Re: [Pixman] Announcement: Patchwork for Pixman

On Tue, 14 Jul 2015 15:48:54 +0200 Thomas Petazzoni wrote: > Hello, > > On Tue, 14 Jul 2015 15:34:30 +0300, Pekka Paalanen wrote: > > > There is one more bit of automation: when a commit gets pushed into the > > upstream git repository, a git hook will run that searches for the > > patch in Pat

Re: [Pixman] Announcement: Patchwork for Pixman

Hello, On Tue, 14 Jul 2015 15:34:30 +0300, Pekka Paalanen wrote: > There is one more bit of automation: when a commit gets pushed into the > upstream git repository, a git hook will run that searches for the > patch in Patchwork. If it finds a match (the patch content must match > exactly, commit

[Pixman] Announcement: Patchwork for Pixman

Hi all, as I promised a long time ago, probably in IRC, Patchwork for Pixman is now up: http://patchwork.freedesktop.org/project/pixman/list/ It has been there for a while, but only recently it was cleaned up from old patch submissions. A huge thank-you to Oded Gabbay for the clean-up, and to Dan

Re: [Pixman] [PATCH 07/12] vmx: implement fast path vmx_composite_over_n_8_8888

On Tue, 14 Jul 2015 11:41:40 +0300 Siarhei Siamashka wrote: > On Thu, 2 Jul 2015 13:04:12 +0300 > Oded Gabbay wrote: > > > POWER8, 8 cores, 3.4GHz, RHEL 7.1 ppc64le. > > > > reference memcpy speed = 24764.8MB/s (6191.2MP/s for 32bpp fills) > > > > Before After

Re: [Pixman] [PATCH 12/12] vmx: implement fast path iterator vmx_fetch_a8

On Thu, 2 Jul 2015 13:04:17 +0300 Oded Gabbay wrote: > no changes were observed when running cairo trimmed benchmarks. > > Signed-off-by: Oded Gabbay This shows good results in "lowlevel-blt-bench src_8_" on my Playstation3: == before == src_8_ = L1: 194.37 L2: 198.4

Re: [Pixman] [PATCH 11/12] vmx: implement fast path iterator vmx_fetch_r5g6b5

On Thu, 2 Jul 2015 13:04:16 +0300 Oded Gabbay wrote: > no changes were observed when running cairo trimmed benchmarks. > > Signed-off-by: Oded Gabbay This seems to cause performance regression in "lowlevel-blt-bench src_0565_" test on my Playstation3: == before == src_0565_88

Re: [Pixman] [PATCH 10/12] vmx: implement fast path iterator vmx_fetch_x8r8g8b8

On Thu, 2 Jul 2015 13:04:15 +0300 Oded Gabbay wrote: > POWER8, 8 cores, 3.4GHz, RHEL 7.1 ppc64le. > > cairo trimmed benchmarks : > > Speedups > > t-firefox-asteroids 533.92 -> 489.94 : 1.09x > > Signed-off-by: Oded Gabbay Acked-by: Siarhei Siamashka -- Best regards, Siarhei S

Re: [Pixman] [PATCH 09/12] vmx: implement fast path scaled nearest vmx_8888_8888_OVER

On Thu, 2 Jul 2015 13:04:14 +0300 Oded Gabbay wrote: > POWER8, 8 cores, 3.4GHz, RHEL 7.1 ppc64le. > reference memcpy speed = 24764.8MB/s (6191.2MP/s for 32bpp fills) > > Before After Change > - > L1

Re: [Pixman] [PATCH 08/12] vmx: implement fast path vmx_composite_src_x888_8888

On Thu, 2 Jul 2015 13:04:13 +0300 Oded Gabbay wrote: > POWER8, 8 cores, 3.4GHz, RHEL 7.1 ppc64le. > reference memcpy speed = 24764.8MB/s (6191.2MP/s for 32bpp fills) > > Before After Change > - > L1

