diff --git a/src/pystencilssfg/ir/call_tree.py b/src/pystencilssfg/ir/call_tree.py index ab84db4fe8452e57b3b3a85b0fb529865e84c5f9..61c5bbbe01cdd42ae7b94b7843aa1e916c3afedb 100644 --- a/src/pystencilssfg/ir/call_tree.py +++ b/src/pystencilssfg/ir/call_tree.py @@ -226,13 +226,12 @@ class SfgGpuKernelInvocation(SfgCallTreeNode): shared_memory_bytes: SfgStatements | None, stream: SfgStatements | None, ): - from pystencils import Target from pystencils.codegen import GpuKernel kernel = kernel_handle.kernel - if not (isinstance(kernel, GpuKernel) and kernel.target == Target.CUDA): + if not isinstance(kernel, GpuKernel): raise ValueError( - "An `SfgCudaKernelInvocation` node can only call a CUDA kernel." + "An `SfgGpuKernelInvocation` node can only call GPU kernels." ) super().__init__() diff --git a/tests/generator_scripts/source/HipKernels.py b/tests/generator_scripts/source/HipKernels.py index 78464b5be1e4b24863a8cab62d96fd55ba60d544..32d9b1dfbabce88517e072cec7c9fa6785968a32 100644 --- a/tests/generator_scripts/source/HipKernels.py +++ b/tests/generator_scripts/source/HipKernels.py @@ -14,7 +14,7 @@ with SourceFileGenerator() as sfg: sfg.use_hip() sfg.namespace("gen") - base_config = ps.CreateKernelConfig(target=ps.Target.CUDA) + base_config = ps.CreateKernelConfig(target=ps.Target.HIP) block_size = sfg.gpu_api.dim3().var("blockSize") grid_size = sfg.gpu_api.dim3().var("gridSize")