experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Fully qualifying uses of tensorflow::int64.

Browse Source 
PiperOrigin-RevId: 321399703
Change-Id: I42732ead99e062444fa5c507f9fce10f1ace765c
This commit is contained in:
A. Unique TensorFlower 2020-07-15 11:17:18 -07:00 committed by TensorFlower Gardener
parent ee781437e8
commit 4c7d80b96a
5 changed files with 127 additions and 113 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

36
tensorflow/compiler/xla/service/cpu/runtime_conv2d.cc
View File

@ -23,16 +23,18 @@ limitations under the License.
#include "tensorflow/core/platform/dynamic_annotations.h"
#include "tensorflow/core/platform/types.h"
using tensorflow::int64;
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_EigenConvF32(
const void* run_options_ptr, float* out, float* lhs, float* rhs,
int64 input_batch, int64 input_rows, int64 input_cols, int64 input_channels,
int64 kernel_rows, int64 kernel_cols, int64 kernel_channels,
int64 kernel_filters, int64 output_rows, int64 output_cols,
int64 row_stride, int64 col_stride, int64 padding_top, int64 padding_bottom,
int64 padding_left, int64 padding_right, int64 lhs_row_dilation,
int64 lhs_col_dilation, int64 rhs_row_dilation, int64 rhs_col_dilation) {
tensorflow::int64 input_batch, tensorflow::int64 input_rows,
tensorflow::int64 input_cols, tensorflow::int64 input_channels,
tensorflow::int64 kernel_rows, tensorflow::int64 kernel_cols,
tensorflow::int64 kernel_channels, tensorflow::int64 kernel_filters,
tensorflow::int64 output_rows, tensorflow::int64 output_cols,
tensorflow::int64 row_stride, tensorflow::int64 col_stride,
tensorflow::int64 padding_top, tensorflow::int64 padding_bottom,
tensorflow::int64 padding_left, tensorflow::int64 padding_right,
tensorflow::int64 lhs_row_dilation, tensorflow::int64 lhs_col_dilation,
tensorflow::int64 rhs_row_dilation, tensorflow::int64 rhs_col_dilation) {
const xla::ExecutableRunOptions* run_options =
static_cast<const xla::ExecutableRunOptions*>(run_options_ptr);
XLA_LIGHTWEIGHT_CHECK(run_options->intra_op_thread_pool() != nullptr);
@ -46,13 +48,17 @@ TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_EigenConvF32(
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_EigenConvF16(
const void* run_options_ptr, Eigen::half* out, Eigen::half* lhs,
Eigen::half* rhs, int64 input_batch, int64 input_rows, int64 input_cols,
int64 input_channels, int64 kernel_rows, int64 kernel_cols,
int64 kernel_channels, int64 kernel_filters, int64 output_rows,
int64 output_cols, int64 row_stride, int64 col_stride, int64 padding_top,
int64 padding_bottom, int64 padding_left, int64 padding_right,
int64 lhs_row_dilation, int64 lhs_col_dilation, int64 rhs_row_dilation,
int64 rhs_col_dilation) {
Eigen::half* rhs, tensorflow::int64 input_batch,
tensorflow::int64 input_rows, tensorflow::int64 input_cols,
tensorflow::int64 input_channels, tensorflow::int64 kernel_rows,
tensorflow::int64 kernel_cols, tensorflow::int64 kernel_channels,
tensorflow::int64 kernel_filters, tensorflow::int64 output_rows,
tensorflow::int64 output_cols, tensorflow::int64 row_stride,
tensorflow::int64 col_stride, tensorflow::int64 padding_top,
tensorflow::int64 padding_bottom, tensorflow::int64 padding_left,
tensorflow::int64 padding_right, tensorflow::int64 lhs_row_dilation,
tensorflow::int64 lhs_col_dilation, tensorflow::int64 rhs_row_dilation,
tensorflow::int64 rhs_col_dilation) {
const xla::ExecutableRunOptions* run_options =
static_cast<const xla::ExecutableRunOptions*>(run_options_ptr);
XLA_LIGHTWEIGHT_CHECK(run_options->intra_op_thread_pool() != nullptr);

