Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Matt]

We currently require a CPU with MMX support (MMX was introduced in 1996).

Would anyone object if we raised this requirement to SSE2?

Intel added SSE2 to the Pentium 4 in 2001.

AMD added SSE2 to the Opteron and Athlon 64 in 2003.

More here:
http://en.wikipedia.org/wiki/SSE2

Does anyone use hardware older than this with Media Center 18?

[mojave]

I don't have anything that old and don't know any other JRiver users locally with computers that old.

[glynor]

As long as most of the Atoms are included, I think you're fine.  And they do.

[Hendrik]

I have been thinking about the same requirement in LAV for a while now, who really runs a PC older then a P4 these days?

[bunglemebaby]

I finally retired my AMD Athlon 2100 system this year. So no objections from me.
To be honest MC was just a glorified file explorer on that machine anyway, so an old version would have sufficed if this change were made to MC18 and I was still using that system.

[kensn]

Nothing older than a P4 sounds pretty resonable to me.

Ken

[AndrewFG]

In my own application (Whitebear) I use SSE2 assembly code for byte order swapping. However I use a conditional to check for SSE2 support, and branch to the original Intel assembly code if not detected. It makes your compiled code a bit bigger, but the conditional switch hardly slows you down at all (you obviously don't call the switch inside of tight loops).

[AndrewFG]

By the way, I think that SSE2 support is a pre-requisitite for running Windows 8.
So you would be in good company...

[InflatableMouse]

I don't even know people that use P4's or older these days.

[Matt]

In my own application (Whitebear) I use SSE2 assembly code for byte order swapping. However I use a conditional to check for SSE2 support, and branch to the original Intel assembly code if not detected. It makes your compiled code a bit bigger, but the conditional switch hardly slows you down at all (you obviously don't call the switch inside of tight loops).

A slow fallback is normally the way to go.

But I'm trying to use SSE in a simple bitdepth converter.  A conditional for an FPU/CPU fallback makes it about 20% slower.

I can push the conditional up a level, but this breaks the encapsulation and would require writing two versions of lots of audio loops (gross).  Trying to use a pure virtual interface with different implementations for SSE and FPU destroys the ability to inline, hurting performance even more than a conditional.

So the fallback has a real cost in this case -- either in code complexity or performance.

[SpeedD408]

Not a short term solution, but how about grabbing the sys info at registration time.  That would tell you what hardware people have at it would be worst case since people typically don't down grade hardware.  This would help you to know what hardware people are running and how to best deal with these type issues.

[MrHaugen]

I do not know of any of my friends running computers older than this, and expecting to run apps like MC on it. So, I would think it's pretty safe.

[Hendrik]

By the way, I think that SSE2 support is a pre-requisitite for running Windows 8.
So you would be in good company...

And i thought i heard Windows 8 runs on any PC with PAE (any since Pentium Pro).. i should go find out.

[leezer3]

Now, a while back, someone brought up being able to use newer licences with older versions.
IMHO, if you make this change, then it'd be an advantage to provide an older build (version) able to be used on SSE2 machines.

I also seem to recall there's been at least a couple of bug reports in the relatively recent past from users running on slightly odd operating systems (Windows 2000 Pro springs to mind, although I'm not looking for the thread now), but I don't know what the hardware in use there would be likely to be?

To put it another way, the minute you make this sort of change, you'll probably acquire a vocal minority of users complaining about it. Whether the speed increases are worth the aggro, you'll have to decide

-Leezer-

[DarkPenguin]

I thought P4's were against the Kyoto treaty.

First you abandon windows 95 and now this.  Don't think that everyone supports your frenetic upgrade pace!

Yeah, it is fine by me.  The cost of a machine that would just barely run this is, what, $100?

[AndrewFG]

And i thought i heard Windows 8 runs on any PC with PAE (any since Pentium Pro).. i should go find out.

See this: http://msdn.microsoft.com/en-us/library/windows/hardware/hh975398.aspx -- Note that it only mentions SSE2 being a prerequisite; which seems to imply that plain SSE is not a prerequisite. Or does anyone read this differently?

[Hendrik]

There is no CPU with SSE2 but not SSE.