import os
import numpy as np
np.random.seed(19)
def gen_golden_data_tshrs(case_name, param):
dtype = param.dtype
src_row, src_col = [param.src_row, param.src_col]
dst_row, dst_col = [param.dst_row, param.dst_col]
h_valid, w_valid = [param.valid_row, param.valid_col]
input1 = np.random.randint(1, 10, size=[src_row, src_col]).astype(dtype)
scalar = np.random.randint(1, 10, size=[1, 1]).astype(dtype)
golden = np.zeros([dst_row, dst_col]).astype(dtype)
for h in range(dst_row):
for w in range(dst_col):
if h < h_valid and w < w_valid:
golden[h][w] = input1[h][w] >> scalar[0][0]
input1.tofile("input.bin")
scalar.tofile("scalar.bin")
golden.tofile("golden.bin")
return input1, golden
class TSHRSParams:
def __init__(self, dtype, dst_row, dst_col, src_row, src_col, valid_row, valid_col):
self.dtype = dtype
self.dst_row = dst_row
self.dst_col = dst_col
self.src_row = src_row
self.src_col = src_col
self.valid_row = valid_row
self.valid_col = valid_col
def generate_case_name(param):
dtype_str = {
np.uint32: 'uint32',
np.uint16: 'uint16',
np.int8: 'int8',
np.int32: 'int32',
np.int16: 'int16'
}[param.dtype]
def substring(a, b) -> str:
return f"_{a}x{b}"
name = f"TSHRSTest.case_{dtype_str}"
name += substring(param.dst_row, param.dst_col)
name += substring(param.src_row, param.src_col)
name += substring(param.valid_row, param.valid_col)
return name
if __name__ == "__main__":
script_dir = os.path.dirname(os.path.abspath(__file__))
testcases_dir = os.path.join(script_dir, "testcases")
if not os.path.exists(testcases_dir):
os.makedirs(testcases_dir)
case_params_list = [
TSHRSParams(np.int16, 64, 64, 64, 64, 64, 64),
TSHRSParams(np.int32, 16, 256, 16, 256, 16, 256)
]
for param in case_params_list:
case_name = generate_case_name(param)
if not os.path.exists(case_name):
os.makedirs(case_name)
original_dir = os.getcwd()
os.chdir(case_name)
gen_golden_data_tshrs(case_name, param)
os.chdir(original_dir)