def fill(arch, dtype, dsize, repeat=10):
repeat = scaled_repeat_times(arch, dsize, repeat)
num_elements = dsize // dtype_size[dtype]
x = ti.field(dtype, shape=num_elements)
@ti.kernel
def fill_const(n: ti.i32):
for i in range(n):
x[i] = ti.cast(0.7, dtype)
return membound_benchmark(fill_const, num_elements, repeat)
def _save_suite_info_as_json(self, suite_path='./'):
info_dict = {
'cases': [func.__name__ for func in self.test_cases],
'dtype': [dtype2str(dtype) for dtype in self.test_dtype_list],
'dsize': [size for size in self.test_dsize_list],
'repeat': [
scaled_repeat_times(self._arch, size, self.basic_repeat_times)
for size in self.test_dsize_list
],
'evaluator': [func.__name__ for func in self.evaluator]
}
info_path = os.path.join(suite_path, '_info.json')
with open(info_path, 'w') as f:
print(dump2json(info_dict), file=f)
def get_results_dict(self):
results_dict = {}
for i in range(len(self._test_dsize_list)):
dsize = self._test_dsize_list[i]
repeat = scaled_repeat_times(self._arch, dsize,
MemoryBound.basic_repeat_times)
elapsed_time = self._min_time_in_us[i]
item_name = size2str(dsize).replace('.0', '')
item_dict = {
'dsize_byte': dsize,
'repeat': repeat,
'elapsed_time_ms': elapsed_time
}
results_dict[item_name] = item_dict
return results_dict
def reduction(arch, dtype, dsize, repeat=10):
repeat = scaled_repeat_times(arch, dsize, repeat)
num_elements = dsize // dtype_size[dtype]
x = ti.field(dtype, shape=num_elements)
y = ti.field(dtype, shape=())
y[None] = 0
@ti.kernel
def reduction(n: ti.i32):
for i in range(n):
y[None] += x[i]
init_const(x, dtype, num_elements)
return membound_benchmark(reduction, num_elements, repeat)
def saxpy(arch, dtype, dsize, repeat=10):
repeat = scaled_repeat_times(arch, dsize, repeat)
num_elements = dsize // dtype_size[dtype] // 3 #z=x+y
x = ti.field(dtype, shape=num_elements)
y = ti.field(dtype, shape=num_elements)
z = ti.field(dtype, shape=num_elements)
@ti.kernel
def saxpy(n: ti.i32):
for i in range(n):
z[i] = 17 * x[i] + y[i]
init_const(x, dtype, num_elements)
init_const(y, dtype, num_elements)
return membound_benchmark(saxpy, num_elements, repeat)