[sv-cc] Re: Next rev of DPI task proposal - imp. tasks are always 'context'

Subject: [sv-cc] Re: Next rev of DPI task proposal - imp. tasks are always 'context'
From: Andrzej Litwiniuk (Andrzej.Litwiniuk@synopsys.com)
Date: Wed Oct 01 2003 - 08:31:15 PDT

• Next message: Michael Rohleder: "[sv-cc] Minutes of the SV-CC conference call at 31-Sep-2003"
• Previous message: Swapnajit Mittra: "[sv-cc] Meeting reminder 10/1"
• In reply to: Warmke, Doug: "[sv-cc] Next rev of DPI task proposal"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

> Having the concept of DPI import tasks makes error-checking such
> situations much easier and more in harmony with the way native
> SV works.
>
> Please review the revised proposal, and we can then go over this
> in a weekly SV-CC meeting pretty soon.

I think that there is no need for specifying 'context' property for
imported tasks. All imported tasks should be deemed 'context'.

In the case of functions the property 'context' instructs a compiler
to instrument a call and to be side-effects aware in code optimizations;
non-context functions may use more efficient call protocol.

None of this applies for imported tasks.

The imported tasks are assumed to call SV functions or tasks, otherwise
a plain imported function could be used instead of an imported task.
Therefore all the instrumentation needed in order to enable calling SV func.
or task from C is needed anyway; similarly side effects must be assumed.
Hence I cannot see how the property 'context' or the lack of it could impact
the way a task enable is implemented for imported tasks.

Thanks,
Andrzej

• Next message: Michael Rohleder: "[sv-cc] Minutes of the SV-CC conference call at 31-Sep-2003"
• Previous message: Swapnajit Mittra: "[sv-cc] Meeting reminder 10/1"
• In reply to: Warmke, Doug: "[sv-cc] Next rev of DPI task proposal"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

This archive was generated by hypermail 2b28 : Wed Oct 01 2003 - 08:35:22 PDT