Joseph Artsimovich writes:
> On 18/07/2011 22:56, Måns Rullgård wrote:
>> Joseph Artsimovich writes:
>>
>>> On 18/07/2011 15:45, Måns Rullgård wrote:
Maxim writes:
> Am 18.07.2011 16:09, schrieb Måns Rullgård:
>> [...]
>>
>>> An integer implementation is surely much f
On 18/07/2011 22:56, Måns Rullgård wrote:
Joseph Artsimovich writes:
On 18/07/2011 15:45, Måns Rullgård wrote:
Maxim writes:
Am 18.07.2011 16:09, schrieb Måns Rullgård:
[...]
An integer implementation is surely much faster than floating-point
one but it's still insufficient for realtim
On 18/07/2011 17:11, Måns Rullgård wrote:
"Ronald S. Bultje" writes:
Hi,
On Mon, Jul 18, 2011 at 8:33 AM, Joseph Artsimovich wrote:
On 18/07/2011 16:29, Kostya wrote:
On Mon, Jul 18, 2011 at 04:20:14PM +0100, Joseph Artsimovich wrote:
On 18/07/2011 16:01, Kostya wrote:
On Mon, Jul 18, 20
On Tue, Jul 19, 2011 at 10:27:06AM +0100, Joseph Artsimovich wrote:
> On 18/07/2011 19:08, Maxim wrote:
> >Am 18.07.2011 18:11, schrieb Måns Rullgård:
> >>"Ronald S. Bultje" writes:
> >>
> >>
> >>>Hi,
> >>>
> >>>On Mon, Jul 18, 2011 at 8:33 AM, Joseph Artsimovich
> >>>wrote:
> >>>
> On 18/07
On 18/07/2011 19:08, Maxim wrote:
Am 18.07.2011 18:11, schrieb Måns Rullgård:
"Ronald S. Bultje" writes:
Hi,
On Mon, Jul 18, 2011 at 8:33 AM, Joseph Artsimovich wrote:
On 18/07/2011 16:29, Kostya wrote:
On Mon, Jul 18, 2011 at 04:20:14PM +0100, Joseph Artsimovich wrote:
On 18/07/2011
Joseph Artsimovich writes:
> On 18/07/2011 15:45, Måns Rullgård wrote:
>> Maxim writes:
>>
>>> Am 18.07.2011 16:09, schrieb Måns Rullgård:
[...]
> An integer implementation is surely much faster than floating-point
> one but it's still insufficient for realtime applications
>>>
Am 18.07.2011 18:11, schrieb Måns Rullgård:
> "Ronald S. Bultje" writes:
>
>
>> Hi,
>>
>> On Mon, Jul 18, 2011 at 8:33 AM, Joseph Artsimovich
>> wrote:
>>
>>> On 18/07/2011 16:29, Kostya wrote:
>>>
On Mon, Jul 18, 2011 at 04:20:14PM +0100, Joseph Artsimovich wrote:
"Ronald S. Bultje" writes:
> Hi,
>
> On Mon, Jul 18, 2011 at 8:33 AM, Joseph Artsimovich
> wrote:
>> On 18/07/2011 16:29, Kostya wrote:
>>> On Mon, Jul 18, 2011 at 04:20:14PM +0100, Joseph Artsimovich wrote:
On 18/07/2011 16:01, Kostya wrote:
> On Mon, Jul 18, 2011 at 03:56:42PM +0100,
Hi,
On Mon, Jul 18, 2011 at 8:33 AM, Joseph Artsimovich wrote:
> On 18/07/2011 16:29, Kostya wrote:
>> On Mon, Jul 18, 2011 at 04:20:14PM +0100, Joseph Artsimovich wrote:
>>> On 18/07/2011 16:01, Kostya wrote:
On Mon, Jul 18, 2011 at 03:56:42PM +0100, Joseph Artsimovich wrote:
> On 18/07
Hi,
On Mon, Jul 18, 2011 at 5:03 AM, Kostya wrote:
> On Mon, Jul 18, 2011 at 12:53:46PM +0100, Joseph Artsimovich wrote:
>> On 18/07/2011 12:06, Kostya wrote:
>> >On Mon, Jul 18, 2011 at 11:54:31AM +0100, Joseph Artsimovich wrote:
>> >>Here is the latest version of my patch set, made against toda
On 18/07/2011 16:29, Kostya wrote:
On Mon, Jul 18, 2011 at 04:20:14PM +0100, Joseph Artsimovich wrote:
On 18/07/2011 16:01, Kostya wrote:
On Mon, Jul 18, 2011 at 03:56:42PM +0100, Joseph Artsimovich wrote:
On 18/07/2011 15:45, Måns Rullgård wrote:
Maximwrites:
Am 18.07.2011 16:09, schri
On Mon, Jul 18, 2011 at 04:20:14PM +0100, Joseph Artsimovich wrote:
> On 18/07/2011 16:01, Kostya wrote:
> >On Mon, Jul 18, 2011 at 03:56:42PM +0100, Joseph Artsimovich wrote:
> >>On 18/07/2011 15:45, Måns Rullgård wrote:
> >>>Maxim writes:
> >>>
> Am 18.07.2011 16:09, schrieb Måns Rullgård:
On 18/07/2011 16:01, Kostya wrote:
On Mon, Jul 18, 2011 at 03:56:42PM +0100, Joseph Artsimovich wrote:
On 18/07/2011 15:45, Måns Rullgård wrote:
Maxim writes:
Am 18.07.2011 16:09, schrieb Måns Rullgård:
[...]
