Skip to content

Read backward reductions freshest-last (closes most of the remaining gap) #45

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ChrisRackauckas merged 1 commit into from
Aug 8, 2026
Merged

Read backward reductions freshest-last (closes most of the remaining gap) #45

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion Project.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
name = "TriangularSolve"
uuid = "d5829a12-d9aa-46ab-831f-fb7c9ab06edf"
authors = ["chriselrod <elrodc@gmail.com> and contributors"]
version = "0.2.2"
version = "0.2.3"

[deps]
CloseOpenIntervals = "fb6a15b2-703c-40df-9091-08a04967cfa9"
Expand Down
23 changes: 18 additions & 5 deletions src/TriangularSolve.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1505,7 +1505,12 @@ end
# kernels' `SafeCloseOpen(n)` loops
spck = gesp(spc, ($z, n + $W))
splk = gesp(spl, (n + $W, n))
for nk ∈ SafeCloseOpen(nend - n - $W) # nmuladd
# nk descends so the most recently solved (just stored) columns are read
# last, giving their stores time to drain — the mirror of the forward
# kernels' ascending reduction, which also ends on the freshest block
nk = nend - n - $W
while nk > 0
nk -= 1
A11 = vload(spck, $(Unroll{1,W,U,1,W,zero(UInt),1})(($(StaticInt(0)), nk)))
Base.Cartesian.@nexprs $W c ->
C11_c = vfnmadd_fast(A11, vload(splk, (nk, $z + (c - 1))), C11_c)
Expand Down Expand Up @@ -1544,7 +1549,9 @@ end
Base.Cartesian.@nexprs $W c -> C11_c = C11[c]
spck = gesp(spc, ($z, n + $W))
splk = gesp(spl, (n + $W, n))
for nk ∈ SafeCloseOpen(nend - n - $W) # nmuladd
nk = nend - n - $W
while nk > 0
nk -= 1
A11 = vload(spck, ($(MM{W}(z)), nk), mask)
Base.Cartesian.@nexprs $W c ->
C11_c = vfnmadd_fast(A11, vload(splk, (nk, $z + (c - 1))), C11_c)
Expand Down Expand Up @@ -1834,7 +1841,9 @@ end
# kernels' `SafeCloseOpen(n)` loops
spck = gesp(spc, ($z, n + $(W * U)))
splk = gesp(spl, (n + $(W * U), n))
for nk ∈ SafeCloseOpen(nend - n - $(W * U)) # nmuladd
nk = nend - n - $(W * U)
while nk > 0
nk -= 1
L_ki = vload(splk, $(Unroll{2,W,U,2,W,zero(UInt),1})((nk, $z)))
Base.Cartesian.@nexprs $W c ->
A11_c = vfnmadd_fast(L_ki, vload(spck, (static(c - 1), nk)), A11_c)
Expand Down Expand Up @@ -1933,7 +1942,9 @@ end
Base.Cartesian.@nexprs $W c -> A11_c = getfield(A11, c)
spck = gesp(spc, ($z, n + $W))
splk = gesp(spl, (n + $W, n))
for nk ∈ SafeCloseOpen(nend - n - $W) # nmuladd
nk = nend - n - $W
while nk > 0
nk -= 1
L_ki = vload(splk, (nk, $(MM{W}(z))))
Base.Cartesian.@nexprs $W c ->
A11_c = vfnmadd_fast(L_ki, vload(spck, (static(c - 1), nk)), A11_c)
Expand Down Expand Up @@ -1966,7 +1977,9 @@ end
Base.Cartesian.@nexprs $R r -> A11_r = getfield(A11, r)
spck = gesp(spc, ($z, n + $W))
splk = gesp(spl, (n + $W, n))
for nk ∈ SafeCloseOpen(nend - n - $W) # nmuladd
nk = nend - n - $W
while nk > 0
nk -= 1
L_ki = vload(splk, (nk, $(MM{W}(z))))
Base.Cartesian.@nexprs $R r ->
A11_r = vfnmadd_fast(L_ki, vload(spck, (static(r - 1), nk)), A11_r)
Expand Down
Loading