Collective Knowledge Aggregator proof-of-concept
Crowd results Raw CK browser Graphs Reports Datasets Models Home

This page is outdated! New version is available here.


Distinct solutions after online classification (auto/crowd-tune GCC compiler flags (minimize execution time))

Scenario UID8289e0cf24346aa7 (experiment.tune.compiler.flags.gcc.e)
Data UID03064a714999bc65
Discuss (optimizations to improve compilers,
semantic/data set/hardware features
to improve predictions
, etc):
GitHub wiki, Google group
Download:[ All solutions in JSON ], [ Solutions' classification in JSON ]
Reproduce all (with reactions):ck replay 8289e0cf24346aa7:03064a714999bc65
CompilerGCC 4.9
CPUstar
Objectivemin
Improvement key IK1Main kernel execution time speedup [min]
Improvement key IK2Code size improvement

Improvements (<4% variation) Distinct workload for highest improvement
# Solution UID IK1 IK2 New distinct optimization choices Ref Best species Worst species Touched Iters Program CMD Dataset Dataset file CPU freq (MHz) Cores Platform OS Replay
S1 4b39833f13707fa7 1.14 0.98 -O3 -fcombine-stack-adjustments -fno-conserve-stack -fno-crossjumping -fno-early-inlining -fgcse-lm -fno-gcse-sm -findirect-inlining -fipa-cp-clone -fno-ira-share-spill-slots -floop-parallelize-all -fmodulo-sched -fno-rerun-cse-after-loop -fno-sched-spec-insn-heuristic -fsched-dep-count-heuristic -ftree-coalesce-vars -ftree-dce -ftree-loop-distribute-patterns -fno-tree-partial-pre -fno-tree-ter -fvpt --param predictable-branch-outcome=13 --param max-variable-expansions-in-unroller=0 --param max-unrolled-insns=80 --param max-unswitch-level=1 --param max-iterations-to-track=1056 --param tracer-max-code-growth=185 --param lim-expensive=3 --param omega-max-eqs=48 --param max-sched-region-blocks=15 --param sched-state-edge-prob-cutoff=50 --param prefetch-latency=246 --param prefetch-min-insn-to-mem-ratio=6 --param lto-min-partition=616 --param tree-reassoc-width=0 --param asan-use-after-return=0 --param uninit-control-dep-attempts=1495 -O3 1 0 2 1 milepost-codelet-mibench-consumer-jpeg-c-src-jchuff-codelet-9-1 default 1000, 1000 1 LGE LG-P990 Android 4.0.4
S2 d6daa222b3df79a2 1.11 1.00 -O3 -fno-auto-inc-dec -fbranch-target-load-optimize -fcse-follow-jumps -fcse-skip-blocks -fno-devirtualize -fno-finite-math-only -fno-function-sections -fno-gcse-las -finline-small-functions -fno-ipa-reference -fno-ira-loop-pressure -fira-share-save-slots -fno-ivopts -fno-keep-static-consts -fmath-errno -foptimize-sibling-calls -fno-sched-critical-path-heuristic -fno-strict-overflow -fno-tree-loop-if-convert-stores -fno-tree-phiprop -fno-tree-switch-conversion -fno-tree-vectorize -funsafe-loop-optimizations -fno-unswitch-loops --param max-inline-insns-recursive=702 --param max-inline-recursive-depth=7 --param max-delay-slot-insn-search=36 --param sms-dfa-history=0 --param min-crossjump-insns=1 --param omega-max-vars=199 --param omega-max-keys=634 --param vect-max-version-for-alias-checks=12 --param max-reload-search-insns=41 --param max-sched-insn-conflict-delay=5 --param ira-max-loops-num=71 --param lra-max-considered-reload-pseudos=3 --param graphite-max-nb-scop-params=17 --param prefetch-min-insn-to-mem-ratio=1 --param ipa-cp-value-list-size=15 --param allow-load-data-races=1 --param asan-memintrin=0 --param asan-use-after-return=1 -O3 1 0 2 1 milepost-codelet-mibench-automotive-basicmath-isqrt-codelet-1-1 default 1000, 1000 1 LGE LG-P990 Android 4.0.4
S3 5ee2f1f993644540 1.09 1.00 -O3 -ffat-lto-objects -fno-modulo-sched -fpeel-loops -fno-ree -fshrink-wrap -fno-split-wide-types -fno-tree-ccp -ftree-ch -fno-tree-tail-merge -ftree-vrp -ffp-contract=fast --param max-inline-insns-single=44 --param max-variable-expansions-in-unroller=0 --param min-vect-loop-bound=1 --param large-unit-insns=11922 --param ipcp-unit-growth=16 --param large-stack-frame=421 --param max-unrolled-insns=284 --param max-completely-peel-times=7 --param max-unswitch-insns=70 --param max-iterations-to-track=1433 --param sms-loop-average-count-threshold=0 --param hot-bb-count-ws-permille=141 --param hot-bb-frequency-fraction=973 --param max-crossjump-edges=28 --param max-goto-duplication-insns=13 --param vect-max-peeling-for-alignment=26 --param max-pipeline-region-blocks=13 --param ssp-buffer-size=4 --param ipa-sra-ptr-growth-factor=2 --param ipa-cp-eval-threshold=228 --param sched-pressure-algorithm=2 --param asan-stack=1 -O3 1 0 2 1 milepost-codelet-mibench-consumer-tiffmedian-src-tiffmedian-codelet-5-1 default 1000, 1000 1 LGE LG-P990 Android 4.0.4



[ Participated users, platforms, OS, CPU, GPU, GPGPU, NN, NPU ] [ How to participate ] [ Motivation (PPT) (PDF) ] [ Papers 1 , 2 , 3] [ Android app ] [ Collective training set ] [ Unified AI ]
View entry in raw format

Developed by Grigori Fursin           
Implemented as a CK workflow
                         
   
                      Hosted at