baseline/source/susan/input.c

#!/usr/bin/python3
from typing import List, Any
import numpy as np
from scipy import stats
import sys
import plotille.plotille as plt
TIMING_ERROR = 1000 #ns
LEVEL_C_ANALYSIS = False
from libSMT import *

def print_usage():
    print("This program takes in the all-pairs and baseline SMT data and computes how much each program is slowed when SMT in enabled.", file=sys.stderr)
    print("Level-A/B usage: {} <file -A> <file -B> <baseline file> --cij".format(sys.argv[0]), file=sys.stderr)
#    print("Level-C usage: {} <continuous pairs> <baseline file>".format(sys.argv[0]), file=sys.stderr)

# Check that we got the right number of parameters
if len(sys.argv) < 3:
    print_usage()
    exit()

if len(sys.argv) > 3:
    print("Analyzing results using Level-A/B methodology...")
else:
    print("Analyzing results using Level-C methodology...")
    LEVEL_C_ANALYSIS = True

assert_valid_input_files(sys.argv[1:-1], print_usage);

# Print Cij values rather than Mij
TIMES_ONLY = len(sys.argv) > 4 and "--cij" in sys.argv[4]
OK_PAIRS_ONLY = len(sys.argv) > 4 and "--cij-ok" in sys.argv[4]

# Pull in the data
if not LEVEL_C_ANALYSIS:
    baseline_times, baseline_sample_cnt, baseline_max_times = load_baseline(sys.argv[3])
    paired_times, paired_offsets, name_to_idx, idx_to_name = load_paired(sys.argv[1], sys.argv[2], len(list(baseline_times.keys())))
#    for key in baseline_times:
#        print(key,max(baseline_times[key]))
else:
    # Paired times use an abuse of the baseline file format
    baseline_times, baseline_sample_cnt, baseline_max_times = load_baseline(sys.argv[2])
    paired_times, name_to_idx, idx_to_name = load_fake_paired(sys.argv[1])

# We work iff the baseline was run for the same set of benchmarks as the pairs were
assert_base_and_pair_keys_match(baseline_times, name_to_idx)

# Only consider benchmarks that are at least an order of magnitude longer than the timing error
reliableNames = []
for i in range(0, len(name_to_idx)):
    benchmark = idx_to_name[i]
    if min(baseline_times[benchmark]) > TIMING_ERROR * 10:
        reliableNames.append(benchmark)

# Compute SMT slowdown for each benchmark
# Output format: table, each row is one benchmark and each column is one benchmark
#                each cell is base1 + base2*m = pair solved for m, aka (pair - base1) / base2
# Print table header
print("Bench          ", end=" ")
for name in reliableNames:
    if not TIMES_ONLY: print("{:<10.10}".format(name), end=" ")
    if TIMES_ONLY: print("{:<12.12}".format(name), end=" ")
print()
# Print rows
sample_f = max # Change this to np.mean to use mean values in Mij generation
M_vals = []
for b1 in reliableNames:
    if not TIMES_ONLY: print("{:<14.14}:".format(b1), end=" ")
    if TIMES_ONLY: print("{:<14.14}:".format(b1), end=" ")
    for b2 in reliableNames:
        if not LEVEL_C_ANALYSIS:
            Ci = max(sample_f(baseline_times[b1]), sample_f(baseline_times[b2]))
            Cj = min(sample_f(baseline_times[b1]), sample_f(baseline_times[b2]))
            Cij = sample_f(paired_times[name_to_idx[b1]][name_to_idx[b2]])
            if False:
                M = np.std(paired_times[name_to_idx[b1]][name_to_idx[b2]]) / np.mean(paired_times[name_to_idx[b1]][name_to_idx[b2]])
            else:
                M = (Cij - Ci) / Cj
            if Cij and Cj * 10 > Ci: # We don't pair tasks with more than a 10x difference in length
                M_vals.append(M)
                if not TIMES_ONLY: print("{:>10.3}".format(M), end=" ")
            else:
                if not TIMES_ONLY: print("{:>10}".format("N/A"), end=" ")

            if TIMES_ONLY and (not OK_PAIRS_ONLY or Cj * 10 > Ci):
                print("{:>12}".format(Cij), end=" ")
            elif OK_PAIRS_ONLY and Cj * 10 <= Ci:
                print("{:>12}".format("0"), end=" ")

