OK, i have it, apologies..

i was benching the complex path with an extra dereference and with such a 
tight loop it makes so much more difference than i thought was possible.
now all makes sense. thanks.
 
On Sunday 10 November 2024 at 19:25:39 UTC Axel Wagner wrote:

> On Mon, 11 Nov 2024 at 03:24, 'simon place' via golang-nuts <
> golan...@googlegroups.com> wrote:
>
>> yes, see the bench and code in the OP.
>>
>
> I don't see any benchmarking code in this thread.
> From the godbolt output <https://go.godbolt.org/z/jbvbrPb6j>, it looks to 
> me as if the compiler generates a single move, which seems as efficient as 
> it can be.
>  
>
>>
>> but since your way is slightly different (just possibly making a 
>> difference to the compiler? unpredictable eliding seems a poor thing to 
>> rely on.) i benched that, exactly the same, that is, about 5x slower than 
>> just operating on a single float. i'd look at the m/c, except at 500% 
>> slower i'm happy it can't be, maybe one day? 
>>
>> i'm going to search for a lib. (if/when speed becomes a problem.) 
>>
>> also unfortunate, complex not being a lib, that wouldn't just be a drop 
>> in.
>>
>>  
>>
>> -- 
>>
> You received this message because you are subscribed to the Google Groups 
>> "golang-nuts" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to golang-nuts...@googlegroups.com.
>>
> To view this discussion visit 
>> https://groups.google.com/d/msgid/golang-nuts/2b65ea5f-3fbd-440d-a6e0-400d61d6a3efn%40googlegroups.com
>>  
>> <https://groups.google.com/d/msgid/golang-nuts/2b65ea5f-3fbd-440d-a6e0-400d61d6a3efn%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
To view this discussion visit 
https://groups.google.com/d/msgid/golang-nuts/f6aea2d3-6fed-40b5-8ce4-92637b4145e0n%40googlegroups.com.

Reply via email to