Re: Oberon and Wirthian languages

[Simon Clubley <clubley@remove_me.eisner.decus.org-Earth.UFP>]

On 2014-04-22, Randy Brukardt <randy@rrsoftware.com> wrote:
> "Simon Clubley" <clubley@remove_me.eisner.decus.org-Earth.UFP> wrote in 
> message news:lj6kfg$jhs$1@dont-email.me...
> ...
>> Ok, now that's interesting. I would not have assumed the compiler would
>> do the read multiple times in this specific case.
>>
>> I would have assumed the Volatile attribute would be satisfied in this
>> specific case by doing the read _once_ at the start of the generated
>> code and only generating code to modify the A, B and C bitfields
>> before writing everything back in one step by using a single store
>> opcode.
>
> C.6(20):
>
> The external effect of a program (see 1.1.3) is defined to include each read 
> and update of a volatile or atomic object. The implementation shall not 
> generate any memory reads or updates of atomic or volatile objects other 
> than those specified by the program.
>
> I see two reads of Rec in the original source code. The above rule means 
> that an Ada compiler must NOT combine those reads for convinience; it must 
> do EXACTLY what's in the source code. If you only want one read, you have to 
> restructure so there is only one explicit read.
>
> There was a long discussion about this some years ago, and the conclusion 
> was that combining operations in particular is not allowed for volatile 
> objects (and all atomic objects are also volatile).
>

Thanks for the feedback.

From a technical viewpoint, I can see exactly what you are saying and
can understand and agree with your reasoning for the general case.

However, from a practical viewpoint the example statement was just used to
update a subset of bitfields in the register at the same time so one read
would be sufficient. I am also very aware of device registers which either
return different results on different reads or treat a read as an
acknowledge.

In the C language replacement discussion I proposed an "atomic using" syntax
which would replace C style bitmasks for updating device registers and would
guarantee a single read register/apply modifications as required/write
register back in one opcode sequence.

I wonder if we need something like that for Ada proper so we don't have
to read the register into a temporary variable first in order to guarantee
a single register read and a single register write while updating a subset
of bitfields in the register.

Having to use a temporary variable to guarantee a update of, say, a couple
of bitfields (out of a possible large number of bitfields) at the same time
just feels as if it isn't elegant.

Simon.

-- 
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Microsoft: Bringing you 1980s technology to a 21st century world