Skip to content

Manual sync from sync from github.com/tensorflow/tensorflow #3601

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
veblush wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Manual sync from sync from github.com/tensorflow/tensorflow #3601

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
294b277
Synced
veblush Jun 18, 2026
6ad2479
No abseil
veblush Jun 18, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 13 additions & 12 deletions ci/tflite_files.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,16 +30,17 @@ tensorflow/lite/core/c/common.h
tensorflow/lite/core/macros.h
tensorflow/lite/kernels/internal/common.h
tensorflow/lite/kernels/internal/compatibility.h
tensorflow/lite/kernels/internal/portable_tensor_utils.h
tensorflow/lite/kernels/internal/portable_tensor_utils.cc
tensorflow/lite/kernels/internal/portable_tensor_utils.h
tensorflow/lite/kernels/internal/quantization_util.h
tensorflow/lite/kernels/internal/reference/add.h
tensorflow/lite/kernels/internal/reference/add_n.h
tensorflow/lite/kernels/internal/reference/add.h
tensorflow/lite/kernels/internal/reference/arg_min_max.h
tensorflow/lite/kernels/internal/reference/batch_to_space_nd.h
tensorflow/lite/kernels/internal/reference/batch_matmul.h
tensorflow/lite/kernels/internal/reference/batch_to_space_nd.h
tensorflow/lite/kernels/internal/reference/binary_function.h
tensorflow/lite/kernels/internal/reference/broadcast_args.h
tensorflow/lite/kernels/internal/reference/broadcast_loop.h
tensorflow/lite/kernels/internal/reference/broadcast_to.h
tensorflow/lite/kernels/internal/reference/ceil.h
tensorflow/lite/kernels/internal/reference/comparisons.h
Expand All @@ -54,17 +55,17 @@ tensorflow/lite/kernels/internal/reference/div.h
tensorflow/lite/kernels/internal/reference/elu.h
tensorflow/lite/kernels/internal/reference/exp.h
tensorflow/lite/kernels/internal/reference/fill.h
tensorflow/lite/kernels/internal/reference/floor.h
tensorflow/lite/kernels/internal/reference/floor_div.h
tensorflow/lite/kernels/internal/reference/floor_mod.h
tensorflow/lite/kernels/internal/reference/floor.h
tensorflow/lite/kernels/internal/reference/fully_connected.h
tensorflow/lite/kernels/internal/reference/hard_swish.h
tensorflow/lite/kernels/internal/reference/integer_ops/add.h
tensorflow/lite/kernels/internal/reference/integer_ops/conv.h
tensorflow/lite/kernels/internal/reference/integer_ops/depthwise_conv.h
tensorflow/lite/kernels/internal/reference/integer_ops/fully_connected.h
tensorflow/lite/kernels/internal/reference/integer_ops/logistic.h
tensorflow/lite/kernels/internal/reference/integer_ops/l2normalization.h
tensorflow/lite/kernels/internal/reference/integer_ops/logistic.h
tensorflow/lite/kernels/internal/reference/integer_ops/mean.h
tensorflow/lite/kernels/internal/reference/integer_ops/mul.h
tensorflow/lite/kernels/internal/reference/integer_ops/pooling.h
Expand All @@ -73,14 +74,16 @@ tensorflow/lite/kernels/internal/reference/integer_ops/transpose_conv.h
tensorflow/lite/kernels/internal/reference/l2normalization.h
tensorflow/lite/kernels/internal/reference/leaky_relu.h
tensorflow/lite/kernels/internal/reference/log_softmax.h
tensorflow/lite/kernels/internal/reference/logistic.h
tensorflow/lite/kernels/internal/reference/lstm_cell.h
tensorflow/lite/kernels/internal/reference/maximum_minimum.h
tensorflow/lite/kernels/internal/reference/mul.h
tensorflow/lite/kernels/internal/reference/neg.h
tensorflow/lite/kernels/internal/reference/pad.h
tensorflow/lite/kernels/internal/reference/pooling.h
tensorflow/lite/kernels/internal/reference/portable_tensor_utils_impl.h
tensorflow/lite/kernels/internal/reference/portable_tensor_utils.cc
tensorflow/lite/kernels/internal/reference/portable_tensor_utils.h
tensorflow/lite/kernels/internal/reference/portable_tensor_utils_impl.h
tensorflow/lite/kernels/internal/reference/prelu.h
tensorflow/lite/kernels/internal/reference/process_broadcast_shapes.h
tensorflow/lite/kernels/internal/reference/quantize.h
Expand All @@ -90,18 +93,16 @@ tensorflow/lite/kernels/internal/reference/resize_bilinear.h
tensorflow/lite/kernels/internal/reference/resize_nearest_neighbor.h
tensorflow/lite/kernels/internal/reference/reverse.h
tensorflow/lite/kernels/internal/reference/round.h
tensorflow/lite/kernels/internal/reference/select.h
tensorflow/lite/kernels/internal/reference/slice.h
tensorflow/lite/kernels/internal/reference/softmax.h
tensorflow/lite/kernels/internal/reference/space_to_batch_nd.h
tensorflow/lite/kernels/internal/reference/space_to_depth.h
tensorflow/lite/kernels/internal/reference/sub.h
tensorflow/lite/kernels/internal/reference/logistic.h
tensorflow/lite/kernels/internal/reference/lstm_cell.h
tensorflow/lite/kernels/internal/reference/select.h
tensorflow/lite/kernels/internal/reference/slice.h
tensorflow/lite/kernels/internal/reference/strided_slice.h
tensorflow/lite/kernels/internal/reference/sub.h
tensorflow/lite/kernels/internal/reference/tanh.h
tensorflow/lite/kernels/internal/reference/transpose.h
tensorflow/lite/kernels/internal/reference/transpose_conv.h
tensorflow/lite/kernels/internal/reference/transpose.h
tensorflow/lite/kernels/internal/cppmath.h
tensorflow/lite/kernels/internal/max.h
tensorflow/lite/kernels/internal/min.h
Expand Down
34 changes: 29 additions & 5 deletions tensorflow/lite/core/c/common.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,7 @@ limitations under the License.
#endif // TF_LITE_STATIC_MEMORY

