diff --git a/topi/python/topi/x86/conv2d_avx_1x1.py b/topi/python/topi/x86/conv2d_avx_1x1.py
index cc264d04ac24ced1de984f8fa4c21b30ad5afef4..afd0be2e2ded3950588fbcd9ae91834f3ac1c7ba 100644
--- a/topi/python/topi/x86/conv2d_avx_1x1.py
+++ b/topi/python/topi/x86/conv2d_avx_1x1.py
@@ -77,9 +77,6 @@ def _schedule_conv(s, data, data_pad, data_vec, kernel, kernel_vec, conv_out, ou
     batch, ic_chunk, ih, ic_block, iw = s[A1].op.axis
     parallel_axis = s[A1].fuse(ic_chunk, ih)
     s[A1].parallel(parallel_axis)
-    s[A1].pragma(batch, "parallel_launch_point")
-    s[A1].pragma(parallel_axis, "parallel_stride_pattern")
-    s[A1].pragma(batch, "parallel_barrier_when_finish")
 
     # schedule kernel pack
     oc_chunk, ic_chunk, oh, ow, ic_block, oc_block = s[W].op.axis
@@ -88,9 +85,6 @@ def _schedule_conv(s, data, data_pad, data_vec, kernel, kernel_vec, conv_out, ou
         s[W].vectorize(oc_block)
     parallel_axis = s[W].fuse(oc_chunk, oh)
     s[W].parallel(parallel_axis)
-    s[W].pragma(parallel_axis, "parallel_launch_point")
-    s[W].pragma(parallel_axis, "parallel_stride_pattern")
-    s[W].pragma(parallel_axis, "parallel_barrier_when_finish")
 
     C, O0, O = conv_out, output, last
     CC = s.cache_write(C, 'global')
@@ -128,8 +122,5 @@ def _schedule_conv(s, data, data_pad, data_vec, kernel, kernel_vec, conv_out, ou
     s[O].vectorize(oc_block)
 
     s[O].parallel(parallel_axis)
-    s[O].pragma(batch, "parallel_launch_point")
-    s[O].pragma(parallel_axis, "parallel_stride_pattern")
-    s[O].pragma(batch, "parallel_barrier_when_finish")
 
     return s
diff --git a/topi/python/topi/x86/conv2d_avx_common.py b/topi/python/topi/x86/conv2d_avx_common.py
index 4f5be019f45a06d6399c06696b1ef150b2f718a8..f4c0e453e643087f7e4f6c5fc2c2249b244643a6 100644
--- a/topi/python/topi/x86/conv2d_avx_common.py
+++ b/topi/python/topi/x86/conv2d_avx_common.py
@@ -90,9 +90,6 @@ def _schedule_conv(s, data, data_pad, data_vec, kernel, kernel_vec, conv_out, ou
     batch, ic_chunk, ih, ic_block, iw = s[A1].op.axis
     parallel_axis = s[A1].fuse(ic_chunk, ih)
     s[A1].parallel(parallel_axis)
-    s[A1].pragma(batch, "parallel_launch_point")
-    s[A1].pragma(parallel_axis, "parallel_stride_pattern")
-    s[A1].pragma(batch, "parallel_barrier_when_finish")
 
     # schedule kernel pack
     oc_chunk, ic_chunk, oh, ow, ic_block, oc_block = s[W].op.axis
@@ -101,9 +98,6 @@ def _schedule_conv(s, data, data_pad, data_vec, kernel, kernel_vec, conv_out, ou
         s[W].vectorize(oc_block)
     parallel_axis = s[W].fuse(oc_chunk, oh)
     s[W].parallel(parallel_axis)
-    s[W].pragma(parallel_axis, "parallel_launch_point")
-    s[W].pragma(parallel_axis, "parallel_stride_pattern")
-    s[W].pragma(parallel_axis, "parallel_barrier_when_finish")
 
     # schedule conv
     C, O0, O = conv_out, output, last
@@ -144,8 +138,5 @@ def _schedule_conv(s, data, data_pad, data_vec, kernel, kernel_vec, conv_out, ou
     s[O].vectorize(oc_block)
 
     s[O].parallel(parallel_axis)
-    s[O].pragma(batch, "parallel_launch_point")
-    s[O].pragma(parallel_axis, "parallel_stride_pattern")
-    s[O].pragma(batch, "parallel_barrier_when_finish")
 
     return s