43
tensorflow/compiler/xla/service/cpu/runtime_key_value_sort.cc
View File

@ -25,21 +25,16 @@ limitations under the License.
#include "tensorflow/core/platform/macros.h"
#include "tensorflow/core/platform/types.h"
namespace {
using tensorflow::int32;
using tensorflow::int64;
} // namespace
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_KeyValueSort(
int64 a, int64 b, int64 c, char** values, int32 values_count,
int32* values_primitive_type_size_in_bytes, bool is_stable,
char* run_options, int64* prof_counters,
tensorflow::int64 a, tensorflow::int64 b, tensorflow::int64 c, char** values, tensorflow::int32 values_count,
tensorflow::int32* values_primitive_type_size_in_bytes, bool is_stable,
char* run_options, tensorflow::int64* prof_counters,
void (*less_than)(char*, char*, char**, char**, tensorflow::int64*)) {
// 'values' and 'values_primitive_type_size_in_bytes' are managed by the JIT
// code, so msan can't tell they are initialized.
TF_ANNOTATE_MEMORY_IS_INITIALIZED(values, values_count * sizeof(char*));
TF_ANNOTATE_MEMORY_IS_INITIALIZED(values_primitive_type_size_in_bytes,
values_count * sizeof(int32));
values_count * sizeof(tensorflow::int32));
// High-level idea of the iteration/sorting logic:
// Conceptually we have a 3-dimensional shape [a, b, c]. b corresponds to the
@ -50,16 +45,16 @@ TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_KeyValueSort(
// 'base_offset' value which points to the first element in that row, and add
// i * c for accessing the 'i'-th element in that row.
int64 sort_dimension_elements = b;
int64 num_iteration_elements = a * c;
int64 sort_dimension_offset = c;
tensorflow::int64 sort_dimension_elements = b;
tensorflow::int64 num_iteration_elements = a * c;
tensorflow::int64 sort_dimension_offset = c;
std::unique_ptr<int64[]> indices(new int64[sort_dimension_elements]);
std::unique_ptr<tensorflow::int64[]> indices(new tensorflow::int64[sort_dimension_elements]);
std::unique_ptr<char*[]> comparison_values(new char*[2 * values_count]);
std::iota(indices.get(), indices.get() + sort_dimension_elements, 0);
std::unique_ptr<std::string[]> reordered_values(
new std::string[sort_dimension_elements]);
for (int64 index = 0; index < num_iteration_elements; ++index) {
for (tensorflow::int64 index = 0; index < num_iteration_elements; ++index) {
// If the sort should be stable, we have to reinitialize indices to iota to
// guarantee that we still keep the relative order in case of ties.
if (is_stable && index > 0) {
@ -71,14 +66,14 @@ TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_KeyValueSort(
// calculating the base offset, we need to multiply the index into the 'a'
// dimension with 'b' * 'c'.
// 'index' / 'c' * 'c' * 'b' = ('index' - 'index' % 'c') * 'b'.
int64 base_offset =
tensorflow::int64 base_offset =
index % sort_dimension_offset +
(index - index % sort_dimension_offset) * sort_dimension_elements;
auto compare_function = [&](int64 a, int64 b) -> bool {
for (int32 i = 0; i < values_count; ++i) {
int64 memory_index_lhs = (base_offset + a * sort_dimension_offset) *
auto compare_function = [&](tensorflow::int64 a, tensorflow::int64 b) -> bool {
for (tensorflow::int32 i = 0; i < values_count; ++i) {
tensorflow::int64 memory_index_lhs = (base_offset + a * sort_dimension_offset) *
values_primitive_type_size_in_bytes[i];
int64 memory_index_rhs = (base_offset + b * sort_dimension_offset) *
tensorflow::int64 memory_index_rhs = (base_offset + b * sort_dimension_offset) *
values_primitive_type_size_in_bytes[i];
comparison_values[i * 2] = values[i] + memory_index_lhs;
comparison_values[i * 2 + 1] = values[i] + memory_index_rhs;
@ -97,9 +92,9 @@ TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_KeyValueSort(
}
// Reorder the values according to the order defined by 'indices'.
for (int32 idx = 0; idx < values_count; ++idx) {
for (int64 i = 0; i < sort_dimension_elements; ++i) {
int64 memory_index =
for (tensorflow::int32 idx = 0; idx < values_count; ++idx) {
for (tensorflow::int64 i = 0; i < sort_dimension_elements; ++i) {
tensorflow::int64 memory_index =
(base_offset + indices[i] * sort_dimension_offset) *
values_primitive_type_size_in_bytes[idx];
@ -107,8 +102,8 @@ TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_KeyValueSort(
std::string(values[idx] + memory_index,
values_primitive_type_size_in_bytes[idx]);
}
for (int64 i = 0; i < sort_dimension_elements; ++i) {
int64 memory_index = (base_offset + i * sort_dimension_offset) *
for (tensorflow::int64 i = 0; i < sort_dimension_elements; ++i) {
tensorflow::int64 memory_index = (base_offset + i * sort_dimension_offset) *
values_primitive_type_size_in_bytes[idx];
memcpy(values[idx] + memory_index, reordered_values[i].c_str(),
values_primitive_type_size_in_bytes[idx]);

