| commit | a7fba11297ecfedcffd443b837645a512761a0b2 | |
| author | Jan Moringen <jmoringe@techfak.uni-bielefeld.de> | |
| Fri, 2 Sep 2016 10:56:00 +0000 (2 12:56 +0200) | ||
| committer | Jan Moringen <jmoringe@techfak.uni-bielefeld.de> | |
| Tue, 6 Sep 2016 12:43:43 +0000 (6 14:43 +0200) | ||
| tree | 965bb56862fe5512c03eb434d2e834ad981bd085 | treesnapshot (tar.gz zip) |
| parent | 66903d73be8d11aa9cd1a7c13995e9470dfe32a3 | commitdiff |
More compact SC-OFFSET coding
The new layout is
ssssoooooooooooooooooooooss (still 27 bits)
where "s" bits encode the SC number and "o" bits encode the offset. This
layout allows SC-OFFSETs for the four most frequently used SCs to be
encoded in 7 bits (as long as the offset takes at most 5 bits to
encode), resulting in a single byte after variable length integer
encoding.
See 2c720b0942f88c973294c67c84193f3615dd024b for making the four most
frequently encoded SC numbers into two bits.
The new layout is
ssssoooooooooooooooooooooss (still 27 bits)
where "s" bits encode the SC number and "o" bits encode the offset. This
layout allows SC-OFFSETs for the four most frequently used SCs to be
encoded in 7 bits (as long as the offset takes at most 5 bits to
encode), resulting in a single byte after variable length integer
encoding.
See 2c720b0942f88c973294c67c84193f3615dd024b for making the four most
frequently encoded SC numbers into two bits.
| src/code/sc-offset.lisp | diffblobblamehistory |