Shabtay’s take 04/28 - Issue resolved. We should consider inserting this clarification in SCEMI 2.0 use model which will require spec review.

Committee agrees

Issues surrounding macro semantics and compatibility with function semantics and guaranteed determinism for streaming

** See consolidated proposal from JohnS in IM208 dealing with IM's 208-211

The streaming is covered in the previous IM. Determinism is guaranteed in an alternating environment. Any optimizations performed by vendors must preserve the exact semantics and determinism of the unoptimized communications.

Shabtay> Does it mean that SCE-MI 2.0 promotes two modes, one that is deterministic and one that is not?

Per> No, it only provides a deterministic mode.

Compatibility between macros and function semantics will be evaluated when firm proposals for each are available. There was a general question if Cadence considers the previous proposal still to be the one on the table for the 2.0 specification, or if an update will be provided?

JohnS>

I think it has already been stated here in the IM but I would like to suggest that we reiterate the long standing requirement for SCE-MI 2:

- That it is deterministic no matter what.

- It should retain deterministic behavior whether using pipes or plain function calls

- It should retain determinism whether running in pure alternating mode or whether vendors have put in "under the hood" optimizations such as batching and concurrency.

I hope the committee still agrees to these requirements. Loosing determinism under any circumstances creates situations that are very difficult for users to debug and my feeling is that the standard should not allow for it.

This is what has and still does distinguish the SCE-MI 2 proposal from the SCE-MI I proposal which is prone to non-determinism if careful modeling practices are not used.

<JohnS