KVR Audio

Basilbasil · Post by **Basilbasil** » Fri Jun 07, 2013 9:22 pm

Good afternoon, it is required to create a package of plug-ins for the new DAW. who want to try their hand?

and post your price for this work please.
is required:
construction of filter circuits
work with the code generation tools(Matlab, etc.)
Filters can be created in C or pseudocode.

ponce · Post by **ponce** » Mon Jun 10, 2013 10:21 am

I'll do it for 1 billion dollars.
Send me the files on my Skype.

Basilbasil · Post by **Basilbasil** » Mon Jun 10, 2013 10:39 am

ponce wrote:I'll do it for 1 billion dollars.
Send me the files on my Skype.

hahaha, very funny.
Come on, prove that the stand of the money, and we will consider your offer.
DSP developer billion cost to be Gandalf and Dumbledore in one.
Show your portfolio, please tell us about your education.

Our proposal is really serious, since we had almost finished the architecture and audio engine.

ponce · Post by **ponce** » Mon Jun 10, 2013 11:00 am

OK I was kidding. I suggest you provide _way_ more information (who are you, what are you doing, what is "the new DAW") to find contractors, else any sort of pricing would be premature. I'm sure you are serious, but a message like you wrote does not really reflect that.

Basilbasil · Post by **Basilbasil** » Mon Jun 10, 2013 11:51 am

ponce wrote:OK I was kidding. I suggest you provide _way_ more information (who are you, what are you doing, what is "the new DAW") to find contractors, else any sort of pricing would be prematurate. I'm sure you are serious, but a message like you wrote does not really reflect that.

Pardon me, we are a young company from Russia, engaged in development of DAW with the engine running on the GPU. Why GPU?
All audio processing algorithms (and synths) are working on the GPU. This gives:
- Scaling (not enough - went to the store bought a more powerful video card., Or just two., Or 3))) Do not need a global upgrade of the entire platform - motherboard + CPU + RAM).
- Best value "performance / money"
- Just power
- Stability (sufficient comfort of up to 99% engine load at minimum latency)
- The release of more powerful hardware is made each year by NVIDIA and AMD

Our successes are: full load of the engine for 2-3 ms asio latency without loss of buffers, partially working protocol "MultiASIO" etc.
Prepares SDK, AMD is preparing to support the video card.

We can not offer a lot of money, but option is possible.

we need to create the following plug-ins: Compressor, EQ, Limiter, Delay, Chorus, Flanger, Reverb (synth + IR), Gate, Phaser, Distortion, Synth.

If you are interested in - ask for more.

Best regard, Basil.

whyterabbyt · Post by **whyterabbyt** » Mon Jun 10, 2013 12:03 pm

Ouch, another proprietary format, and nonstandard hardware platform. Erm, good luck....

Basilbasil · Post by **Basilbasil** » Mon Jun 10, 2013 12:10 pm

whyterabbyt wrote:Ouch, another proprietary format, and nonstandard hardware platform. Erm, good luck....

no difficulty to design, simple SDK
yes, there will be some pitfalls in the optimization of parallel programming, but we will resolve these issues together, and we have people with the necessary skills.

ponce · Post by **ponce** » Mon Jun 10, 2013 2:43 pm

Basilbasil wrote: Our successes are: full load of the engine for 2-3 ms asio latency without loss of buffers, partially working protocol "MultiASIO" etc.

If you achieve such latencies through GPU it's quite an achievement indeed.

I think it's a good opportunity to try things that are impractical on a general purpose CPU.
Large FIR kernels, resampling through 1D mipmaps comes to mind.

Are you using shaders for sound effects?
Since GL_ARB_shader_image_load_store some GPUs (NVIDIA and AMD ones) seem to support any load/store scheme within a single pass, so a lot of things should be possible indeed.

Best regards

PurpleSunray · Post by **PurpleSunray** » Mon Jun 10, 2013 3:10 pm

sound likes an interesting project.. but unfortunately no time to apply.. :/
Would be interesting to find a way to keep the gpu pipe filled, especially with the modules that need historic samples by definition (if you'r on shader code). Was already some years ago I played around with that stuff last time.. maybe they introduced some more pipe-controlling-functions in the meanwhile but back than, if your algorithm doesn't allow you to sequentially process data your basically lost as a GPU pipe is quite a bit longer than the one of a CPU and so re-filling it because of random memory access almost immediately pushes the throughput below of what you get on the CPU.

anyway.. good luck with searching for devs.. would love to work on that but there is so much other stuff on my "todo list" already.. :/

DaveHoskins · Post by **DaveHoskins** » Mon Jun 10, 2013 4:47 pm

Basilbasil wrote:
ponce wrote:OK I was kidding. I suggest you provide _way_ more information (who are you, what are you doing, what is "the new DAW") to find contractors, else any sort of pricing would be prematurate. I'm sure you are serious, but a message like you wrote does not really reflect that.
Pardon me, we are a young company from Russia, engaged in development of DAW with the engine running on the GPU. Why GPU?
All audio processing algorithms (and synths) are working on the GPU. This gives:
- Scaling (not enough - went to the store bought a more powerful video card., Or just two., Or 3))) Do not need a global upgrade of the entire platform - motherboard + CPU + RAM).
- Best value "performance / money"
- Just power
- Stability (sufficient comfort of up to 99% engine load at minimum latency)
- The release of more powerful hardware is made each year by NVIDIA and AMD

Our successes are: full load of the engine for 2-3 ms asio latency without loss of buffers, partially working protocol "MultiASIO" etc.
Prepares SDK, AMD is preparing to support the video card.

We can not offer a lot of money, but option is possible.

we need to create the following plug-ins: Compressor, EQ, Limiter, Delay, Chorus, Flanger, Reverb (synth + IR), Gate, Phaser, Distortion, Synth.

If you are interested in - ask for more.
Best regard, Basil.

That's interesting, is that using Cuda or OpenCL, or are you using straight GPU shader code in DirectX or OpenGL?

Basilbasil · Post by **Basilbasil** » Mon Jun 10, 2013 7:48 pm

shaders is no longer fashionable to use))).
before it was comfortable and somehow possible to work.
Now it's "CUDA" and the level below

wavepunk · Post by **wavepunk** » Tue Jun 11, 2013 11:38 am

We can not offer a lot of money

Hi, how much you offer?

MeldaProduction · Post by **MeldaProduction** » Tue Jun 11, 2013 11:50 am

Well, I don't know almost anything about CUDA, but I can hardly see this possible - if you want latency of 3ms in 44k, that would be about 100 samples. That doesn't seem to be very optimal for optimization via GPU, or can it? When I imagine the data flow from CPU to GPU and back for every process in series, this seems at least impractical. Or do you have some statistics?

Basilbasil · Post by **Basilbasil** » Tue Jun 11, 2013 12:12 pm

wavepunk wrote:
We can not offer a lot of money
Hi, how much you offer?

1. replied in PM
2. Please tell me about your skill and portfolio

Basilbasil · Post by **Basilbasil** » Tue Jun 11, 2013 12:28 pm

MeldaProduction wrote:Well, I don't know almost anything about CUDA, but I can hardly see this possible - if you want latency of 3ms in 44k, that would be about 100 samples. That doesn't seem to be very optimal for optimization via GPU, or can it? When I imagine the data flow from CPU to GPU and back for every process in series, this seems at least impractical. Or do you have some statistics?

data flow is an exact scheduled time according to the specifications PCI-E ))))

execution of multiple parallel working filters takes time on the order of less than 1 ms (including 2 way transaction).
We already have some simple filters, latency of which we are measuring.
Interested? - Let's work together

required to create plugin bundle