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This page is outdated!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 UID2cec2d4064764b2f
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 8289e0cf24346aa7:2cec2d4064764b2f
CompilerGCC 4.9
CPU0x41-8-0x1-0xd44-0
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 7de52a942be9eb7e 5.27 0.85 -O3 -fno-dse -fno-fat-lto-objects -fno-gcse-las -fno-ipa-cp-clone -fira-share-spill-slots -fno-loop-parallelize-all -fbranch-count-reg -fno-peephole2 -freciprocal-math -freorder-blocks -fno-sched-spec-load-dangerous -fno-sel-sched-pipelining -fsignaling-nans -ftracer -fno-tree-bit-ccp -ftree-copy-prop -ftree-dse -fno-tree-loop-distribution -ftree-slsr -fno-tree-vrp -finline-limit=0 -fsched-stalled-insns=0 --param max-inline-insns-auto=48 --param max-variable-expansions-in-unroller=0 --param inline-unit-growth=32 --param ipcp-unit-growth=11 --param max-unroll-times=9 --param max-peeled-insns=92 --param max-iterations-computation-cost=0 --param unlikely-bb-count-fraction=1089 --param iv-consider-all-candidates-bound=47 --param scev-max-expr-size=194 --param selsched-insns-to-rename=2 --param prefetch-latency=304 --param max-partial-antic-length=52 --param sccvn-max-scc-size=16644 --param sccvn-max-alias-queries-per-access=740 --param max-vartrack-expr-depth=4 --param max-stores-to-sink=3 -O3 1 0 2 1 milepost-codelet-mibench-consumer-mad-src-layer3-codelet-5-1 default 851, 851, 851, 851, 851, 851, 851, 851 1 GOOGLE PIXEL 6 Android 12
S2 1e14874b17fcdd10 1.62 0.77 -O3 -fno-branch-probabilities -fbranch-target-load-optimize -fno-cx-fortran-rules -ffloat-store -fno-indirect-inlining -fno-ipa-reference -fno-keep-inline-functions -frename-registers -fno-tree-copyrename -ftree-dse -ftree-forwprop -ftree-loop-im -ftree-loop-linear -fno-tree-pta -fno-tree-switch-conversion --param partial-inlining-entry-probability=11 --param max-variable-expansions-in-unroller=0 --param max-delay-slot-live-search=14 --param inline-unit-growth=39 --param large-stack-frame-growth=1934 --param tracer-min-branch-ratio=79 --param max-crossjump-edges=100 --param max-goto-duplication-insns=10 --param integer-share-limit=400 --param min-size-for-stack-sharing=28 --param ira-max-loops-num=79 --param ira-max-conflict-table-size=915 --param lra-max-considered-reload-pseudos=115 --param allow-store-data-races=1 --param allow-packed-store-data-races=0 --param max-slsr-cand-scan=514810 -O3 1 0 2 1 milepost-codelet-mibench-telecomm-fft-fourierf-codelet-3-1 default 500, 500, 500, 500, 500, 500, 500, 500 1 GOOGLE PIXEL 6 Android 12
S3 2d71f47268a64de8 1.30 1.03 -O3 -fbranch-target-load-optimize -fno-branch-target-load-optimize2 -fno-cse-skip-blocks -fno-function-sections -fgcse-lm -finline-small-functions -fisolate-erroneous-paths-attribute -floop-block -fdefer-pop -fno-function-cse -fno-toplevel-reorder -fno-reorder-functions -fno-thread-jumps -fno-tracer -ftree-copy-prop -ftree-loop-distribution -fno-tree-loop-optimize -fno-web -falign-jumps=0 --param max-delay-slot-live-search=114 --param large-function-growth=122 --param large-stack-frame-growth=338 --param max-average-unrolled-insns=2 --param max-unroll-times=1 --param max-iterations-computation-cost=8 --param sms-dfa-history=0 --param unlikely-bb-count-fraction=4745 --param omega-max-geqs=491 --param omega-max-wild-cards=7 --param omega-eliminate-redundant-constraints=0 --param max-last-value-rtl=18445 --param l1-cache-line-size=14 --param ira-max-conflict-table-size=785 --param loop-block-tile-size=73 --param graphite-max-nb-scop-params=9 --param loop-invariant-max-bbs-in-loop=17684 --param max-vartrack-size=67056745 --param ipa-sra-ptr-growth-factor=0 --param allow-packed-store-data-races=0 --param asan-stack=1 --param asan-use-after-return=1 -O3 1 0 2 1 milepost-codelet-mibench-consumer-jpeg-c-src-jchuff-codelet-9-1 default 500, 500, 500, 500, 500, 500, 500, 500 1 GOOGLE PIXEL 6 Android 12
S4 4de48a64991bd012 1.13 1.00 -O3 -fno-associative-math -fgraphite-identity -fif-conversion -fno-ipa-pure-const -floop-parallelize-all -fno-modulo-sched-allow-regmoves -fno-trapping-math -fno-omit-frame-pointer -fpredictive-commoning -fno-sched-spec-load-dangerous -fno-tree-dce -fno-tree-loop-if-convert-stores -fno-tree-loop-distribute-patterns -fno-tree-pre -ftree-reassoc -fno-tree-switch-conversion -fvect-cost-model -fno-whole-program -fira-algorithm=CB --param max-delay-slot-insn-search=35 --param large-function-insns=564 --param gcse-unrestricted-cost=2 --param max-unroll-times=14 --param max-cse-path-length=11 --param omega-max-geqs=16 --param ssp-buffer-size=10 --param loop-invariant-max-bbs-in-loop=3712 --param sched-pressure-algorithm=2 --param asan-instrument-reads=0 -O3 1 0 2 1 milepost-codelet-mibench-telecomm-gsm-src-short-term-codelet-2-1 default 500, 500, 500, 500, 500, 500, 500, 500 1 GOOGLE PIXEL 6 Android 12
