#
#   This file is valid gnuplot input.
#
# 
# Attached, please find the latest canopy.  There are four plots.  Note
# that a "NOP" below means "set flags a=a, b=b".  The "a" flag was
# CLEARED throughout all experiments.
# 
#   norep-noskip:                          a MOVE followed by n+1 unconditional NOPs
#   norep-skip:                            a MOVE followed by n+1 "only if a is set" NOPs
#     rep-noskip: a set ILC=31 followed by a MOVE followed by n   unconditional NOPs
#     rep-skip:   a set ILC=31 followed by a MOVE followed by n   "only if a is set" NOPs
# 
# Based on these data, I conclude that the "pipeline-parallel skip
# capability" is certainly effective IF the instruction ahead of the
# instruction being skip requires STRICTLY MORE THAN one GasP cycle to
# execute.  However -- for reasons which I will explain on the telephone
# -- I believe this will not actually happen very often in Fleet,
# particularly if the compiler takes care to re-order instructions when
# possible.  I think we should certainly highlight this fascinating
# feature in our writeup of this experiment, but I suggest we consider
# omitting the capability from Dock2 in the interest of simplicity.
# 

set terminal pdf
set title "Marina Canopy Plot, Pipeline-Parallel Skips"
set output "pipeline-parallel-skip.pdf"
set xlabel "number of instructions in ring"
set ylabel "BILLIONS of instructions executed per second"

#set arrow 1 from 1.5,3.716 to 12,3.716 nohead linetype 4 lc rgb "#000000"
#set label 1 "3.7 GOP/s" at 0.3,3.7 left

#set arrow 2 from 1.5,7.67 to 10,7.67 nohead linetype 4 lc rgb "#000000"
#set label 2 "7.6 GOP/s" at 0.3,7.67 left

plot \
     '-' using 2:3 title "rep skip"     with linespoints lw 6 pt 6  lc rgb "#00dddd", \
     '-' using 2:3 title "rep noskip"   with linespoints lw 6 pt 6  lc rgb "#dd0000", \
     '-' using 2:3 title "norep skip"   with linespoints lw 6 pt 6  lc rgb "#0000dd", \
     '-' using 2:3 title "norep noskip" with linespoints lw 6 pt 6  lc rgb "#00dd00"

1.0 0 0.1096289284
1.0 1 0.16435084520000004
1.0 2 0.21920466
1.0 3 0.274011772
1.0 4 0.3288507906
1.0 5 0.38359662660000005
1.0 6 0.4385278694
1.0 7 0.4933551658
1.0 8 0.5475802388
1.0 9 0.5744030024
1.0 10 0.0
1.0 11 0.0
1.0 12 0.0
e

1.0 0 0.10963330019999999
1.0 1 0.15925804440000002
1.0 2 0.20586563360000004
1.0 3 0.2498035266
1.0 4 0.2910410806
1.0 5 0.329998314
1.0 6 0.367113143
1.0 7 0.4019697946
1.0 8 0.4110008090000001
1.0 9 0.3354915276
1.0 10 0.0
1.0 11 0.0
1.0 12 0.0
e

1.0 0 0.8537532144
1.0 1 1.2720878452
1.0 2 1.6987759322000002
1.0 3 1.7110399514
1.0 4 1.716044852
1.0 5 1.7201504912
1.0 6 1.7232913534
1.0 7 1.7256210036000001
1.0 8 1.3012082484
1.0 9 0.6538585483999999
1.0 10 0.0
1.0 11 0.0
1.0 12 0.0
e

1.0 0 0.8442314019999999
1.0 1 1.2784833081999998
1.0 2 1.6776937616
1.0 3 1.6792106535999998
1.0 4 1.6814821098000003
1.0 5 1.6825615937999998
1.0 6 1.6839861894
1.0 7 1.6847104672
1.0 8 1.3131872088
1.0 9 0.6592053592
1.0 10 0.0
1.0 11 0.0
1.0 12 0.0
e