Re: Slow memory access, Not understood when using integer@(mem address)

[Jussi Lahtinen <jussi.lahtinen@xxxxxxxxx>]

Now I need to ask if this is just testing for interest, or is this related
to any real life problem? Because I'm quite sure this will not be the real
problem is practice.


Jussi

On Fri, Sep 25, 2026 at 3:31 AM brian <brian@xxxxxxxxxxxxxxxx> wrote:

> Using the jit with integer@ makes it 1442 times faster. To not using
> fast. It takes 4 seconds. It is hard to believe that the checks need to be
> that significant  a slowdown.
>
>
> On September 24, 2026 1:12:47 PM PDT, Jussi Lahtinen <
> jussi.lahtinen@xxxxxxxxx> wrote:
>
>> Note that assigning address to structure is not checked at all.
>> Thus example "Dim bomb As Myint = CPointer(0)" is accepted by Gambas. The
>> crash happens only when accessing the address (example "Inc bomb").
>> So, perhaps Integer@() could cache its checks for it to be faster... but
>> this could be a lot of trouble for little benefit. Or it could be unsafe as
>> default like many memory access functions.
>>
>>
>> Jussi
>>
>>
>> On Thu, Sep 24, 2026 at 10:10 PM Jussi Lahtinen <jussi.lahtinen@xxxxxxxxx>
>> wrote:
>>
>>> This:  Integer@(mptr) = Integer@(mptr) + 1
>>> does a lot more than this: vptr.counter = vptr.counter + 1
>>>
>>> They are not quite equal.
>>> First one fetches Gambas integer from the memory address twice and then
>>> adds one. The other one does everything completely within Gambas. IE adds
>>> one to the variable in Gambas.
>>>
>>>
>>> Jussi
>>>
>>>
>>>
>>> On Thu, Sep 24, 2026 at 12:03 AM Brian G <brian@xxxxxxxxxxxxxxxx> wrote:
>>>
>>>> On 9/23/26 13:36, Brian G wrote:
>>>> > I have attached a small project that tests the access speed to memory
>>>> > using integer@(memory address), as used for Linux library interfaces.
>>>> > I am sorry but 44 times slower than access through a struct pointer
>>>> is
>>>> > just a bit much to accept!
>>>> > Using this structure and allocating a data space
>>>> > Public Struct myint
>>>> >     counter As Integer
>>>> > End Struct
>>>> > mptr=alloc(100)
>>>> > Here is an example using integer@(xx) compared to accessing thru a
>>>> > structure 1 million times
>>>> >
>>>> > Test Speed to Access script Variables using integer@ptr access
>>>> method,
>>>> > no locking
>>>> > using: Integer@(mptr) = Integer@(mptr) + 1
>>>> > Total time 4.40942561195698
>>>> >
>>>> > Test Speed to Access script Variables using struct direct access
>>>> > method, no locking
>>>> > dim vptr as myint = mptr
>>>> > using: Inc vptr.counter and tested with vptr.counter = vptr.counter +
>>>> > 1, same results
>>>> > Total time 0.098687710007653
>>>> >
>>>> > Struct access is 44.6810829482192 times faster than integer @ access
>>>> > to the same memory address
>>>> >
>>>> > Can this be fixed, it is pretty bad!
>>>> >
>>>> Update
>>>>
>>>> using fast unsafe the values are interesting so normal is about 1442
>>>> times slower than fast unsafe version
>>>>
>>>> Hello world
>>>> Test Speed to Access script Variables using integer@ptr access method,
>>>> no locking
>>>> Total time 0.003050090977922
>>>> Test Speed to Access script Variables using struct direct access
>>>> method,
>>>> no locking
>>>> Total time 0.041397490072995
>>>> Struct access is now 13.5725427118894 times slower..... confusing such
>>>> a
>>>> big difference
>>>>
>>>> --
>>>> ~~~~ Brian
>>>>
>>>> ~~~Brian
>