Merge branch 'feature/performance-fixes' into 'master'

Fix performance bugs with the sumfact kernels in functions merge

See merge request dominic/dune-perftool!246

python/dune/perftool/sumfact/realization.py

@@ -275,7 +275,7 @@ def realize_sumfact_kernel_function(sf):
     # Construct a loopy kernel object
     from dune.perftool.pdelab.localoperator import extract_kernel_from_cache
     args = ("const char* buffer0", "const char* buffer1") + sf.interface.signature_args
-    signature = "void {}({}) const".format(sf.function_name, ", ".join(args))
+    signature = "void {}({}) const __attribute__((always_inline))".format(sf.function_name, ", ".join(args))
     kernel = extract_kernel_from_cache("kernel_default", sf.function_name, [signature], add_timings=False)
     delete_cache_items("kernel_default")
     return kernel