        else:
            time_with_smt = sample_f(paired_times[name_to_idx[b1]][name_to_idx[b2]])
            time_wout_smt = sample_f(baseline_times[b1])
            M = time_with_smt / time_wout_smt
            M_vals.append(M)
            print("{:>10.3}".format(M), end=" ")
    print("")
# Print some statistics about the distribution
print("Overall average is {:>5.3} with standard deviation {:>5.3} using `{}`".format(np.mean(M_vals), np.std(M_vals), sample_f.__name__))
Ms = np.asarray(M_vals, dtype=np.float32)
if not LEVEL_C_ANALYSIS:
    print(np.sum(Ms <= 0), "of", len(M_vals), "M_i:j values are at most zero -", 100*np.sum(Ms <= 0)/len(M_vals), "percent")
    print(np.sum(Ms > 1), "of", len(M_vals), "M_i:j values are greater than one -", 100*np.sum(Ms > 1)/len(M_vals), "percent")
    M_vals_to_plot = Ms[np.logical_and(Ms > 0, Ms <= 1)]
    # Sims' analysis
    mean = np.mean(list(M_vals_to_plot))
    std = np.std(list(M_vals_to_plot))
    print("For {} of {} M_i:j values in (0, 1], average: {:>5.3} with std. dev. {:>4.3} (coeff. var. {:>4.3}) using `{}`".format(len(M_vals_to_plot), len(M_vals), mean, std, std/mean, sample_f.__name__))
else:
    print(np.sum(Ms <= 1), "of", len(M_vals), "M_i:j values are at most one -", 100*np.sum(Ms <= 1)/len(M_vals), "percent")
    print(np.sum(Ms > 2), "of", len(M_vals), "M_i:j values are greater than two -", 100*np.sum(Ms > 2)/len(M_vals), "percent")
    M_vals_to_plot = Ms
    # Sims' analysis
    print("For {} of {} M_i:j values in (1, 2], average: {:>5.3} with std. dev. {:>4.3} (coeff. var. {:>4.3}) using `{}`".format(len(M_vals_to_plot), len(M_vals), np.mean(list(M_vals_to_plot)), np.std(list(M_vals_to_plot)), np.mean(list(M_vals_to_plot))/np.std(list(M_vals_to_plot)), sample_f.__name__))

print(plt.hist(M_vals_to_plot, bins=10))

##### BELOW TEXT IS OLD OFFSET CODE (patched) #####
## This still works, but is hacky and deprecated ##
## PearsonR doesn't work though                  ##
if not LEVEL_C_ANALYSIS and False:
    benchmarkNames = idx_to_name
    benchmarkCount = len(benchmarkNames)
    numJobs = len(paired_times[0][0])

    reliableNames=["ndes", "cjpeg_wrbmp", "adpcm_enc", "cjpeg_transupp", "epic", "gsm_dec", "h264_dec", "huff_enc", "rijndael_enc", "rijndael_dec", "gsm_enc", "ammunition", "mpeg2"]

    #stats.pearsonr(time[b1][b2], oList),

    with open("weakRelPairs_offset.csv", mode="w+") as f3:
        print("Benchmark1", "Benchmark2", "minOffset", "maxOffset", "meanOffset", "meddOffset", "stdOffset", "minLength", "maxLength", sep=",", file=f3)
        for b1 in range (0, benchmarkCount):
            for b2 in range (0, benchmarkCount):
                if benchmarkNames[b1] in reliableNames and benchmarkNames[b2] in reliableNames:
                    #exclude last job due to inccurate timing
                    oList = paired_offsets[b1][b2][:numJobs-1]
                    jList = paired_times[b1][b2][:numJobs-1]
#                   plt.scatter(oList, jList)
#                   plt.title(benchmarkNames[b1] + ", " + benchmarkNames[b2])
#                   plt.show()
#                   print(benchmarkNames[b1], benchmarkNames[b2], min(oList), max(oList), np.mean(oList), np.median(oList), np.std(oList), stats.pearsonr(jList, oList), stats.spearmanr(jList, oList),  sep=",", file=f3)
                    print(benchmarkNames[b1], benchmarkNames[b2], min(oList), max(oList), np.mean(oList), np.median(oList), np.std(oList), min(jList), max(jList),  sep=",", file=f3)
"""
#with open("reliableGraphs.csv", mode="x") as f3:
        for b1 in range(0, benchmarkCount):
            for b2 in range(0, benchmarkCount):
                if benchmarkNames[b1] in reliableNames and benchmarkNames[b2] in reliableNames:
                    oList = offset[b1][b2][:numJobs - 1]
                    jList=time[b1][b2][:numJobs-1]
                    # offset, time scatterplot
                    plt.scatter(oList, jList)
                    plt.title(benchmarkNames[b1] + " " + benchmarkNames[b2] + " Offsets v. Time")
                    plt.show()
                    #time histogram
                    #plt.hist(jList, bins=10)
                    #plt.title(benchmarkNames[b1] + benchmarkNames[b2] + "Completion Times")
                    #plt.show()
                    #offset histogram
                    #plt.hist(oList, bins=10)
                    #plt.title(benchmarkNames[b1] + benchmarkNames[b2] + "Offsets")
                    #plt.show()
"""