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 UID7ffa1fadc5bc6801
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 8289e0cf24346aa7:7ffa1fadc5bc6801
CompilerGCC 4.9
CPUQualcomm Technologies, Inc MSM8939
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 4124c928176ebda7 4.72 0.24 -O3 -fbranch-target-load-optimize -fno-branch-target-load-optimize2 -fno-caller-saves -fira-share-save-slots -fkeep-inline-functions -floop-interchange -flto -fpartial-inlining -fpredictive-commoning -freschedule-modulo-scheduled-loops -fsched2-use-superblocks -fno-tree-ccp -fno-tree-loop-distribute-patterns -ftree-loop-ivcanon -fno-tree-sra -fvpt -falign-jumps=0 -falign-loops=0 -fexcess-precision=standard --param gcse-unrestricted-cost=0 --param max-average-unrolled-insns=32 --param max-unswitch-insns=93 --param sms-max-ii-factor=176 --param builtin-expect-probability=37 --param min-crossjump-insns=6 --param lim-expensive=33 --param scev-max-expr-size=144 --param omega-max-geqs=71 --param vect-max-peeling-for-alignment=22 --param simultaneous-prefetches=3 --param max-partial-antic-length=30 --param sccvn-max-scc-size=7224 --param ipa-cp-array-index-hint-bonus=33 --param lto-min-partition=1947 --param asan-globals=0 -O3 1 0 2 1 milepost-codelet-mibench-office-ghostscript-src-gdevpbm-codelet-1-1 default image-tiff-0001-bw data.tiff 533.333, 533.333, 533.333, 533.333, 533.333, 533.333, 533.333, 533.333 1 SAMSUNG SM-A700YD Android 5.0.2
S2 7dabafe2cc003502 1.39 1.01 -O3 -fno-cse-skip-blocks -fno-function-sections -fgcse-lm -fno-ipa-pure-const -fno-ira-share-spill-slots -fmodulo-sched-allow-regmoves -fpeephole2 -fsched-spec -fno-signed-zeros -fno-toplevel-reorder -fno-omit-frame-pointer -foptimize-sibling-calls -fno-sched-critical-path-heuristic -fno-sched-last-insn-heuristic -fselective-scheduling -fno-selective-scheduling2 -fsignaling-nans -fno-strict-overflow -fno-tree-phiprop -ftree-loop-distribute-patterns -fno-tree-pre -ftree-vrp -finline-limit=0 --param min-inline-recursive-probability=5 --param max-pending-list-length=43 --param max-peel-times=30 --param max-completely-peeled-insns=150 --param max-iterations-to-track=1015 --param max-predicted-iterations=24 --param omega-eliminate-redundant-constraints=0 --param max-dse-active-local-stores=7900 --param simultaneous-prefetches=6 --param l1-cache-line-size=57 --param loop-block-tile-size=24 --param max-tracked-strlens=952 -O3 1 0 2 1 milepost-codelet-mibench-automotive-susan-e-src-susan-codelet-10-1 default image-tiff-0001-bw data.tiff 800, 800, 800, 800, 800, 800, 800, 800 1 SAMSUNG SM-A700YD Android 5.0.2
S3 641e1849eb271443 1.14 0.78 -O3 -fconserve-stack -fno-dse -ffast-math -fgcse -fgcse-las -fno-hoist-adjacent-loads -fif-conversion2 -fkeep-static-consts -fno-sched-interblock -fno-optimize-sibling-calls -fpredictive-commoning -fno-sched-rank-heuristic -fno-split-wide-types -fno-tree-sink -ftree-ter -funroll-loops -fno-unswitch-loops -fweb -falign-labels=0 -fira-region=one --param max-inline-recursive-depth=4 --param gcse-after-reload-critical-fraction=18 --param max-peel-branches=27 --param max-completely-peel-times=6 --param max-once-peeled-insns=353 --param sms-dfa-history=0 --param max-goto-duplication-insns=15 --param max-sched-extend-regions-iters=0 --param sccvn-max-scc-size=9248 --param max-vartrack-size=99792998 --param ipa-sra-ptr-growth-factor=0 --param ipa-cp-value-list-size=14 --param case-values-threshold=0 --param max-tracked-strlens=832 --param asan-instrument-writes=0 -O3 1 0 2 1 milepost-codelet-mibench-security-pgp-e-src-mpilib-codelet-4-1 default audio-wav-0001 data.wav 533.333, 533.333, 533.333, 533.333, 533.333, 533.333, 533.333, 533.333 1 SAMSUNG SM-A700YD Android 5.0.2
S4 ea226391a03e4112 1.11 1.00 -O3 -fcombine-stack-adjustments -fno-finite-math-only -finline-functions -fno-ira-share-spill-slots -floop-strip-mine -fdefer-pop -fno-peephole2 -fno-toplevel-reorder -fno-omit-frame-pointer -fpredictive-commoning -fsched-group-heuristic -fno-schedule-insns -fno-section-anchors -fno-sel-sched-pipelining -fno-tree-ch -ftree-loop-vectorize -fno-tree-reassoc -ftree-switch-conversion -falign-functions=0 --param max-early-inliner-iterations=1 --param sms-max-ii-factor=156 --param iv-always-prune-cand-set-bound=13 --param max-pipeline-region-insns=8 --param max-sched-insn-conflict-delay=8 --param sched-state-edge-prob-cutoff=1 --param selsched-max-lookahead=0 --param simultaneous-prefetches=5 --param max-tracked-strlens=1294 --param asan-memintrin=0 -O3 1 0 2 1 milepost-codelet-mibench-consumer-tiffdither-src-tiffdither-codelet-1-1 default SGEMM_NT SGEMM_NT_4x1.json 800, 800, 800, 800, 800, 800, 800, 800 1 SAMSUNG SM-A700YD Android 5.0.2



[ Participated users, platforms, OS, CPU, GPU, GPGPU, NN ] [ How to participate ] [ Slides ] [ Paper ] [ Android app ] [ dividiti ] [ 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