On Mon, May 24, 2010 at 8:41 PM, thierry wrote:

> I've download next branch (with modification suppressing rounding) and
> did "./acprep opt update" (no use of boost mt)
>

I downloaded today the next branch source code


> $ ledger -v
> Ledger next-0-gefcede3, the command-line accounting tool
>

$ ledger -v
Ledger 3.0.0-20100623, the command-line accounting tool


>
> Here are results:
>
> 1. My first use case (something like "ledger Assets:Invest --exchange=
> € -J")
> - 5m59.399s (previous ledger version, with rounding)
> - 3m31.564s (this ledger version, without rounding)
> - 0m1.314s (this ledger version, without rounding, without --exchange
> option)
>

- 0m7.001s (Ledger 3.0.0-20100623, with rounding)
- 0m6.289s (Ledger 3.0.0-20100623, without rounding)
- 0m1.291s (Ledger 3.0.0-20100623, without rounding, without --exchange
option)


>
> 2. My second use case (something like "ledger Assets Liabilities --
> exchange=€ -J")
> - 39m34.209s (previous ledger version, with rounding)
> - 27m1.114s (this ledger version, without rounding)
> - 0m5.279s (this ledger version, without rounding, without --exchange
> option)
>

- 1m0.702s (Ledger 3.0.0-20100623, with rounding)
- 0m54.790s (Ledger 3.0.0-20100623, without rounding)
- 0m5.558s (Ledger 3.0.0-20100623, without rounding, without --exchange
option)


>
> There is definitely an amelioration by dropping the exact rounding
> (41% for 1st, 32% for 2nd).
>

The --no-rounding is a 10% speed improvement with Ledger 3.0.0-20100623 on
my machine.
6.289/7.001=.898
54.790/60.702=.902


> But may I ask to dig further about --exchange impact?
>

I definitely acknowledge the performance boost, and I am very happy of it!


One question :

Your results are :

> using an "acprep opt" build on a 2.26GHz Core i7 (all output going to
/dev/null):
>
>    reg, no flags       5.00s
>    reg, -V            13.39s
>    reg, -X            12.91s

My results are, on a AMD Athlon(tm) Dual Core Processor 4850e, 2500MHz, 64
bits

$ time ledger reg        > /dev/null
real    0m12.727s
$ time ledger reg -X € > /dev/null
real    2m8.355s

The ratio are 12.7/5=2.54 and 128.355/12.91=9.94

When my ledger executable is running I can see that only one core is used. I
suspect you used a compilation option that enables ledger executable to use
several cores when running. Am I right ? What is this option ?


>
> Thanks
>

Thank you very much for the speed improvement.


> Thierry
>
>

Reply via email to