found one bug, store_factorization_as_float still not converging

OPM · Oct 22, 2024 · e672209 · e672209
1 parent b6d2394
commit e672209
Show file tree

Hide file tree

Showing 2 changed files with 11 additions and 13 deletions.
diff --git a/opm/simulators/linalg/gpuistl/OpmGpuILU0.cpp b/opm/simulators/linalg/gpuistl/OpmGpuILU0.cpp
@@ -190,7 +190,7 @@ OpmGpuILU0<M, X, Y, l>::apply(X& v, const Y& d, int lowerSolveThreadBlockSize, i
                         v.data(),
                         upperSolveThreadBlockSize);
             }
-            if (m_mixedPrecisionScheme == MixedPrecisionScheme::STORE_ENTIRE_FACTORIZATION_AS_FLOAT) {
+            if (m_mixedPrecisionScheme == MixedPrecisionScheme::STORE_ONLY_FACTORIZED_DIAGONAL_AS_DOUBLE) {
                     detail::ILU0::solveUpperLevelSetSplit<blocksize_, field_type, float, field_type>(
                         m_gpuMatrixReorderedUpperFloat->getNonZeroValues().data(),
                         m_gpuMatrixReorderedUpperFloat->getRowIndices().data(),

diff --git a/opm/simulators/linalg/gpuistl/detail/preconditionerKernels/ILU0Kernels.cu b/opm/simulators/linalg/gpuistl/detail/preconditionerKernels/ILU0Kernels.cu
@@ -505,18 +505,16 @@ LUFactorizationSplit(InputScalar* srcReorderedLowerMat,
     template void LUFactorizationSplit<blocksize, T, double, MixedPrecisionScheme::STORE_ONLY_FACTORIZED_DIAGONAL_AS_DOUBLE>(                                                   \
         T*, int*, int*, T*, int*, int*, T*, double*, double*, double*, int*, int*, const int, int, int); 
 
-INSTANTIATE_KERNEL_WRAPPERS(float, 1);
-INSTANTIATE_KERNEL_WRAPPERS(float, 2);
-INSTANTIATE_KERNEL_WRAPPERS(float, 3);
-INSTANTIATE_KERNEL_WRAPPERS(float, 4);
-INSTANTIATE_KERNEL_WRAPPERS(float, 5);
-INSTANTIATE_KERNEL_WRAPPERS(float, 6);
-INSTANTIATE_KERNEL_WRAPPERS(double, 1);
-INSTANTIATE_KERNEL_WRAPPERS(double, 2);
-INSTANTIATE_KERNEL_WRAPPERS(double, 3);
-INSTANTIATE_KERNEL_WRAPPERS(double, 4);
-INSTANTIATE_KERNEL_WRAPPERS(double, 5);
-INSTANTIATE_KERNEL_WRAPPERS(double, 6);
+#define INSTANTIATE_BLOCK_SIZED_KERNEL_WRAPPERS(T) \
+    INSTANTIATE_KERNEL_WRAPPERS(T, 1); \
+    INSTANTIATE_KERNEL_WRAPPERS(T, 2); \
+    INSTANTIATE_KERNEL_WRAPPERS(T, 3); \
+    INSTANTIATE_KERNEL_WRAPPERS(T, 4); \
+    INSTANTIATE_KERNEL_WRAPPERS(T, 5); \
+    INSTANTIATE_KERNEL_WRAPPERS(T, 6);
+
+INSTANTIATE_BLOCK_SIZED_KERNEL_WRAPPERS(float)
+INSTANTIATE_BLOCK_SIZED_KERNEL_WRAPPERS(double)
 
 #define INSTANTIATE_MIXED_PRECISION_KERNEL_WRAPPERS(blocksize)                                                         \
     /* double preconditioner */                                                                                        \