Sean,

Many factors affect performance at a given clock speed,
including the hardware optimizations of the processor, the
compiler's ability to optimize, and the parameters fed to
LAME.  On an Athlon 800, a specific example might show a
throughput of 8 times real time.  However, I doubt any
current FPGA microprocessor can compare to a superscaler
Athlon with three fully pipelined floating point units.

The MIPS benchmark offers little information, as a compiled
application may heavily emphasize specific instructions or
patterns mostly ignored by the benchmark.  Additionally,
the instructions used and their throughput depend upon the
target architecture and the compiler.

An implementation of LAME in a 17 MHz FPGA microprocessor
will benefit greatly from custom DSP extensions tuned for
LAME.  However, with the associated expense of a large FPGA
or FPGA array, simply porting LAME to a DSP processor may
cost fewer resources and effort.  Either way, developing 
processor specific code for critical FFT, MDCT, or filter
functions in LAME will be required for real-time operation
on certain chips or processors.

Creating a more extensively rewritten version of LAME may
further reduce the minimum chip requirements.  For example,
a fixed point version of LAME may require fewer gates or
transistors than a floating point version.

Kind regards,

- John


sean jaye wrote:
> 
> I need any one to answer me Urgently  I want to know
> the MIPS  of lame ,I want to use LAME  with a
> microprocessor  on an FPGA  but of_course  you know on
> FPGA  I get low freq about  17 MHZ Is that sufficient
> to lame to  work correctly
> So I need to know its max & min required freq.
> 
> thanks for your help
>                                         Sean  Jaye
_______________________________________________
mp3encoder mailing list
[EMAIL PROTECTED]
http://minnie.tuhs.org/mailman/listinfo/mp3encoder

Reply via email to