Bump compat for Trixi to 0.10 to adapt the latest version of CUDA (#141)

huiyuxie · web-flow · commit fc98d41a5d6f · 2025-03-19T12:03:41.000-10:00
* Bump Trixi

* Fix 1D

* Fix typo

* Fix 2D

* Update readme
diff --git a/Project.toml b/Project.toml
@@ -15,7 +15,7 @@ BenchmarkTools = "1"
 CUDA = "5"
 SciMLBase = "2"
 StaticArrays = "1"
-Trixi = "0.9.10 - 0.9.15"
+Trixi = "0.10.0 - 0.10.2"
 julia = "1.10"
 
 [extras]
diff --git a/README.md b/README.md
@@ -12,14 +12,14 @@ General docs: https://trixi-gpu.github.io
 > [!WARNING]
 > The package may not always be updated with the latest updates in Trixi.jl. Forcing an update of Trixi.jl as a dependency for TrixiCUDA.jl beyond the version bounds specified in Project.toml may cause unexpected errors.
 
+*Update on Mar 19, 2025*:
+- The issue between the latest version of CUDA.jl and Trixi.jl has been resolved. The package is now compatible with CUDA.jl v5.7.0 and Trixi.jl v0.10 (see [TrixiCUDA.jl PR #141](https://github.com/trixi-gpu/TrixiCUDA.jl/pull/141)).
+
 *Update on Jan 28, 2025*:
 - It is recommended to update your Julia version to 1.10.8 (the latest LTS release) to avoid the issue of circular dependencies during package precompilation, which is present in Julia 1.10.7 (see [issue](https://discourse.julialang.org/t/circular-dependency-warning/123388)).
 
 *Update on Dec 31, 2024*:
-- The kernel optimization starts with the volume integral kernels (see [TrixiCUDA.jl PR #102](https://github.com/trixi-gpu/TrixiCUDA.jl/pull/102)) and will extend to all existing kernels used in the semidiscretization. The approach includes improving global memory access patterns, using shared memory, and selecting appropriate launch sizes to reduce warp stalls and achieve better occupancy.
-
-*Update on Nov 21, 2024*:
-- Due to the [issue](https://github.com/trixi-framework/Trixi.jl/issues/2108) from upstream with Trixi.jl and CUDA.jl in Julia v1.11, this package now supports only Julia v1.10. Using or developing this package with Julia v1.11 will result in precompilation errors. To fix this, downgrade to Julia v1.10. If you have any other problems, please file issues [here](https://github.com/trixi-gpu/TrixiCUDA.jl/issues).
+- The kernel optimization starts with the volume integral kernels (see [TrixiCUDA.jl PR #102](https://github.com/trixi-gpu/TrixiCUDA.jl/pull/102)) and will extend to all existing kernels used in the semidiscretization.
 
 [Archived Update](https://trixi-gpu.github.io/update/)
 
diff --git a/src/solvers/dg_1d.jl b/src/solvers/dg_1d.jl
@@ -424,7 +424,8 @@ end
 # Pack kernels for calculating boundary fluxes
 function cuda_boundary_flux!(t, mesh::TreeMesh{1}, boundary_conditions::NamedTuple,
                              nonconservative_terms::True, equations, dg::DGSEM, cache)
-    surface_flux, nonconservative_flux = dg.surface_integral.surface_flux
+    # Contain both conservative and nonconservative fluxes
+    surface_flux = dg.surface_integral.surface_flux
 
     n_boundaries_per_direction = cache.boundaries.n_boundaries_per_direction
     neighbor_ids = cache.boundaries.neighbor_ids
@@ -456,11 +457,11 @@ function cuda_boundary_flux!(t, mesh::TreeMesh{1}, boundary_conditions::NamedTup
                                                                     boundary_conditions_callable,
                                                                     equations,
                                                                     surface_flux,
-                                                                    nonconservative_flux)
+                                                                    nonconservative_terms)
     boundary_flux_kernel(surface_flux_values, boundaries_u, node_coordinates, t, boundary_arr,
                          indices_arr, neighbor_ids, neighbor_sides, orientations,
                          boundary_conditions_callable, equations, surface_flux,
-                         nonconservative_flux;
+                         nonconservative_terms;
                          kernel_configurator_1d(boundary_flux_kernel, length(boundary_arr))...)
 
     return nothing
diff --git a/src/solvers/dg_1d_kernel.jl b/src/solvers/dg_1d_kernel.jl
@@ -836,7 +836,7 @@ end
 function boundary_flux_kernel!(surface_flux_values, boundaries_u, node_coordinates, t, boundary_arr,
                                indices_arr, neighbor_ids, neighbor_sides, orientations,
                                boundary_conditions::NamedTuple, equations::AbstractEquations{1},
-                               surface_flux::Any, nonconservative_flux::Any)
+                               surface_flux::Any, nonconservative_terms::True)
     k = (blockIdx().x - 1) * blockDim().x + threadIdx().x
 
     if (k <= length(boundary_arr))
@@ -855,15 +855,11 @@ function boundary_flux_kernel!(surface_flux_values, boundaries_u, node_coordinat
 
