[Cfp-interest 2766] Re: CFP review of NB comments on CD2 - 20230602

[Jim Thomas]

> On Jun 5, 2023, at 7:05 AM, Damian McGuckin <damianm at esi.com.au> wrote:
> 
> On Sun, 4 Jun 2023, Fred J. Tydeman wrote:
> 
>> On Sun, 4 Jun 2023 11:58:44 -0700 Jim Thomas wrote:
>>> 
>>> Each thread has its own floating-point environment. The floating-point environment has thread storage duration. The floating-point environment for a thread is initialized from the parent thread's floating-point environment when the thread is created which is by a function call.
>>> 
>> 
>> I am thinking about the values of variables in shared memory between the threads. I assume that if exceptions are delayed, then the value from those exceptions are also delayed (so storage is not updated, so other threads get the wrong value).

The state of the exception flags is local to the thread, not in shared memory. Am I missing something?

>> 
>> That only matters if threads can sync without a function call.
> 
> To answer my email correctly, you have both (quite correctly) used lots of words to explain the scenario.
> 
> My question was, and I did explain myself well enough, was probably
> 
> 	Is the simple CD2 change suggested for the C23 document of
> 
> 		between function calls -> between sequence points
> 
> 	adequate for US-200?
> 
> Is the term "sequence points" adequate or precise enough to cover all that is needed to be explained here.  Maybe discuss this on Wednesday.

We don’t want to say that the floating-point environment must be precise at sequence points. That would disallow optimizations like code motion and common subexpression elimination that can be safely done between function calls.

- Jim Thomas

> 
> Thanks - Damian
> 
> Pacific Engineering Systems International ..... 20D Grose St, Glebe NSW 2037
> Ph:+61-2-8571-0847 .. Fx:+61-2-9692-9623 | unsolicited email not wanted here
> Views & opinions here are mine and not those of any past or present employer