def time_ak_coargsort(N_per_locale, trials, dtype, seed):
print(">>> arkouda {} coargsort".format(dtype))
cfg = ak.get_config()
N = N_per_locale * cfg["numLocales"]
print("numLocales = {}, N = {:,}".format(cfg["numLocales"], N))
for numArrays in (1, 2, 8, 16):
if seed is None:
seeds = [None for _ in range(numArrays)]
else:
seeds = [seed+i for i in range(numArrays)]
if dtype == 'int64':
arrs = [ak.randint(0, 2**32, N//numArrays, seed=s) for s in seeds]
nbytes = sum(a.size * a.itemsize for a in arrs)
elif dtype == 'float64':
arrs = [ak.randint(0, 1, N//numArrays, dtype=ak.float64, seed=s) for s in seeds]
nbytes = sum(a.size * a.itemsize for a in arrs)
elif dtype == 'str':
arrs = [ak.random_strings_uniform(1, 8, N//numArrays, seed=s) for s in seeds]
nbytes = sum(a.bytes.size * a.bytes.itemsize for a in arrs)
timings = []
for i in range(trials):
start = time.time()
perm = ak.coargsort(arrs)
end = time.time()
timings.append(end - start)
tavg = sum(timings) / trials
a = arrs[0][perm]
if dtype in ('int64', 'float64'):
assert ak.is_sorted(a)
print("{}-array Average time = {:.4f} sec".format(numArrays, tavg))
bytes_per_sec = nbytes / tavg
print("{}-array Average rate = {:.4f} GiB/sec".format(numArrays, bytes_per_sec/2**30))
def time_ak_scatter(isize, vsize, trials, dtype, random):
print(">>> arkouda scatter")
cfg = ak.get_config()
Ni = isize * cfg["numLocales"]
Nv = vsize * cfg["numLocales"]
print("numLocales = {}, num_indices = {:,} ; num_values = {:,}".format(
cfg["numLocales"], Ni, Nv))
# Index vector is always random
i = ak.randint(0, Nv, Ni)
c = ak.zeros(Nv, dtype=dtype)
if random:
if dtype == 'int64':
v = ak.randint(0, 2**32, Ni)
elif dtype == 'float64':
v = ak.randint(0, 1, Ni, dtype=ak.float64)
else:
v = ak.ones(Ni, dtype=dtype)
timings = []
for _ in range(trials):
start = time.time()
c[i] = v
end = time.time()
timings.append(end - start)
tavg = sum(timings) / trials
print("Average time = {:.4f} sec".format(tavg))
bytes_per_sec = (i.size * i.itemsize * 3) / tavg
print("Average rate = {:.2f} GiB/sec".format(bytes_per_sec / 2**30))
def time_ak_argsort(N_per_locale, trials, dtype, scale_by_locales):
print(">>> arkouda argsort")
cfg = ak.get_config()
if scale_by_locales:
N = N_per_locale * cfg["numLocales"]
else:
N = N_per_locale
print("numLocales = {}, N = {:,}".format(cfg["numLocales"], N))
if dtype == 'int64':
a = ak.randint(0, 2**32, N)
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64)
timings = []
for i in range(trials):
start = time.time()
perm = ak.argsort(a)
end = time.time()
timings.append(end - start)
tavg = sum(timings) / trials
assert ak.is_sorted(a[perm])
print("Average time = {:.4f} sec".format(tavg))
bytes_per_sec = (a.size * a.itemsize) / tavg
print("Average rate = {:.4f} GiB/sec".format(bytes_per_sec/2**30))
def time_ak_reduce(N_per_locale, trials, dtype, random):
print(">>> arkouda reduce")
cfg = ak.get_config()
N = N_per_locale * cfg["numLocales"]
print("numLocales = {}, N = {:,}".format(cfg["numLocales"], N))
if random:
if dtype == 'int64':
a = ak.randint(0, 2**32, N)
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64)
else:
a = ak.arange(0, N, 1)
if dtype == 'float64':
a = 1.0 * a
timings = {op: [] for op in OPS}
results = {}
for i in range(trials):
for op in timings.keys():
fxn = getattr(a, op)
start = time.time()
r = fxn()
end = time.time()
timings[op].append(end - start)
results[op] = r
tavg = {op: sum(t) / trials for op, t in timings.items()}
