DAVEM TPF grant#2 report #103

I continued working on the context stack stuff, then having got fed up with
it, I decided (as a bit of relaxation and diversion) to look at whether
perl's arithmetic operators could be easily optimised. The results are now
smoking as smoke-me/davem/fast_arith2. This branch optimises +,-,* and
++/-- for the common cases, and makes the nbody benchmark (lots of
floating-point vector arithmetic) about a third faster.

2015/10/19
     0:36 process p5p mailbox
     5:24 #124156: death during unwinding causes crash

2015/10/20
     1:22 process p5p mailbox
     0:48 #124156: death during unwinding causes crash
     2:27 [perl #126309] 'x' operator on list causes segfault and confuses valgrind,  64-bit version
     0:09 [perl #117341] av_undef's POD is confusing

2015/10/21
     0:24 process p5p mailbox
     4:58 #124156: death during unwinding causes crash

2015/10/22
     7:00 make arithmetic faster

2015/10/24
     3:00 make arithmetic faster

2015/10/25
     3:00 make arithmetic faster

SUMMARY:
     11:10 #124156: death during unwinding causes crash
      0:09 [perl #117341] av_undef's POD is confusing
      2:27 [perl #126309] 'x' operator on list causes segfault and confuses valgrind, 64-bit version
     13:00 make arithmetic faster
      2:22 process p5p mailbox
    ------
     29:08 TOTAL (HH::MM)


-- 
Never do today what you can put off till tomorrow.

• DAVEM TPF grant#2 report #103 by Dave Mitchell