56
tensorflow/compiler/xla/service/cpu/runtime_matmul.cc
View File

@ -27,9 +27,6 @@ limitations under the License.
#include "tensorflow/core/kernels/eigen_contraction_kernel.h"
#endif
using tensorflow::int32;
using tensorflow::int64;
namespace {
bool Is16BytesAligned(void* ptr) {
@ -37,19 +34,20 @@ bool Is16BytesAligned(void* ptr) {
}
template <typename T, Eigen::AlignmentType Alignment>
void MatMul(const void* run_options_ptr, T* out, T* lhs, T* rhs, int64 m,
int64 n, int64 k, int32 transpose_lhs, int32 transpose_rhs) {
void MatMul(const void* run_options_ptr, T* out, T* lhs, T* rhs,
tensorflow::int64 m, tensorflow::int64 n, tensorflow::int64 k,
tensorflow::int32 transpose_lhs, tensorflow::int32 transpose_rhs) {
const xla::ExecutableRunOptions* run_options =
static_cast<const xla::ExecutableRunOptions*>(run_options_ptr);
int64 lhs_rows = m;
int64 lhs_cols = k;
tensorflow::int64 lhs_rows = m;
tensorflow::int64 lhs_cols = k;
if (transpose_lhs) {
std::swap(lhs_rows, lhs_cols);
}
int64 rhs_rows = k;
int64 rhs_cols = n;
tensorflow::int64 rhs_rows = k;
tensorflow::int64 rhs_cols = n;
if (transpose_rhs) {
std::swap(rhs_rows, rhs_cols);
}
@ -75,8 +73,9 @@ void MatMul(const void* run_options_ptr, T* out, T* lhs, T* rhs, int64 m,
template <typename T>
void MatMulDispatch(const void* run_options_ptr, T* out, T* lhs, T* rhs,
int64 m, int64 n, int64 k, int32 transpose_lhs,
int32 transpose_rhs) {
tensorflow::int64 m, tensorflow::int64 n,
tensorflow::int64 k, tensorflow::int32 transpose_lhs,
tensorflow::int32 transpose_rhs) {
bool all_buffers_16b_aligned =
Is16BytesAligned(out) && Is16BytesAligned(lhs) && Is16BytesAligned(rhs);
@ -94,45 +93,52 @@ void MatMulDispatch(const void* run_options_ptr, T* out, T* lhs, T* rhs,
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_EigenMatMulF16(
const void* run_options_ptr, Eigen::half* out, Eigen::half* lhs,
Eigen::half* rhs, int64 m, int64 n, int64 k, int32 transpose_lhs,
int32 transpose_rhs) {
Eigen::half* rhs, tensorflow::int64 m, tensorflow::int64 n,
tensorflow::int64 k, tensorflow::int32 transpose_lhs,
tensorflow::int32 transpose_rhs) {
MatMulDispatch<Eigen::half>(run_options_ptr, out, lhs, rhs, m, n, k,
transpose_lhs, transpose_rhs);
}
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_EigenMatMulF32(
const void* run_options_ptr, float* out, float* lhs, float* rhs, int64 m,
int64 n, int64 k, int32 transpose_lhs, int32 transpose_rhs) {
const void* run_options_ptr, float* out, float* lhs, float* rhs,
tensorflow::int64 m, tensorflow::int64 n, tensorflow::int64 k,
tensorflow::int32 transpose_lhs, tensorflow::int32 transpose_rhs) {
MatMulDispatch<float>(run_options_ptr, out, lhs, rhs, m, n, k, transpose_lhs,
transpose_rhs);
}
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_EigenMatMulF64(
const void* run_options_ptr, double* out, double* lhs, double* rhs, int64 m,
int64 n, int64 k, int32 transpose_lhs, int32 transpose_rhs) {
const void* run_options_ptr, double* out, double* lhs, double* rhs,
tensorflow::int64 m, tensorflow::int64 n, tensorflow::int64 k,
tensorflow::int32 transpose_lhs, tensorflow::int32 transpose_rhs) {
MatMulDispatch<double>(run_options_ptr, out, lhs, rhs, m, n, k, transpose_lhs,
transpose_rhs);
}
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_EigenMatMulC64(
const void* run_options_ptr, std::complex<float>* out,
std::complex<float>* lhs, std::complex<float>* rhs, int64 m, int64 n,
int64 k, int32 transpose_lhs, int32 transpose_rhs) {
std::complex<float>* lhs, std::complex<float>* rhs, tensorflow::int64 m,
tensorflow::int64 n, tensorflow::int64 k, tensorflow::int32 transpose_lhs,
tensorflow::int32 transpose_rhs) {
MatMulDispatch<std::complex<float>>(run_options_ptr, out, lhs, rhs, m, n, k,
transpose_lhs, transpose_rhs);
}
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_EigenMatMulC128(
const void* run_options_ptr, std::complex<double>* out,
std::complex<double>* lhs, std::complex<double>* rhs, int64 m, int64 n,
int64 k, int32 transpose_lhs, int32 transpose_rhs) {
std::complex<double>* lhs, std::complex<double>* rhs, tensorflow::int64 m,
tensorflow::int64 n, tensorflow::int64 k, tensorflow::int32 transpose_lhs,
tensorflow::int32 transpose_rhs) {
MatMulDispatch<std::complex<double>>(run_options_ptr, out, lhs, rhs, m, n, k,
transpose_lhs, transpose_rhs);
}
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void __xla_cpu_runtime_EigenMatMulS32(
const void* run_options_ptr, int32* out, int32* lhs, int32* rhs, int64 m,
int64 n, int64 k, int32 transpose_lhs, int32 transpose_rhs) {
MatMulDispatch<int32>(run_options_ptr, out, lhs, rhs, m, n, k, transpose_lhs,
transpose_rhs);
const void* run_options_ptr, tensorflow::int32* out, tensorflow::int32* lhs,
tensorflow::int32* rhs, tensorflow::int64 m, tensorflow::int64 n,
tensorflow::int64 k, tensorflow::int32 transpose_lhs,
tensorflow::int32 transpose_rhs) {
MatMulDispatch<tensorflow::int32>(run_options_ptr, out, lhs, rhs, m, n, k,
transpose_lhs, transpose_rhs);
}