         # TODO: Improve this part
         if direction == 1
-            flux_node = boundary_conditions[1](u_inner, orientation, direction, x, t, surface_flux,
-                                               equations)
-            noncons_flux_node = boundary_conditions[1](u_inner, orientation, direction, x, t,
-                                                       nonconservative_flux, equations)
+            flux_node, noncons_flux_node = boundary_conditions[1](u_inner, orientation, direction,
+                                                                  x, t, surface_flux, equations)
         else
-            flux_node = boundary_conditions[2](u_inner, orientation, direction, x, t, surface_flux,
-                                               equations)
-            noncons_flux_node = boundary_conditions[2](u_inner, orientation, direction, x, t,
-                                                       nonconservative_flux, equations)
+            flux_node, noncons_flux_node = boundary_conditions[2](u_inner, orientation, direction,
+                                                                  x, t, surface_flux, equations)
         end
 
         for ii in axes(surface_flux_values, 1)
diff --git a/src/solvers/dg_2d.jl b/src/solvers/dg_2d.jl
@@ -491,7 +491,8 @@ end
 # Pack kernels for calculating boundary fluxes
 function cuda_boundary_flux!(t, mesh::TreeMesh{2}, boundary_conditions::NamedTuple,
                              nonconservative_terms::True, equations, dg::DGSEM, cache)
-    surface_flux, nonconservative_flux = dg.surface_integral.surface_flux
+    # Contain both conservative and nonconservative fluxes
+    surface_flux = dg.surface_integral.surface_flux
 
     n_boundaries_per_direction = cache.boundaries.n_boundaries_per_direction
     neighbor_ids = cache.boundaries.neighbor_ids
@@ -523,11 +524,11 @@ function cuda_boundary_flux!(t, mesh::TreeMesh{2}, boundary_conditions::NamedTup
                                                                     boundary_conditions_callable,
                                                                     equations,
                                                                     surface_flux,
-                                                                    nonconservative_flux)
+                                                                    nonconservative_terms)
     boundary_flux_kernel(surface_flux_values, boundaries_u, node_coordinates, t, boundary_arr,
                          indices_arr, neighbor_ids, neighbor_sides, orientations,
                          boundary_conditions_callable, equations, surface_flux,
-                         nonconservative_flux;
+                         nonconservative_terms;
                          kernel_configurator_2d(boundary_flux_kernel, size(surface_flux_values, 2),
                                                 length(boundary_arr))...)
 
diff --git a/src/solvers/dg_2d_kernel.jl b/src/solvers/dg_2d_kernel.jl
@@ -1120,7 +1120,7 @@ end
 function boundary_flux_kernel!(surface_flux_values, boundaries_u, node_coordinates, t, boundary_arr,
                                indices_arr, neighbor_ids, neighbor_sides, orientations,
                                boundary_conditions::NamedTuple, equations::AbstractEquations{2},
-                               surface_flux::Any, nonconservative_flux::Any)
+                               surface_flux::Any, nonconservative_terms::True)
     j = (blockIdx().x - 1) * blockDim().x + threadIdx().x
     k = (blockIdx().y - 1) * blockDim().y + threadIdx().y
 
@@ -1141,25 +1141,17 @@ function boundary_flux_kernel!(surface_flux_values, boundaries_u, node_coordinat
 
         # TODO: Improve this part
         if direction == 1
-            flux_node = boundary_conditions[1](u_inner, orientation, direction, x, t, surface_flux,
-                                               equations)
-            noncons_flux_node = boundary_conditions[1](u_inner, orientation, direction, x, t,
-                                                       nonconservative_flux, equations)
+            flux_node, noncons_flux_node = boundary_conditions[1](u_inner, orientation, direction,
+                                                                  x, t, surface_flux, equations)
         elseif direction == 2
-            flux_node = boundary_conditions[2](u_inner, orientation, direction, x, t, surface_flux,
-                                               equations)
-            noncons_flux_node = boundary_conditions[2](u_inner, orientation, direction, x, t,
-                                                       nonconservative_flux, equations)
+            flux_node, noncons_flux_node = boundary_conditions[2](u_inner, orientation, direction,
+                                                                  x, t, surface_flux, equations)
         elseif direction == 3
-            flux_node = boundary_conditions[3](u_inner, orientation, direction, x, t, surface_flux,
-                                               equations)
-            noncons_flux_node = boundary_conditions[3](u_inner, orientation, direction, x, t,
-                                                       nonconservative_flux, equations)
+            flux_node, noncons_flux_node = boundary_conditions[3](u_inner, orientation, direction,
+                                                                  x, t, surface_flux, equations)
         else
-            flux_node = boundary_conditions[4](u_inner, orientation, direction, x, t, surface_flux,
-                                               equations)
-            noncons_flux_node = boundary_conditions[4](u_inner, orientation, direction, x, t,
-                                                       nonconservative_flux, equations)
+            flux_node, noncons_flux_node = boundary_conditions[4](u_inner, orientation, direction,
+                                                                  x, t, surface_flux, equations)
         end
 
         for ii in axes(surface_flux_values, 1)
