topic FPU unit(s) in Archives Discussions
https://community.amd.com/t5/archives-discussions/fpu-unit-s/m-p/357682#M29058
<HTML><HEAD></HEAD><BODY><P>I was wondering why a separate FPU, ie X_87 was necessary.</P><P></P><P>As all numbers ARE Integers; you just take the decimal out and put it back in at the end of the integer op.</P><P></P><P>The numbers have to be padded with zeros left and right of the decimal so both int's and dec's are the same length.</P><P>After that you just take the decimal out and proceed with the integer function , and then stick the decimal in at the end.</P><P></P><P></P><P>A micro-code prep instruction could allow for floating point operations to be done in the same clock-ticks as integer math.</P><P>So it might take a few extra clock-ticks to pad the numbers for integer math and a reg to hold the ending decimal place.</P></BODY></HTML>Thu, 16 Feb 2012 01:04:30 GMTalbert_redditt2012-02-16T01:04:30ZFPU unit(s)
https://community.amd.com/t5/archives-discussions/fpu-unit-s/m-p/357682#M29058
<HTML><HEAD></HEAD><BODY><P>I was wondering why a separate FPU, ie X_87 was necessary.</P><P></P><P>As all numbers ARE Integers; you just take the decimal out and put it back in at the end of the integer op.</P><P></P><P>The numbers have to be padded with zeros left and right of the decimal so both int's and dec's are the same length.</P><P>After that you just take the decimal out and proceed with the integer function , and then stick the decimal in at the end.</P><P></P><P></P><P>A micro-code prep instruction could allow for floating point operations to be done in the same clock-ticks as integer math.</P><P>So it might take a few extra clock-ticks to pad the numbers for integer math and a reg to hold the ending decimal place.</P></BODY></HTML>Thu, 16 Feb 2012 01:04:30 GMThttps://community.amd.com/t5/archives-discussions/fpu-unit-s/m-p/357682#M29058albert_redditt2012-02-16T01:04:30Z