Questions for the DAC Pavilion Panel on multicore design
Steve Leibson - June 11, 2010
As I wrote yesterday, I’ll be chairing a Multicore panel at DAC in Anaheim on Monday in the Panel Pavilion at DAC. This morning, I worked on creating some questions for the panelists with my good friend Markus Levy, President of the Multicore Association, who was the original person scheduled to be the moderator.
Here are the questions so far:
1. What comes first with multicore, the hardware or the software?
2. What do you do about transitioning single-core software to multicore?
3. Do vendors provide enough software tools to support multicore programming’s special needs?
4. Is there something magic about adding special features enabled by multicore?
5. Is there a particular way to organize software teams in an optimal way for writing multicore software?
6. To Marvell: Tell me about your multicore server chips. Do they have all the capabilities of the Xeon? How do they differ? (Like OOO execution, 3-channel SDRAM, etc.)
7. How are you supporting or do you use industry-standard interfaces as opposed to Multicore Association Standard interfaces Multicore Communications API (an IPC protocol)?
8. Where will we be two years from now with respect to support for multicore software development? What about five years from now?
9. When does multicore development become the status quo?
10. What percentage of future chips will be homogeneous multicores versus heterogeneous multicores?
11. What are the ramifications of the current memory bottleneck, where all the on-chip processors share on SDRAM interface? Is that really a problem yet? When do you think it will become a problem?
If you have a question you’d like asked of the panel you might consider writing it down and asking it yourself at the panel on Monday or sending it to me via LinkedIn.
See you at DAC.
Share your thoughts.
Currently no items