>From: "Bresticker, Shalom" <shalom.bresticker@intel.com> >Karen specifically said that she preferred not forcing side-effects to >occur. She did not mention the other side, but I believe there might be >cases where forcing short-circuiting could hurt the quality of synthesis >results. However, it seems likely that forcing side-effects would >usually be more costly than forcing them not to occur. If I understood her explanations correctly, she mentioned both sides. In her second example, >>If (a || b || .... || f(x)) >> >>in the event that a is a critical path signal, then there are a lot of >>ors that the signal will have to go through to prove that the side >>effects of f(x) can happen, putting that side effect now on the critical >>path. I believe she was saying that she wanted to be able to execute f(x) regardless of the previous conditions. The motivation was to be able to avoid waiting for the result of the OR of the previous conditions before starting the side effect. This would be forcing side-effects to occur, as opposed to short-circuiting them. >I believe Karen was thinking of cases where the user would not care if >the side-effect did not always occur in the synthesized gate-level >netlist. I am not clear how the tool would determine that the user would not care. If the tool could determine that the side-effects make no difference to the behavior of the circuit, then it would be free to do them regardless of what the LRM says. Otherwise, I don't see how the tool is supposed to know what parts of the circuit behavior the user cares about. Steven Sharp sharp@cadence.comReceived on Thu Aug 17 17:12:54 2006
This archive was generated by hypermail 2.1.8 : Thu Aug 17 2006 - 17:13:15 PDT