Sylvain Le Gall <sylvain@le-gall.net> writes:

> On 29-10-2009, Goswin von Brederlow <goswin-v-b@web.de> wrote:
>> Xavier Leroy <Xavier.Leroy@inria.fr> writes:
>>> Goswin von Brederlow wrote:
>>
>> Here are some benchmark results:
>>
>> get an int out of a string:
>>                 C               Ocaml
>>   uint8  le     19.496          17.433
>>    int8  le     19.298          17.850
>>   uint16 le     19.427          25.046
>>    int16 le     19.383          27.664
>>   uint16 be     20.502          23.200
>>    int16 be     20.350          27.535
>>
>> get an int out of a Bigarray.Array1.t:
>> 		safe		unsafe
>>   uint8  le	55.194s		54.508s
>>   uint64 le     80.51s		81.46s
>>
>
> Can you provide us with the corresponding code and benchmark? 
>
> Maybe you can just commit this in libaio/test/bench.ml.
>
> Regards,
> Sylvain Le Gall

As Christophe guessed the problem was polymorphic functions. If I
specify a fixed Array1 type then the compiler uses the optimized
access functions. Makes unsafe Bigarray slightly faster than unsafe
string actually (must not optimize int_of_char/Char.unsafe_chr away)
and that independent of argument size (on set, on get allocating
int32/int64 costs time so they are slower).

So Bigarray is the fastest but getting different types out of a
Bigarray will be tricky. Unaligned even more so if not impossible.

I have to sleep on this. Maybe in my use case I can have all
structures int64 aligned and then split the int64 up in ocaml where
structures have smaller members. Would have been too much to have a
Bigarray with access functions for any type. Maybe some little wrapper
with Obj.Magic will do *hide*.

As for libaio it should be easy to make it create and use any Bigarray
type.

MfG
        Goswin