SoftFloat
|
SoftFloat
|
|
|
This provides a software implementation of functions performing floating point
|
This provides a software implementation of functions performing floating point
|
operations according to the IEEE 754 standard.
|
operations according to the IEEE 754 standard.
|
|
|
It is used in or1ksim due to problems arising when using the simulator host
|
It is used in or1ksim due to problems arising when using the simulator host
|
machine's floating point unit to perform single precision floating point
|
machine's floating point unit to perform single precision floating point
|
calculations (or even double, in some cases.) Typically the host machine's FPU
|
calculations (or even double, in some cases.) Typically the host machine's FPU
|
will support extended (80-bit), or higher, precision, resulting in values which
|
will support extended (80-bit), or higher, precision, resulting in values which
|
can differ from those which are calculated on a true single or double precision
|
can differ from those which are calculated on a true single or double precision
|
floating point unit (very small differences when in rounding and generating
|
floating point unit (very small differences when in rounding and generating
|
status flags.)
|
status flags.)
|
|
|
Although using a software implementation is slow, it ensures an accurate
|
Although using a software implementation is slow, it ensures an accurate
|
simulation of a single or double precision floating point unit.
|
simulation of a single or double precision floating point unit.
|
|
|
This software was sourced from the author's (John Hauser) site:
|
This software was sourced from the author's (John Hauser) site:
|
|
|
http://www.jhauser.us/arithmetic/SoftFloat.html
|
http://www.jhauser.us/arithmetic/SoftFloat.html
|
|
|
|
Note that the code cannot be compiled with -O0 for debugging purpose, because
|
|
of the approach used to code inlining. Additional macros have been added to
|
|
ease this problem. For debugging purposes, Or1ksim should be configured adding
|
|
CFLAGS="-O0 -DINLINE=static -DNO_SOFTFLOAT_UNUSED" to the configure line.
|
|
|
Julius Baxter, August 1, 2010
|
Julius Baxter, August 1, 2010
|
|
Jeremy Bennett, August 4, 2010 (-O0 compilation)
|
|
|