def sums

bend-parallel@0.1.0.0 · sort.bend:132

import Base

132# Pass 2b's forks: d levels down to groups of 2^w columns under group j,
133# with two handles each to the table, the counts and the offsets. Group j
134# reads (and when wr writes) only its columns of the table and writes
135# only its own entries of counts and offs (j * 2^w to j * 2^w + 2^w - 1).
136def sums(+d: Nat, +wr: Bool, +j: U32, +w: Nat, +k: U32, +o: U32, +step: U32, +nb: Nat, ft: Array<U32> & Array<U32>, fc: Array<U32> & Array<U32>, fo: Array<U32> & Array<U32>) -> Three:
137  match d:
138    case 0n:
139      (t, t2) = ft
140      (cs, cs2) = fc
141      (os, os2) = fo
142      three.rejoin(t2, cs2, os2, sum.go(U32.to_nat((U32.shln(1, w) - 1 : U32)), wr, U32.shln(j, w), k, o, step, nb, Three{t, cs, os}))
143    case 1n+p:
144      (t0, t1) = ft
145      (c0, c1) = fc
146      (o0, o1) = fo
147      +i = (j * 2 : U32)
148      l r = sums(p, wr, i, w, k, o, step, nb, Array.fork(U32, t0), Array.fork(U32, c0), Array.fork(U32, o0))
149        sums(p, wr, (i + 1 : U32), w, k, o, step, nb, Array.fork(U32, t1), Array.fork(U32, c1), Array.fork(U32, o1))
150      three.join(l, r)