Yesterday, there was Alea.cuBase live coding session. Don Syme led live tweets from inside. This post is dedicated to those who missed it.
I’ll be live tweeting from the .NET in the City event where Daniel Egloff presents F# for GPGPU quantalea.net/news/29/
— Don Syme (@dsyme) May 16, 2013
Daniel Egloff overviewing the challenges of quant software development #fsharp #cuda
— Don Syme (@dsyme) May 16, 2013
Why F# for quant? Strong types, high productivity, highly flexible, extensible, ideal for #gpgpu, lots more too quantalea.net/news/29/
— Don Syme (@dsyme) May 16, 2013
QuantAlea F# success story #1: computing greeks using Monte Carlo and PDEs in derivatives pricing library – simpler with#fsharp #cuda
— Don Syme (@dsyme) May 16, 2013
QuantAlea F# success story #2: scalable compute infrastructure for derivatives pricing, integrated with #sunguard.
— Don Syme (@dsyme) May 16, 2013
QuantAlea F# success story #3: dynamically compile F# to #cuda code, based on LLVM, with NSight GPU debugging,#fsharp #cuda
— Don Syme (@dsyme) May 16, 2013
QuantAlea have code generator for exposing F# code for use from #python. Sounds fantastically useful. #fsharp
— Don Syme (@dsyme) May 16, 2013
QuantAlea’s lessons from F# success stories: productivity improvement, less code, fewer bugs, better code quality, smaller teams, more agile
— Don Syme (@dsyme) May 16, 2013
QuantAlea say their project-cost-estimations are 2x better than their competitors because of the productivity they get from #fsharp
— Don Syme (@dsyme) May 16, 2013
Beer is definitely the drink of choice at the F#-for-financial-GPU-programming event. #fsharp #cuda #aleacubase #gpgpu
— Don Syme (@dsyme) May 16, 2013
Daniel explains latest GPUs in good detail. Now moving on to Alea.cuBase, which “turns F# into a first-class CUDA language” #fsharp #cuda
— Don Syme (@dsyme) May 16, 2013
Daniel explains how Alea.cuBase lets you do interactive, rapid, full-performance CUDA programming with F#,#fsharp #cuda
— Don Syme (@dsyme) May 16, 2013
Alea.cuBase uses F# quotations as composition blocks. No need to install NVIDIA nvcc compiler tools
— Don Syme (@dsyme) May 16, 2013
You can use Alea.cuBase F# GPU components from #python. Awesome! #fsharp #cuda
— Don Syme (@dsyme) May 16, 2013
Daniel describes compositional CUDA programming in Alea.cuBase with the “cuda workflows”. (ok, he used the “monad” word, don’t tell anyone)
— Don Syme (@dsyme) May 16, 2013
Alea.cuBase also supports CUDA’s “fast math” functions – not IEEE but faster. #fsharp #cuda
— Don Syme (@dsyme) May 16, 2013
4 steps to a #CUDA kernel with Alea.cuBase 1 describe your resources, 2. create device worker, 3. generate LLVM IR code. 4 Wrap into PModule
— Don Syme (@dsyme) May 16, 2013
Daniel keeps hinting he will show #fsharp GPU CUDA scripting integrated with Excel in the live coding session. #cuda #excel #fsharp
— Don Syme (@dsyme) May 16, 2013
Daniel now live coding Alea.cuBase #CUDA programming #fsharp
— Don Syme (@dsyme) May 16, 2013
With Alea.cuBase you see the exact PTX CUDA code in the output window of F# interactive. Interactive GPU programming, awesome
— Don Syme (@dsyme) May 16, 2013
Alea.cuBase CUDA workflows can be run in different ways, with and without diagnostics.
— Don Syme (@dsyme) May 16, 2013
OMG, Alea.cuBase F# GPU cuda scripting, integrated with Excel as UDFs, solving heat equation PDEs on gpu, with 3D plotting in excel #fsharp
— Don Syme (@dsyme) May 16, 2013
Free personal licenses for Alea.cuBase at quantalea.net/news/22 #fsharp #gpu #cuda
— Don Syme (@dsyme) May 16, 2013
Great summary. Was this session recorded on video and do you have a link?
No, as I know. https://twitter.com/dsyme/status/335097711315533825