#include <cstring>
#include <new>
#include <type_traits>
#include <utility>

Expand Down Expand Up @@ -111,6 +112,7 @@ TfLiteSparsity TfLiteSparsityClone(const TfLiteSparsity& src) {
if (src.dim_metadata) {
dst.dim_metadata = reinterpret_cast<TfLiteDimensionMetadata*>(
calloc(1, sizeof(TfLiteDimensionMetadata) * src.dim_metadata_size));
if (src.dim_metadata_size > 0 && !dst.dim_metadata) return TfLiteSparsity();
for (int i = 0; i < src.dim_metadata_size; ++i) {
dst.dim_metadata[i] = src.dim_metadata[i];
dst.dim_metadata[i].array_segments =
Expand All @@ -129,6 +131,7 @@ TfLiteSparsity* TfLiteSparsityClone(const TfLiteSparsity* const src) {
}
TfLiteSparsity* dst =
reinterpret_cast<TfLiteSparsity*>(calloc(1, sizeof(TfLiteSparsity)));
if (!dst) return nullptr;
*dst = TfLiteSparsityClone(*src);
return dst;
}
Expand All @@ -147,6 +150,7 @@ TfLiteQuantization TfLiteQuantizationClone(const TfLiteQuantization& src) {
break;
case kTfLiteAffineQuantization: {
dst.params = calloc(1, sizeof(TfLiteAffineQuantization));
if (!dst.params) return TfLiteQuantization();
const TfLiteAffineQuantization* const src_params =
reinterpret_cast<TfLiteAffineQuantization*>(src.params);
TfLiteAffineQuantization* const dst_params =
Expand All @@ -158,6 +162,7 @@ TfLiteQuantization TfLiteQuantizationClone(const TfLiteQuantization& src) {
}
case kTfLiteBlockwiseQuantization: {
dst.params = calloc(1, sizeof(TfLiteBlockwiseQuantization));
if (!dst.params) return TfLiteQuantization();
const TfLiteBlockwiseQuantization* const src_params =
(TfLiteBlockwiseQuantization*)(src.params);
TfLiteBlockwiseQuantization* const dst_params =
Expand Down Expand Up @@ -219,6 +224,9 @@ TfLiteFloatArray* TfLiteFloatArrayCopy(const TfLiteFloatArray* src) {
void TfLiteFloatArrayFree(TfLiteFloatArray* a) { TfLiteVarArrayFree(a); }

