GC_DEBUG & timings

Date view Thread view Subject view Author view

From: Godmar Back (gback@cs.utah.edu)
Date: Tue Nov 10 1998 - 14:16:01 EST


Note: these questions relate to the merging of Jason Baker's new block
allocation scheme which reduces cache conflict, VM behavior and speeds
up pointer identification.

About GC_DEBUG:

    Currently, it adds a "GC_MAGIC" to each gc_block header and checks its
    consistency. I would like to get rid of it in favor of simply DEBUG.

About timing:

    Jason added an getrusage-based timing facility that tells you the
    time spent in scanning/marking, but the routines can actually be
    used for all kinds of timings. So, I propose another configure
    options, --enable-timing, to enable these internal timers.
    Right now, he has it linked with the debugging option, which is
    clearly suboptimal since we don't usually want to pay the price of the
    debug statements when doing timings, even when these timings do not
    have the accuracy required to do benchmarking etc.

        - Godmar


Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Sat Sep 23 2000 - 19:57:02 EDT