|
From: <sv...@va...> - 2005-12-25 03:33:22
|
Author: njn Date: 2005-12-25 03:33:12 +0000 (Sun, 25 Dec 2005) New Revision: 5434 Log: update Modified: trunk/docs/internals/performance.txt Modified: trunk/docs/internals/performance.txt =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D --- trunk/docs/internals/performance.txt 2005-12-25 02:53:02 UTC (rev 543= 3) +++ trunk/docs/internals/performance.txt 2005-12-25 03:33:12 UTC (rev 543= 4) @@ -23,6 +23,9 @@ - Julian changed findSb to slowly move superblocks to the front of the l= ist as they were accessed. This sped up perf/heap by 25--50%, and some bi= g programs (eg. ktuberling) programs by a couple of percent. +- Nick reduced the iteration count of the loop in swizzle() from 20 to 5= , + which gave almost identical results while saving 2% in perf/tinycc and= 10% + in perf/heap on a 3GHz Prescott P4. =20 COMPVBITS branch: - Nick converted to compress V bits, initial version saved 0--5% on most |