[Info-vax] Why it is a good idea that OpenVMS isn't on x86-64 just yet

[abrsvc]

> I also think the 730 was the last machine to do H-floating in real hardware.  I remember seeing some crude benchmarks between the 730 and 750 for REAL*16 showing the 730 to be much faster.

Per the VAX Performance Summary dated 1983:

Note: The G and H instructions were not part of the FPA in the 780 or 750

Instruction timings in microseconds: (Without FPA)

            780     750      730      730
           no FPA  No FPA   no FPA    FPA
ADDH3     153.19   227.10   496.68   140.95
SUBH3       9.64    47.81    39.09    22.82


The SUBH3 instruction was the only one that was faster on the  730 without the FPA than the 750 without the FPA.  The instruction timings for the 780 and 750 are not in the table I have as these were not executed within the FPA.

Dan