On 2023-08-07 19:05, Anders Magnusson wrote:
Den 2023-08-07 kl. 18:49, skrev Johnny Billquist:On 2023-08-07 13:34, Anders Magnusson wrote:Den 2023-08-07 kl. 13:31, skrev Maciej W. Rozycki:Overall I think we ought to support both, but it might be a bit tricky onThe 780 had H-float in hardware. Actually, it had all instructions in hardware (microcode).the toolchain side, as it's an ABI issue to sort out. Did any machineactually implement H_floating in hardware, or was it only ever emulated?The 780 had H-float as an option, I think. The 730 might always have had it, but otherwise the default answer for me for any of these questions is always the 86x0. It had everything in hardware (well, ok, microcode), except things like vector instructions, which hadn't been defined when the machine was sold. Fairly sure it was the machine with the most of all VAX stuff done in hardware.Hm, maybe, long time ago.
Paul commented on it as well, I think that he was right. The G- and H-float wasn't there originally, and was an option on the 780 and 750.
I remember there was a floating point accelerator (on both 780 and 750), but everything was in microcode if the accelerator was not present.
The FPA as such didn't add any capabilities as such, it just made FP operations run faster by throwing more hardware at it. But there is still additional microcode for the FPA as well. So it's microcode either way. Just with more gates added...
Everything everywhere, on every VAX, is always microcode. The machine is just too complex to implement without microcode.
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt%softjar.se@localhost || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol