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This page is outdated!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 UID0ea214b436591a7f
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 8289e0cf24346aa7:0ea214b436591a7f
CompilerGCC 4.9
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 588424cba8018bc8 1.35 0.79 -O3 -fassociative-math -fcheck-data-deps -fcx-fortran-rules -fno-delayed-branch -fgcse-after-reload -findirect-inlining -fno-inline-functions -finline-small-functions -fno-ipa-cp-clone -fisolate-erroneous-paths-attribute -fno-move-loop-invariants -fno-function-cse -freorder-blocks-and-partition -fno-selective-scheduling2 -ftree-loop-if-convert-stores -ftree-phiprop -ftree-loop-optimize -fno-tree-pre -funroll-all-loops --param large-function-insns=4293 --param gcse-after-reload-critical-fraction=12 --param gcse-cost-distance-ratio=12 --param max-hoist-depth=17 --param sms-max-ii-factor=196 --param sms-dfa-history=0 --param tracer-min-branch-probability-feedback=74 --param iv-max-considered-uses=300 --param scev-max-expr-size=138 --param min-spec-prob=5 --param sched-state-edge-prob-cutoff=64 --param max-jump-thread-duplication-stmts=24 --param l1-cache-line-size=25 --param loop-block-tile-size=97 --param allow-store-data-races=1 --param max-tail-merge-iterations=0 -O3 1 0 2 1 milepost-codelet-mibench-telecomm-adpcm-d-src-adpcm-codelet-1-1 default 1300, 1300 1 ALPS LP N-37 Android 6.0

[ 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