for op, t in tavg.items():
print("{} = {}".format(op, results[op]))
print(" Average time = {:.4f} sec".format(t))
bytes_per_sec = (a.size * a.itemsize) / t
print(" Average rate = {:.2f} GiB/sec".format(bytes_per_sec / 2**30))
def time_ak_write_read(N_per_locale, trials, dtype, path, seed):
print(">>> arkouda {} write/read".format(dtype))
cfg = ak.get_config()
N = N_per_locale * cfg["numLocales"]
print("numLocales = {}, N = {:,}".format(cfg["numLocales"], N))
if dtype == 'int64':
a = ak.randint(0, 2**32, N, seed=seed)
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64, seed=seed)
writetimes = []
readtimes = []
for i in range(trials):
start = time.time()
a.save(path)
end = time.time()
writetimes.append(end - start)
start = time.time()
b = ak.load(path)
end = time.time()
readtimes.append(end - start)
for f in glob(path + '_LOCALE*'):
os.remove(f)
avgwrite = sum(writetimes) / trials
avgread = sum(readtimes) / trials
print("write Average time = {:.4f} sec".format(avgwrite))
print("read Average time = {:.4f} sec".format(avgread))
nb = a.size * a.itemsize
print("write Average rate = {:.2f} GiB/sec".format(nb / 2**30 / avgwrite))
print("read Average rate = {:.2f} GiB/sec".format(nb / 2**30 / avgread))
def time_ak_coargsort(N_per_locale, trials, dtype):
print(">>> arkouda coargsort")
cfg = ak.get_config()
N = N_per_locale * cfg["numLocales"]
print("numLocales = {}, N = {:,}".format(cfg["numLocales"], N))
for numArrays in (1, 2, 8, 16):
if dtype == 'int64':
arrs = [
ak.randint(0, 2**32, N // numArrays) for _ in range(numArrays)
]
elif dtype == 'float64':
arrs = [
ak.randint(0, 1, N // numArrays, dtype=ak.float64)
for _ in range(numArrays)
]
timings = []
for i in range(trials):
start = time.time()
perm = ak.coargsort(arrs)
end = time.time()
timings.append(end - start)
tavg = sum(timings) / trials
a = arrs[0][perm]
assert ak.is_sorted(a)
print("{}-array Average time = {:.4f} sec".format(numArrays, tavg))
bytes_per_sec = sum(a.size * a.itemsize for a in arrs) / tavg
print("{}-array Average rate = {:.4f} GiB/sec".format(
numArrays, bytes_per_sec / 2**30))
def time_ak_argsort(N_per_locale, trials, dtype, seed):
print(">>> arkouda {} argsort".format(dtype))
cfg = ak.get_config()
N = N_per_locale * cfg["numLocales"]
print("numLocales = {}, N = {:,}".format(cfg["numLocales"], N))
if dtype == 'int64':
a = ak.randint(0, 2**32, N, seed=seed)
nbytes = a.size * a.itemsize
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64, seed=seed)
nbytes = a.size * a.itemsize
elif dtype == 'str':
a = ak.random_strings_uniform(1, 16, N, seed=seed)
nbytes = a.nbytes * a.entry.itemsize
timings = []
for i in range(trials):
start = time.time()
perm = ak.argsort(a)
end = time.time()
timings.append(end - start)
tavg = sum(timings) / trials
if dtype in ('int64', 'float64'):
assert ak.is_sorted(a[perm])
print("Average time = {:.4f} sec".format(tavg))
bytes_per_sec = nbytes / tavg
print("Average rate = {:.4f} GiB/sec".format(bytes_per_sec / 2**30))
def time_ak_stream(N_per_locale, trials, alpha, dtype, random):
print(">>> arkouda stream")
cfg = ak.get_config()
N = N_per_locale * cfg["numLocales"]
print("numLocales = {}, N = {:,}".format(cfg["numLocales"], N))
if random:
if dtype == 'int64':
a = ak.randint(0, 2**32, N)
b = ak.randint(0, 2**32, N)
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64)
b = ak.randint(0, 1, N, dtype=ak.float64)
else:
a = ak.ones(N, dtype=dtype)
b = ak.ones(N, dtype=dtype)
