emacs28
Joined 5 years ago
Karma
345
Submissions
84
Member since
Oct 18
Recent submissions
Comment · 1 year ago
Comment · 2 years ago
It produces identical/bit-equivalent results as conventional/naive matrix multiplication for integer/fixed-point data types
Comment · 2 years ago
For everyone discussing the reduced accuracy/numerical stability of the algorithms in floating-point, this is true. But note that the application of the algorithms in the work is explored for fixed-point MM/quantized integer NN inference, not floating-point MM/inference. Hence, there is no reduction in accuracy for that application of it compared to using conventional fixed-point MM.
Comment · 2 years ago
> you have to build hardware that matches the dimensions of the algorithm
Yes the benefits are realized in custom hardware designs as opposed to software, however, the hardware architectures work for multiplying matrices of arbitrary dimensions by splitting up larger matrices into smaller tiles, then summing up the tile products to form the final larger matrix products (i.e. GEMM)
Comment · 2 years ago
Thanks, good summary. Regarding numerical stability, the application is for fixed-point arithmetic, and therefore numerical stability is not an issue (the result is identical compared to using the conventional inner-product)
Comment · 2 years ago
IMHO, for fixed-point MM accelerators, there is no catch, I think it's an overlooked algorithm. It's based on an algorithm by Winograd who coincidentally also proposed another unrelated algorithm that later became very popular for CNN acceleration which would take some visibility away from this other algorithm by Winograd... But that is speculative
https://github.com/trevorpogue/algebraic-nnhw
The hardware architectures focused on in the paper are systolic array designs, an efficient type of hardware design for matrix multiplication (e.g., the Google TPU uses this), as opposed to more SIMD-like vector architectures like GPUs. It may be possible to extend the proposed KMM algorithm to other types of hardware architectures also in future work. Regarding floating point - this work is applicable for integer matrix multiplication acceleration, it may be possible to extend the concept to floating point data types in future work also.