[Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-16 Thread Kristian Høgsberg
The I915_EXEC_NO_RELOC flag lets us tell the kernel that the offset we provide in the validate list entry is what we've used in all relocations to the bo in question. If the bo hasn't moved, the kernel can skip relocations completely. Signed-off-by: Kristian Høgsberg --- intel/intel_bufmgr_gem.

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-16 Thread Daniel Vetter
On Fri, Jan 16, 2015 at 05:46:00PM -0800, Kristian Høgsberg wrote: > The I915_EXEC_NO_RELOC flag lets us tell the kernel that the offset we > provide in the validate list entry is what we've used in all relocations > to the bo in question. If the bo hasn't moved, the kernel can skip > relocations

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-17 Thread Chris Wilson
On Sat, Jan 17, 2015 at 05:23:59AM +0100, Daniel Vetter wrote: > On Fri, Jan 16, 2015 at 05:46:00PM -0800, Kristian Høgsberg wrote: > > The I915_EXEC_NO_RELOC flag lets us tell the kernel that the offset we > > provide in the validate list entry is what we've used in all relocations > > to the bo i

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-19 Thread Kristian Høgsberg
On Fri, Jan 16, 2015 at 8:23 PM, Daniel Vetter wrote: > On Fri, Jan 16, 2015 at 05:46:00PM -0800, Kristian Høgsberg wrote: >> The I915_EXEC_NO_RELOC flag lets us tell the kernel that the offset we >> provide in the validate list entry is what we've used in all relocations >> to the bo in question.

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-19 Thread Kristian Høgsberg
On Sat, Jan 17, 2015 at 1:49 AM, Chris Wilson wrote: > On Sat, Jan 17, 2015 at 05:23:59AM +0100, Daniel Vetter wrote: >> On Fri, Jan 16, 2015 at 05:46:00PM -0800, Kristian Høgsberg wrote: >> > The I915_EXEC_NO_RELOC flag lets us tell the kernel that the offset we >> > provide in the validate list

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-20 Thread Daniel Vetter
On Mon, Jan 19, 2015 at 09:58:55PM -0800, Kristian Høgsberg wrote: > On Sat, Jan 17, 2015 at 1:49 AM, Chris Wilson > wrote: > > On Sat, Jan 17, 2015 at 05:23:59AM +0100, Daniel Vetter wrote: > >> On Fri, Jan 16, 2015 at 05:46:00PM -0800, Kristian Høgsberg wrote: > >> > The I915_EXEC_NO_RELOC flag

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-20 Thread Daniel Vetter
On Mon, Jan 19, 2015 at 09:45:35PM -0800, Kristian Høgsberg wrote: > On Fri, Jan 16, 2015 at 8:23 PM, Daniel Vetter wrote: > > On Fri, Jan 16, 2015 at 05:46:00PM -0800, Kristian Høgsberg wrote: > >> The I915_EXEC_NO_RELOC flag lets us tell the kernel that the offset we > >> provide in the validate

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-20 Thread Kristian Høgsberg
On Tue, Jan 20, 2015 at 12:42 AM, Daniel Vetter wrote: > On Mon, Jan 19, 2015 at 09:45:35PM -0800, Kristian Høgsberg wrote: >> On Fri, Jan 16, 2015 at 8:23 PM, Daniel Vetter wrote: >> > On Fri, Jan 16, 2015 at 05:46:00PM -0800, Kristian Høgsberg wrote: >> >> The I915_EXEC_NO_RELOC flag lets us te

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-20 Thread Chris Wilson
On Tue, Jan 20, 2015 at 12:53:35PM -0800, Kristian Høgsberg wrote: > On Tue, Jan 20, 2015 at 12:42 AM, Daniel Vetter wrote: > > My idea for all this would have been to create a per-thread execbuf > > relocation context with a hashtab to map buffer pointers to execbuf index > > and a bunch of array

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-20 Thread Kristian Høgsberg
On Tue, Jan 20, 2015 at 1:46 PM, Chris Wilson wrote: > On Tue, Jan 20, 2015 at 12:53:35PM -0800, Kristian Høgsberg wrote: >> On Tue, Jan 20, 2015 at 12:42 AM, Daniel Vetter wrote: >> > My idea for all this would have been to create a per-thread execbuf >> > relocation context with a hashtab to ma

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-21 Thread Daniel Vetter
On Tue, Jan 20, 2015 at 10:10:57PM -0800, Kristian Høgsberg wrote: > On Tue, Jan 20, 2015 at 1:46 PM, Chris Wilson > wrote: > > On Tue, Jan 20, 2015 at 12:53:35PM -0800, Kristian Høgsberg wrote: > >> On Tue, Jan 20, 2015 at 12:42 AM, Daniel Vetter wrote: > >> > My idea for all this would have be

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-21 Thread Daniel Vetter
On Tue, Jan 20, 2015 at 10:10:57PM -0800, Kristian Høgsberg wrote: > However, drm_intel_bo_reference/unreference() is showing up on the > profiles now that relocations are cheaper, but I think the better way > to make those cheaper is to move the ref count to the public struct > and make the ref/un

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-21 Thread Daniel Vetter
On Wed, Jan 21, 2015 at 10:17:01AM +0100, Daniel Vetter wrote: > On Tue, Jan 20, 2015 at 10:10:57PM -0800, Kristian Høgsberg wrote: > > However, drm_intel_bo_reference/unreference() is showing up on the > > profiles now that relocations are cheaper, but I think the better way > > to make those chea

Re: [Intel-gfx] [PATCH 2/2] intel: Use I915_EXEC_NO_RELOC when available

2015-01-21 Thread Chris Wilson
On Tue, Jan 20, 2015 at 10:10:57PM -0800, Kristian Høgsberg wrote: > However, drm_intel_bo_reference/unreference() is showing up on the > profiles now that relocations are cheaper, but I think the better way > to make those cheaper is to move the ref count to the public struct > and make the ref/un