. This is one of the context saved registers so doing the LRI just once per context should be enough. I don't like that solution. For one, it's impossible - you can't LRI from userspace batches, even if you add it to the kernel command parser's whitelist, because the hardware scanner is still enabled. Given that I've been waiting two years for this capability, I want to find a more immediate solution. Ah. I've somehow convinced myself the cmd parser might actually be doing something besides just eating CPU cycles these days. But I guess not. Another option is to have some sort of execbuf flag...maybe a 3D/Media usage flag. If set to 3D, write 0x6000200...if media, write 0x600. Or something specific. I do hate adding more junk to the execbuf path, though. Other ideas? Fast vs. slow flag? :) More seriously, one somewhat crappy option would be to initialize that bit to 1 for all explicit contexts, and then have the kernel always turn it off before executing something with the default context. It's not unlike how we imagined the RS stuff would work since old userspace doesn't know to turn RS off when using the default context. But if we would do something like this, I think it would be nice if we could just add it as a temporary hack and potentially drop it if we manage to lose the restore inhibit flag and someone finishes the cmd parser. So maybe Mesa could also try to set it from a batch, which currently would be a nop or rejected by the cmd parser, but when the cmd parser starts to work it might actually do something? But always initializing it to 1 for all explicit context in the kernel obviously requires that libva doesn't use an explicit context itself. In case it does, another idea would be to just add a hack to the cmd parser to emit the LRI on the ring when it enconters this register. Not very pretty, but should work somewhat. It wouldn't allow mid-batch changes to the register value though, but enough for setting it once for each context in Mesa. And of course we'd still need the kernel to turn it off before any default context will see it. Well that's assuming libva doesn't use the
Re: [Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
On Thursday, October 30, 2014 09:26:01 PM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 10:32:38AM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 01:01:30 PM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 02:32:40AM -0700, Kenneth Graunke wrote: On Thursday, October 30, 2014 11:00:51 AM Ville Syrjälä wrote: On Thu, Oct 30, 2014 at 10:50:03AM +0200, Ville Syrjälä wrote: On Wed, Oct 29, 2014 at 03:12:43PM -0700, Kenneth Graunke wrote: Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Nice. We need more bits like this ;) Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77fce96..340821a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5952,6 +5952,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN3 0xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS(18) #define GEN8_SAMPLER_POWER_BYPASS_DIS (11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..50c72a7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5736,6 +5736,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + /* Make sample_c messages faster. */ I found a name for it in the w/a database. WaSampleCChickenBitEnable:hsw Reviewed-by: Ville Syrjälä ville.syrj...@linux.intel.com Oh actually it says palette won't work when this bit is on. I'm assuming that's the texture palette. Do we have any use of that anywhere? That's a good point. 3DSTATE_SAMPLER_PALETTE_LOAD and the A8P8/indexed formats aren't used by Mesa or xf86-video-intel, but it looks like they might be used by libva. Can someone confirm that libva does use the sampler palette? If they do, what do we do about it? I suppose the best option then would be to use an LRI from a batch, which means the register would need to be added to the cmd parser white list. This is one of the context saved registers so doing the LRI just once per context should be enough. I don't like that solution. For one, it's impossible - you can't LRI from userspace batches, even if you add it to the kernel command parser's whitelist, because the hardware scanner is still enabled. Given that I've been waiting two years for this capability, I want to find a more immediate solution. Ah. I've somehow convinced myself the cmd parser might actually be doing something besides just eating CPU cycles these days. But I guess not. Another option is to have some sort of execbuf flag...maybe a 3D/Media usage flag. If set to 3D, write 0x6000200...if media, write 0x600. Or something specific. I do hate adding more junk to the execbuf path, though. Other ideas? Fast vs. slow flag? :) More seriously, one somewhat crappy option would be to initialize that bit to 1 for all explicit contexts, and then have the kernel always turn it off before executing something with the default context. It's not unlike how we imagined the RS stuff would work since old userspace doesn't know to turn RS off when using the default context. Interesting idea - that might work. We don't need mid-batch changes either. I don't think HALF_SLICE_CHICKEN3 is part of the logical context, FWIW. Before we get too much further...we should check if libva is actually broken. I don't know if this means the sampler palette completely doesn't work, or if it just means sample_c doesn't work with the palette. If it's the latter, we're probably fine, because I doubt libva uses sample_c. --Ken signature.asc Description: This is a digitally signed message part. ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
[Intel-gfx] [PATCH] drm/i915: Make sample_c messages go faster on Haswell.
Haswell significantly improved the performance of sampler_c messages, but the optimization appears to be off by default. Later platforms remove this bit, and apparently always enable the optimization. Improves performance in Counter Strike: Global Offensive by 18% at default settings on Iris Pro. No Piglit regressions. Signed-off-by: Kenneth Graunke kenn...@whitecape.org --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 4 2 files changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 77fce96..340821a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -5952,6 +5952,7 @@ enum punit_power_well { #define HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE(1 6) #define HALF_SLICE_CHICKEN30xe184 +#define HSW_SAMPLE_C_PERFORMANCE (19) #define GEN8_CENTROID_PIXEL_OPT_DIS (18) #define GEN8_SAMPLER_POWER_BYPASS_DIS(11) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7a69eba..50c72a7 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5736,6 +5736,10 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_WRITE(GEN7_GT_MODE, GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4); + /* Make sample_c messages faster. */ + I915_WRITE(HALF_SLICE_CHICKEN3, + _MASKED_BIT_ENABLE(HSW_SAMPLE_C_PERFORMANCE)); + /* WaSwitchSolVfFArbitrationPriority:hsw */ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL); -- 2.1.2 ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx