NVIDIA/nvbench
1
0
Fork 0
mirror of https://github.com/NVIDIA/nvbench.git synced 2026-04-20 06:48:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

Change test and examples from using camelCase to using snake_case as implementation changed

Browse Source
This commit is contained in:
Oleksandr Pavlyk
2025-07-21 12:50:28 -05:00
parent 81fff085b9
commit e589518376
9 changed files with 101 additions and 105 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
python/test/run_1.py
View File

@@ -14,18 +14,18 @@ def kernel(a, b, c):
c[tid] = a[tid] + b[tid]
def getNumbaStream(launch):
return cuda.external_stream(launch.getStream().addressof())
def get_numba_stream(launch):
return cuda.external_stream(launch.get_stream().addressof())
def add_two(state):
# state.skip("Skipping this benchmark for no reason")
N = state.getInt64("elements")
N = state.get_int64("elements")
a = cuda.to_device(np.random.random(N))
c = cuda.device_array_like(a)
state.addGlobalMemoryReads(a.nbytes)
state.addGlobalMemoryWrites(c.nbytes)
state.add_global_memory_reads(a.nbytes)
state.add_global_memory_writes(c.nbytes)
nthreads = 256
nblocks = (len(a) + nthreads - 1) // nthreads
@@ -35,22 +35,22 @@ def add_two(state):
cuda.synchronize()
def kernel_launcher(launch):
stream = getNumbaStream(launch)
stream = get_numba_stream(launch)
kernel[nblocks, nthreads, stream](a, a, c)
state.exec(kernel_launcher, batched=True, sync=True)
def add_float(state):
N = state.getInt64("elements")
v = state.getFloat64("v")
name = state.getString("name")
N = state.get_int64("elements")
v = state.get_gloat64("v")
name = state.get_string("name")
a = cuda.to_device(np.random.random(N).astype(np.float32))
b = cuda.to_device(np.random.random(N).astype(np.float32))
c = cuda.device_array_like(a)
state.addGlobalMemoryReads(a.nbytes + b.nbytes)
state.addGlobalMemoryWrites(c.nbytes)
state.add_global_memory_reads(a.nbytes + b.nbytes)
state.add_global_memory_writes(c.nbytes)
nthreads = 64
nblocks = (len(a) + nthreads - 1) // nthreads
@@ -58,26 +58,26 @@ def add_float(state):
def kernel_launcher(launch):
_ = v
_ = name
stream = getNumbaStream(launch)
stream = get_numba_stream(launch)
kernel[nblocks, nthreads, stream](a, b, c)
state.exec(kernel_launcher, batched=True, sync=True)
def add_three(state):
N = state.getInt64("elements")
N = state.get_int64("elements")
a = cuda.to_device(np.random.random(N).astype(np.float32))
b = cuda.to_device(np.random.random(N).astype(np.float32))
c = cuda.device_array_like(a)
state.addGlobalMemoryReads(a.nbytes + b.nbytes)
state.addGlobalMemoryWrites(c.nbytes)
state.add_global_memory_reads(a.nbytes + b.nbytes)
state.add_global_memory_writes(c.nbytes)
nthreads = 256
nblocks = (len(a) + nthreads - 1) // nthreads
def kernel_launcher(launch):
stream = getNumbaStream(launch)
stream = get_numba_stream(launch)
kernel[nblocks, nthreads, stream](a, b, c)
state.exec(kernel_launcher, batched=True, sync=True)
@@ -86,18 +86,18 @@ def add_three(state):
def register_benchmarks():
(
nvbench.register(add_two).addInt64Axis(
nvbench.register(add_two).add_int64_axis(
"elements", [2**pow2 for pow2 in range(20, 23)]
)
)
(
nvbench.register(add_float)
.addFloat64Axis("v", [0.1, 0.3])
.addStringAxis("name", ["Anne", "Lynda"])
.addInt64Axis("elements", [2**pow2 for pow2 in range(20, 23)])
.add_float64_axis("v", [0.1, 0.3])
.add_string_axis("name", ["Anne", "Lynda"])
.add_int64_axis("elements", [2**pow2 for pow2 in range(20, 23)])
)
(
nvbench.register(add_three).addInt64Axis(
nvbench.register(add_three).add_int64_axis(
"elements", [2**pow2 for pow2 in range(20, 22)]
)
)