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This page is outdated!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 LLVM compiler flags (minimize execution time))
Scenario UID
2aaed4c520956635 (experiment.tune.compiler.flags.llvm.e)
Data UID
e1b341196a5887aa
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 2aaed4c520956635:e1b341196a5887aa
Compiler
LLVM 4.0.0
CPU
AMD E1-2500 APU with Radeon(TM) HD Graphics
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
24b951bc94d4cab9
1.63
1.16
-O3 -ffast-math -fslp-vectorize-aggressive -msoft-float
-O3
1
1
6
35
milepost-codelet-mibench-automotive-basicmath-cubic-codelet-3-1
default
1400
2
Acer V1.06 (Aspire ES1-520)
Windows 10
S2
82250e92b4e55a3b
1.22
1.11
-O3 -fno-unroll-loops
-O3
1
0
8
36
milepost-codelet-mibench-network-dijkstra-src-dijkstra-large-codelet-5-1
default
1400
2
Acer V1.06 (Aspire ES1-520)
Windows 10
[ 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