| .. |
|
integer_ops
|
Support quantized Mul with multiplier>1, even if that's a pathological case.
|
2019-08-01 10:15:35 -07:00 |
|
cpu_check.h
|
Fix for SparkFun Edge build problem
|
2019-06-25 23:48:36 -07:00 |
|
depthwiseconv_3x3_filter_common.h
|
Checking for defined(USE_NEON) && defined(__aarch64__) is redundant: A64 always has NEON support.
|
2019-06-21 12:38:53 -07:00 |
|
depthwiseconv_float.h
|
Move USE_NEON declaration to cpu_check.h, include that header everywhere USE_NEON is used.
|
2019-06-19 10:47:12 -07:00 |
|
depthwiseconv_multithread.h
|
Don't access CpuBackendContext concurrently from multiple threads.
|
2019-05-30 12:38:39 -07:00 |
|
depthwiseconv_uint8_3x3_filter.h
|
Disable dot-product depthwise conv when not compiling for Android with Clang.
|
2019-06-26 11:42:13 -07:00 |
|
depthwiseconv_uint8_transitional.h
|
Move USE_NEON declaration to cpu_check.h, include that header everywhere USE_NEON is used.
|
2019-06-19 10:47:12 -07:00 |
|
depthwiseconv_uint8.h
|
fix regression
|
2019-06-28 06:52:58 -07:00 |
|
eigen_spatial_convolutions.h
|
Ensure the implementation of Eigen spatial convolutions in TF and TFLite in sync.
|
2019-03-11 23:48:32 -07:00 |
|
eigen_tensor_reduced_instantiations_google.h
|
Upgrade the version of Eigen to commit b4890dc6bc34.
|
2018-11-06 15:09:21 -08:00 |
|
eigen_tensor_reduced_instantiations_oss.h
|
Upgrade the version of Eigen to commit b4890dc6bc34.
|
2018-11-06 15:09:21 -08:00 |
|
im2col_utils.h
|
No longer depend directly on the general gemmlowp library,
|
2019-05-07 11:46:46 -07:00 |
|
legacy_optimized_ops.h
|
Optimize 8bit Softmax op handler
|
2019-08-02 17:52:28 -07:00 |
|
multithreaded_conv.h
|
Avoid im2col creation for multi-threaded conv
|
2019-04-01 16:39:57 -07:00 |
|
neon_check.h
|
Fix for SparkFun Edge build problem
|
2019-06-25 23:48:36 -07:00 |
|
neon_tensor_utils_impl.h
|
Don't define Neon* functions when USE_NEON define is missing.
|
2019-06-26 03:19:46 -07:00 |
|
neon_tensor_utils.cc
|
Fix compilation error, use vreinterpreq instead.
|
2019-08-04 22:28:49 -07:00 |
|
neon_tensor_utils.h
|
Automated rollback of commit 1e1535bbcd
|
2019-06-25 19:07:25 -07:00 |
|
optimized_ops.h
|
Optimize requantize methods for int8 to int8 and uint8 to uint8
|
2019-08-04 22:11:37 -07:00 |
|
sse_check.h
|
Add x86 vector code using SSE 4.1 for MatrixBatchVectorMultiplyAccumulate and SparseMatrixBatchVectorMultiplyAccumulate, int8 versions only.
|
2019-06-26 17:55:26 -07:00 |
|
sse_tensor_utils_impl.h
|
Add x86 vector code using SSE 4.1 for MatrixBatchVectorMultiplyAccumulate and SparseMatrixBatchVectorMultiplyAccumulate, int8 versions only.
|
2019-06-26 17:55:26 -07:00 |
|
sse_tensor_utils.cc
|
Add x86 vector code using SSE 4.1 for MatrixBatchVectorMultiplyAccumulate and SparseMatrixBatchVectorMultiplyAccumulate, int8 versions only.
|
2019-06-26 17:55:26 -07:00 |
|
sse_tensor_utils.h
|
Add x86 vector code using SSE 4.1 for MatrixBatchVectorMultiplyAccumulate and SparseMatrixBatchVectorMultiplyAccumulate, int8 versions only.
|
2019-06-26 17:55:26 -07:00 |
