National Labs & Public Research
Debugging and Profiling for the World’s Largest Supercomputers
Linaro Forge helps national laboratories and research computing centres find bugs and performance problems in parallel codes, from one node to hundreds of thousands of processes.
Hundreds of Thousands
of processes in one Linaro DDT session, limited only by your licence.
Parallel Tools for Diverse Research Workloads
Research centres support many users running Fortran, C and C++ with MPI, OpenMP and GPU offload. Linaro Forge is used at US Department of Energy national laboratories, European national and EuroHPC centres, and university research computing services, including NERSC, Los Alamos, TACC, Jülich, RIKEN and EPCC.
Find Deadlocks and Memory Errors at Scale
See every process’s stack at once to find the ranks that are stuck, spot deadlock loops in MPI message queues, and catch out-of-bounds writes, leaks and double frees with memory debugging.
NVIDIA, AMD and Intel GPUs on x86 and Arm
Debug CUDA, HIP, SYCL and OpenMP offload code on NVIDIA, AMD and Intel GPUs. Profile GPU code on NVIDIA and AMD. GPU support is a licence option.
Profile Production Codes
Linaro MAP shows which lines take the time and why: compute, MPI, I/O or memory. Overhead is usually under 5%, and a program built with -g rarely needs recompiling.
Site-Wide Floating Licences
Serve floating licences to every user on the cluster from Linaro Forge Licence Server. Works with Slurm, Flux and HPE Cray PALS.
Equip Your Users with Linaro Forge
Talk to us about site licences, training and support for your centre.