On Wed, Jun 8, 2011 at 2:03 AM, S. Erisman <seris...@serisman.com> wrote:
> Marc,
>
>
> On 6/7/2011 11:35 PM, Marc-André Moreau wrote:
>
>> Hi Steve,
>>
>> I just tried your patch - awesome!
>>
>> Thanks. That was the first SSE code I have ever written and it ended up
> being pretty easy. Once we have high level agreement on the structure
> needed around these optimizations there is definitely more that can probably
> be optimized using SSE in the RemoteFX code, and I should be able to keep
> working on it.
>
>
> I can almost watch a youtube video in fullscreen now with a resolution of
>> 1680x1050.
>>
> What kind of computer are you using for your testing? I have been mainly
> testing with an Atom D510 and so far haven't seen much improvement. :(
Ah, we're using totally different hardware. Comparison must be relative to
the same hardware to actually make sense. I have a Core i7-950 with 16GB of
DDR3 RAM, which is obviously non-representative of lower-end systems. I have
virtualbox running one instance of Windows Server 2008 R2 SP1 with RemoteFX
in RDSH. I also have another vm running Windows 7 SP1 with the version of
mstsc.exe that supports RemoteFX for comparison. mstsc.exe in the vm is
still relatively faster than xfreerdp at this point for me.
>
> For testing performance, I think we can go with relative comparison of
>> performance, unless someone wants to implement some way to properly measure
>> performance.
>>
> Well that might work if the change was big enough to be easily visible. On
> my test system (see above) I actually haven't been able to tell any real
> difference between any of the optimizations that have been made so far.
> That includes all the memory allocation work that was done the other day as
> well.
>
> I also want to try a second version of the SSE method that I already wrote
> (using integer logic instead of floats). Without being able to see subtle
> differences I won't know which one actually performs better.
>
> There has to be an easy way to do this. I have used several different
> tools/techniques on the Windows side of the world in the past (most of my
> experience has been in the Windows world so far). I will keep thinking
> about it and looking for a solution.
Ok, we need metrics, just comparing how performance feels is not enough. I'm
thinking that we could record the time it takes to decode each tile and make
the "average tile decoding time" a good way to compare performance using
numbers.
>
>
> It's easy to find ways to stress the implementation and see if it works
>> faster or slower. In my case, I try increasingly larger resolutions for
>> youtube videos of different qualities put in full screen viewing.
>> Non-fullscreen youtube is already very smooth, but it's not smooth yet when
>> you want to go fullscreen, especially as resolutions get larger. The ideal
>> goal would be to get a full hd video smoothly played over RemoteFX, in a
>> full hd resolution RDP session.
>>
> My ideal goal would be to get smooth application refreshes (including
> dragging/scrolling/minimizing/maximizing windows) on a ~$100-$200 basic PC
> (think Intel Atom or AMD Zacate system) with a dual-monitor setup running at
> up to 1920x1200. A secondary goal for me would be smooth HD video running
> on at least one of the monitors. I have no idea if we will get there, but
> every little bit will help.
>
I have no idea as well, but we'll try out best to get every possible bit of
power out of the hardware available :P
>
> Thanks,
> Steve
>
------------------------------------------------------------------------------
EditLive Enterprise is the world's most technically advanced content
authoring tool. Experience the power of Track Changes, Inline Image
Editing and ensure content is compliant with Accessibility Checking.
http://p.sf.net/sfu/ephox-dev2dev
_______________________________________________
Freerdp-devel mailing list
Freerdp-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/freerdp-devel