36
tensorflow/compiler/xla/service/cpu/runtime_single_threaded_conv2d.cc
View File

@ -19,18 +19,20 @@ limitations under the License.
#include "tensorflow/core/platform/dynamic_annotations.h"
#include "tensorflow/core/platform/types.h"
using tensorflow::int64;
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void
__xla_cpu_runtime_EigenSingleThreadedConvF16(
const void* run_options_ptr, Eigen::half* out, Eigen::half* lhs,
Eigen::half* rhs, int64 input_batch, int64 input_rows, int64 input_cols,
int64 input_channels, int64 kernel_rows, int64 kernel_cols,
int64 kernel_channels, int64 kernel_filters, int64 output_rows,
int64 output_cols, int64 row_stride, int64 col_stride, int64 padding_top,
int64 padding_bottom, int64 padding_left, int64 padding_right,
int64 lhs_row_dilation, int64 lhs_col_dilation, int64 rhs_row_dilation,
int64 rhs_col_dilation) {
Eigen::half* rhs, tensorflow::int64 input_batch,
tensorflow::int64 input_rows, tensorflow::int64 input_cols,
tensorflow::int64 input_channels, tensorflow::int64 kernel_rows,
tensorflow::int64 kernel_cols, tensorflow::int64 kernel_channels,
tensorflow::int64 kernel_filters, tensorflow::int64 output_rows,
tensorflow::int64 output_cols, tensorflow::int64 row_stride,
tensorflow::int64 col_stride, tensorflow::int64 padding_top,
tensorflow::int64 padding_bottom, tensorflow::int64 padding_left,
tensorflow::int64 padding_right, tensorflow::int64 lhs_row_dilation,
tensorflow::int64 lhs_col_dilation, tensorflow::int64 rhs_row_dilation,
tensorflow::int64 rhs_col_dilation) {
tensorflow::xla::EigenConvImpl(
Eigen::DefaultDevice(), out, lhs, rhs, input_batch, input_rows,
input_cols, input_channels, kernel_rows, kernel_cols, kernel_channels,
@ -42,12 +44,16 @@ __xla_cpu_runtime_EigenSingleThreadedConvF16(
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void
__xla_cpu_runtime_EigenSingleThreadedConvF32(
const void* run_options_ptr, float* out, float* lhs, float* rhs,
int64 input_batch, int64 input_rows, int64 input_cols, int64 input_channels,
int64 kernel_rows, int64 kernel_cols, int64 kernel_channels,
int64 kernel_filters, int64 output_rows, int64 output_cols,
int64 row_stride, int64 col_stride, int64 padding_top, int64 padding_bottom,
int64 padding_left, int64 padding_right, int64 lhs_row_dilation,
int64 lhs_col_dilation, int64 rhs_row_dilation, int64 rhs_col_dilation) {
tensorflow::int64 input_batch, tensorflow::int64 input_rows,
tensorflow::int64 input_cols, tensorflow::int64 input_channels,
tensorflow::int64 kernel_rows, tensorflow::int64 kernel_cols,
tensorflow::int64 kernel_channels, tensorflow::int64 kernel_filters,
tensorflow::int64 output_rows, tensorflow::int64 output_cols,
tensorflow::int64 row_stride, tensorflow::int64 col_stride,
tensorflow::int64 padding_top, tensorflow::int64 padding_bottom,
tensorflow::int64 padding_left, tensorflow::int64 padding_right,
tensorflow::int64 lhs_row_dilation, tensorflow::int64 lhs_col_dilation,
tensorflow::int64 rhs_row_dilation, tensorflow::int64 rhs_col_dilation) {
tensorflow::xla::EigenConvImpl(
Eigen::DefaultDevice(), out, lhs, rhs, input_batch, input_rows,
input_cols, input_channels, kernel_rows, kernel_cols, kernel_channels,