timings = []
for i in range(trials):
start = time.time()
c = a + b * alpha
end = time.time()
timings.append(end - start)
tavg = sum(timings) / trials
print("Average time = {:.4f} sec".format(tavg))
bytes_per_sec = (c.size * c.itemsize * 3) / tavg
print("Average rate = {:.2f} GiB/sec".format(bytes_per_sec / 2**30))
def time_ak_write(N_per_locale, numfiles, trials, dtype, path, seed, parquet):
print(">>> arkouda {} write".format(dtype))
cfg = ak.get_config()
N = N_per_locale * cfg["numLocales"]
print("numLocales = {}, N = {:,}, filesPerLoc = {}".format(
cfg["numLocales"], N, numfiles))
if dtype == 'int64':
a = ak.randint(0, 2**32, N, seed=seed)
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64, seed=seed)
writetimes = []
for i in range(trials):
for j in range(numfiles):
start = time.time()
a.save(f"{path}{j:04}") if not parquet else a.save_parquet(
f"{path}{j:04}")
end = time.time()
writetimes.append(end - start)
avgwrite = sum(writetimes) / trials
print("write Average time = {:.4f} sec".format(avgwrite))
nb = a.size * a.itemsize * numfiles
print("write Average rate = {:.2f} GiB/sec".format(nb / 2**30 / avgwrite))
def compare_strategies(length, ncat, op, dtype):
keys = ak.randint(0, ncat, length)
if dtype == 'int64':
vals = ak.randint(0, length // ncat, length)
elif dtype == 'bool':
vals = ak.zeros(length, dtype='bool')
for i in np.random.randint(0, length, ncat // 2):
vals[i] = True
else:
vals = ak.linspace(-1, 1, length)
print("Global groupby", end=' ')
start = time()
gg = ak.GroupBy(keys, False)
ggtime = time() - start
print(ggtime)
print("Global reduce", end=' ')
start = time()
gk, gv = gg.aggregate(vals, op)
grtime = time() - start
print(grtime)
print("Local groupby", end=' ')
start = time()
lg = ak.GroupBy(keys, True)
lgtime = time() - start
print(lgtime)
print("Local reduce", end=' ')
start = time()
lk, lv = lg.aggregate(vals, op)
lrtime = time() - start
print(lrtime)
print(f"Keys match? {(gk == lk).all()}")
print(f"Absolute diff of vals = {ak.abs(gv - lv).sum()}")
return ggtime, grtime, lgtime, lrtime
def time_ak_setops(N_per_locale, trials, dtype, seed):
print(">>> arkouda {} setops".format(dtype))
cfg = ak.get_config()
N = N_per_locale * cfg["numLocales"]
print("numLocales = {}, N = {:,}".format(cfg["numLocales"], N))
if dtype == 'int64':
a = ak.randint(0, 2**32, N, seed=seed)
b = ak.randint(0, 2**32, N, seed=seed)
timings = {op: [] for op in OPS}
results = {}
for i in range(trials):
for op in timings.keys():
fxn = getattr(ak, op)
start = time.time()
r = fxn(a, b)
end = time.time()
timings[op].append(end - start)
results[op] = r
tavg = {op: sum(t) / trials for op, t in timings.items()}
for op, t in tavg.items():
print(" {} Average time = {:.4f} sec".format(op, t))
bytes_per_sec = (a.size * a.itemsize * 2) / t
print(" {} Average rate = {:.2f} GiB/sec".format(
op, bytes_per_sec / 2**30))
def generate_arrays(N, seed):
# Sort keys so that aggregations will not have to permute values
# We just want to measure aggregation time, not gather
keys = ak.sort(ak.randint(0, 2**32, N, seed=seed))
if seed is not None: seed += 1
intvals = ak.randint(0, 2**16, N, seed=seed)
boolvals = (intvals % 2) == 0
return keys, intvals, boolvals
def check_correctness(dtype):
N = 10**4
if dtype == 'int64':
a = ak.randint(0, 2**32, N)
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64)
perm = ak.argsort(a)
assert ak.is_sorted(a[perm])
def check_int_float(N):
