blas

Python interface to GPU-powered libraries

[DEPRECATED] Moved to ROCm/rocm-libraries repo

Information on optimizing python libraries specifically for oobabooga to take advantage of Apple Silicon and Accelerate Framework.

Linnea is an experimental tool for the automatic generation of optimized code for linear algebra problems.

A lightweight triton-based General Matrix Multiplication (GEMM) library.

Fast lattice reduction using segmentation, multithreading, Seysen reduction and BLAS

Experimental Linear Algebra Performance Studies

Fast cosine similarity for Python

Concurrent NumPy in Python

📱 🐍 Run spaCy and its C-extension stack (numpy, blis, thinc) on Android.

NumPy & SciPy for GPU as well as Ascend NPU (80% math API supported, alpha)

Auxiliary tools: distributed SVD, QR, sample mesh, etc

Compute the sum of the absolute values of the real and imaginary components of a single-precision complex floating-point number.

Error handler for BLAS/LAPACK routines.

Compute the sum of the absolute values of the real and imaginary components of a double-precision complex floating-point number.

Zig-native numerical runtime with full BLAS compatibility, typed vector/matrix views, and optimized GEMM kernels.

📱 🐍 run spaCy and its C-extension stack (numpy, blis, thinc) on iOS.

Compute the sum of a one-dimensional double-precision floating-point ndarray, ignoring NaN values.

Compute the sum of all elements in a one-dimensional single-precision floating-point ndarray.

Strict-FP32 CUDA SGEMM with shape-dispatched SM80 cp.async kernels and reproducible A100 evidence