void TfLiteTensorDataFree(TfLiteTensor* t) {
if (t == nullptr) {
return;
}
if (t->allocation_type == kTfLiteVariantObject && t->data.data) {
delete static_cast<VariantData*>(t->data.data);
} else if (t->allocation_type == kTfLiteDynamic ||
Expand All @@ -238,6 +246,9 @@ void TfLiteTensorDataFree(TfLiteTensor* t) {
}

void TfLiteQuantizationFree(TfLiteQuantization* quantization) {
if (quantization == nullptr) {
return;
}
if (quantization->type == kTfLiteAffineQuantization) {
TfLiteAffineQuantization* q_params =
reinterpret_cast<TfLiteAffineQuantization*>(quantization->params);
Expand Down Expand Up @@ -294,6 +305,9 @@ void TfLiteSparsityFree(TfLiteSparsity* sparsity) {
}

void TfLiteTensorFree(TfLiteTensor* t) {
if (t == nullptr) {
return;
}
TfLiteTensorDataFree(t);
if (t->dims) TfLiteIntArrayFree(t->dims);
t->dims = nullptr;
Expand All @@ -308,7 +322,7 @@ void TfLiteTensorFree(TfLiteTensor* t) {
t->sparsity = nullptr;
}

TfLiteTensor TfLiteTensorClone(const TfLiteTensor src) {
TfLiteTensor TfLiteTensorClone(TfLiteTensor src) {
// We copy all of the source data first, then we clone the fields that can't
// be shared between two tensor instances.
TfLiteTensor dst = src;
Expand All @@ -335,16 +349,18 @@ TfLiteTensor TfLiteTensorClone(const TfLiteTensor src) {
break;
case kTfLiteAllocationStrategyMalloc:
dst.data.data = malloc(src.bytes);
if (src.bytes > 0 && !dst.data.data) return TfLiteTensor();
std::memcpy(dst.data.data, src.data.data, src.bytes);
break;
case kTfLiteAllocationStrategyNew:
// Special case for variant objects. They are allocated using new/delete
// but require using the `CloneTo` function.
if (src.allocation_type == kTfLiteVariantObject) {
dst.data.data = reinterpret_cast<const VariantData*>(src.data.data)
->CloneTo(nullptr);
dst.data.data =
static_cast<const VariantData*>(src.data.data)->CloneTo(nullptr);
} else {
dst.data.data = new char[src.bytes];
dst.data.data = new (std::nothrow) char[src.bytes];
if (src.bytes > 0 && !dst.data.data) return TfLiteTensor();
std::memcpy(dst.data.data, src.data.data, src.bytes);
}
break;
Expand Down Expand Up @@ -394,13 +410,21 @@ TfLiteStatus TfLiteTensorCopy(const TfLiteTensor* src, TfLiteTensor* dst) {
}
auto* dst_vd = static_cast<VariantData*>(dst->data.data);
auto* src_vd = static_cast<VariantData*>(src->data.data);
if (!src_vd) return kTfLiteError;

// `CloneTo` will handle the case when `dst_vd` is nullptr, so it is safe
// to `CloneTo` something which was "freed". Also, returning from `CloneTo`
// will implicitly cast to `VariantData`; don't need static cast here.
dst->data.data = src_vd->CloneTo(dst_vd);
} else {
memcpy(dst->data.raw, src->data.raw, src->bytes);
if (dst->allocation_type == kTfLiteVariantObject) {
TfLiteTensorDataFree(dst);
dst->allocation_type = src->allocation_type;
}
if (src->bytes > 0) {
if (!dst->data.raw || !src->data.raw) return kTfLiteError;
memcpy(dst->data.raw, src->data.raw, src->bytes);
}
}
dst->buffer_handle = src->buffer_handle;
dst->data_is_stale = src->data_is_stale;
Expand Down
41 changes: 35 additions & 6 deletions tensorflow/lite/core/c/common.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,6 +56,7 @@ limitations under the License.
#include <stdbool.h>
#include <stddef.h>
#include <stdint.h>
#include <stdio.h>

