Hi.
I'm trying to implement a fast 128-bit barrel shifter in a Virtex4 device.
The obvious way is a 7-level cascade, each level shifting by another power of 2. The result is rather slow. Adding one pipeline stop I can achieve just above 125 MHz, but the result is available after 2 clocks instead of 1. For my application I'd prefer to get the full result within a single clock period.
I saw that it's possible to use the DSP48 slices as 18-bit barrel shifters, but I couldn't figure out how to combine them to form a
128-bit shifter - and less so in a single clock cycle (the DSP48 output is registered).Is there anything I can do to speed up the 7-level approach?
Regards, Marc