This small series provides another ~6% speedup on top of Bandan's patches. It does so by speeding up instruction fetches. The three tricks, one per patch, are to help the compiler a bit by redistributing the work between do_insn_fetch_byte and do_insn_fetch, to avoid repeated checks on the boundary of the fetch cache, and to exploit the processor's support for unaligned accesses. Before: 819 cycles/emulated jump instruction 1020 cycles/emulated move instruction 1026 cycles/emulated arithmetic instruction 1264 cycles/emulated memory load instruction 1182 cycles/emulated memory store instruction 1299 cycles/emulated memory RMW instruction After: 771 cycles/emulated jump instruction 963 cycles/emulated move instruction 960 cycles/emulated arithmetic instruction 1192 cycles/emulated memory load instruction 1110 cycles/emulated memory store instruction 1228 cycles/emulated memory RMW instruction Paolo Bonzini (4): KVM: emulate: speed up do_insn_fetch KVM: emulate: avoid repeated calls to do_insn_fetch_bytes KVM: emulate: avoid per-byte copying in instruction fetches KVM: emulate: put pointers in the fetch_cache arch/x86/include/asm/kvm_emulate.h | 4 +- arch/x86/kvm/emulate.c | 97 ++++++++++++++++++++------------------ arch/x86/kvm/trace.h | 6 +-- 3 files changed, 57 insertions(+), 50 deletions(-) -- 1.8.3.1 -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