Re: [Pixman] [PATCH 07/12] vmx: implement fast path vmx_composite_over_n_8_8888

On Thu, 2 Jul 2015 13:04:12 +0300 Oded Gabbay wrote: > POWER8, 8 cores, 3.4GHz, RHEL 7.1 ppc64le. > > reference memcpy speed = 24764.8MB/s (6191.2MP/s for 32bpp fills) > > Before After Change > - > L1

Re: [Pixman] [PATCH 06/12] vmx: implement fast path vmx_composite_over_n_8888_8888_ca

On Thu, 2 Jul 2015 13:04:11 +0300 Oded Gabbay wrote: > POWER8, 8 cores, 3.4GHz, RHEL 7.1 ppc64le. > > reference memcpy speed = 24764.8MB/s (6191.2MP/s for 32bpp fills) > > Before After Change > - > L1

Re: [Pixman] [PATCH 04/12] vmx: implement fast path vmx_blt

On Thu, 2 Jul 2015 13:04:09 +0300 Oded Gabbay wrote: > No changes were observed when running cairo trimmed benchmarks. Maybe mention that the performance improvements can be observed after applying another "vmx: implement fast path vmx_composite_copy_area" patch? Or even squash these two patche

Re: [Pixman] [PATCH 05/12] vmx: implement fast path vmx_composite_copy_area

On Thu, 2 Jul 2015 13:04:10 +0300 Oded Gabbay wrote: > No changes were observed when running cairo trimmed benchmarks. > > Signed-off-by: Oded Gabbay > --- > pixman/pixman-vmx.c | 26 ++ > 1 file changed, 26 insertions(+) > > diff --git a/pixman/pixman-vmx.c b/pixman/

Re: [Pixman] [PATCH 03/12] vmx: implement fast path vmx_fill

On Thu, 2 Jul 2015 13:04:08 +0300 Oded Gabbay wrote: > Based on sse2 impl. > > Tested cairo trimmed benchmarks on POWER8, 8 cores, 3.4GHz, > RHEL 7.1 ppc64le : > > speedups > > t-swfdec-giant-steps 1382.86 -> 719.65 : 1.92x speedup >t-gnome-system-monitor 1405.45 -> 923

Re: [Pixman] [PATCH 02/12] vmx: add helper functions

On Thu, 2 Jul 2015 13:04:07 +0300 Oded Gabbay wrote: > This patch adds the following helper functions for reuse of code, > hiding BE/LE differences and maintainability. > > All of the functions were defined as static force_inline. > > Names were copied from pixman-sse2.c so conversion of fast-

Re: [Pixman] [PATCH 01/12] vmx: add LOAD_VECTOR macro

On Thu, 2 Jul 2015 13:04:06 +0300 Oded Gabbay wrote: > This patch adds a macro for loading a single vector. > It also make the other LOAD_VECTORx macros use this macro as a base so > code would be re-used. > > In addition, I fixed minor coding style issues. > > Signed-off-by: Oded Gabbay Ack

Re: [Pixman] [PATCH 3/3] vmx: implement fast path composite_add_8888_8888

On Tue, 23 Jun 2015 12:54:41 +0300 Oded Gabbay wrote: > Copied impl. from sse2 file and edited to use vmx functions > > POWER8, 16 cores, 3.4GHz, ppc64le : > > reference memcpy speed = 27036.4MB/s (6759.1MP/s for 32bpp fills) > > Before After Change >

Re: [Pixman] [PATCH 2/3] vmx: implement fast path composite_add_8_8

On Tue, 23 Jun 2015 12:54:40 +0300 Oded Gabbay wrote: > Copied impl. from sse2 file and edited to use vmx functions > > POWER8, 16 cores, 3.4GHz, ppc64le : > > reference memcpy speed = 27036.4MB/s (6759.1MP/s for 32bpp fills) > > Before After Change >

Re: [Pixman] [PATCH 1/3] vmx: implement fast path composite_over_8888_8888

On Tue, 23 Jun 2015 12:54:39 +0300 Oded Gabbay wrote: > Copied impl. from sse2 file and edited to use vmx functions > > POWER8, 16 cores, 3.4GHz, ppc64le : > > reference memcpy speed = 27036.4MB/s (6759.1MP/s for 32bpp fills) > > Before After Change >