Wavelet Compression Using FMA Dot Products for Faster HPC I/O
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data compression methods in High Performance Computer (HPC) systems are not sufficiently fast for processing large volumes of data, leading to increased computing costs and reduced attractiveness during I/O processing.
Innovation Solution
A method utilizing vector computing units with fused multiply-add (FMA) instructions to perform dot products during wavelet transforms, optimizing data compression by storing and processing data in cache memories to minimize reliance on central memory access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If existing compression routines are used to compress data during I/O processing, then compression ratio is improved, but processing speed deteriorates
Solution Approach 1:
The patent changes the computational parameters by using FMA instructions instead of traditional multiplication and addition separate operations. This parameter change in the computational approach allows achieving both high compression ratio and high processing speed by efficiently computing wavelet transforms through fused operations that reduce instruction overhead and improve computational throughput.
Solution Approach 2:
The patent substitutes traditional scalar processing mechanics with vectorized FMA instruction mechanics. By replacing the conventional approach of separate multiply and add operations with fused multiply-add vector instructions, the system achieves faster processing while maintaining compression effectiveness, thus resolving the speed-compression ratio tradeoff.
2Quantity of substance
If wavelet transform is applied to compress large volumes of data, then compression effectiveness is improved, but computing cost increases
Solution Approach 1:
The patent changes the computational parameters by utilizing FMA instructions that perform multiply and add operations in a single fused instruction. This parameter change reduces the total number of operations required for wavelet transform, thereby maintaining compression effectiveness while significantly reducing computing cost and energy consumption for large data volumes.
Solution Approach 2:
The patent ensures continuous useful action by using vectorized FMA instructions that process multiple data elements simultaneously in a continuous stream. This eliminates idle cycles and keeps the computing units fully utilized, maintaining high compression effectiveness while reducing overall computing cost through efficient continuous processing.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
The present invention concerns a method of data compression performed by at least one core communicating with a central memory. The input data presents a two-dimensional input array formed by a plurality data items stored contiguously in the central memory according to a contiguous direction. The method comprises a step (110) of wavelet transform comprising the following substeps: A) forming from the input array at least one tile comprising a plurality of consecutive data block columns, each data block column being formed by a plurality of lines of consecutive data items according to the contiguous direction, the length of each line being a multiple of the cache line length; and C) for each data block column computing dot products between a filter vector and each group of N lines using fused multiply-add instructions for the core.