Improve performance for branchy code
authorAndrew Waterman <waterman@eecs.berkeley.edu>
Tue, 14 Jan 2014 00:42:02 +0000 (16:42 -0800)
committerAndrew Waterman <waterman@eecs.berkeley.edu>
Tue, 14 Jan 2014 00:42:02 +0000 (16:42 -0800)
commit77f281580773fe5217ae40bcf8a0f8bc05c28ded
tree79885362aa51ce842ec912ba4523f36620b1dce8
parent7f457c47b339cc7c79f56bb277ed8ed989e88ae1
Improve performance for branchy code

We now use a heavily unrolled loop as the software I$, which allows the
host machine's branch target prediction to associate target PCs with
unique-ish host PCs.
14 files changed:
Makefile.in
riscv/decode.h
riscv/encoding.h
riscv/gen_icache [new file with mode: 0755]
riscv/htif.cc
riscv/htif.h
riscv/interactive.cc
riscv/mmu.cc
riscv/mmu.h
riscv/processor.cc
riscv/processor.h
riscv/riscv.mk.in
riscv/sim.cc
riscv/sim.h