Survey of Cache Compression Outline Background Motivation Block

Survey of Cache Compression

Outline Background & Motivation Block based cache compression FPC ZCA BDI SC 2 Hy. Comp Stream based cache compression MORC

Background Cache is important… Conflicts: Larger LLC Limited LLC off-chip access more area, more latency, more energy Compression: Trade off latency for less off-chip access

Frequent Pattern Compression An example: 1 1 2 3 5 8 13 21 … 8: 0000 0000 1000 Pattern: 4 -bits sign-extended

Frequent Pattern Compression Patterns :

Zero-content Augmentation Common lines of code: Several “blank” blocks

Base Delta Immediate A simple example: 0 x 8048004 0 x 8048008 0 x 80480 c 0 0 x 8048004 +, 0 x 0 +, 0 x 8 +, 0 xc 0 -, 0 x 4

Base Delta Immediate Multiple Bases:

Base Delta Immediate Organization:

Base Delta Immediate Decompression

Statistical Cache Compression Huffman encoding “Heap” for sampling The most mathematical method so far, in my opinion

Statistical Cache Compression

Hycomp FP-H compression Floating-point number specified compression method, based on SC 2

Hycomp FP-H paralleled decompression

Hycomp Hybrid compression

MORC Log-based cache

MORC LMT

MORC

MORC Throughput oriented Manycore-Oriented-Compressed-Cache When cores accumulate, off-chip bandwidth limits performance For throughput oriented works, reducing off-chip access is more important than reducing latency. Less off-chip access saves energy