Electronic Design Automation (EDA)

EDA involves a diverse set of software algorithms and applications that are required for the design of complex next generation semiconductor and electronics products. The increase in VLSI design complexity poses a significant challenge to EDA; application performance is not scaling effectively since microprocessor performance gains have been hampered due to increases in power and manufacturability issues, which accompany scaling. Digital systems are typically validated by distributing logic simulation tasks among huge compute farms for weeks at a time. Yet, the performance of simulation often falls behind, leading to incomplete verification and missed functional bugs. It is indeed no surprise that the semiconductor industry is always seeking for faster simulation solutions. Recent trends in HPC are increasingly exploiting many-core GPUs to a competitive advantage through the use of such GPUs as a massively-parallel CPU co-processor to achieve speed up of computationally intensive EDA simulations including Verilog simulation, Signal Integrity & Electromagnetics, Computational Lithography, SPICE circuit simulation and more.
Verilog Simulation on GPUs with RocketSim [learn more] (Source: Tomer Ben-David, Rocketick, Israel)
GPU accelerated full wave EM simulation to analyze crosstalk on the opposite side of the package (Source: Martin Timm, CST, Germany)

ISV Applications Using CUDA

* Expected Speed Up vs a quad-core x64 CPU based system. Speed-ups as per NVIDIA in house testing or application provider   documentation.

Other Relevant Software Using CUDA

• Acceleware FDTD Solvers
• FMSlib: GPU Parallel Out-of-core Matrix Algebra from Multipath
• Acceleware Electromagnetic Solutions
• Accelerating simulations of light scattering based on FDTD method with general purpose GPUs

Technical Reports on EDA on CUDA

• High Performance Gate-level Simulation with GP-GPU Computing
• Acceleration of Functional Validation using GPGPU
• Parallel Equivalence Checking with GPUs
• Towards EDA Computing on GPUs
• Hardware Acceleration of EDA Algorithms – Custom ICs, FPGAs, GPUs
• It's About Time (EM simulations in EDA PCB Design)

Solvers and Core Kernels for EDA on CUDA GPUs

• Acceleware Matrix Solvers
• Dense Linear Algebra library (MAGMA)
• GPU accelerated linear algebra library (CULA)
• Sparse Matrix Linear Solvers : Iterative solvers
• SpMV from NVIDIA: Code
• Paper 1
• Paper 2
• Iterative CUDA
• Sparse Matrix Linear Solvers : Direct solvers
• PARDISO with CUDA

See Also

• AccelerEyes Jacket™ for MATLAB®
• MATLAB®
• Tesla/CUDA Success stories
• Other Tesla Vertical Solutions
• CUDA Software development tools & libraries
• Buy Tesla

MATLAB is a registered trademark of The MathWorks, Inc.
Jacket is a trademark of AccelerEyes