Feature overview

  • CPUs (Intel, AMD, ARM)
  • GPUs (NVidia, AMD)
  • Multi-node architectures via MPI
  • Lazy data streaming from/to GPUs
  • Lossy-compression of data
  • Serialization of data to disk
  • Optimized MPI communications on GPUs
  • Adaptive (expanding/contracting) box
  • Comprehensive performance autotuning

Features comparison

Devito (Open-source)

DevitoPRO

Platform support
Cloud-readiness
CPUs (Intel, AMD, ARM)
AMD GPUs - OpenMP
NVidia GPUs - OpenMP, OpenACC
AMD GPUs - HIP (beta)
NVidia GPUs - CUDA (beta)
Multi-node parallel via MPI
Optimized MPI communications on GPUs
Lazy data streaming from/to GPUs
Algorithms
Domain-specific language for finite-differences
Unstructured operations
Custom operators (for, e.g., BCs)
Sub-domains, staggered-grids, sub-sampling
Extensive examples and documentation
Checkpointing
Reduced-precision
Floating-point (lossy) compression
Serialization of data to disk
Calls to external C/C++ functions
Adaptive (expanding/contracting) box
Comprehensive performance autotuning
Solver examples
Isotropic (visco)acoustic + adjoints
Isotropic (visco)elastic
TTI acoustic + adjoints
VTI viscoacoustic + adjoints
Support
Projects
Training
Consultancy
Support/maintenance