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This page is outdated!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 UID
8289e0cf24346aa7 (experiment.tune.compiler.flags.gcc.e)
Data UID
74eb196246a8b2a7
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 8289e0cf24346aa7:74eb196246a8b2a7
Compiler
GCC 4.9
CPU
sc8830
Objective
min
Improvement key IK1
Main kernel execution time speedup [min]
Improvement key IK2
Code 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
aad352e7f43ee591
2.41
0.27
-O3 -fno-associative-math -fno-branch-target-load-optimize2 -fipa-pta -fira-hoist-pressure -fira-share-save-slots -floop-block -fno-loop-interchange -flto -fmodulo-sched-allow-regmoves -fno-move-loop-invariants -fsched-interblock -fno-reciprocal-math -fno-rerun-cse-after-loop -fno-sched-spec-load -fsched-rank-heuristic -fsched-dep-count-heuristic -fsingle-precision-constant -fno-tree-partial-pre -ftree-sink -fno-use-linker-plugin -falign-jumps=0 --param predictable-branch-outcome=38 --param max-inline-insns-auto=29 --param max-inline-insns-recursive-auto=477 --param gcse-unrestricted-cost=0 --param max-average-unrolled-insns=125 --param max-peeled-insns=66 --param max-peel-branches=13 --param max-once-peeled-insns=763 --param max-iterations-to-track=1408 --param max-crossjump-edges=82 --param max-goto-duplication-insns=13 --param integer-share-limit=467 --param ira-max-conflict-table-size=1136 --param switch-conversion-max-branch-ratio=11 --param graphite-max-nb-scop-params=7
-O3
1
0
2
1
milepost-codelet-mibench-office-rsynth-src-nsynth-codelet-5-1
default
1
SAMSUNG GT-I9060I
Android 4.4.4
S2
7d0469cc814e68a7
1.63
1.07
-O3 -fno-dse -findirect-inlining -fira-hoist-pressure -fisolate-erroneous-paths-attribute -fno-keep-inline-functions -fno-merge-constants -fsched-spec -ftoplevel-reorder -fpeel-loops -fno-reorder-blocks-and-partition -fsched2-use-superblocks -fsched-last-insn-heuristic -fschedule-insns2 -fsel-sched-pipelining -fsingle-precision-constant -fno-strict-overflow -fno-tree-loop-im -fno-unroll-loops -fno-unsafe-loop-optimizations --param max-inline-insns-recursive-auto=141 --param min-vect-loop-bound=2 --param max-delay-slot-live-search=559 --param ipcp-unit-growth=18 --param max-peel-times=18 --param hot-bb-count-ws-permille=705 --param max-goto-duplication-insns=6 --param lim-expensive=33 --param omega-max-geqs=440 --param omega-max-wild-cards=8 --param sched-state-edge-prob-cutoff=29 --param selsched-max-lookahead=52 --param selsched-insns-to-rename=2 --param l1-cache-line-size=13 --param switch-conversion-max-branch-ratio=10 --param loop-block-tile-size=55 --param asan-use-after-return=0
-O3
1
0
2
1
milepost-codelet-mibench-automotive-susan-e-src-susan-codelet-10-1
default
1
SAMSUNG GT-I9060I
Android 4.4.4
S3
38055b840eb32a99
1.50
0.23
