[WIP][AUTOTVM][TOPI] Port x86 NCHWc to AutoTVM for Task Extraction (a…

…pache#2664)

[AUTOTVM][TOPI] Port x86 NCHWc to AutoTVM for Task Extraction

python/tvm/autotvm/task/relay_integration.py

@@ -50,7 +50,7 @@ def extract_from_program(func, params, ops, target, target_host=None):
     # relay op -> topi compute
     OP2TOPI = {
         tvm.relay.op.nn.conv2d: [topi.nn.conv2d, topi.nn.depthwise_conv2d_nchw,
-                                 topi.nn.group_conv2d_nchw],
+                                 topi.nn.group_conv2d_nchw, topi.nn.conv2d_NCHWc],
         tvm.relay.op.nn.conv2d_transpose: [topi.nn.conv2d_transpose_nchw],
         tvm.relay.op.nn.dense: [topi.nn.dense],
         tvm.relay.op.nn.deformable_conv2d: [topi.nn.deformable_conv2d_nchw],

python/tvm/autotvm/task/topi_integration.py

@@ -67,6 +67,7 @@ def __init__(self):
             topi.nn.depthwise_conv2d_nchw: "topi_nn_depthwise_conv2d_nchw",
             topi.nn.group_conv2d_nchw: "topi_nn_group_conv2d_nchw",
             topi.nn.conv2d_transpose_nchw: "topi_nn_conv2d_transpose_nchw",
+            topi.nn.conv2d_NCHWc: "topi_x86_conv2d_NCHWc",
             topi.nn.dense: "topi_nn_dense",
             topi.nn.bitserial_conv2d_nchw: "topi_nn_bitserial_conv2d_nchw",
             topi.nn.bitserial_conv2d_nhwc: "topi_nn_bitserial_conv2d_nhwc",
@@ -80,6 +81,7 @@ def __init__(self):
                                             topi.generic.schedule_depthwise_conv2d_nhwc],
             topi.nn.group_conv2d_nchw: [topi.generic.schedule_group_conv2d_nchw],
             topi.nn.conv2d_transpose_nchw: [topi.generic.schedule_conv2d_transpose_nchw],
+            topi.nn.conv2d_NCHWc: [topi.generic.schedule_conv2d_NCHWc],
             topi.nn.dense: [topi.generic.schedule_dense],
             topi.nn.bitserial_conv2d_nchw: [topi.generic.schedule_bitserial_conv2d_nchw],
             topi.nn.bitserial_conv2d_nhwc: [topi.generic.schedule_bitserial_conv2d_nhwc],
@@ -108,7 +110,6 @@ def _tracing_topi_compute(*args, **kwargs):
                         key = (self.topi_to_task[compute_func], serialize_args(args))
                         if key not in self.task_collection:
                             self.task_collection.append(key)
-
                     return compute_func.fdefault(*args)
             _local_scope(topi_compute)
 
@@ -205,6 +206,15 @@ def _topi_nn_deformable_conv2d_nchw(*args, **kwargs):
             s = topi.generic.schedule_deformable_conv2d_nchw([C])
             return s, [A, Offset, W, C]
 
+        @register("topi_nn_conv2d_NCHWc")
+        def _topi_nn_conv2d_NCHWc(*args, **kwargs):
+            assert not kwargs, "Do not support kwargs in template function call"
+            args = deserialize_args(args)
+            A, W = args[:2]
+            C = topi.nn.conv2d_NCHWc(*args, **kwargs)
+            s = topi.generic.schedule_conv2d_NCHWc([C])
+            return s, [A, W, C]
+
     def reset(self, wanted_topi_funcs):
         """Reset task collections
 