S5 6e296f37ef47f094 1.12 0.30 -O3 -fexpensive-optimizations -ffinite-math-only -fno-gcse-after-reload -fif-conversion -findirect-inlining -finline-functions-called-once -fno-ipa-cp-clone -fira-loop-pressure -floop-parallelize-all -flto -fno-inline -fmath-errno -free -freorder-blocks -fno-sched2-use-superblocks -fno-sched-critical-path-heuristic -fno-signaling-nans -fno-tree-coalesce-vars -ftree-loop-vectorize -fno-tree-pre -funswitch-loops -fno-vect-cost-model -fweb --param max-inline-insns-recursive=119 --param min-inline-recursive-probability=18 --param max-early-inliner-iterations=0 --param max-gcse-insertion-ratio=24 --param gcse-cost-distance-ratio=17 --param max-completely-peeled-insns=92 --param max-completely-peel-loop-nest-depth=5 --param sms-loop-average-count-threshold=0 --param omega-eliminate-redundant-constraints=1 --param vect-max-version-for-alias-checks=10 --param max-sched-extend-regions-iters=0 --param selsched-max-lookahead=63 --param l1-cache-size=81 --param ipa-sra-ptr-growth-factor=4 --param ipa-cp-eval-threshold=589 --param max-stores-to-sink=4 --param sched-pressure-algorithm=2 -O3 1 0 2 1 milepost-codelet-mibench-consumer-tiffmedian-src-tiffmedian-codelet-5-1 default 500, 500, 500, 500, 500, 500, 500, 500 1 GOOGLE PIXEL 6 Android 12
S6 41f67d4491abd89f 1.12 1.00 -O3 -fcheck-data-deps -fno-compare-elim -fno-cx-fortran-rules -fforward-propagate -fgcse -fif-conversion2 -fno-indirect-inlining -finline-functions-called-once -fno-ipa-reference -fkeep-static-consts -ftrapping-math -fsched-spec-insn-heuristic -fsched-dep-count-heuristic -fshrink-wrap -fsplit-wide-types -ftree-bit-ccp -ftree-loop-linear -falign-jumps=0 -falign-loops=0 -fira-algorithm=CB --param max-inline-insns-auto=43 --param min-vect-loop-bound=1 --param max-iterations-to-track=1376 --param tracer-min-branch-probability-feedback=91 --param iv-max-considered-uses=234 --param omega-max-geqs=464 --param max-reload-search-insns=163 --param sink-frequency-threshold=32 --param min-spec-prob=52 --param simultaneous-prefetches=5 --param l1-cache-size=4 --param sccvn-max-scc-size=2810 --param loop-block-tile-size=20 --param ipa-cp-value-list-size=2 --param lto-min-partition=53 --param allow-load-data-races=0 --param max-tracked-strlens=736 -O3 1 0 2 1 milepost-codelet-mibench-consumer-lame-src-takehiro-codelet-16-1 default 500, 500, 500, 500, 500, 500, 500, 500 1 GOOGLE PIXEL 6 Android 12
S7 b0daf20c2e712a90 1.11 0.77 -O3 -fno-cx-fortran-rules -fdevirtualize-speculatively -fexpensive-optimizations -fgcse -fgcse-after-reload -fgcse-las -fno-ira-hoist-pressure -fira-share-spill-slots -floop-block -floop-interchange -floop-strip-mine -ftrapping-math -freorder-blocks-and-partition -fno-sel-sched-pipelining -fsel-sched-pipelining-outer-loops -fno-tree-loop-if-convert -ftree-loop-vectorize -funit-at-a-time -funroll-all-loops -fno-vect-cost-model -ffp-contract=off -fira-region=one --param max-inline-recursive-depth-auto=10 --param partial-inlining-entry-probability=124 --param max-delay-slot-live-search=420 --param max-peel-times=28 --param max-once-peeled-insns=465 --param max-goto-duplication-insns=5 --param iv-consider-all-candidates-bound=29 --param scev-max-expr-complexity=18 --param sched-mem-true-dep-cost=1 --param graphite-max-bbs-per-function=37 --param loop-max-datarefs-for-datadeps=1377 --param min-insn-to-prefetch-ratio=4 --param tm-max-aggregate-size=3 --param max-tracked-strlens=633 -O3 1 0 2 1 milepost-codelet-mibench-security-pgp-e-src-mpilib-codelet-1-1 default 500, 500, 500, 500, 500, 500, 500, 500 1 GOOGLE PIXEL 6 Android 12
S8 0392f8faf8ad8ab7 1.10 1.00 -O3 -fcse-follow-jumps -fno-cse-skip-blocks -fno-dse -fno-fast-math -fno-forward-propagate -fno-isolate-erroneous-paths-dereference -fno-loop-interchange -fno-defer-pop -ffunction-cse -fmath-errno -fno-signed-zeros -fpeel-loops -frename-registers -freorder-functions -fsched-last-insn-heuristic -fselective-scheduling -fno-single-precision-constant -ftree-fre -fno-tree-loop-if-convert -fno-tree-ter -fno-web -fsched-stalled-insns=0 --param large-function-growth=11 --param iv-consider-all-candidates-bound=35 --param iv-max-considered-uses=42 --param l1-cache-line-size=11 --param max-vartrack-size=59276653 --param ipa-sra-ptr-growth-factor=0 --param allow-packed-load-data-races=0 --param asan-use-after-return=0 -O3 1 0 2 1 milepost-codelet-mibench-security-pgp-d-src-mpilib-codelet-1-1 default 500, 500, 500, 500, 500, 500, 500, 500 1 GOOGLE PIXEL 6 Android 12



[ 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