69
tensorflow/compiler/xla/service/cpu/runtime_single_threaded_matmul.cc
View File

@ -23,9 +23,6 @@ limitations under the License.
#include "tensorflow/core/kernels/eigen_contraction_kernel.h"
#endif
using tensorflow::int32;
using tensorflow::int64;
namespace {
bool Is16BytesAligned(void* ptr) {
@ -33,16 +30,17 @@ bool Is16BytesAligned(void* ptr) {
}
template <typename T, Eigen::AlignmentType Alignment>
void MatMul(const void* run_options_ptr, T* out, T* lhs, T* rhs, int64 m,
int64 n, int64 k, int32 transpose_lhs, int32 transpose_rhs) {
int64 lhs_rows = m;
int64 lhs_cols = k;
void MatMul(const void* run_options_ptr, T* out, T* lhs, T* rhs,
tensorflow::int64 m, tensorflow::int64 n, tensorflow::int64 k,
tensorflow::int32 transpose_lhs, tensorflow::int32 transpose_rhs) {
tensorflow::int64 lhs_rows = m;
tensorflow::int64 lhs_cols = k;
if (transpose_lhs) {
std::swap(lhs_rows, lhs_cols);
}
int64 rhs_rows = k;
int64 rhs_cols = n;
tensorflow::int64 rhs_rows = k;
tensorflow::int64 rhs_cols = n;
if (transpose_rhs) {
std::swap(rhs_rows, rhs_cols);
}
@ -67,8 +65,10 @@ void MatMul(const void* run_options_ptr, T* out, T* lhs, T* rhs, int64 m,
template <typename T>
void SingleThreadedMatMulDispatch(const void* run_options_ptr, T* out, T* lhs,
T* rhs, int64 m, int64 n, int64 k,
int32 transpose_lhs, int32 transpose_rhs) {
T* rhs, tensorflow::int64 m,
tensorflow::int64 n, tensorflow::int64 k,
tensorflow::int32 transpose_lhs,
tensorflow::int32 transpose_rhs) {
bool all_buffers_16b_aligned =
Is16BytesAligned(out) && Is16BytesAligned(lhs) && Is16BytesAligned(rhs);
@ -86,28 +86,27 @@ void SingleThreadedMatMulDispatch(const void* run_options_ptr, T* out, T* lhs,
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void
__xla_cpu_runtime_EigenSingleThreadedMatMulF16(
const void* run_options_ptr, Eigen::half* out, Eigen::half* lhs,
Eigen::half* rhs, int64 m, int64 n, int64 k, int32 transpose_lhs,
int32 transpose_rhs) {
Eigen::half* rhs, tensorflow::int64 m, tensorflow::int64 n,
tensorflow::int64 k, tensorflow::int32 transpose_lhs,
tensorflow::int32 transpose_rhs) {
SingleThreadedMatMulDispatch<Eigen::half>(run_options_ptr, out, lhs, rhs, m,
n, k, transpose_lhs, transpose_rhs);
}
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void
__xla_cpu_runtime_EigenSingleThreadedMatMulF32(const void* run_options_ptr,
float* out, float* lhs,
float* rhs, int64 m, int64 n,
int64 k, int32 transpose_lhs,
int32 transpose_rhs) {
__xla_cpu_runtime_EigenSingleThreadedMatMulF32(
const void* run_options_ptr, float* out, float* lhs, float* rhs,
tensorflow::int64 m, tensorflow::int64 n, tensorflow::int64 k,
tensorflow::int32 transpose_lhs, tensorflow::int32 transpose_rhs) {
SingleThreadedMatMulDispatch<float>(run_options_ptr, out, lhs, rhs, m, n, k,