f = ak.randint(0, 2**63, N, dtype=ak.float64)
i = ak.randint(0, 2**63, N, dtype=ak.int64)
perm = ak.coargsort([f, i])
assert ak.is_sorted(f[perm])
perm = ak.coargsort([i, f])
assert ak.is_sorted(i[perm])
def generate_arrays(length, nkeys, nvals, dtype='int64'):
keys = ak.randint(0, nkeys, length)
if dtype == 'int64':
vals = ak.randint(0, nvals, length)
elif dtype == 'bool':
vals = ak.zeros(length, dtype='bool')
for i in np.random.randint(0, length, nkeys // 2):
vals[i] = True
else:
vals = ak.linspace(-1, 1, length)
return keys, vals
def check_correctness(dtype, seed):
N = 10**4
if dtype == 'int64':
a = ak.randint(0, 2**32, N, seed=seed)
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64, seed=seed)
elif dtype == 'str':
a = ak.random_strings_uniform(1, 16, N, seed=seed)
perm = ak.argsort(a)
if dtype in ('int64', 'float64'):
assert ak.is_sorted(a[perm])
def check_correctness(dtype, path, seed):
N = 10**4
if dtype == 'int64':
a = ak.randint(0, 2**32, N, seed=seed)
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64, seed=seed)
a.save(path)
b = ak.load(path)
for f in glob(path + "_LOCALE*"):
os.remove(f)
assert (a == b).all()
def check_correctness(dtype):
N = 10**4
if dtype == 'int64':
a = ak.randint(0, 2**32, N)
z = ak.zeros(N, dtype=dtype)
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64)
z = ak.zeros(N, dtype=dtype)
perm = ak.coargsort([a, z])
assert ak.is_sorted(a[perm])
perm = ak.coargsort([z, a])
assert ak.is_sorted(a[perm])
def time_ak_gather(isize, vsize, trials, dtype, random):
print(">>> arkouda gather")
cfg = ak.get_config()
Ni = isize * cfg["numLocales"]
Nv = vsize * cfg["numLocales"]
print("numLocales = {}, num_indices = {:,} ; num_values = {:,}".format(cfg["numLocales"], Ni, Nv))
# Index vector is always random
i = ak.randint(0, Nv, Ni)
if random:
if dtype == 'int64':
v = ak.randint(0, 2**32, Nv)
elif dtype == 'float64':
v = ak.randint(0, 1, Nv, dtype=ak.float64)
elif dtype == 'str':
v = ak.random_strings_uniform(1, 16, Nv)
else:
if dtype == 'str':
v = ak.random_strings_uniform(1, 16, Nv)
else:
v = ak.ones(Nv, dtype=dtype)
print("v={}".format(v))
print("v.offsets={}".format(v.offsets))
print("v.nbytes={}".format(v.nbytes))
print("v[1]={}".format(v[1]))
print("In Gather size={}".format(v.size))
print("In Gather nbytes={}".format(v.nbytes))
print("In Gather ndim={}".format(v.ndim))
print("In Gather shape={}".format(v.shape))
print("In Gather offsets name ={}".format(v.offsets.name))
print("In Gather offsets size={}".format(v.offsets.size))
print("In Gather bytes name ={}".format(v.bytes.name))
print("In Gather bytes size={}".format(v.bytes.size))
timings = []
for _ in range(trials):
print("In Gather loop i={}".format(i))
print("In Gather v[i]={}".format(v[i]))
start = time.time()
c = v[i]
end = time.time()
print("In Gather loop c={}".format(c))
timings.append(end - start)
tavg = sum(timings) / trials
print("Average time = {:.4f} sec".format(tavg))
if dtype == 'str':
offsets_transferred = 3 * c.offsets.size * c.offsets.itemsize
bytes_transferred = (c.offsets.size * c.offsets.itemsize) + (2 * c.bytes.size)
bytes_per_sec = (offsets_transferred + bytes_transferred) / tavg
else:
bytes_per_sec = (c.size * c.itemsize * 3) / tavg
print("Average rate = {:.2f} GiB/sec".format(bytes_per_sec/2**30))
def time_ak_array_transfer(N, trials, dtype, random, seed):
print(">>> arkouda {} array creation".format(dtype))
cfg = ak.get_config()
print("numLocales = {}, N = {:,}".format(cfg["numLocales"], N))
