From 85db0326817188021683a24d3848967cbc391e0d Mon Sep 17 00:00:00 2001 From: Tom Smeding Date: Mon, 13 Jan 2025 21:55:19 +0100 Subject: [PATCH] Add TODO comment about stencils with unequal rows --- src/Data/Array/Accelerate/Smart.hs | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/src/Data/Array/Accelerate/Smart.hs b/src/Data/Array/Accelerate/Smart.hs index a734200b8..bf0e85873 100644 --- a/src/Data/Array/Accelerate/Smart.hs +++ b/src/Data/Array/Accelerate/Smart.hs @@ -689,6 +689,10 @@ type instance StencilR Sugar.DIM1 (Exp e, Exp e, Exp e, Exp e, Exp e, Exp e, Exp = EltR (e, e, e, e, e, e, e, e, e) -- DIM(n+1) +-- TODO: These instances allow stencils with unequal sub-stencil sizes (i.e. +-- unequal row lengths for a 2D stencil). This is rather odd, probably unused, +-- and makes type inference a little less eager to deduce the stencil type in +-- case of underspecified rows. Can we just require the rows to be equal? instance (Stencil (sh:.Int) a row2, Stencil (sh:.Int) a row1, Stencil (sh:.Int) a row0,