import time
import random
import torch

from torch.testing._internal.common_utils import TestCase, run_tests
from hypothesis import given, settings
from hypothesis import strategies as st

class indexPutTest(TestCase):
  @settings(deadline=None)
  @given(batch_size=st.sampled_from([1000, 10000, 100000, 1000000]))
  def test_index_put(self, batch_size):
    print('\ntest_index_put:')
    random.seed(100)
    index0 = random.sample(range(batch_size), batch_size//10)
    index1 = random.sample(range(64), 30)
    index2 = [(x, y) for x in index0 for y in index1]
    indices = torch.LongTensor(index2)
    indices = torch.cat([indices, indices], dim=0)
    src = torch.zeros(batch_size, 64, device='cpu')
    value = torch.ones(indices.shape[0])
    indices_tuple = tuple(indices.t())
    start = time.time()
    for i in range(100):
      src.index_put_(indices_tuple , value, accumulate=True)
    end = time.time()
    i = 0
    sizeofList = len(index2)
    while i < sizeofList :
      x, y = index2[i]
      if (src[x, y] != 200.0):
        print('not match !! src[{},{}] = {:.2f}'.format(x, y, src[x, y]))
      i += 1
    print('indices size:{}, time: {:.6f} ms/per_index_put'.format(len(indices), (end-start)*10))

if __name__=='__main__':
  run_tests()

ncores=28
max_core=$((ncores-1))
export KMP_BLOCKTIME=1
export KMP_AFFINITY="granularity=fine,compact,1,0"
numactl --physcpubind=0-$max_core python ./test_index_put.py

... /usr/lib/clang/8.0.1/include/mmintrin.h:47:5: error: use of undeclared identifier '__builtin_ia32_emms'; did you mean '__builtin_isless'?
    __builtin_ia32_emms();
    ^
... /usr/lib/clang/8.0.1/include/mmintrin.h:47:25: error: too few arguments to function call, expected 2, have 0
    __builtin_ia32_emms();
                        ^
... /usr/lib/clang/8.0.1/include/mmintrin.h:64:19: error: use of undeclared identifier '__builtin_ia32_vec_init_v2si'
    return (__m64)__builtin_ia32_vec_init_v2si(__i, 0);