Yes. This was intentional. In the microACE model, the ACE component on the SA core is critical in that it configures the data structures needed by the microcode, patches symbols, handles exception packets, exports an API for user applications etc. Without the core component the functionality of the microACE is quite incomplete. While it is possible to write ind files to get the microcode running, it is relatively difficult to test/use all the features without the core component. It is also easy to change and misconfigure something using the ind files. For this reason, the ACE samples were extensively QA'ed on hardware only. We do ship the workbench projects running on hardware. All customers of the SDK were supposed to get associated reference hardware. So not shipping the ind files was not considered an issue. Uday -----Original Message----- From: Mirko Benz [mailto:benz@mephisto.inf.tu-dresden.de] Sent: Wednesday, April 10, 2002 2:30 AM To: ixp1200@CS.Princeton.EDU Subject: [ixp1200] SDK 2 ACE sample simulation Hi, The ACE samples supplied with SDK 2 do not contain any simulation initialization scripts (*.ind). Therefore, one can not use the workbench for simulation and evaluation of these samples. Is this intended? Thanks, Mirko University of Technology Dresden, Germany
Uday, Of course it is not sufficient to simulate alone, but: - Much of the functionality of the IDE is useless (e.g. for optimization). - How should one select a specific NPU (speed grade, memory ...) when one can not simply evaluate them in a simulation? - What about forthcoming chips? - What about a customer that simply wants to evaluate different NPU architectures for a specific project? - How should one use external simulation models? - What about a group of developers - one evaluation system for every developer is a bit expensive. - Most samples supplied require very little initialization from the SA part. What we did was to copy and modify old init scripts from SDK 1.x samples and modify them as appropriate. Then we developed our new code in the IDE and tested it afterwards in hardware (IXP12EB). Having to analyze every memory access and reinventing the wheel slows down. I hope this situation changes - where possible - for SDK 3. Regards, Mirko -----Ursprüngliche Nachricht----- Von: owner-ixp1200@CS.Princeton.EDU [mailto:owner-ixp1200@CS.Princeton.EDU] Im Auftrag von Naik, Uday Gesendet: Mittwoch, 10. April 2002 18:18 An: 'ixp1200@CS.Princeton.EDU' Betreff: RE: [ixp1200] SDK 2 ACE sample simulation Yes. This was intentional. In the microACE model, the ACE component on the SA core is critical in that it configures the data structures needed by the microcode, patches symbols, handles exception packets, exports an API for user applications etc. Without the core component the functionality of the microACE is quite incomplete. While it is possible to write ind files to get the microcode running, it is relatively difficult to test/use all the features without the core component. It is also easy to change and misconfigure something using the ind files. For this reason, the ACE samples were extensively QA'ed on hardware only. We do ship the workbench projects running on hardware. All customers of the SDK were supposed to get associated reference hardware. So not shipping the ind files was not considered an issue. Uday -----Original Message----- From: Mirko Benz [mailto:benz@mephisto.inf.tu-dresden.de] Sent: Wednesday, April 10, 2002 2:30 AM To: ixp1200@CS.Princeton.EDU Subject: [ixp1200] SDK 2 ACE sample simulation Hi, The ACE samples supplied with SDK 2 do not contain any simulation initialization scripts (*.ind). Therefore, one can not use the workbench for simulation and evaluation of these samples. Is this intended? Thanks, Mirko University of Technology Dresden, Germany
For this reason, the ACE samples were extensively QA'ed on hardware only.
Uday
Uday, Since the ACE samples were tested extensively on hardware, I was wondering if the documentation includes the performance numbers for SlowIngress/FastIngress code with maybe L3 forwarding. (or NAT or just plain bridging) Or if it is present in the documentation, can you point me to it? Since ACE includes source code, I can always pick up the Slow Ingress files and put it in my application and measure it. Not so easy, if my application is not written to the ACE framework, but _doable_. But just wondering if it is already done for some sample ACE configurations. Thanks. - Abhijeet
participants (3)
-
Abhijeet Joglekar -
Mirko Benz -
Naik, Uday