transpose_lhs, transpose_rhs);
}
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void
__xla_cpu_runtime_EigenSingleThreadedMatMulF64(const void* run_options_ptr,
double* out, double* lhs,
double* rhs, int64 m, int64 n,
int64 k, int32 transpose_lhs,
int32 transpose_rhs) {
__xla_cpu_runtime_EigenSingleThreadedMatMulF64(
const void* run_options_ptr, double* out, double* lhs, double* rhs,
tensorflow::int64 m, tensorflow::int64 n, tensorflow::int64 k,
tensorflow::int32 transpose_lhs, tensorflow::int32 transpose_rhs) {
SingleThreadedMatMulDispatch<double>(run_options_ptr, out, lhs, rhs, m, n, k,
transpose_lhs, transpose_rhs);
}
@ -115,8 +114,9 @@ __xla_cpu_runtime_EigenSingleThreadedMatMulF64(const void* run_options_ptr,
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void
__xla_cpu_runtime_EigenSingleThreadedMatMulC64(
const void* run_options_ptr, std::complex<float>* out,
std::complex<float>* lhs, std::complex<float>* rhs, int64 m, int64 n,
int64 k, int32 transpose_lhs, int32 transpose_rhs) {
std::complex<float>* lhs, std::complex<float>* rhs, tensorflow::int64 m,
tensorflow::int64 n, tensorflow::int64 k, tensorflow::int32 transpose_lhs,
tensorflow::int32 transpose_rhs) {
SingleThreadedMatMulDispatch<std::complex<float>>(
run_options_ptr, out, lhs, rhs, m, n, k, transpose_lhs, transpose_rhs);
}
@ -124,18 +124,19 @@ __xla_cpu_runtime_EigenSingleThreadedMatMulC64(
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void
__xla_cpu_runtime_EigenSingleThreadedMatMulC128(
const void* run_options_ptr, std::complex<double>* out,
std::complex<double>* lhs, std::complex<double>* rhs, int64 m, int64 n,
int64 k, int32 transpose_lhs, int32 transpose_rhs) {
std::complex<double>* lhs, std::complex<double>* rhs, tensorflow::int64 m,
tensorflow::int64 n, tensorflow::int64 k, tensorflow::int32 transpose_lhs,
tensorflow::int32 transpose_rhs) {
SingleThreadedMatMulDispatch<std::complex<double>>(
run_options_ptr, out, lhs, rhs, m, n, k, transpose_lhs, transpose_rhs);
}
TF_ATTRIBUTE_NO_SANITIZE_MEMORY void
__xla_cpu_runtime_EigenSingleThreadedMatMulS32(const void* run_options_ptr,
int32* out, int32* lhs,
int32* rhs, int64 m, int64 n,
int64 k, int32 transpose_lhs,
int32 transpose_rhs) {
SingleThreadedMatMulDispatch<int32>(run_options_ptr, out, lhs, rhs, m, n, k,
transpose_lhs, transpose_rhs);
__xla_cpu_runtime_EigenSingleThreadedMatMulS32(
const void* run_options_ptr, tensorflow::int32* out, tensorflow::int32* lhs,
tensorflow::int32* rhs, tensorflow::int64 m, tensorflow::int64 n,
tensorflow::int64 k, tensorflow::int32 transpose_lhs,
tensorflow::int32 transpose_rhs) {
SingleThreadedMatMulDispatch<tensorflow::int32>(
run_options_ptr, out, lhs, rhs, m, n, k, transpose_lhs, transpose_rhs);
}