+ SSE2=64bit audio, SSE3=faster FFT... please

[medanby]

Funnily enough, infuzion, my last PC was about 7 years old!!!!!!! Until the new PC I could only dream about a CODE and ASSEMBLER module. It was a hard time !

I had been using a freely available maths packacge (Scilab) for non realtime audio DSP and found uses for a lot of functions.

First of all sin1,cos1 and tan1 could have their inverses.
Second log10 could be accompanied by loge and exp.
Third a bit of power is good ^ and follows on from 2, sqrt can be done with ^ but always looks nicer as sqrt(x)

These of course can be coded from a little knowledge of Maclaurin & Taylor expansions, but is, however rather long winded. I guess that really means if you like beans the inevitible will happen more frequently whilst coding. Thinking of M & T reminds me of;

Fourth Factorial n! ie 1*2*3*...n

1,2,3 would be a great starting place I find the CODE module far friendlier than all other 'high-level' IDES which change syntax so often with version that, unless you use them all the while, you're forever reading the help files which is CR|) (censored for US friends).

I intend to write some Spectral processing software which I have made some algorythms for obviously that brings me to the next request;

Fifth FFT & IFFT Real to complex & complex to real The HOLY GRAIL V

Of course I have posted elsewhere for the inclusion of an IFFT to accompany the GraphFFT component.

The CODE component with these would be like mana from heave? (not to offend any atheists)

What does everybody thing?

Michael

[infuzion]

I know you guys have alot on your plate, but I hope this is in the burner somewhere... Now that is is easy to find a SSE2 computer for less than $400USD new, $70 used, there shouldn't be much of an issue of compatibility, unless maybe one is making free VSTs for broke people. Also, 64bit audio (not talking OS here) is becoming standard, like in Live7, so it is becoming expected (at least with the clients I'm trying to attract). New computers can handle SSE3, & newest some sort of SSE4...

So please add SSE2 & maybe 3 ASM soon; shouldn't be too hard.

A primitive that output what level & kind of SSE the CPU can handle (or that outputs the raw CPU ID & the module would sort that out). Not a big deal though, I don't mind releasing an extra SSE2+ only version & people can use CPU-Z

cheers

[TomC]

Oh, I missed that posting.

Bumpidy bumpidiy bump

Cheers.

Tom

[infuzion]

...pretty please? Even if only in the Assembler? Live & many DAWs have 64bit audio, & many of the new SSEx instructions can help make even faster code.

BTW, the SSSE3 & SSE4 instructions look nice, even if they are Intel-only.

[infuzion]

A few more "votes" fore a fuller 64bit support:
1 sample delay for double stream?
Truncation math

[infuzion]

Did I forget to mention that SSE3 speeds up FFT?
http://www.fftw.org/
http://ieeexplore.ieee.org/xpl/freeabs_ ... mber=34759
http://developer.amd.com/documentation/ ... 07122.aspx
http://www.springerlink.com/content/6481607675376w42/

[aliasant]

Could it be so that that the developers is saving this for SM 2.0 ?

That is if SM 2.0 isnt just a fairy tale...

[infuzion]

Could it be so that that the developers is saving this for SM 2.0 ?

Since the time of the original posting, we did get some Stream "Double" math primitives (really just for the sample playback modules), so that is a step in the right direction. But of course, we/I want more!!
Really, I'd be happy with just all of the ASM commands open in the ASM-code, or atleast SSE2-4.1.

[kirkagur]

Could it be so that that the developers is saving this for SM 2.0 ?

That is if SM 2.0 isnt just a fairy tale...

+1 For the 64 Bit before SM 2.0

[TheAudiophileDutchman]

About timing: Wouldnt it be great to have an SM v2.0 beta as soon as we hit 2000 total members @this SM forum? (currently #1932)

[infuzion]

About timing: Wouldn't it be great to have an SM v2.0 beta as soon as we hit 2000 total members @this SM forum? (currently #1932)

That's about when a 1.1.3 beta will be out, & SSE2+ is not in there.

Nice to see an old face back here Mr Audiophile Dutchman !

[TheAudiophileDutchman]

Nice to see an old face back here Mr Audiophile Dutchman !

Thanks, infuzion. I hope to share some 'fireworks' again in 2009!

[RuleT]

It seems I don't have permission to vote, but verbally I vote for the first option "Yes, I have atleast one SSE2 computer now". And I completely agree with infuzion's offer! And even more, I offer to add SSEX (X is version), MMX & 3DNow!(maybe useful), etc.. and expand current SSE1 & previous x86 opcodes list.
Of course, it brings many difficulties, and I think that any improvements should not decrease current simplicity and reliability. Therefore, it may take much efforts.

[infuzion]

It seems I don't have permission to vote...

I made it limited time. Pointless to re-open it now, since one can buy a brand new SSE3 compatible netbook for $250.

Honestly, I've given up on the idea of SM having SSE2+ this decade. When I get to the point of needing it bad enough, I'll just learn C++ & port.

[infuzion]

...since one can buy a brand new SSE3 compatible netbook for $250.

If one really looks hard enough, [urlhttp://dealnews.com/categories/Computer/PC-Computers/PC-Laptop/Netbooks/535.html]netbook sales are $200USD[/url] now, so SSE2+ is for the masses. I've been poking around SM's internal ASM, & see that there are many opcodes not available for the ASM primitive... please share some Christmas love here & let us have SSE2 & 3.