baum@Apple.COM (Allen J. Baum) (07/26/88)
[] >In article <1988Jul22.162623.5353@utzoo.uucp> henry@utzoo.uucp (Henry Spencer) writes: >Let us not forget that IBM's Stretch machine (the 7030, late 50s) had a bit >which told the floating-point processor whether to round correctly or >randomly. I don't think that the "noisy" floating point mode rounded randomly. I think it rounded in the opposite direction that the normal rounding did. -- {decwrl,hplabs,ihnp4}!nsc!apple!baum (408)973-3385
root@cca.ucsf.edu (Computer Center) (07/26/88)
In article <14659@apple.Apple.COM>, baum@Apple.COM (Allen J. Baum) writes: > > I don't think that the "noisy" floating point mode rounded randomly. I think > it rounded in the opposite direction that the normal rounding did. > The "Noisy Mode" of floating point on the Stretch is described on p. 84 of the Reference Manual, 7030 Data Processing System, A22-6530-1, August 1960. Essentially it involves using 1-bit fill instead of the 0-bit fill usually occuring during normalization. In addition certain operand extensions when using single precision (48 bit mantissa) use 1-bits in extending to the intermediate (96 bit) operand format. Thos Sumner (thos@cca.ucsf.edu) BITNET: thos@ucsfcca (The I.G.) (...ucbvax!ucsfcgl!cca.ucsf!thos) OS|2 -- an Operating System for puppets. #include <disclaimer.std>