Huawei’s Da Vinci AI architecture, introduced with the Kirin 810 SoC, combines a 3D Cube matrix‑multiply engine, vector and scalar units, and flexible scaling to deliver high‑performance, energy‑efficient AI compute across devices from low‑power IoT to high‑end cloud servers.
This article explains Python's @ operator for matrix multiplication, shows basic usage with NumPy, contrasts it with element‑wise *, demonstrates matrix‑vector multiplication, highlights common dimension‑mismatch errors, and provides a concise summary for efficient linear‑algebra calculations.
The article explains why standard GEMM libraries may fall short for special matrix shapes, introduces NVIDIA’s open‑source CUTLASS library, details its hierarchical tiling architecture, and walks through a complete device‑API example that customizes tile sizes and data layouts to achieve near‑hand‑written kernel performance on modern GPUs.
The article details MNN’s CPU backend dynamic quantization for Transformer‑type models, describing runtime int8 conversion, block‑wise matrix‑multiply optimizations using ARM SMMLA/SDOT and AVX‑512 VNNI, weight‑group and batch‑wise quantization techniques, and reports up to three‑fold speed‑ups on Snapdragon 8 Gen 3.
This article explains the CPU cache locality principle, shows how to restructure Go data access patterns—including data structures, field ordering, memory allocation, and false sharing avoidance—and demonstrates measurable performance gains with a matrix‑multiplication benchmark.
The article examines Modular's new Mojo language, its claim of massive performance gains over Python for AI workloads, presents benchmark code and results, discusses its origins, investment interest, and current beta status, concluding that while impressive, the 36,000× claim is overstated.
This article demonstrates how reorganizing Java matrix‑multiplication loops and understanding Java 2‑D array storage and CPU cache hierarchies can turn a naïve implementation into a version that runs up to a hundred times faster, backed by JMH benchmark results.
The article explains how ARMv86’s new SMMLA and BFMMLA GEMM instructions are integrated into MNN to accelerate INT8 and BF16 matrix multiplication, delivering up to 90% speedup over ARMv82’s SDOT and FP16‑FMLA kernels through optimized kernels, tiling, and compatibility handling.
DeepMind’s AlphaTensor, built on AlphaZero and reinforcement learning, automatically discovers novel, provably correct matrix multiplication algorithms that outperform classic methods like Strassen’s, demonstrating how modern AI can automate algorithm discovery and significantly accelerate computations across many fields.
Learn how to translate common mathematical symbols such as summation, product, factorial, conditional expressions, and matrix multiplication into clear Python code, revealing the underlying computations and helping data scientists and ML practitioners deepen their mathematical intuition through practical examples.
This article walks through three typical interview coding rounds—implementing a robust string‑to‑int parser, designing a high‑concurrency trace‑storage API, and building a matrix multiplication routine with performance and sparse‑matrix optimizations—providing code examples, corner‑case handling, and improvement ideas.
The article explores why mathematically equal expressions like 5×3 and 5+5+5 are not necessarily equivalent, explains the distinction between equality and equivalence, illustrates with real‑world and programming examples such as JavaScript’s == versus === and matrix multiplication rules, and discusses teaching implications.
This article describes how the English fluency app leverages deep neural network (DNN) models for real‑time speech scoring on smartphones, detailing offline inference challenges, BLAS‑based matrix‑vector optimizations, sparsity exploitation, cache‑friendly implementations, fixed‑point and NEON acceleration, as well as model compression techniques to improve accuracy and latency.
