question about 32-bit interger arithmetic vs. 32-bit float arithmetic

  • Thread starter Thread starter zhihang.wang
  • Start date Start date
Z

zhihang.wang

I have a question.
Is the implemention of 32-bit integer arithmetic more complexity than
32-bit float arithmetic?
And which is more simple?
Perhaps 32-bit integer arithmetic is simple, but how much?
 
I have a question.
Is the implemention of 32-bit integer arithmetic more complexity than
32-bit float arithmetic?
And which is more simple?
Perhaps 32-bit integer arithmetic is simple, but how much?
Click to expand...

Integer is of course simpler.

Michael
 
I have a question.
Is the implemention of 32-bit integer arithmetic more complexity than
32-bit float arithmetic?
And which is more simple?
Perhaps 32-bit integer arithmetic is simple, but how much?
Click to expand...

Floating point has a mantissa and an exponent.
An FP multiply means integer multiplying the two mantissa parts,
and integer adding the two exponent parts. Normalization is needed
to fix up the decimal point location and make a final correction
to the exponent part. So yes, FP is more complex. You can see
someone's attempt to build an FPU here, which hints at the
operations required.

http://www.opencores.org/cvsweb.shtml/fpu/FPU.pdf (page 4)
http://www.opencores.org/cvsweb.shtml/fpu/verilog/

Paul
 
Back
Top