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This page is outdated!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 UID0d1f6de4662d4d69
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 8289e0cf24346aa7:0d1f6de4662d4d69
CompilerGCC 4.9
CPUQualcomm Technologies, Inc MSM8940
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 ffc237c4c82971c2 1.29 1.00 -O3 -fdce -fno-function-sections -fno-gcse-after-reload -fno-graphite-identity -fno-hoist-adjacent-loads -fno-indirect-inlining -fno-inline-functions -fira-share-spill-slots -fkeep-static-consts -fno-modulo-sched-allow-regmoves -fbranch-count-reg -fpredictive-commoning -fno-prefetch-loop-arrays -fno-reschedule-modulo-scheduled-loops -fno-sched-spec-load-dangerous -fno-sched-rank-heuristic -fno-schedule-insns2 -fno-tree-fre -ftree-pre -ftree-sink -fvect-cost-model --param max-variable-expansions-in-unroller=0 --param max-pending-list-length=15 --param max-completely-peel-loop-nest-depth=1 --param builtin-expect-probability=21 --param iv-consider-all-candidates-bound=52 --param iv-always-prune-cand-set-bound=20 --param omega-max-keys=224 --param sink-frequency-threshold=65 --param min-spec-prob=2 --param sched-state-edge-prob-cutoff=58 --param max-last-value-rtl=18759 --param l1-cache-size=79 --param switch-conversion-max-branch-ratio=8 -O3 1 0 2 1 milepost-codelet-mibench-automotive-basicmath-cubic-codelet-3-1 default 1094.4, 1094.4, 1094.4, 1094.4, 1094.4, 1094.4, 1094.4, 1094.4 1 UNKNOWN GENERIC_A53_32 Android 7.0
S2 136cb29f7c43c190 1.26 0.81 -O3 -fassociative-math -fcombine-stack-adjustments -fno-cprop-registers -fdse -fisolate-erroneous-paths-dereference -fno-live-range-shrinkage -floop-interchange -fno-modulo-sched-allow-regmoves -fno-peephole2 -fno-zero-initialized-in-bss -fpartial-inlining -fno-peel-loops -fprefetch-loop-arrays -fselective-scheduling -ftree-ch -ftree-partial-pre -fweb --param inline-min-speedup=20 --param comdat-sharing-probability=29 --param large-function-growth=54 --param max-iterations-to-track=399 --param sms-dfa-history=0 --param align-loop-iterations=1 --param max-cse-insns=653 --param omega-max-wild-cards=13 --param vect-max-version-for-alignment-checks=3 --param min-spec-prob=59 --param max-jump-thread-duplication-stmts=10 --param loop-block-tile-size=61 --param prefetch-min-insn-to-mem-ratio=4 --param max-vartrack-size=75899376 --param ipa-sra-ptr-growth-factor=2 --param asan-instrument-reads=1 -O3 1 0 2 1 milepost-codelet-mibench-consumer-jpeg-c-src-jchuff-codelet-9-1 default 768, 768, 768, 768, 768, 768, 768, 768 1 LGE LG-LS777 Android 7.0



[ 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