On Tue, 07 Oct 2008, Phil Barnett wrote:
Hi Phil,
> Didn't think it would get done, but here is the same test after the
> recompile.
> I'll do it once more without the -gc3 in a minute. Don't know if it will get
> done before I leave for work.
> [EMAIL PROTECTED] tests]$ for i in 1 2 4 6 8;
> do ./speedtst --exclude=mem --scale --thread=$i; done
[...]
> 1 th. 2 th. factor
> ============================================================================
> [ T001: x := L_C ]____________________________________ 0.19 0.09 -> 2.05
> [ T002: x := L_N ]____________________________________ 0.15 0.08 -> 1.89
> [ T003: x := L_D ]____________________________________ 0.16 0.07 -> 2.14
> [ T004: x := S_C ]____________________________________ 0.21 0.26 -> 0.78
> [ T024: eval( bc := { || i % 16 } ) ]_________________ 0.80 0.56 -> 1.43
> [ T026: eval( bc := { |x| x % 16 }, i ) ]_____________ 0.77 0.55 -> 1.40
> [ T028: eval( bc := { |x| f1( x ) }, i ) ]____________ 0.92 0.64 -> 1.43
> ============================================================================
> [ TOTAL ]_________________________________________ 18.97 10.30 -> 1.84
> ============================================================================
[...]
> 1 th. 4 th. factor
> ============================================================================
> [ T001: x := L_C ]____________________________________ 0.38 0.11 -> 3.53
> [ T002: x := L_N ]____________________________________ 0.27 0.09 -> 2.88
> [ T003: x := L_D ]____________________________________ 0.28 0.09 -> 3.09
> [ T004: x := S_C ]____________________________________ 0.40 0.49 -> 0.81
> [ T017: x := o[8] ]___________________________________ 0.64 0.18 -> 3.55
> [ T018: round( i / 1000, 2 ) ]________________________ 1.28 0.33 -> 3.85
> [ T020: val( s ) ]____________________________________ 1.42 0.34 -> 4.13
> [ T021: val( a [ i % 16 + 1 ] ) ]_____________________ 2.07 0.54 -> 3.85
> [ T024: eval( bc := { || i % 16 } ) ]_________________ 1.58 1.22 -> 1.29
> [ T026: eval( bc := { |x| x % 16 }, i ) ]_____________ 1.50 0.99 -> 1.51
> [ T028: eval( bc := { |x| f1( x ) }, i ) ]____________ 1.83 1.11 -> 1.65
> [ T051: f2( a ) ]_____________________________________ 0.83 0.23 -> 3.60
> ============================================================================
> [ TOTAL ]_________________________________________ 37.65 13.05 -> 2.88
Thank you very much. Immediately better. Just like in min computer
accessing the same string item in [ T004: x := S_C ] simultaneously
by many threads is slow due to concurrent updating the same counter.
As you can see also eval() reduce over all performance but it's expected
because codeblocks body are compiled like for -gc0 even if you use -gc3
and HVM was compiled without -HB_GUI.
Recompiling harbour with -DHB_GUI should make above eval() tests results
much better and increase overall performance.
best regards,
Przemek
_______________________________________________
Harbour mailing list
Harbour@harbour-project.org
http://lists.harbour-project.org/mailman/listinfo/harbour