-O3 -fdelayed-branch -fdevirtualize-speculatively -ffunction-sections -fno-gcse -floop-interchange -floop-strip-mine -flto -fno-defer-pop -fguess-branch-probability -fno-peephole -fsched-spec -fzero-initialized-in-bss -fno-sched-critical-path-heuristic -fselective-scheduling -fno-single-precision-constant -fno-split-ivs-in-unroller -fno-tree-loop-distribution -fno-tree-pre -fno-tree-switch-conversion -falign-loops=0 -ffp-contract=off --param max-inline-insns-auto=66 --param min-inline-recursive-probability=9 --param max-variable-expansions-in-unroller=2 --param min-vect-loop-bound=2 --param large-function-growth=16 --param early-inlining-insns=11 --param gcse-after-reload-partial-fraction=2 --param gcse-cost-distance-ratio=15 --param max-completely-peel-times=23 --param hot-bb-frequency-fraction=1593 --param max-cse-path-length=17 --param omega-hash-table-size=894 --param max-sched-ready-insns=107 --param sccvn-max-alias-queries-per-access=534 --param slp-max-insns-in-bb=402 --param ipa-cp-loop-hint-bonus=85 --param ipa-cp-array-index-hint-bonus=85 --param lto-min-partition=848 --param allow-load-data-races=0 --param asan-stack=1
-O3
1
0
2
1
milepost-codelet-mibench-consumer-lame-src-takehiro-codelet-16-1
default
1
SAMSUNG GT-I9060I
Android 4.4.4
S4
d98b246379bab434
1.18
0.98
-O3 -fno-associative-math -fcombine-stack-adjustments -fconserve-stack -fcx-fortran-rules -fdata-sections -finline-functions -fisolate-erroneous-paths-attribute -fno-live-range-shrinkage -fno-loop-interchange -fmerge-all-constants -fbranch-count-reg -fno-sched-spec -fno-rename-registers -fno-selective-scheduling2 -fno-tracer -fno-tree-bit-ccp -ftree-ccp -fno-tree-vrp -fexcess-precision=standard --param comdat-sharing-probability=29 --param partial-inlining-entry-probability=111 --param max-pending-list-length=47 --param early-inlining-insns=9 --param max-hoist-depth=29 --param max-unrolled-insns=282 --param max-average-unrolled-insns=152 --param max-predicted-iterations=72 --param iv-consider-all-candidates-bound=17 --param min-spec-prob=18 --param min-size-for-stack-sharing=58 --param max-dse-active-local-stores=3139 --param max-vartrack-reverse-op-size=1 --param ipa-sra-ptr-growth-factor=3 --param ipa-cp-loop-hint-bonus=97 --param case-values-threshold=0 --param max-tail-merge-comparisons=19 --param asan-globals=0
-O3
1
0
2
1
milepost-codelet-mibench-automotive-susan-e-src-susan-codelet-2-1
default
1
SAMSUNG GT-I9060I
Android 4.4.4
S5
f3e1439792298e66
1.15
0.87
-O3 -fassociative-math -fcombine-stack-adjustments -fcrossjumping -fcse-skip-blocks -fdce -fira-share-spill-slots -fbranch-count-reg -fno-inline -fno-sched-last-insn-heuristic -ftracer -fno-tree-copy-prop -ftree-loop-ivcanon -ftree-loop-linear -fno-tree-slsr -fno-tree-switch-conversion -funsafe-loop-optimizations -funswitch-loops -ffp-contract=on -ftree-parallelize-loops=0 --param max-peel-times=9 --param sms-min-sc=4 --param max-goto-duplication-insns=11 --param vect-max-peeling-for-alignment=31 --param max-sched-region-insns=83 --param lra-max-considered-reload-pseudos=490 --param prefetch-min-insn-to-mem-ratio=3 --param ipa-sra-ptr-growth-factor=3 --param tree-reassoc-width=0 --param max-tracked-strlens=564 --param uninit-control-dep-attempts=1129
-O3
1
0
2
1
milepost-codelet-mibench-consumer-tiffmedian-src-tiffmedian-codelet-5-1
default
1
SAMSUNG GT-I9060I
Android 4.4.4
S6
c4982a809d3dc992
1.15
0.79
