From 4231740e48c3bcc274eca01c77c22b7092fd0f76 Mon Sep 17 00:00:00 2001 From: Tim Besard Date: Fri, 23 Oct 2026 15:11:22 +0200 Subject: [PATCH 1/2] Load the ordered reduction's vectors with `_vloada` #165 and #168 were merged on the same day. #165 added a vector load to the ordered reduction kernel with SIMD.jl's `vloada`, which #168 had just replaced with an `llvmcall`-based `_vloada` and stopped importing. The kernel thus failed to compile for dense sources with `commutative=false` on every GPU back-end, which breaks 0.5.8 and the existing vector-load tests in test/generic/reduce.jl. --- src/reduce/mapreduce_ordered.jl | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/reduce/mapreduce_ordered.jl b/src/reduce/mapreduce_ordered.jl index 6bbd3f1d..89af5c58 100644 --- a/src/reduce/mapreduce_ordered.jl +++ b/src/reduce/mapreduce_ordered.jl @@ -44,7 +44,7 @@ acc = neutral if W > 0 && (tile + 1) * tile_size <= reduce_size for k in 0x0:W:(K - 0x1) - v = vloada(Vec{W, eltype(src)}, pointer(src, input_base + run + k + 0x1)) + v = _vloada(Val(W), pointer(src, input_base + run + k + 0x1)) acc = _acc_lanes(f_lanes, op_lanes, acc, v) end elseif (tile + 1) * tile_size <= reduce_size From 9e970909b8befa9088b379b7cd97c2e6125bc2c0 Mon Sep 17 00:00:00 2001 From: Tim Besard Date: Fri, 23 Oct 2026 15:29:21 +0200 Subject: [PATCH 2/2] Bump version. --- Project.toml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Project.toml b/Project.toml index 464ec222..aa7cd335 100644 --- a/Project.toml +++ b/Project.toml @@ -1,6 +1,6 @@ name = "AcceleratedKernels" uuid = "6a4ca0a5-0e36-4168-a932-d9be78d558f1" -version = "0.5.8" +version = "0.5.9" authors = ["Andrei-Leonard Nicusan and contributors"] [deps]