Re: [Intel-gfx] [PATCH] drm/i915: swizzling support for snb/ivb

2012-02-08 Thread Daniel Vetter
On Tue, Feb 07, 2012 at 11:56:36AM -0800, Eric Anholt wrote: > On Mon, 6 Feb 2012 16:45:00 +0100, Daniel Vetter wrote: > > On Sat, Feb 04, 2012 at 09:59:57PM +0100, Eric Anholt wrote: > > > On Thu, 2 Feb 2012 09:58:12 +0100, Daniel Vetter > > > wrote: > > > It looks to me like you're making the

Re: [Intel-gfx] [PATCH] drm/i915: swizzling support for snb/ivb

2012-02-07 Thread Eric Anholt
On Mon, 6 Feb 2012 16:45:00 +0100, Daniel Vetter wrote: > On Sat, Feb 04, 2012 at 09:59:57PM +0100, Eric Anholt wrote: > > On Thu, 2 Feb 2012 09:58:12 +0100, Daniel Vetter > > wrote: > > > We have to do this manually. Somebody had a Great Idea. > > > > > > I've measured speed-ups just a few pe

Re: [Intel-gfx] [PATCH] drm/i915: swizzling support for snb/ivb

2012-02-06 Thread Daniel Vetter
On Sat, Feb 04, 2012 at 07:13:41PM -0800, Ben Widawsky wrote: > On 02/02/12 00:40, Daniel Vetter wrote: > >On Wed, Feb 01, 2012 at 09:30:57PM -0800, Ben Widawsky wrote: > >[...] > >>I'd also say it's not a bad idea to elaborate the assumption that we > >>never have less than 256MB of memory WARN_ON

Re: [Intel-gfx] [PATCH] drm/i915: swizzling support for snb/ivb

2012-02-06 Thread Daniel Vetter
On Sat, Feb 04, 2012 at 09:59:57PM +0100, Eric Anholt wrote: > On Thu, 2 Feb 2012 09:58:12 +0100, Daniel Vetter > wrote: > > We have to do this manually. Somebody had a Great Idea. > > > > I've measured speed-ups just a few percent above the noise level > > (below 5% for the best case), but no

Re: [Intel-gfx] [PATCH] drm/i915: swizzling support for snb/ivb

2012-02-04 Thread Ben Widawsky
On 02/02/12 00:40, Daniel Vetter wrote: On Wed, Feb 01, 2012 at 09:30:57PM -0800, Ben Widawsky wrote: [...] I'd also say it's not a bad idea to elaborate the assumption that we never have less than 256MB of memory WARN_ON(dimm_c0 + dimm_c1 == 0). I'm pretty sure that we can't boot with 0mb of

Re: [Intel-gfx] [PATCH] drm/i915: swizzling support for snb/ivb

2012-02-04 Thread Eric Anholt
On Thu, 2 Feb 2012 09:58:12 +0100, Daniel Vetter wrote: > We have to do this manually. Somebody had a Great Idea. > > I've measured speed-ups just a few percent above the noise level > (below 5% for the best case), but no slowdows. Chris Wilson measured > quite a bit more (10-20% above the usua

[Intel-gfx] [PATCH] drm/i915: swizzling support for snb/ivb

2012-02-02 Thread Daniel Vetter
We have to do this manually. Somebody had a Great Idea. I've measured speed-ups just a few percent above the noise level (below 5% for the best case), but no slowdows. Chris Wilson measured quite a bit more (10-20% above the usual snb variance) on a more recent and better tuned version of sna, but

Re: [Intel-gfx] [PATCH] drm/i915: swizzling support for snb/ivb

2012-02-02 Thread Daniel Vetter
On Wed, Feb 01, 2012 at 09:30:57PM -0800, Ben Widawsky wrote: > Hmm, for my curiosity, why doesn't swizzling make sense with 3 channels? > I did some searching and it appears that you're right about no product > shipping with the configuration, but I suspect the comment will help in > case any prod

Re: [Intel-gfx] [PATCH] drm/i915: swizzling support for snb/ivb

2012-02-01 Thread Ben Widawsky
On Thu, Feb 02, 2012 at 12:15:38AM +0100, Daniel Vetter wrote: > We have to do this manually. Somebody had a Great Idea. > > I've measured speed-ups just a few percent above the noise level > (below 5% for the best case), but no slowdows. Chris Wilson measured > quite a bit more (10-20% above the

[Intel-gfx] [PATCH] drm/i915: swizzling support for snb/ivb

2012-02-01 Thread Daniel Vetter
We have to do this manually. Somebody had a Great Idea. I've measured speed-ups just a few percent above the noise level (below 5% for the best case), but no slowdows. Chris Wilson measured quite a bit more (10-20% above the usual snb variance) on a more recent and better tuned version of sna, but