From 77bb8bd5030088673e51ebaa3b3f5f4b6d2e01e6 Mon Sep 17 00:00:00 2001
From: ChrisRackauckas-Claude <accounts@chrisrackauckas.com>
Date: Sat, 9 May 2026 07:13:27 -0400
Subject: [PATCH] compat: restore CUDA v5 lower bound
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Drop the upper-only CUDA = "6" pin in favor of CUDA = "5, 6". CUDA 6
ships with no breaking changes from 5 (per the v6.0.0 release notes),
the CUDAExt path only touches CUDA.CUDABackend and
CUDA.CUBLAS.getrf_strided_batched! — both stable across both majors —
and the test utilities use only CUDA.CUDABackend(). Restoring the v5
lower bound unblocks consumers (e.g. NeuralLyapunov via AdvancedHMC)
that haven't yet upgraded their CUDA compat.

Resolves #449.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
---
 Project.toml      | 2 +-
 docs/Project.toml | 2 +-
 test/Project.toml | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/Project.toml b/Project.toml
index fedcac02..65bf16b9 100644
--- a/Project.toml
+++ b/Project.toml
@@ -45,7 +45,7 @@ oneAPIExt = ["oneAPI"]
 [compat]
 AMDGPU = "2"
 Adapt = "4"
-CUDA = "6"
+CUDA = "5, 6"
 ChainRulesCore = "1"
 DiffEqBase = "7"
 DocStringExtensions = "0.9"
diff --git a/docs/Project.toml b/docs/Project.toml
index 0f9b2a5e..bc1ba879 100644
--- a/docs/Project.toml
+++ b/docs/Project.toml
@@ -25,7 +25,7 @@ Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 [compat]
 Adapt = "4"
 BenchmarkTools = "1"
-CUDA = "6"
+CUDA = "5, 6"
 DiffEqBase = "7"
 DiffEqGPU = "3"
 Documenter = "1"
diff --git a/test/Project.toml b/test/Project.toml
index 70556c07..558c0313 100644
--- a/test/Project.toml
+++ b/test/Project.toml
@@ -34,7 +34,7 @@ pocl_jll = "627d6b7a-bbe6-5189-83e7-98cc0a5aeadd"
 [compat]
 Adapt = "4"
 BenchmarkTools = "1"
-CUDA = "6"
+CUDA = "5, 6"
 DiffEqDevTools = "3, 4.0"
 DiffEqGPU = "3"
 ForwardDiff = "1"