An integer implementation is surely much faster than floating-point
one but it'
On Mon, Jul 18, 2011 at 03:56:42PM +0100, Joseph Artsimovich wrote:
> On 18/07/2011 15:45, Måns Rullgård wrote:
> >Maxim writes:
> >
> >>Am 18.07.2011 16:09, schrieb Måns Rullgård:
> >>>[...]
> >>>
> An integer implementation is surely much faster than floating-point
> one but it's still i
On 18/07/2011 15:45, Måns Rullgård wrote:
Maxim writes:
Am 18.07.2011 16:09, schrieb Måns Rullgård:
[...]
An integer implementation is surely much faster than floating-point
one but it's still insufficient for realtime applications
The point?
You'll be able to obtain reasonable performa
Maxim writes:
> Am 18.07.2011 16:09, schrieb Måns Rullgård:
>> [...]
>>
>>> An integer implementation is surely much faster than floating-point
>>> one but it's still insufficient for realtime applications
>>>
>> The point?
>>
>
> You'll be able to obtain reasonable performance benefit
Am 18.07.2011 16:09, schrieb Måns Rullgård:
> [...]
>
>> An integer implementation is surely much faster than floating-point
>> one but it's still insufficient for realtime applications
>>
> The point?
>
You'll be able to obtain reasonable performance benefits only by using
SIMD-optimiz
Maxim writes:
> Am 18.07.2011 14:10, schrieb Måns Rullgård:
>> [...]
>> Why are you using the incredibly slow floating-point dct anyway? It
>> can't be that hard to extend one of the integer ones to 10-bit. It
>> shouldn't take more than updating the coefficients and some shift
>> values.
>>
On Mon, Jul 18, 2011 at 03:45:30PM +0200, Maxim wrote:
>
> I was able to update the existing Altivec code (ppc/idct_altivec.c) to
> support 10bit and would submit a patch if someone cares. Unfortunately I
> don't have any solution for x86 etc.
There's never a reason not to submit such a patch - g
On Mon, Jul 18, 2011 at 11:54:31AM +0100, Joseph Artsimovich wrote:
> Here is the latest version of my patch set, made against today's Git.
>
> --- a/libavcodec/dnxhddec.c
> +++ b/libavcodec/dnxhddec.c
> @@ -43,20 +47,33 @@ typedef struct {
>
> +static void dnxhd_8bit_idct_put(DNXHDContext* ctx,
Am 18.07.2011 14:10, schrieb Måns Rullgård:
> [...]
> Why are you using the incredibly slow floating-point dct anyway? It
> can't be that hard to extend one of the integer ones to 10-bit. It
> shouldn't take more than updating the coefficients and some shift
> values.
>
Unfortunately such a t
Joseph Artsimovich writes:
> Here is the latest version of my patch set, made against today's Git.
>
> I think I addressed all raised issues, except I decided not to mess
> with dsputil.c, at least until we actually have some
> assembly-accelerated routines for 10-bit DNxHD.
It would be much cle
On Mon, Jul 18, 2011 at 12:53:46PM +0100, Joseph Artsimovich wrote:
> On 18/07/2011 12:06, Kostya wrote:
> >On Mon, Jul 18, 2011 at 11:54:31AM +0100, Joseph Artsimovich wrote:
> >>Here is the latest version of my patch set, made against today's Git.
> >>
> >>I think I addressed all raised issues, e
On 18/07/2011 12:06, Kostya wrote:
On Mon, Jul 18, 2011 at 11:54:31AM +0100, Joseph Artsimovich wrote:
Here is the latest version of my patch set, made against today's Git.
I think I addressed all raised issues, except I decided not to mess
with dsputil.c, at least until we actually have some
a
On Mon, Jul 18, 2011 at 11:54:31AM +0100, Joseph Artsimovich wrote:
> Here is the latest version of my patch set, made against today's Git.
>
> I think I addressed all raised issues, except I decided not to mess
> with dsputil.c, at least until we actually have some
> assembly-accelerated routines
25 matches
Mail list logo