Re: [PATCH] cangen: use long long in time computation for -g to support >2.1s

[Konrad Anton <konrad.anton@awinia.de>]

Hello,

On 07.02.2013 11:41, Marc Kleine-Budde wrote:
> we're a bit oldschool and discuss the code on the
> linux-can@vger.kernel.org mailinglist. Feel free to subscribe.

Can you please put that information on the Gitorious page for the next 
casual committer?


>>   	ts.tv_sec = gap / 1000;
>> -	ts.tv_nsec = ((int)(gap * 1000000)) % 1000000000;
>> +	ts.tv_nsec = (long)(((long long)(gap * 1000000)) % 1000000000ll);
>
> What's going on here exactly?
>
> gap is a double - Will it be converted into an int before multiplication
> with 1000000? IIRC there is a difference between "gap * 1000000" and
> "gap * 1000000.0". In the latter the multiplication is done in floating
> point, because of the ".0".

gap is a double.
(gap * 1000000) will cause 1000000 to be converted to intermediate 
temp1 of type double,
temp1 will be truncated to long long temp2, losing its decimals,
temp2 will be moduloed by 1000000000ll in a long-long mod operation, 
giving temp3.
temp3 will be truncated to as many bits as long has, and stored in long 
tv_nsec.

I think it is safe.

--Konrad



-- 
------------------------------------------------------------------------
Konrad Anton
awinia gmbh, Software Consulting and Development
Max-Josef-Metzger-Str. 2, D-79111 Freiburg
Tel. +49 761 52 07 449 19
Fax. +49 761 52 07 449 10
EMail: konrad.anton@awinia.de
Handelsregister Nr.: HRB 706296, eingetragen beim Amtsgericht Freiburg
USt.-IdNr.: DE 275018159
Geschäftsführer: Marc Gossweiler, Bernd Wiedle
------------------------------------------------------------------------