Re: Notes... The case for an open client
jbass-/[email protected] Mon, 16 Aug 2004 20:00:56 -0600
| Newsgroups | gmane.comp.distributed-net.hardware |
|---|---|
| Message-ID | <[email protected]> |
"Dan Oetting" <[email protected]> > The rotates unfortunately require many more LUTS and have at least a 32 > clock latency. For Xilinx Virtex devices, a 3 bit serial adder requires 2 LUT's, and the associated carry logic and registers. With care (and some luck) you can build one of the two adders from the carry logic not used by shift registers, but that is dependent on the FPGA. The ROTL3 function also requires 2 LUTs (dual 16 bit shift registers) and a mux which is part of the support logic in a CLB. There is a 29 clock latency, (32-3). The ROTL is a bit more complex, and on Virtex Pro devices requires 2 LUT's (dual 16 bit shifters) plus 5 registers and two muxes, which generally can be taken from other support logic on the S box chain, and an inverter from the BY input from some LUT. Virtex Pro LUT's have a shift-out port, that prior Xilinx FPGAs do not have, which saves considerable logic for this function. On other FPGA's, it requires either more control logic and mux'es, or an additional shift register LUT. In just about all cases, there is at least (32-n) clocks additional latency, and some easy implementations will always have 32 clocks latency. John _______________________________________________ Hardware mailing list Hardware-Ra3b/QYEcJ3d140v2zMXi0fjHoOT/h/[email protected] http://lists.distributed.net/mailman/listinfo/hardware