On 20/08/18 19:51, Luc Van Oostenryck wrote: [snip] >> So, are you saying that the 'mask' given above is M' (or omask >> in the code) and M is the function parameter 'mask'. ;-) >> >> So src2->value in the above instructions are all 'mask' values >> rather than (say) the shift count S? Hmm, so, Huh! :-D > > No, sorry. > In comments & explanations I use uppercase letters for constants: > 'M' for a constant mask, 'S' for a constant shift, 'N' for a constant > number of bits and 'C' for others constants. I also sometimes > use '$mask(N)' for ((1 << N) -1), and (-1 << S) as a shorthand for > '(-1 << S) truncated to the size of the instruction'. > Also M is (normaly) reserved for the outer mask, I use M', M'', ... > for other masks if present. > Lowercase letters 'a', 'b', 'x', 'y' are for other variables/pseudos. > > For the code here, the variable 'mask' n the code correspond to a > known/constant mask and is represented as 'M' in the formulae. > * if OP(_,_) is AND(x, M), obviously M is M is mask. > * if OP(_,_) is TRUNC(x, N) then mask is ((1 << N) -1), > that I also note as '$mask(N)', where N is the new size. > * if OP(_,_) is LSR(x, S) then mask is (-1 << S) (truncated to > the size of the instruction). > * if OP(_,_) is SHL(x, S) then mask is (-1 >> S). > > This is done because: > * TRUNC(x, N) is always equivalent to TRUNC(AND(x, ((1 << N) -1))) > * LSR(X, S) is always equivalent to LSR(AND(x, (-1 << S)), S) > * SHL(X, S) is always equivalent to SHL(AND(x, (-1 >> S)), S) > In other words: > * TRUNC(x, N) simplify the same as AND(x, ((1 << N) - 1)) > * LSR(x, S) simplify the same as AND(x, (-1 << N)) > * SHL(x, S) simplify the same as AND(x, (-1 >> N)) > or equivalently: > * to TRUNC(x, N) correspond an equivalent mask ((1 << N) -1) > * to LSR(x, S) correspond an equivalent mask (-1 << S) > * to SHL(x, S) correspond an equivalent mask (-1 >> S) > and so I use the notation 'OP(x, C)' for all these operations > and sometimes I abuse the notation and directly use 'OP(x, M)' > where M is the the mask corresponding to OP(x, C). OK, so just to be sure: the mask parameter to the function is the 'equivalent mask' or 'outer mask' or the 'mask associated with the outer operation', or M in the formulas. The M' mask is given by the 'omask' variable in the code, right? I suspect that I'm confusing myself because I am trying to see this as a graph coverage problem, where the 'outer' operation is a node 'higher-up' the 'tree'. ;-) So, the notation/formulas are confusing to me (partly because the M mask used is some function of OP and C: M = F(OP,C) and that is not stated anywhere). However, you understand them, so I will just have to squint a little harder! :-P ATB, Ramsay Jones
