cpu-to-gpu-offloading good first issues (C)

Tests for GPU contention, out-of-memory and multi-threaded callers 20 days ago

help wanted

0

gpu-offload: transparent CPU-to-GPU offloading. Runs the BLAS/LAPACK calls of unmodified programs (NumPy, R, Octave, C, Fortran) on an NVIDIA GPU when a per-machine calibration says it is faster, with CPU fallback. LD_PRELOAD + cuBLAS/cuSOLVER.

C

#auto-tuning#blas#cublas#cuda#cusolver#drop-in#gpu#gpu-acceleration#gpu-offloading#heterogeneous-computing#hpc#lapack#ld-preload#linear-algebra#numpy#nvidia#openblas#performance#scientific-computing#wsl2

A second GPU backend (AMD ROCm, Intel oneAPI, or a portable one) 20 days ago

enhancement help wanted

jaifar530/cpu-to-gpu-offloading

0

gpu-offload: transparent CPU-to-GPU offloading. Runs the BLAS/LAPACK calls of unmodified programs (NumPy, R, Octave, C, Fortran) on an NVIDIA GPU when a per-machine calibration says it is faster, with CPU fallback. LD_PRELOAD + cuBLAS/cuSOLVER.

C

#auto-tuning#blas#cublas#cuda#cusolver#drop-in#gpu#gpu-acceleration#gpu-offloading#heterogeneous-computing#hpc#lapack#ld-preload#linear-algebra#numpy#nvidia#openblas#performance#scientific-computing#wsl2

Design: keep data on the GPU between calls (data residency) 20 days ago

enhancement help wanted

jaifar530/cpu-to-gpu-offloading

0

gpu-offload: transparent CPU-to-GPU offloading. Runs the BLAS/LAPACK calls of unmodified programs (NumPy, R, Octave, C, Fortran) on an NVIDIA GPU when a per-machine calibration says it is faster, with CPU fallback. LD_PRELOAD + cuBLAS/cuSOLVER.

C

#auto-tuning#blas#cublas#cuda#cusolver#drop-in#gpu#gpu-acceleration#gpu-offloading#heterogeneous-computing#hpc#lapack#ld-preload#linear-algebra#numpy#nvidia#openblas#performance#scientific-computing#wsl2

Offload FFTW transforms through cuFFT 20 days ago

enhancement help wanted

jaifar530/cpu-to-gpu-offloading

0

gpu-offload: transparent CPU-to-GPU offloading. Runs the BLAS/LAPACK calls of unmodified programs (NumPy, R, Octave, C, Fortran) on an NVIDIA GPU when a per-machine calibration says it is faster, with CPU fallback. LD_PRELOAD + cuBLAS/cuSOLVER.

C

#auto-tuning#blas#cublas#cuda#cusolver#drop-in#gpu#gpu-acceleration#gpu-offloading#heterogeneous-computing#hpc#lapack#ld-preload#linear-algebra#numpy#nvidia#openblas#performance#scientific-computing#wsl2

Offload syevr / heevr (used by R's eigen() and SciPy's eigh) 20 days ago

enhancement help wanted

jaifar530/cpu-to-gpu-offloading

0

gpu-offload: transparent CPU-to-GPU offloading. Runs the BLAS/LAPACK calls of unmodified programs (NumPy, R, Octave, C, Fortran) on an NVIDIA GPU when a per-machine calibration says it is faster, with CPU fallback. LD_PRELOAD + cuBLAS/cuSOLVER.

C

#auto-tuning#blas#cublas#cuda#cusolver#drop-in#gpu#gpu-acceleration#gpu-offloading#heterogeneous-computing#hpc#lapack#ld-preload#linear-algebra#numpy#nvidia#openblas#performance#scientific-computing#wsl2

Offload QR factorisation (geqrf / orgqr / ormqr) 20 days ago

enhancement help wanted

jaifar530/cpu-to-gpu-offloading

0

gpu-offload: transparent CPU-to-GPU offloading. Runs the BLAS/LAPACK calls of unmodified programs (NumPy, R, Octave, C, Fortran) on an NVIDIA GPU when a per-machine calibration says it is faster, with CPU fallback. LD_PRELOAD + cuBLAS/cuSOLVER.

C

#auto-tuning#blas#cublas#cuda#cusolver#drop-in#gpu#gpu-acceleration#gpu-offloading#heterogeneous-computing#hpc#lapack#ld-preload#linear-algebra#numpy#nvidia#openblas#performance#scientific-computing#wsl2

Complete the end-to-end benchmark on a quiet machine 20 days ago

good first issue help wanted

jaifar530/cpu-to-gpu-offloading

0

gpu-offload: transparent CPU-to-GPU offloading. Runs the BLAS/LAPACK calls of unmodified programs (NumPy, R, Octave, C, Fortran) on an NVIDIA GPU when a per-machine calibration says it is faster, with CPU fallback. LD_PRELOAD + cuBLAS/cuSOLVER.

C

#auto-tuning#blas#cublas#cuda#cusolver#drop-in#gpu#gpu-acceleration#gpu-offloading#heterogeneous-computing#hpc#lapack#ld-preload#linear-algebra#numpy#nvidia#openblas#performance#scientific-computing#wsl2

Share your machine's calibration table 20 days ago

good first issue help wanted machine report

jaifar530/cpu-to-gpu-offloading

0

gpu-offload: transparent CPU-to-GPU offloading. Runs the BLAS/LAPACK calls of unmodified programs (NumPy, R, Octave, C, Fortran) on an NVIDIA GPU when a per-machine calibration says it is faster, with CPU fallback. LD_PRELOAD + cuBLAS/cuSOLVER.

C

#auto-tuning#blas#cublas#cuda#cusolver#drop-in#gpu#gpu-acceleration#gpu-offloading#heterogeneous-computing#hpc#lapack#ld-preload#linear-algebra#numpy#nvidia#openblas#performance#scientific-computing#wsl2

jaifar530/cpu-to-gpu-offloading

Open Issues Need Help