a = ak.randint(0, 2**32, N, dtype=dtype, seed=seed)
nb = a.size * a.itemsize
ak.client.maxTransferBytes = nb
to_ndarray_times = []
to_pdarray_times = []
for i in range(trials):
start = time.time()
npa = a.to_ndarray()
end = time.time()
to_ndarray_times.append(end - start)
start = time.time()
aka = ak.array(npa)
end = time.time()
to_pdarray_times.append(end - start)
gc.collect()
avgnd = sum(to_ndarray_times) / trials
avgpd = sum(to_pdarray_times) / trials
print("to_ndarray Average time = {:.4f} sec".format(avgnd))
print("ak.array Average time = {:.4f} sec".format(avgpd))
print("to_ndarray Average rate = {:.4f} GiB/sec".format(nb / 2**30 /
avgnd))
print("ak.array Average rate = {:.4f} GiB/sec".format(nb / 2**30 / avgpd))
def check_correctness(dtype, seed):
arrays, totalbytes = generate_arrays(1000, 2, dtype, seed)
g = ak.GroupBy(arrays)
perm = ak.argsort(ak.randint(0, 2**32, arrays[0].size))
g2 = ak.GroupBy([a[perm] for a in arrays])
assert all((uk == uk2).all() for uk, uk2 in zip(g.unique_keys, g2.unique_keys))
assert (g.segments == g2.segments).all()
def time_ak_scan(N_per_locale, trials, dtype, random, seed):
print(">>> arkouda {} scan".format(dtype))
cfg = ak.get_config()
N = N_per_locale * cfg["numLocales"]
print("numLocales = {}, N = {:,}".format(cfg["numLocales"], N))
if random or args.seed is not None:
if dtype == 'int64':
a = ak.randint(1, N, N, seed=seed)
elif dtype == 'float64':
a = ak.uniform(N, seed=seed) + 0.5
else:
a = ak.arange(1, N, 1)
if dtype == 'float64':
a = 1.0 * a
timings = {op: [] for op in OPS}
final_values = {}
for i in range(trials):
for op in timings.keys():
fxn = getattr(ak, op)
start = time.time()
r = fxn(a)
end = time.time()
timings[op].append(end - start)
final_values[op] = r[r.size-1]
tavg = {op: sum(t) / trials for op, t in timings.items()}
for op, t in tavg.items():
print("{}, final value = {}".format(op, final_values[op]))
print(" {} Average time = {:.4f} sec".format(op, t))
bytes_per_sec = (a.size * a.itemsize * 2) / t
print(" {} Average rate = {:.2f} GiB/sec".format(op, bytes_per_sec/2**30))
def create_ak_array(N, op, dtype, seed):
if op == 'zeros':
a = ak.zeros(N, dtype=dtype)
elif op == 'ones':
a = ak.ones(N, dtype=dtype)
elif op == 'randint':
a = ak.randint(0, 2**32, N, dtype=dtype, seed=seed)
return a
def check_correctness(dtype, path, seed):
N = 10**4
a = ak.randint(0, 2**32, N, seed=seed)
a.save_parquet(path)
b = ak.read_parquet(path+'*')
for f in glob(path + '_LOCALE*'):
os.remove(f)
assert (a == b).all()
def check_correctness(dtype, seed):
N = 10**4
if dtype == 'int64':
a = ak.randint(0, 2**32, N, seed=seed)
z = ak.zeros(N, dtype=dtype)
elif dtype == 'float64':
a = ak.randint(0, 1, N, dtype=ak.float64, seed=seed)
z = ak.zeros(N, dtype=dtype)
elif dtype == 'str':
a = ak.random_strings_uniform(1, 16, N, seed=seed)
z = ak.cast(ak.zeros(N), 'str')
perm = ak.coargsort([a, z])
if dtype in ('int64', 'float64'):
assert ak.is_sorted(a[perm])
perm = ak.coargsort([z, a])
if dtype in ('int64', 'float64'):
assert ak.is_sorted(a[perm])
def check_float(N):
a = ak.randint(0, 1, N, dtype=ak.float64)
n = ak.randint(-1, 1, N, dtype=ak.float64)
z = ak.zeros(N, dtype=ak.float64)
perm = ak.coargsort([a])
assert ak.is_sorted(a[perm])
perm = ak.coargsort([a, n])
assert ak.is_sorted(a[perm])
perm = ak.coargsort([n, a])
assert ak.is_sorted(n[perm])
perm = ak.coargsort([z, a])
assert ak.is_sorted(a[perm])
perm = ak.coargsort([z, n])
