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This page is outdated!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 UID5e81cccf744e58b9
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 8289e0cf24346aa7:5e81cccf744e58b9
CompilerGCC 4.9
CPU----
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 b05a460cb8244b37 1.64 1.00 -O3 -fno-cx-fortran-rules -fdata-sections -fhoist-adjacent-loads -fno-ipa-pta -fno-branch-count-reg -fno-function-cse -fno-partial-inlining -fno-peel-loops -fno-prefetch-loop-arrays -fno-reschedule-modulo-scheduled-loops -fno-shrink-wrap -fno-tree-loop-ivcanon -ftree-pta -ftree-reassoc -ftree-slsr -ftree-sra -fvariable-expansion-in-unroller --param max-modulo-backtrack-attempts=73 --param large-stack-frame-growth=1628 --param gcse-cost-distance-ratio=16 --param sms-dfa-history=0 --param unlikely-bb-count-fraction=6996 --param align-loop-iterations=3 --param max-predicted-iterations=87 --param scev-max-expr-size=167 --param scev-max-expr-complexity=13 --param omega-max-vars=18 --param max-dse-active-local-stores=1454 --param cxx-max-namespaces-for-diagnostic-help=550 -O3 1 0 2 1 milepost-codelet-mibench-automotive-bitcount-src-bitcnt-1-codelet-2-1 default 1 LGE NEXUS 5X Android 6.0.1
S2 4e719f826a77ab3b 1.57 1.00 -O3 -fbranch-probabilities -fno-dse -fforward-propagate -fno-if-conversion2 -fno-ipa-cp-clone -fmerge-constants -fno-modulo-sched-allow-regmoves -fpeel-loops -fno-sched-spec-insn-heuristic -fno-thread-jumps -ftree-bit-ccp -fno-tree-dominator-opts -fno-tree-phiprop -ftree-vectorize -falign-labels=0 -fexcess-precision=fast -finline-limit=0 --param max-variable-expansions-in-unroller=0 --param gcse-after-reload-partial-fraction=1 --param max-once-peeled-insns=692 --param hot-bb-frequency-fraction=132 --param min-crossjump-insns=4 --param omega-max-geqs=42 --param max-sched-region-insns=87 --param loop-max-datarefs-for-datadeps=1897 --param asan-memintrin=0 -O3 1 0 2 1 milepost-codelet-mibench-consumer-tiffmedian-src-tiffmedian-codelet-3-1 default 1 LGE NEXUS 5X Android 6.0.1
S3 d8d09bdf4bf667ca 1.13 1.00 -O3 -fno-branch-target-load-optimize -fdata-sections -fno-delayed-branch -fearly-inlining -fno-expensive-optimizations -finline-functions-called-once -fipa-cp -fipa-pure-const -fno-ira-share-save-slots -fno-isolate-erroneous-paths-dereference -floop-parallelize-all -fmath-errno -fno-signed-zeros -fno-rounding-math -fno-sched2-use-superblocks -fsched-critical-path-heuristic -fno-schedule-insns2 -fno-sel-sched-pipelining-outer-loops -fno-tree-dce -fno-tree-dominator-opts -fno-tree-loop-linear -ftree-vectorize --param max-peel-times=5 --param max-peel-branches=42 --param max-completely-peel-loop-nest-depth=12 --param max-iterations-to-track=1040 --param align-threshold=136 --param tracer-min-branch-probability-feedback=53 --param max-sched-region-insns=37 --param max-pipeline-region-insns=365 --param max-sched-extend-regions-iters=0 --param selsched-insns-to-rename=0 --param l1-cache-size=65 --param sccvn-max-alias-queries-per-access=1724 --param lra-max-considered-reload-pseudos=464 --param loop-max-datarefs-for-datadeps=307 --param max-vartrack-reverse-op-size=23 --param ipa-sra-ptr-growth-factor=4 --param asan-globals=0 -O3 1 0 2 1 milepost-codelet-mibench-automotive-susan-s-src-susan-codelet-1-1 default 1 LGE NEXUS 5X Android 6.0.1
S4 699c557e4041159a 1.10 1.00 -O3 -fno-check-data-deps -fcprop-registers -fexpensive-optimizations -fno-float-store -fno-function-sections -fgcse -fmerge-constants -fno-modulo-sched -fno-peephole2 -fsched-interblock -fsched-spec -ftrapping-math -fzero-initialized-in-bss -fno-rename-registers -fno-sched-spec-load -fsched-group-heuristic -fsection-anchors -fno-tree-bit-ccp -ftree-copy-prop -ftree-loop-optimize -fweb -fno-whole-program -ffp-contract=fast -fira-algorithm=priority -fsched-stalled-insns=0 --param max-inline-recursive-depth=11 --param large-function-growth=179 --param max-hoist-depth=0 --param max-peel-times=28 --param max-once-peeled-insns=569 --param hot-bb-count-ws-permille=829 --param min-crossjump-insns=10 --param vect-max-version-for-alignment-checks=7 --param max-dse-active-local-stores=4181 --param simultaneous-prefetches=0 --param graphite-max-nb-scop-params=12 --param prefetch-min-insn-to-mem-ratio=1 --param cxx-max-namespaces-for-diagnostic-help=651 -O3 1 0 2 1 milepost-codelet-mibench-security-pgp-d-src-mpilib-codelet-1-1 default 1 LGE NEXUS 5X Android 6.0.1
S5 2b913f6ee24d0945 1.10 1.00 -O3 -fcprop-registers -fgcse-las -fno-hoist-adjacent-loads -fif-conversion2 -findirect-inlining -fno-zero-initialized-in-bss -fno-omit-frame-pointer -fsched-spec-load -fno-sched-dep-count-heuristic -fsection-anchors -ftree-ccp -fno-tree-copyrename -ftree-dse -ftree-fre -fno-tree-loop-if-convert -falign-functions=0 -fira-algorithm=CB --param max-delay-slot-live-search=343 --param max-hoist-depth=50 --param max-peel-times=30 --param hot-bb-count-ws-permille=757 --param builtin-expect-probability=36 --param lim-expensive=20 --param omega-hash-table-size=266 --param vect-max-version-for-alignment-checks=12 --param max-sched-extend-regions-iters=0 --param sched-spec-prob-cutoff=55 --param ssp-buffer-size=7 --param ipa-max-agg-items=2 --param allow-packed-load-data-races=0 --param tree-reassoc-width=0 -O3 1 0 2 1 milepost-codelet-mibench-consumer-tiff2rgba-src-tif-predict-codelet-4-1 default 1 LGE NEXUS 5X Android 6.0.1



[ 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