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This page is outdated!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 UIDbc6fc34b77848d4a
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:bc6fc34b77848d4a
CompilerGCC 4.9
CPUQualcomm MSM8974PRO-AC
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 b354f3806b5245ff 1.17 1.00 -O3 -fno-auto-inc-dec -fno-cse-follow-jumps -fno-inline-small-functions -fisolate-erroneous-paths-attribute -floop-strip-mine -fno-loop-parallelize-all -fno-defer-pop -fno-tree-ch -fno-tree-loop-im -ftree-partial-pre -fno-tree-vrp -fno-unroll-loops -finline-limit=0 -fira-algorithm=priority --param inline-min-speedup=19 --param max-inline-insns-auto=31 --param partial-inlining-entry-probability=43 --param max-variable-expansions-in-unroller=2 --param max-pending-list-length=37 --param inline-unit-growth=46 --param max-completely-peel-loop-nest-depth=16 --param max-unswitch-insns=25 --param sms-dfa-history=0 --param hot-bb-count-ws-permille=342 --param tracer-min-branch-probability=98 --param lim-expensive=19 --param omega-max-eqs=2 --param omega-max-keys=17 --param omega-eliminate-redundant-constraints=0 --param max-pipeline-region-insns=296 --param selsched-insns-to-rename=0 --param max-fields-for-field-sensitive=0 --param use-canonical-types=1 --param loop-invariant-max-bbs-in-loop=13517 --param max-vartrack-expr-depth=19 --param tree-reassoc-width=0 --param max-tail-merge-iterations=4 --param asan-use-after-return=0 -O3 1 0 2 1 milepost-codelet-mibench-security-pgp-e-src-mpilib-codelet-3-1 default 2457.6, 2457.6, 2457.6, 2457.6 1 SAMSUNG SM-G900F Android 5.0
S2 4f797f27b4b27a53 1.16 1.00 -O3 -fno-compare-elim -fcprop-registers -fdata-sections -fearly-inlining -fno-gcse-las -fno-hoist-adjacent-loads -fno-ipa-pure-const -fno-loop-interchange -fno-math-errno -fno-peephole2 -frename-registers -fsched2-use-superblocks -fno-sched-pressure -fschedule-insns -fno-tree-vrp --param inline-min-speedup=12 --param min-inline-recursive-probability=0 --param min-vect-loop-bound=1 --param max-modulo-backtrack-attempts=28 --param large-unit-insns=3672 --param large-stack-frame=286 --param gcse-unrestricted-cost=4 --param sms-min-sc=1 --param sms-loop-average-count-threshold=0 --param align-threshold=136 --param builtin-expect-probability=43 --param tracer-min-branch-ratio=81 --param tracer-min-branch-probability-feedback=2 --param max-crossjump-edges=39 --param max-goto-duplication-insns=16 --param max-cse-insns=186 --param iv-consider-all-candidates-bound=37 --param max-cselib-memory-locations=14 --param sched-state-edge-prob-cutoff=57 --param selsched-max-sched-times=3 --param min-size-for-stack-sharing=33 --param max-jump-thread-duplication-stmts=27 --param max-partial-antic-length=44 --param max-vartrack-size=36744139 --param tm-max-aggregate-size=4 --param ipa-cp-array-index-hint-bonus=47 --param lto-partitions=58 -O3 1 0 2 1 milepost-codelet-mibench-automotive-susan-s-src-susan-codelet-1-1 default 2457.6, 2457.6, 2457.6, 2457.6 1 ONEPLUS A0001 Android 7.0
S3 680dbd84e4fb7e57 1.09 1.00 -O3 -fno-caller-saves -fno-cx-limited-range -fdelete-null-pointer-checks -fno-if-conversion -fipa-cp-clone -fno-function-cse -fpeephole2 -fno-ree -fno-reorder-blocks -fsched-spec-insn-heuristic -fno-section-anchors -fsel-sched-pipelining -fno-split-wide-types -fno-tree-copy-prop -fno-tree-loop-distribute-patterns -fno-tree-loop-optimize -fno-tree-switch-conversion -ftree-tail-merge -fno-unroll-loops -falign-labels=0 --param max-inline-recursive-depth=1 --param min-vect-loop-bound=1 --param max-delay-slot-live-search=281 --param large-function-insns=5399 --param gcse-cost-distance-ratio=1 --param max-unrolled-insns=314 --param max-average-unrolled-insns=55 --param max-peel-times=12 --param sms-dfa-history=0 --param sms-loop-average-count-threshold=0 --param tracer-min-branch-probability=87 --param max-cse-insns=936 --param omega-max-keys=197 --param sched-mem-true-dep-cost=0 --param ira-max-conflict-table-size=1821 --param tm-max-aggregate-size=15 --param max-stores-to-sink=2 --param tree-reassoc-width=0 --param sched-pressure-algorithm=2 --param max-slsr-cand-scan=545000 --param asan-globals=1 -O3 1 0 2 1 milepost-codelet-mibench-consumer-tiffmedian-src-tiffmedian-codelet-5-1 default 2457.6, 2457.6, 2457.6, 2457.6 1 SAMSUNG SM-G900F Android 5.0



[ 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