The MOAIS teamproject Objective Endtoend parallel programming solutions

The MOAIS team-project Ø Objective: End-to-end parallel programming solutions for high-performance interactive computing with provable performances. optimization computational steering, VR embedded input output Option pricing on grid [LIG Moais / LJK Math. Fi] INRIA Grimage platform [LIG/LJK MOAIS, PERCEPTION, EVASION] Ø Performance is multi-objective Ø Adaptive to the platform Streaming on MPSo. Cs [ST]

Grids Evolution of parallel programming and MPSo. C Moais’softwares Cluster SMP multi-core GPU AWS / Kaapi-C Multi-objective Adaptive Performance • Applications are time-consuming but essential to validate scientific approach 2

Highlights Kaapi/Taktuk – 1 st prize Plugtest Nov. 2007 and Nov 2008 Nov. 2007: 1 st prize • Nqueens(23) in 2107 s with 3654 cores Nov 2008: 1 st Prize • Portolio Pricing (LIG/Moais + LJK/Math. Fi) – KAAPI winner of V GRID PLUGTEST 2008 Super Quant Monte Carlo Challenge • Parallelization of a code for pricing high dimensional European options (Monte Carlo method) • 1 st rank : 988 options priced in the 1 hour allocated computation time on a grid with 3609 cores (machines in Japan and France) • Kaapi/Taktuk: programming/deployment tool [for GCM standard] 3

Highlights Flow. VR / Grimage http: //grimage. inrialpes. fr/ – GRIMAGE platform: SIGGRAPH Aug. 2007 Emerging Tech. Demo – … fête de la science / Grand Palais Paris Nov. 2008 ~4000 visitors – Valorization: start-up 4

Highlight HPC Kaapi: SOFA http: //www. sofa-framework. org/ 5

Contacts • Moais: Jean-Louis. Roch@imag. fr • Kaapi: Thierry. Gautier@imag. fr • Flow. VR / Grimage: Bruno. Raffin@imag. fr