#include "tensorflow/lite/core/c/c_api_types.h" // IWYU pragma: export

Expand Down Expand Up @@ -277,13 +278,34 @@ void TfLiteFloatArrayFree(TfLiteFloatArray* a);
} \
} while (0)

#define TF_LITE_ENSURE_OK(context, status) \
do { \
const TfLiteStatus s = (status); \
if ((s) != kTfLiteOk) { \
return s; \
} \
#ifndef TF_LITE_STRIP_ERROR_STRINGS
#define TF_LITE_VAR_ARG_HEAD(FIRST, ...) FIRST
#define TF_LITE_STRINGIFY_HELPER(x) #x
#define TF_LITE_STRINGIFY(x) TF_LITE_STRINGIFY_HELPER(x)
// Checks that `status` evaluates to `kTfLiteOk`.
//
// Can take a printf style log message and its parameters after the status. The
// message will be printed using `TF_LITE_KERNEL_LOG` in case of error.
#define TF_LITE_ENSURE_OK(context, status, ...) \
do { \
const TfLiteStatus s = (status); \
if (s != kTfLiteOk) { \
if (sizeof(TF_LITE_VAR_ARG_HEAD("" __VA_ARGS__)) > sizeof("")) { \
TF_LITE_MAYBE_KERNEL_LOG((context), __FILE__ ":" TF_LITE_STRINGIFY( \
__LINE__) ": " __VA_ARGS__); \
} \
return s; \
} \
} while (0)
#else
#define TF_LITE_ENSURE_OK(context, status, ...) \
do { \
const TfLiteStatus s = (status); \
if ((s) != kTfLiteOk) { \
return s; \
} \
} while (0)
#endif

// `std::unreachable` not available until CC23.
#ifdef __GNUC__ // GCC, Clang, ICC
Expand Down Expand Up @@ -1060,6 +1082,13 @@ typedef struct TfLiteContext {
/// WARNING: This is an experimental interface that is subject to change.
TfLiteStatus (*ReleaseSubgraphContext)(struct TfLiteContext* context,
int subgraph_index);
#if defined(_WIN32)
/// Create a array of a given `size` (uninitialized entries).
TfLiteIntArray* (*TfLiteIntArrayCreate)(int size); // NOLINT

/// Free memory of array `a`.
void (*TfLiteIntArrayFree)(TfLiteIntArray* a); // NOLINT
#endif // defined(_WIN32)
} TfLiteContext;

/// `TfLiteOperator` is an external version of `TfLiteRegistration`
Expand Down
1 change: 1 addition & 0 deletions tensorflow/lite/kernels/internal/BUILD
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -146,6 +146,7 @@ cc_library(
copts = tflite_copts(),
deps = [
":compatibility",
"//tensorflow/lite/types:half",
],
)

