| ----------------------------------------------------------------------------- |
| Notes on performance |
| ----------------------------------------------------------------------------- |
| The intent of this file is to record progress in improving performance. |
| |
| ----------------------------------------------------------------------------- |
| Just before 3.1.0: |
| - Julian made LibVEX_Alloc() inlinable. Saved a couple of percent. |
| - Julian started building Vex at -O2. Saved up to 8% or so(?) in some |
| cases. |
| |
| Post 3.1.0: |
| - Julian made the tree builder linear. Saved 2--13% on a range of programs. |
| - Nick improved vg_SP_update_pass() to identify more small constant |
| increments/decrements of SP, so the fast cases can be used more often. |
| Saved 1--3% on a few programs. |
| - r5345,r5346,r5352: Julian improved the dispatcher so that x86 and |
| AMD64 use jumps instead of call/return for calling translations. |
| Also, on x86, amd64, ppc32 and ppc64, --profile-flags style profiling was |
| removed from the despatch loop unless --profile-flags is being used. |
| Improved Nulgrind performance typically by 10--20%, and Memcheck |
| performance typically by 2--20%. |
| - Julian changed findSb to slowly move superblocks to the front of the list |
| as they were accessed. This sped up perf/heap by 25--50%, and some big |
| programs (eg. ktuberling) programs by a couple of percent. |
| - Nick reduced the iteration count of the loop in swizzle() from 20 to 5, |
| which gave almost identical results while saving 2% in perf/tinycc and 10% |
| in perf/heap on a 3GHz Prescott P4. |
| - Nick changed ExeContext gathering to not record/save extra zeroes at the |
| end. Saved 7% on perf/heap with --num-callers=50, and about 1% on |
| perf/tinycc. |
| - Julian vectorised copy_address_range_perms for common cases, which |
| gives about 40% speedup on artificial programs which just do |
| realloc() and nothing else, and about a 3-4% speedup on starting |
| kpresenter-1.5.0 and loading a 16-slide presentation. |
| |
| COMPVBITS branch: |
| - Nick converted to compress V bits, initial version saved 0--5% on most |
| cases, with a 30% improvement on one case (tsim_arch) which calls |
| set_address_range_perms() a lot. |
| - Nick rewrote set_address_range_perms(), which gained 0--3% typically, |
| and 22% on tsim_arch. |
| |