topi/python/topi/nn/conv2d.py

@@ -329,7 +329,68 @@ def conv2d_NCHWc(data, kernel, stride, padding, dilation, layout, out_layout, ou
     """
     # search platform specific declaration first
     # default declaration
-    raise ValueError("missing register for topi.nn.conv2d_NCHWc")
+    # layout and out_layout are not used here,
+    # we keep them for debug convenience when dumping autotvm workload
+    pad_top, pad_left, pad_down, pad_right = get_pad_tuple(padding,
+                                                           (dilated_kernel_h,
+                                                            dilated_kernel_w))
+    HPAD = pad_top + pad_down
+    WPAD = pad_left + pad_right
+    HSTR, WSTR = stride if isinstance(stride, (tuple, list)) else (stride, stride)
+    dh, dw = dilation if isinstance(dilation, (tuple, list)) else (dilation, dilation)
+    assert (dh, dw) == (1, 1), "Does not support dilation"
+
+    n, ic_chunk, ih, iw, ic_bn = get_const_tuple(data.shape)
+    in_channel = ic_chunk * ic_bn
+    if data.dtype == 'uint8':
+        oc_chunk, _, kernel_height, kernel_width, _, oc_bn, _ = get_const_tuple(kernel.shape)
+    else:
+        oc_chunk, _, kernel_height, kernel_width, _, oc_bn = get_const_tuple(kernel.shape)
+    num_filter = oc_chunk * oc_bn
+
+    # output shape
+    out_height = (ih + 2 * HPAD - kernel_height) // HSTR + 1
+    out_width = (iw + 2 * WPAD - kernel_width) // WSTR + 1
+    oshape = (n, oc_chunk, out_height, out_width, oc_bn)
+
+    # DOPAD
+    DOPAD = (HPAD != 0 or WPAD != 0)
+    if DOPAD:
+        data_pad = pad(data, (0, 0, HPAD, WPAD, 0), name="data_pad")
+    else:
+        data_pad = data
+
+    ic = tvm.reduce_axis((0, in_channel), name='ic')
+    kh = tvm.reduce_axis((0, kernel_height), name='kh')
+    kw = tvm.reduce_axis((0, kernel_width), name='kw')
+
+    if data.dtype == 'uint8':
+        assert out_dtype == "int32", \
+            "INT8 convolution requires input dtype = uint8 and output dtype=int32"
+        # Intel performs dot product of 2 "4" Int8 values
+        # Current implementation requires ic_bn to be a multiple of 4
+        n_elems = 4
+        assert ic_bn % n_elems == 0
+
+        ic_outer = tvm.reduce_axis((0, in_channel//ic_bn), name='ic_outer')
+        ic_f_inner = tvm.reduce_axis((0, ic_bn//n_elems), name='ic_f_inner')
+        ic_s_inner = tvm.reduce_axis((0, n_elems), name='ic_s_inner')
+        return tvm.compute(oshape, lambda n, oc_chunk, oh, ow, oc_block:
+                           tvm.sum(data_pad[n, ic_outer, oh*HSTR+kh, ow*WSTR+kw,
+                                            ic_f_inner * n_elems +  ic_s_inner]
+                                   .astype(out_dtype) *
+                                   kernel[oc_chunk, ic_outer, kh, kw, ic_f_inner,
+                                          oc_block, ic_s_inner].astype(out_dtype),
+                                   axis=[kh, kw, ic_outer, ic_f_inner, ic_s_inner]),
+                           name='conv2d_NCHWc_int8', tag="conv2d_NCHWc_int8")
+    # else: fp implementation
+    return tvm.compute(oshape, lambda n, oc_chunk, oh, ow, oc_block:
+                       tvm.sum(data_pad[n, ic//ic_bn, oh*HSTR+kh, ow*WSTR+kw,
+                                        ic%ic_bn].astype(out_dtype) *
+                               kernel[oc_chunk, ic//ic_bn, kh, kw, ic%ic_bn, oc_block],
+                               axis=[ic, kh, kw]),
+                       name='conv2d_NCHWc', tag="conv2d_NCHWc")
+
 
 
 def conv2d_winograd_weight_transform(kernel, tile_size):

topi/python/topi/x86/conv2d.py

@@ -2,6 +2,7 @@
 """Conv2D schedule on x86"""
 
 import logging
+import re
 
 import tvm
 from tvm import autotvm
@@ -41,9 +42,22 @@ def _create_tuning_space(cfg, data, kernel, strides, padding, dilation, layout):
     """Create schedule configuration from input arguments"""
     dshape = get_const_tuple(data.shape)
     kshape = get_const_tuple(kernel.shape)
+    pat = re.compile(r'NCHW.+(\d+)c')
     if layout == 'NCHW':
         n, ic, h, w = dshape
         oc, _, kh, kw = kshape
+    elif pat.match(layout) is not None:
+        n, ic_chunk, h, w, ic_bn = dshape
+        if data.dtype == 'uint8':
+            oc_chunk, k_ic, kh, kw, k_ic_f, oc_bn, k_ic_s = kshape
+            ic = ic_chunk*ic_bn
+            assert ic == k_ic*k_ic_f*kic_s
+        else:
+            oc_chunk, k_ic_chunk, kh, kw, k_ic_bn, oc_bn = kshape
+            assert ic_chunk == k_ic_chunk
+            assert ic_bn == k_ic_bn
+            ic = ic_chunk*ic_bn
+        oc = oc_chunk*oc_bn
     else:
         raise ValueError("Not support this layout {} with "
                          "schedule template.".format(layout))
@@ -258,7 +272,14 @@ def traverse(op):
 @autotvm.task.register("topi_x86_conv2d_NCHWc")
 def _topi_nn_conv2d_NCHWc(*args, **kwargs):
     assert not kwargs, "Do not support kwargs in template function call"
-    data, kernel, strides, padding, dilation, origin_layout, dtype = deserialize_args(args)
+    args = deserialize_args(args)
+
+    if len(args) == 7:
+        data, kernel, strides, padding, dilation, origin_layout, dtype = args
+    else:
+        assert len(args) == 8
+        data, kernel, strides, padding, dilation, origin_layout, out_layout, dtype = args
+
     raw_data_shape = get_const_tuple(data.shape)
     raw_kernel_shape = get_const_tuple(kernel.shape)