J.Wexler@edinburgh.ac.uk (06/07/90)
fpldnldbi. Ds nyne knw hw t prnnc ths gbbldygk? For those who don't recognise it, it means "floating point load non-local double indexed", it's part of the T800 instruction set, and it's not worth reading this posting any further. If, however, somebody at Inmos can explain what it does, then I would be pleased to hear. I should explain that fpldnldbi is lucidly and concisely described in the "Transputer Instruction Set: a compiler writer's guide" as published by Prentice Hall. In fact it's described twice, on pages 52 and 152. The two descriptions are - you've guessed it - contradictory. Experience shows that page 52 is correct in general when it says that fpldnldbi is equivalent to wsubdb followed by fpldnldb. Page 152 is clearly wrong in its description of the indexing (you would have to supply an index of 2n if you wanted to load arr[n]). What I cannot figure out, and what I am too lazy to discover by experiment, is which description is correct about the final state of the (non-FP) stack. Is there anybody out there who can give an authoritative ruling on this? John Wexler Edinburgh Parallel Computing Centre