blob: dcf1225071342ba446ee8e88a0bcade4e21801a3 [file] [log] [blame]
njne1b349b2005-11-28 17:32:49 +00001-----------------------------------------------------------------------------
2Notes on performance
3-----------------------------------------------------------------------------
4The intent of this file is to record progress in improving performance.
5
6-----------------------------------------------------------------------------
7Just before 3.1.0:
8- Julian made LibVEX_Alloc() inlinable. Saved a couple of percent.
9- Julian started building Vex at -O2. Saved up to 8% or so(?) in some
10 cases.
11
12Post 3.1.0:
13- Julian made the tree builder linear. Saved 2--13% on a range of programs.
njn5096a392005-12-13 20:05:00 +000014- Nick improved vg_SP_update_pass() to identify more small constant
15 increments/decrements of SP, so the fast cases can be used more often.
16 Saved 1--3% on a few programs.
njn265bf442005-12-15 19:41:14 +000017- r5345,r5346: Julian improved the dispatcher so that x86 and AMD64 use
18 jumps instead of call/return for calling translations, and also removed
19 the --profile-flags profiling from the dispatcher unless --profile-flags
20 is being used. Improved Nulgrind performance typically by 10--20%,
21 and Memcheck performance typically by 2--20%.
njne1b349b2005-11-28 17:32:49 +000022
23COMPVBITS branch:
24- Nick converted to compress V bits, initial version saved 0--5% on most
njn47fb6502005-12-02 23:09:49 +000025 cases, with a 30% improvement on one case (tsim_arch) which calls
njne1b349b2005-11-28 17:32:49 +000026 set_address_range_perms() a lot.
njn47fb6502005-12-02 23:09:49 +000027- Nick rewrote set_address_range_perms(), which gained 0--3% typically,
28 and 22% on tsim_arch.
njne1b349b2005-11-28 17:32:49 +000029