Why is there no companion instruction to leave?

Question

Why is there no companion instruction to leave on the x86? That way,

pushl   %ebp
movl    %esp,%ebp
pushl   $3
popl    %eax
leave
ret

could become:

enter #or something
pushl   $3
popl    %eax
leave
ret

Isn't that just faster in general?

first code is faster . see http://stackoverflow.com/questions/5959890/enter-vs-push-ebp-mov-ebp-esp-sub-esp-imm-and-leave-vs-mov-esp-ebp — qwr, Jul 05 '13 at 06:16

score 5 · Accepted Answer · answered Jul 05 '13 at 06:20

5

In fact, there is an enter instruction. As to why it's not seeing much use, "Intel® 64 and IA-32 Architectures Optimization Reference Manual" offers a hint:

Assembly/ Compiler Coding Rule 32. (MH impact, M generality) Use push/pop to manage stack space and address adjustments between function calls/returns instead of enter/leave. Using enter instruction with non-zero immediates can experience significant delays in the pipeline in addition to misprediction.

answered Jul 05 '13 at 06:20

NPE

486,780
108
951
1,012

`leave` is not slow so GCC *does* use it when it made a frame pointer, if ESP isn't already pointing at a saved EBP. `leave` is 3 uops, vs. 2 uops for separate `mov %ebp,%esp` / `pop %ebp`. But yes, `enter` is terrible, even `enter 0,0` is 12 uops on Haswell/Skylake. (https://agner.org/optimize/ ; https://uops.info/) – Peter Cordes Feb 10 '20 at 21:30

score 4 · Answer 2 · answered Jul 05 '13 at 06:20

4

There is indeed a matching ENTER (Art of Assembly Language) instruction. It's seldom used however, because it is very slow, as explained in the link.

answered Jul 05 '13 at 06:20

David Heffernan

601,492
42
1,072
1,490

score 1 · Answer 3 · answered Jul 05 '13 at 06:20

1

There is enter, for some processors.

answered Jul 05 '13 at 06:20

alex

479,566
201
878
984

3

IIRC, it was introduced on x86 with the 80186/80188 processors. – Brian Knoblauch Jul 05 '13 at 13:22

Why is there no companion instruction to leave?

3 Answers3