[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 Ivo Raisr changed: What|Removed |Added Status|CONFIRMED |RESOLVED Resolution|---

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #15 from Julian Seward --- Ivo, can we close this? I assume that all of these improvements have long since landed in the v3 allocator (and also, that we're now shipping v3 by default!) But do let me know if I assume wrongly. -- You are

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 Ivo Raisr changed: What|Removed |Added See Also||https://bugs.kde.org/show_b

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 Ivo Raisr changed: What|Removed |Added Attachment #107674|0 |1 is obsolete|

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 Ivo Raisr changed: What|Removed |Added See Also||https://bugs.kde.org/show_b

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 Ivo Raisr changed: What|Removed |Added See Also||https://bugs.kde.org/show_b

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #13 from Ivo Raisr --- Interestingly enough, VEX register allocator v2 also benefits from register reordering: For Memcheck on perf/bz2: ra2-baseline: 45,214 M instructions total; 209 M doRegisterAllocation_v2

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #12 from Ivo Raisr --- (In reply to Ivo Raisr from comment #10) Included is also total cost of running doRegisterAllocation_v2/3: Here are my findings for running Memcheck on perf/bz2: v2 baseline: 45.214 G instructions

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 Julian Seward changed: What|Removed |Added CC||jsew...@acm.org --- Comment

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 Ivo Raisr changed: What|Removed |Added Assignee|jsew...@acm.org |iv...@ivosh.net Ever

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #9 from Ivo Raisr --- Created attachment 107711 --> https://bugs.kde.org/attachment.cgi?id=107711=edit Reorder allocatable registers for AMD64 and X86 archs Reorder allocatable registers for AMD64 and X86 so that the

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #8 from Julian Seward --- One thing you could try is this: ./vg-in-place --vex-regalloc-version=2 \ --profile-flags=0010 [whatever tool and args you like] ./perf/bz2 x That both shows you the hot blocks and also

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #7 from Ivo Raisr --- What I would like to do next is to compare how v2 and v3 allocate registers on some hot paths in perf/bz2. However I have difficulty how to locate such SBs. Although Callgrind shows addresses of such

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #6 from Ivo Raisr --- Philippe, please could you measure these guys on your performance tests. Although I was able to determine total instruction count on Memcheck+perf/bz2, I was not able to measure real time in a

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #5 from Ivo Raisr --- Created attachment 107678 --> https://bugs.kde.org/attachment.cgi?id=107678=edit Small performance enhancement to v2. Also makes it the default for testing. Small performance enhancement to VEX

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #4 from Ivo Raisr --- Created attachment 107677 --> https://bugs.kde.org/attachment.cgi?id=107677=edit Different algorithm for find_vreg_to_spill() in v3. Different algorithm for find_vreg_to_spill() in v3. Also scans 10

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #3 from Ivo Raisr --- Created attachment 107676 --> https://bugs.kde.org/attachment.cgi?id=107676=edit Scan 10 instructions ahead instead of just 5 Scans 10 instructions ahead in function find_vreg_to_spill() instead of

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 --- Comment #2 from Ivo Raisr --- Created attachment 107675 --> https://bugs.kde.org/attachment.cgi?id=107675=edit Track vreg usage on per-instruction precision in v3 Tracks vreg usage on per-instruction basis, using bitset. Should

[valgrind] [Bug 384337] performance improvements to VEX register allocator v2 and v3

https://bugs.kde.org/show_bug.cgi?id=384337 Ivo Raisr changed: What|Removed |Added CC||iv...@ivosh.net --- Comment #1