-O3 -fconserve-stack -fno-fat-lto-objects -fforward-propagate -finline-functions-called-once -fira-share-save-slots -fno-rerun-cse-after-loop -fsingle-precision-constant -fno-split-wide-types -fno-tree-reassoc -fno-tree-ter -funit-at-a-time -funroll-all-loops --param predictable-branch-outcome=11 --param max-inline-insns-recursive=720 --param gcse-after-reload-critical-fraction=17 --param max-unswitch-insns=57 --param max-grow-copy-bb-insns=16 --param lim-expensive=12 --param omega-max-vars=13 --param max-sched-region-insns=181 --param sched-state-edge-prob-cutoff=98 --param ssp-buffer-size=7 --param use-canonical-types=0 --param max-partial-antic-length=41 --param graphite-max-nb-scop-params=16 --param ipa-cp-eval-threshold=643 --param ipa-cp-array-index-hint-bonus=45
-O3
1
0
2
1
milepost-codelet-mibench-security-pgp-e-src-mpilib-codelet-3-1
default
1
SAMSUNG GT-I9060I
Android 4.4.4
S7
4838ad546ab0cfe3
1.11
0.81
-O3 -fcombine-stack-adjustments -fcse-skip-blocks -fdelete-null-pointer-checks -ffloat-store -finline-functions-called-once -fno-ira-loop-pressure -fno-inline -fno-peel-loops -fno-reorder-blocks-and-partition -fsched-spec-insn-heuristic -fsched-rank-heuristic -fschedule-insns -fno-single-precision-constant -ftree-bit-ccp -fno-tree-ccp -ftree-loop-if-convert-stores -ftree-ter -funroll-loops -fira-algorithm=priority --param max-modulo-backtrack-attempts=54 --param max-unrolled-insns=350 --param unlikely-bb-count-fraction=9753 --param lim-expensive=6 --param iv-max-considered-uses=43 --param simultaneous-prefetches=2 --param min-insn-to-prefetch-ratio=12 --param max-vartrack-reverse-op-size=94 --param ipa-sra-ptr-growth-factor=4 --param lto-min-partition=1312 --param allow-load-data-races=0 --param max-slsr-cand-scan=125968
-O3
1
0
2
1
milepost-codelet-mibench-automotive-basicmath-isqrt-codelet-1-1
default
1
SAMSUNG GT-I9060I
Android 4.4.4
S8
0a724957a565b254
1.11
1.23
-O3 -fno-branch-target-load-optimize -fbranch-target-load-optimize2 -fdelayed-branch -fno-fat-lto-objects -ffast-math -fif-conversion -fipa-pta -fisolate-erroneous-paths-attribute -fno-move-loop-invariants -fno-defer-pop -fno-inline -fno-rerun-cse-after-loop -frounding-math -fselective-scheduling -fno-sel-sched-pipelining -fsingle-precision-constant -ftree-copyrename -fno-unsafe-loop-optimizations -falign-functions=0 -falign-loops=0 -fexcess-precision=standard --param inline-min-speedup=10 --param max-early-inliner-iterations=0 --param large-function-growth=95 --param early-inlining-insns=15 --param max-completely-peeled-insns=43 --param max-unswitch-level=2 --param hot-bb-count-ws-permille=324 --param hot-bb-frequency-fraction=1699 --param align-threshold=58 --param max-grow-copy-bb-insns=2 --param omega-max-geqs=478 --param omega-max-wild-cards=29 --param omega-eliminate-redundant-constraints=0 --param sink-frequency-threshold=23 --param sched-mem-true-dep-cost=0 --param ssp-buffer-size=10 --param max-jump-thread-duplication-stmts=19 --param simultaneous-prefetches=3 --param tm-max-aggregate-size=11 --param allow-load-data-races=1 --param sched-pressure-algorithm=1
-O3
1
0
2
1
milepost-codelet-mibench-telecomm-gsm-src-short-term-codelet-2-1
default
1
SAMSUNG GT-I9060I
Android 4.4.4
[ 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
Go Back
Developed by
Grigori Fursin
Implemented as a
CK workflow
Hosted at