|
![](static/image/common/ico_lz.png)
楼主 |
发表于 2012-7-24 10:10:33
|
显示全部楼层
p.nicholas 发表于 2012-7-19 11:23 ![](static/image/common/back.gif)
1.硬件乘法器
2.latency少并不代表跑得快,能跑多快看fmax
多谢!
PS:来自altera工程师的回复——
1. Actually the float add operation is harder to perform than float multiply opeation, and Altera uses dedicated multiplier circuitry for float multiply operation. So the latency of multiply operation is less.
2. Less latency because of less pipeline level, which means we use less register. Of course resource utilization is less. But you can find that the performance is poor too.
|
|