Collective Knowledge Aggregator proof-of-concept
Crowdsourced experiments CK project Partners Open AI powered by CK Reusable AI artifacts Get CK

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

Scenario UID8289e0cf24346aa7 (experiment.tune.compiler.flags.gcc.e)
Data UID4cc0b5206483eba4
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 8289e0cf24346aa7:4cc0b5206483eba4
CompilerGCC 4.9
CPUMT6735
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 5f0e5690a346fc92 1.33 0.31 -O3 -fauto-inc-dec -fdce -fdevirtualize-speculatively -ffast-math -fno-ipa-cp -fno-loop-block -flto -fmove-loop-invariants -fno-branch-count-reg -fno-defer-pop -fpeephole -fpartial-inlining -fno-prefetch-loop-arrays -fno-signaling-nans -fno-split-wide-types -ftree-dce -fno-tree-fre -ftree-reassoc -falign-jumps=0 -falign-loops=0 -finline-limit=0 -fsched-stalled-insns-dep=0 -ftree-parallelize-loops=0 --param max-inline-insns-recursive=357 --param max-inline-insns-recursive-auto=341 --param max-modulo-backtrack-attempts=1 --param large-stack-frame=103 --param max-peel-times=29 --param max-peel-branches=38 --param omega-max-geqs=461 --param max-cselib-memory-locations=825 --param max-sched-insn-conflict-delay=3 --param selsched-max-lookahead=48 --param sched-mem-true-dep-cost=2 --param max-fields-for-field-sensitive=0 --param max-sched-ready-insns=74 --param l1-cache-size=40 --param ipa-cp-loop-hint-bonus=113 --param max-tail-merge-comparisons=14 --param max-tracked-strlens=930 -O3 1 0 2 1 milepost-codelet-mibench-office-rsynth-src-nsynth-codelet-5-1 default 1 CUBOT H2 Android 5.1
S2 63b5b67c2a504b87 1.20 0.83 -O3 -fbtr-bb-exclusive -fcompare-elim -ffinite-math-only -fno-ivopts -fno-function-cse -fno-partial-inlining -fsched-spec-load -fselective-scheduling -fno-sel-sched-pipelining -fno-sel-sched-pipelining-outer-loops -fno-thread-jumps -ftree-dce -fno-tree-dominator-opts -fno-tree-partial-pre -funroll-loops -fno-unswitch-loops -fexcess-precision=fast --param predictable-branch-outcome=39 --param max-inline-insns-recursive-auto=378 --param max-variable-expansions-in-unroller=2 --param ipcp-unit-growth=14 --param large-stack-frame=458 --param max-peel-times=7 --param sms-max-ii-factor=67 --param sms-loop-average-count-threshold=0 --param max-crossjump-edges=135 --param iv-always-prune-cand-set-bound=3 --param scev-max-expr-size=139 --param omega-max-geqs=509 --param sink-frequency-threshold=55 --param selsched-max-lookahead=35 --param max-dse-active-local-stores=7149 --param prefetch-latency=311 --param ira-max-conflict-table-size=1547 --param max-vartrack-expr-depth=8 --param ipa-max-agg-items=3 --param allow-packed-load-data-races=1 --param asan-instrument-reads=1 -O3 1 0 2 1 milepost-codelet-mibench-automotive-basicmath-isqrt-codelet-1-1 default image-tiff-0001-nocomp data.tiff 1 ZTE BLADE V6 Android 5.0.2
S3 da5ad077d154f3e8 1.12 0.99 -O3 -fcompare-elim -fcprop-registers -fdevirtualize-speculatively -fmodulo-sched -fdefer-pop -fpeephole2 -fno-reciprocal-math -fno-reorder-blocks-and-partition -fno-reorder-functions -fno-tracer -ftree-reassoc -fno-tree-switch-conversion -falign-jumps=0 --param max-early-inliner-iterations=2 --param max-delay-slot-insn-search=140 --param large-unit-insns=15825 --param max-unswitch-insns=48 --param tracer-min-branch-probability=45 --param max-crossjump-edges=28 --param scev-max-expr-complexity=10 --param vect-max-version-for-alias-checks=12 --param vect-max-peeling-for-alignment=59 --param max-pipeline-region-blocks=29 --param selsched-insns-to-rename=0 --param simultaneous-prefetches=3 --param use-canonical-types=0 --param ira-max-conflict-table-size=861 --param case-values-threshold=0 --param allow-store-data-races=0 --param max-slsr-cand-scan=915924 --param asan-stack=1 --param asan-instrument-reads=0 --param asan-memintrin=1 -O3 1 0 2 1 milepost-codelet-mibench-security-pgp-d-src-mpilib-codelet-1-1 default 1300, 1300, 1300, 1300 1 CUBOT H2 Android 5.1



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

Developed by dividiti,
cTuning foundation,
and the community
          
Implemented as a CK workflow
                     
   
   
                      Hosted at