Expand Down
20 changes: 20 additions & 0 deletions tensorflow/lite/kernels/internal/common.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,15 +78,23 @@ bool ReduceDimensionsForBroadcast(const RuntimeShape& input1_shape,
if (!broadcast_input1) {
broadcast_input1 = true;
broadcast_input2 = false;
if (num_compressed_dims >= MAX_DIM) return false;
num_compressed_dims++;
if (num_compressed_dims > MAX_DIM) {
return false;
}
}
compressed_input2_shape[num_compressed_dims - 1] *= input2_dim;
compressed_output_shape[num_compressed_dims - 1] *= input2_dim;
} else if (input2_dim == 1) {
if (!broadcast_input2) {
broadcast_input1 = false;
broadcast_input2 = true;
if (num_compressed_dims >= MAX_DIM) return false;
num_compressed_dims++;
if (num_compressed_dims > MAX_DIM) {
return false;
}
}
compressed_input1_shape[num_compressed_dims - 1] *= input1_dim;
compressed_output_shape[num_compressed_dims - 1] *= input1_dim;
Expand All @@ -95,7 +103,11 @@ bool ReduceDimensionsForBroadcast(const RuntimeShape& input1_shape,
if (broadcast_input1 || broadcast_input2 || first_nonunit) {
broadcast_input1 = false;
broadcast_input2 = false;
if (num_compressed_dims >= MAX_DIM) return false;
num_compressed_dims++;
if (num_compressed_dims > MAX_DIM) {
return false;
}
}
compressed_input1_shape[num_compressed_dims - 1] *= input1_dim;
compressed_input2_shape[num_compressed_dims - 1] *= input1_dim;
Expand All @@ -105,7 +117,11 @@ bool ReduceDimensionsForBroadcast(const RuntimeShape& input1_shape,
}
if (num_input1_dims > num_input2_dims) {
if (!broadcast_input2) {
if (num_compressed_dims >= MAX_DIM) return false;
num_compressed_dims++;
if (num_compressed_dims > MAX_DIM) {
return false;
}
}
for (size_t i = 0; i < num_input1_dims - num_input2_dims; i++) {
const size_t input1_dim = input1_dims[i];
Expand All @@ -117,7 +133,11 @@ bool ReduceDimensionsForBroadcast(const RuntimeShape& input1_shape,
}
} else if (num_input2_dims > num_input1_dims) {
if (!broadcast_input1) {
if (num_compressed_dims >= MAX_DIM) return false;
num_compressed_dims++;
if (num_compressed_dims > MAX_DIM) {
return false;
}
}
for (size_t i = 0; i < num_input2_dims - num_input1_dims; i++) {
const size_t input2_dim = input2_dims[i];
Expand Down
30 changes: 29 additions & 1 deletion tensorflow/lite/kernels/internal/reference/add.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@ limitations under the License.
#include "fixedpoint/fixedpoint.h"
#include "tensorflow/lite/kernels/internal/common.h"
#include "tensorflow/lite/kernels/internal/compatibility.h"
#include "tensorflow/lite/kernels/internal/reference/broadcast_loop.h"

namespace tflite {

Expand All @@ -39,7 +40,7 @@ inline void Add(const ArithmeticParams& params,
const int flat_size =
MatchingElementsSize(input1_shape, input2_shape, output_shape);
for (int i = 0; i < flat_size; ++i) {
output_data[i] = ActivationFunctionWithMinMax(
output_data[i] = ActivationFunctionWithMinMax<T>(
input1_data[i] + input2_data[i], activation_min, activation_max);
}
}
Expand Down Expand Up @@ -328,6 +329,20 @@ BroadcastAdd6DSlow(const ArithmeticParams& params,
constexpr int kMaxBroadcastDim = 6;
T activation_min, activation_max;
GetActivationParams(params, &activation_min, &activation_max);
const int broadcast_rank = std::max(
output_shape.DimensionsCount(),
std::max(input1_shape.DimensionsCount(), input2_shape.DimensionsCount()));
if (broadcast_rank > kMaxBroadcastDim) {
ForEachBroadcastedElement(
input1_shape, input2_shape, output_shape,
[&](int output_index, int input1_index, int input2_index) {
output_data[output_index] = ActivationFunctionWithMinMax(
static_cast<T>(input1_data[input1_index] +
input2_data[input2_index]),
activation_min, activation_max);
});
return;
}

// In Tensorflow, the dimensions are canonically named (batch_number, row,
// col, channel), with extents (batches, height, width, depth), with the
Expand Down Expand Up @@ -421,6 +436,19 @@ BroadcastAdd6DSlow(const ArithmeticParams& params,
const RuntimeShape& input2_shape, const T* input2_data,
const RuntimeShape& output_shape, T* output_data) {
constexpr int kMaxBroadcastDim = 6;
const int broadcast_rank = std::max(
output_shape.DimensionsCount(),
std::max(input1_shape.DimensionsCount(), input2_shape.DimensionsCount()));
if (broadcast_rank > kMaxBroadcastDim) {
ForEachBroadcastedElement(
input1_shape, input2_shape, output_shape,
[&](int output_index, int input1_index, int input2_index) {
AddElementwise(1, params, input1_data + input1_index,
input2_data + input2_index,
output_data + output_index);
});
return;
}

// In Tensorflow, the dimensions are canonically named (batch_number, row,
// col, channel), with extents (batches, height, width, depth), with the
Expand Down
Loading
Loading