assert ak.is_sorted(n[perm])
def gen_rmat_edges(lgNv, Ne_per_v, p, perm=False):
# number of vertices
Nv = 2**lgNv
# number of edges
Ne = Ne_per_v * Nv
# probabilities
a = p
b = (1.0 - a) / 3.0
c = b
d = b
# init edge arrays
ii = ak.ones(Ne, dtype=ak.int64)
jj = ak.ones(Ne, dtype=ak.int64)
# quantites to use in edge generation loop
ab = a + b
c_norm = c / (c + d)
a_norm = a / (a + b)
# generate edges
for ib in range(1, lgNv):
ii_bit = (ak.randint(0, 1, Ne, dtype=ak.float64) > ab)
jj_bit = (ak.randint(0, 1, Ne, dtype=ak.float64) >
(c_norm * ii_bit + a_norm * (~ii_bit)))
ii = ii + ((2**(ib - 1)) * ii_bit)
jj = jj + ((2**(ib - 1)) * jj_bit)
# sort all based on ii and jj using coargsort
# all edges should be sorted based on both vertices of the edge
iv = ak.coargsort((ii, jj))
# permute into sorted order
ii = ii[iv] # permute first vertex into sorted order
jj = jj[iv] # permute second vertex into sorted order
# to premute/rename vertices
if perm:
# generate permutation for new vertex numbers(names)
ir = ak.argsort(ak.randint(0, 1, Nv, dtype=ak.float64))
# renumber(rename) vertices
ii = ir[ii] # rename first vertex
jj = ir[jj] # rename second vertex
#
# maybe: remove edges which are self-loops???
#
# return pair of pdarrays
return (ii, jj)
def refinement(N):
'''
Coargsort of two arrays, where the first is already sorted
but has many repeated values
'''
groupsize = 100
a = ak.arange(N // 2) // groupsize
factor = 2**32 // a.max()
a *= factor
b = ak.randint(0, 2**32, N // 2)
yield 'refinement int64', (a, b)
def IP_like(N):
'''
Data like a 90/10 mix of IPv4 and IPv6 addresses
'''
multiplicity = 10
nunique = N // (2 * multiplicity)
# First generate unique addresses, then sample with replacement
u1 = ak.zeros(nunique, dtype=ak.int64)
u2 = ak.zeros(nunique, dtype=ak.int64)
v4 = ak.uniform(nunique) < 0.9
n4 = v4.sum()
v6 = ~v4
n6 = v4.size - n4
u1[v4] = ak.randint(0, 2**32, n4)
u1[v6] = ak.randint(-2**63, 2**63, n6)
u2[v6] = ak.randint(-2**63, 2**63, n6)
sample = ak.randint(0, nunique, N // 2)
IP1 = u1[sample]
IP2 = u2[sample]
yield 'IP-like 2*int64', (IP1, IP2)
def time_ak_gather(isize, vsize, trials, dtype, random, seed):
print(">>> arkouda {} gather".format(dtype))
cfg = ak.get_config()
Ni = isize * cfg["numLocales"]
Nv = vsize * cfg["numLocales"]
print("numLocales = {}, num_indices = {:,} ; num_values = {:,}".format(cfg["numLocales"], Ni, Nv))
# Index vector is always random
i = ak.randint(0, Nv, Ni, seed=seed)
if seed is not None:
seed += 1
if random or seed is not None:
if dtype == 'int64':
v = ak.randint(0, 2**32, Nv, seed=seed)
elif dtype == 'float64':
v = ak.randint(0, 1, Nv, dtype=ak.float64, seed=seed)
elif dtype == 'bool':
v = ak.randint(0, 1, Nv, dtype=ak.bool, seed=seed)
elif dtype == 'str':
v = ak.random_strings_uniform(1, 16, Nv, seed=seed)
else:
if dtype == 'str':
v = ak.cast(ak.arange(Nv), 'str')
else:
v = ak.ones(Nv, dtype=dtype)
timings = []
for _ in range(trials):
start = time.time()
c = v[i]
end = time.time()
timings.append(end - start)
tavg = sum(timings) / trials
print("Average time = {:.4f} sec".format(tavg))
if dtype == 'str':
offsets_transferred = 3 * c.offsets.size * c.offsets.itemsize
bytes_transferred = (c.offsets.size * c.offsets.itemsize) + (2 * c.bytes.size)
bytes_per_sec = (offsets_transferred + bytes_transferred) / tavg
else:
bytes_per_sec = (c.size * c.itemsize * 3) / tavg
print("Average rate = {:.2f} GiB/sec".format(bytes_per_sec/2**30))