This set of patches adds support for running benchmarks and tests using simple make commands like make check, make bench and make (to run both check and then a bench). The test runs are pretty much complete and just need to be synced up with the way bench is run, so most of the changes are in bench.
The benchmarks can now be run in two ways, one via a benchmark driver linked against libluajit.a (provide a path to it using LUAJIT_PATH) or directly via an interpreter specified using the LUA_BIN make variable. Detailed descriptions are in bench/README.
This should also make LuaJIT-test-cleanup good enough for incorporation into LuaJIT proper with added copyright and licensing. Given that the code is in public domain (except code that I added, which is MIT licensed), it should be pretty straightforward to do.
This set of patches adds support for running benchmarks and tests using simple make commands like make check, make bench and make (to run both check and then a bench). The test runs are pretty much complete and just need to be synced up with the way bench is run, so most of the changes are in bench.
The benchmarks can now be run in two ways, one via a benchmark driver linked against libluajit.a (provide a path to it using LUAJIT_PATH) or directly via an interpreter specified using the LUA_BIN make variable. Detailed descriptions are in bench/README.
This should also make LuaJIT-test-cleanup good enough for incorporation into LuaJIT proper with added copyright and licensing. Given that the code is in public domain (except code that I added, which is MIT licensed), it should be pretty straightforward to do.