jverma-quic commented on code in PR #12340: URL: https://github.com/apache/tvm/pull/12340#discussion_r950511701
########## python/tvm/topi/hexagon/qnn/avg_pool2d.py: ########## @@ -0,0 +1,205 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# pylint: disable=invalid-name, unused-variable, unused-argument, too-many-locals + +""" Compute and schedule for quantized avg_pool2d op + +Please note the following assumptions made by the implementation: + +1) The input must be padded in advance to account for 'padding'. In addition, + both input and output must be padded as per the physical buffer layout. +2) The current implementation assumes 'count_include_pad' to be 'True'. It can be + modified to support 'False' case but the element count for the pooling window + must be pre-computed and provided as an input to reduce the run-time overhead. +3) 'padding' is ignored. It must be handled outside of the sliced op. +4) Please note that this implementation will not work if the output includes any + physical layout related padding as it can result into out-of-bound access + for the input. +""" + +from tvm import te +from tvm import tir +from ..utils import get_layout_transform_fn, get_fixed_point_value + + +def validate_out_shape(out_shape: list, in_shape: list, kernel: list, stride: list, dilation: list): + """Validate output shape""" + _, oh, ow, _ = out_shape + _, ih, iw, _ = in_shape + kh, kw = kernel + sh, sw = stride + dh, dw = dilation + if ih < (oh - 1) * sh + dh * (kh - 1) + 1: + raise RuntimeError("Output height is too large") + if iw < (ow - 1) * sw + dw * (kw - 1) + 1: + raise RuntimeError("Output width is too large") + + +def saturate(x: te.Tensor, dtype: str): + """Saturate value for the specified data type""" + return te.max(te.min_value(dtype), te.min(x, te.max_value(dtype))) Review Comment: I agree. I think it will be better to do it in TVM as we can generate the appropriate LLVM saturating instruction during TVM codegen which can then be lowered into target specific instructions in the LLVM backend. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
