Experimental results

Best optimization sequences (Don’t figure in the paper)

Below we present, in order, the list of involved optimizations as they have been applied through CLI:

fautoincdec fcompareelim fcprop-registers –fdce fdefer-pop –fdse fguess-branch-probability –fif-conversion2 –fif-conversion –fipa-pure-const fipa-profile –fipa-reference –fmerge-constants –fsplit-wide-types –ftree-bit-ccp ftreebuiltin-call-dce ftreeccp ftreech ftreecopyrename ftreedce ftreedominator-opts –ftreedse ftreeforwprop ftreefre ftreephiprop ftreeslsr ftreesra ftreepta ftreeter funit-at-a-time –fomit-frame-pointer –fthread-jumps –falign-functions –falign-jumps –falign-loops –falign-labels –fcaller-saves –fcrossjumping fcse-follow-jumps –fcse-skip-blocks –fdelete-null-pointer-checks –fdevirtualize fexpensive-optimizations –fgcse fgcselm fhoist-adjacent-loads –finline-small-functions –findirectinlining fipasra foptimize-sibling-calls –fpartialinlining -fpeephole2 –fregmove freorder-blocks –freorder-functions –freruncse-after-loop –fschedinterblock fschedspec fscheduleinsns fschedule-insns2 –fstrict-aliasing –fstrict-overflow –ftree-switch-conversion –ftree-tail-merge –ftreepre ftreevrp finline-functions –funswitch-loops –fpredictivecommoning fgcse-after-reload –ftreevectorize fvect-cost-model –ftree-partial-pre fipacp-clone –ffast-math

Thus, for each meta-heuristic we state the best discovered sequences. Each sequence is presented as a binary vector where 1 means optimization-on and 0 means optimization-off. The order as they have been described above is important.

In the following, we present the list of best optimization sets that led to best scores. We report the results of the mono-objective (RQ1) and multi-objective (RQ4) experiments:

RQ1: best optimization sequences

Time

Time RS:

110011110010011101010101101111101000101000100111000000100111001101000111000

Time GA

010010010010011011101101101011011100001000100101100011110011100110100011010

Time NS

100010110110110111100100000011000111000010110111010001000111000010101101111

Memory

Memory RS:

000110110110010001111101011101011101100001100101011110101110001001110101110

Memory GA:

110110010000110011010111100101010000010101110010100111001000010101011011110

Memory NS:

000110111110011101011111000110001110001000111110110100000010001001011100000

CPU

CPU RS:

010101101000011111111111101111001100110000010010010010101011011010001000001

CPU GA:

010101010111101011111110001110011110001011110011000101010010011011100001000

CPU NS:

11010110100101000000010110111010011000010000111011101101110110000010100010

RQ4: best optimization sequences

Pareto Front Solution NSGA-II

000111101000001111101111011111101000000010111011110111110100101111011001001

110101101101111000011110110011010101101010011010001011111110111100000110100

011111110011111111101111010100101011111111111011111010101110010111000110000

001111101001011100111111010101110100100110010110110001001110101010000101010

101101011010111100001111111010111100111001111111100011101111111101000011010

011110001001011001000111110101000111101110011110010111001011011110010011110

Pareto Front Solution NS-II

01110100111111100000011110000101000110110101010101101010110100101010110000

100001110110110011010000000011101100100001010110010010111011010010110110110

110001011110100011111101000001000000111000010011101101011110001010110101110

111101011100110011101110001001110011110001000011111001100010011111000000000