Improve time measuring

master
Max Ehrlicher-Schmidt 4 years ago
parent ea3b090f6d
commit 7eb96a69ec

2
.gitignore vendored

@ -2,3 +2,5 @@
/shelf/
/workspace.xml
.idea
primes.txt
timings.csv

@ -33,10 +33,10 @@ check_prime(unsigned long long *input, bool *output)
""")
ker2 = SourceModule("""
__global__ void check_prime2(const unsigned long *IN, bool *OUT) {
__global__ void check_prime2(const unsigned long long *IN, bool *OUT) {
int id = threadIdx.x + blockDim.x * blockIdx.x;
unsigned long num = IN[id];
unsigned long limit = (unsigned long) sqrt((double) num) + 1;
unsigned long long num = IN[id];
unsigned long long limit = (unsigned long long) sqrt((double) num) + 1;
if (num == 2 || num == 3) {
OUT[id] = true;
@ -45,7 +45,7 @@ __global__ void check_prime2(const unsigned long *IN, bool *OUT) {
return;
}
if (limit < 9) {
for (unsigned long i = 3; i <= limit; i++) {
for (unsigned long long i = 3; i <= limit; i++) {
if (num % i == 0) {
return;
}
@ -54,7 +54,7 @@ __global__ void check_prime2(const unsigned long *IN, bool *OUT) {
if (num > 3 && num % 3 == 0) {
return;
}
for (unsigned long i = 9; i <= (limit + 6); i += 6) {
for (unsigned long long i = 9; i <= (limit + 6); i += 6) {
if (num % (i - 2) == 0 || num % (i - 4) == 0) {
return;
}
@ -76,26 +76,32 @@ def calc_primes(start: int = 1, grid_size: int = 1000, block_size: int = 1024):
if start % 2 == 0:
start = start + 1
testvec = np.arange(start, block_size * grid_size * 2 + start, step=2).astype(np.uint)
startEvent = drv.Event()
endEvent = drv.Event()
testvec = np.arange(start, block_size * grid_size * 2 + start, step=2).astype(np.ulonglong)
testvec_gpu = gpuarray.to_gpu(testvec)
outvec_gpu = gpuarray.to_gpu(np.full(block_size * grid_size, False, dtype=bool))
t1 = time()
startEvent.record()
check_prime(testvec_gpu, outvec_gpu, block=(block_size, 1, 1), grid=(grid_size, 1, 1))
endEvent.record()
endEvent.synchronize()
kernel_execution_time = startEvent.time_till(endEvent)
result = outvec_gpu.get()
t2 = time()
for idx, val in enumerate(result):
if val:
primes.append(testvec[idx])
print(len(primes))
print('checked ' + str(block_size * grid_size) + ' numbers')
print('checked ' + str(block_size * grid_size) + ' numbers' + ' (' + str(start) + ' - ' + str(
start + block_size * grid_size) + ')')
print('last prime: ' + str(primes[-1]))
print('The GPU needed ' + str(t2 - t1) + ' seconds')
print('The GPU needed ' + str(kernel_execution_time) + ' milliseconds')
with open(options.timings_output, 'a') as file:
file.write(str(start) + "," + str((t2 - t1) * 1000) + "\n")
file.write(str(start) + "," + str(kernel_execution_time) + "\n")
return primes
@ -103,7 +109,7 @@ def calc_primes(start: int = 1, grid_size: int = 1000, block_size: int = 1024):
if __name__ == "__main__":
parser = OptionParser()
parser.add_option("-e", "--end", dest="end",
help="numbers to check without even numbers", default="1000000000", type="int")
help="numbers to check without even numbers", default="5000000000", type="int")
parser.add_option("--numbers-per-step", dest="numbers_per_step",
help="amount of uneven numbers checked in each step (even number are skipped)", default="8388608",
type="int")
@ -122,9 +128,11 @@ if __name__ == "__main__":
with open(options.timings_output, 'w') as file:
file.write("offset, duration\n")
with open(options.output, 'w') as file:
file.write("")
while last_number_checked < options.end:
calculated_primes = calc_primes(last_number_checked + 1, grid_size, block_size)
with open(options.output, 'w') as file:
with open(options.output, 'a') as file:
file.write("\n".join([str(p) for p in calculated_primes]))
last_number_checked = last_number_checked + resulting_numbers_per_step * 2

Loading…
Cancel
Save