< BACK TO NEWS
importantSYS.SOURCE: Lei Mao's Log Book2026-08-13T14:13:41Z

CUDA Shared Memory Swizzling Techniques and Applications

The article explains CUDA shared memory swizzling as a technique to avoid bank conflicts without padding, using a matrix transpose example. It provides a mathematical proof of the swizzling formula's one-to-one mapping properties.

Comments

Read original article

*** END OF TRANSMISSION ***