Comment

Comment by nneonneo

FWIW, I get very different results from inside my Linux VM (MacBook Pro, 2.9 GHz Core i7):

    jit: 1.433856 ns/call
    plt: 1.700611 ns/call
    ind: 1.404397 ns/call

The PLT is double-indirected (call->jmp->func), but the single-indirect call performs the same as the JIT. This backs up @tlb's note that modern x86 machines will predict indirect calls.

parent nneonneo 8 years ago

Comment by nneonneo

Discussion