This looks like the right approach to me.
Note that RNDNE is in fact the default rounding mode for IEEE_754-2008
(in fact it was the default for earlier ieee-754 too).
So I see no reason to deviate from that unless there would be some very
good reason, even if it may look "wrong" for humans.
Rolan
>From f76d23a57996eea6862d3bd899e08f9cb3ac5dec Mon Sep 17 00:00:00 2001
From: Lauri Kasanen
Date: Mon, 8 Aug 2011 12:20:27 +0300
Subject: [PATCH] r600g: Add support for ROUND, v2
This is a GLSL 1.3 feature, but also used by MLAA.
Signed-off-by: Lauri Kasanen
---
src/gallium/drivers/r600/r600_a
> > On 8 August 2011 03:58, Jose Fonseca wrote:
> > > It's subjective. It depends on the expected input distribution,
> > > which is effectively impossible to characterize in general. One
> > > can easily find datasets where one method gives biased results and
> > > the other not, and vice versa.
- Original Message -
> On 8 August 2011 03:58, Jose Fonseca wrote:
> > It's subjective. It depends on the expected input distribution,
> > which is effectively impossible to characterize in general. One
> > can easily find datasets where one method gives biased results and
> > the other
On 8 August 2011 03:58, Jose Fonseca wrote:
> It's subjective. It depends on the expected input distribution, which is
> effectively impossible to characterize in general. One can easily find
> datasets where one method gives biased results and the other not, and vice
> versa. And if one takes
- Original Message -
> On 8 August 2011 02:24, Jose Fonseca wrote:
> > There's no "wrong" or "right" when there are two equidistant
> > integers -- it's all a matter of convention.
> >
> But note that rounding to nearest even is a slightly better
> convention
> in terms of rounding bias. I
On 8 August 2011 02:24, Jose Fonseca wrote:
> There's no "wrong" or "right" when there are two equidistant integers -- it's
> all a matter of convention.
>
But note that rounding to nearest even is a slightly better convention
in terms of rounding bias. I.e., not using RNDNE is both likely to be
- Original Message -
> On Sun, 7 Aug 2011 19:22:06 +0200
> Henri Verbeet wrote:
>
> > On 7 August 2011 19:03, Lauri Kasanen wrote:
> > > + /* floor(a + 0.5) */
> > Why not use RNDNE?
>
> I feel it would cause more "wrong" results, with two numbers
> resolving to the same (3.5 = 4.
On Sun, 7 Aug 2011 19:22:06 +0200
Henri Verbeet wrote:
> On 7 August 2011 19:03, Lauri Kasanen wrote:
> > + /* floor(a + 0.5) */
> Why not use RNDNE?
I feel it would cause more "wrong" results, with two numbers resolving to the
same (3.5 = 4.5 = 4).
Of course for best rounding one could
On 7 August 2011 19:03, Lauri Kasanen wrote:
> + /* floor(a + 0.5) */
Why not use RNDNE?
___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
>From 4b04e2b45e3bb2acaca838453f999bc53ba35fef Mon Sep 17 00:00:00 2001
From: Lauri Kasanen
Date: Sun, 7 Aug 2011 18:19:17 +0300
Subject: [PATCH] r600g: Add support for ROUND
This is a GLSL 1.3 feature, but also used by MLAA.
Signed-off-by: Lauri Kasanen
---
src/gallium/drivers/r600/r600_shad
11 matches
Mail list logo
