l.mulu is not implemented in the OR1200 core. The ALU opcode 0xb reserved for l.mulu in the specification document is actually implemented such that l.mulu rd,ra,rb results in rd <- rb. (Ref. line 198 in or1200_alu.v, line 468 in or1200_defines.v)
All fixed now. ALU opcodes arranged correctly.
Software test implemented:
http://opencores.org/ocsvn/openrisc/openrisc/trunk/orpsocv2/sw/tests/or1200/sim/or1200-mul.c