irangeify most of runtime/core/exec_aten (#8612)

Had to skip scalar_type_util.h because it needs the constexpr patch to
land in PyTorch (and for us to update our pin to pick it up).

Author: swolchok

runtime/core/exec_aten/testing_util/targets.bzl

@@ -44,6 +44,7 @@ def define_common_targets():
                 "//executorch/runtime/core/exec_aten/util:scalar_type_util" + aten_suffix,
                 "//executorch/runtime/core/exec_aten/util:tensor_util" + aten_suffix,
                 "//executorch/runtime/core/exec_aten/util:tensor_dimension_limit",
+                "//executorch/runtime/core/portable_type/c10/c10:c10",
             ],
             exported_external_deps = [
                 "gmock" + aten_suffix,

runtime/core/exec_aten/testing_util/tensor_factory.h

@@ -5,6 +5,7 @@
 #include <algorithm>
 #include <cstdint>
 
+#include <c10/util/irange.h>
 #include <executorch/runtime/core/exec_aten/exec_aten.h>
 #include <executorch/runtime/core/exec_aten/util/dim_order_util.h>
 #include <executorch/runtime/core/exec_aten/util/scalar_type_util.h>
@@ -78,7 +79,7 @@ inline bool check_strides(
   // a.strides == (1, 1, 2). We want to sort create a mapping to make the
   // sorted_stride as (2, 1, 1) while sorted_size == (3, 2, 1)
   std::vector<std::int32_t> sorted_idx(sizes.size());
-  for (size_t i = 0; i < sizes.size(); i++) {
+  for (const auto i : c10::irange(sizes.size())) {
     sorted_idx[i] = i;
   }
   std::sort(
@@ -98,7 +99,7 @@ inline bool check_strides(
   // Use the mapping to rearrange the sizes and strides
   std::vector<std::int32_t> sorted_sizes(sizes.size());
   std::vector<std::int32_t> sorted_strides(sizes.size());
-  for (size_t i = 0; i < sizes.size(); i++) {
+  for (const auto i : c10::irange(sizes.size())) {
     sorted_sizes[i] = sizes[sorted_idx[i]] == 0 ? 1 : sizes[sorted_idx[i]];
     sorted_strides[i] = strides[sorted_idx[i]];
   }
@@ -132,7 +133,7 @@ inline bool check_dim_order(
   }
   size_t gauss_sum = 0;
   std::vector<int> count(dim_order.size(), 0);
-  for (int i = 0; i < dim_order.size(); i++) {
+  for (const auto i : c10::irange(dim_order.size())) {
     if (dim_order[i] >= sizes.size()) {
       return false;
     }
@@ -378,7 +379,7 @@ class TensorFactory {
     std::vector<executorch::aten::StridesType> contiguous_strides =
         internal::strides_from_dim_order(sizes, contiguous_dim_order);
 
-    for (int32_t i = 0; i < input.dim(); i++) {
+    for (const auto i : c10::irange(input.dim())) {
       ET_CHECK_MSG(
           input.strides()[i] == contiguous_strides[i],
           "Input tensor is not contiguous");
@@ -394,10 +395,10 @@ class TensorFactory {
     std::vector<ctype> channels_last_data(
         N * C * H * W); // Create a new blob with the same total size to contain
                         // channels_last data
-    for (int32_t n = 0; n < N; ++n) {
-      for (int32_t c = 0; c < C; ++c) {
-        for (int32_t h = 0; h < H; ++h) {
-          for (int32_t w = 0; w < W; ++w) {
+    for (const auto n : c10::irange(N)) {
+      for (const auto c : c10::irange(C)) {
+        for (const auto h : c10::irange(H)) {
+          for (const auto w : c10::irange(W)) {
             // Calculate the index in the original blob
             int32_t old_index = ((n * C + c) * H + h) * W + w;
             // Calculate the index in the new blob
@@ -614,7 +615,7 @@ inline void validate_strides(
     }
   }
   // No two dimensions can have same stride value
-  for (int32_t i = 0; i < strides.size(); ++i) {
+  for (const auto i : c10::irange(strides.size())) {
     for (int32_t j = i + 1; j < strides.size(); ++j) {
       if ((sizes[i] == 0) || (sizes[j] == 0) ||
           ((sizes[i] == 1) || (sizes[j] == 1))) {
@@ -830,7 +831,7 @@ class TensorFactory {
     // given strides is empty.
     if (!sizes.empty() && dim_order.empty()) {
       default_dim_order.resize(sizes.size(), 1);
-      for (size_t i = 0; i < sizes.size(); ++i) {
+      for (const auto i : c10::irange(sizes.size())) {
         default_dim_order[i] = i;
       }
     }
@@ -904,10 +905,10 @@ class TensorFactory {
     std::vector<ctype> channels_last_data(
         N * C * H * W); // Create a new blob with the same total size to contain
                         // channels_last data
-    for (int32_t n = 0; n < N; ++n) {
-      for (int32_t c = 0; c < C; ++c) {
-        for (int32_t h = 0; h < H; ++h) {
-          for (int32_t w = 0; w < W; ++w) {
+    for (const auto n : c10::irange(N)) {
+      for (const auto c : c10::irange(C)) {
+        for (const auto h : c10::irange(H)) {
+          for (const auto w : c10::irange(W)) {
             // Calculate the index in the original blob
             int32_t old_index = ((n * C + c) * H + h) * W + w;
             // Calculate the index in the new blob

runtime/core/exec_aten/testing_util/tensor_util.cpp

@@ -10,6 +10,7 @@
 #include <cstring>
 #include <ostream>
 
+#include <c10/util/irange.h>
 #include <executorch/runtime/core/exec_aten/exec_aten.h>
 #include <executorch/runtime/core/exec_aten/testing_util/tensor_factory.h>
 #include <executorch/runtime/core/exec_aten/testing_util/tensor_util.h>
@@ -50,7 +51,7 @@ bool data_is_close(
   if (a == b) {
     return true;
   }
-  for (size_t i = 0; i < numel; i++) {
+  for (const auto i : c10::irange(numel)) {
     const auto ai = a[i];
     const auto bi = b[i];
 
@@ -201,7 +202,7 @@ bool tensor_lists_are_close(
   if (num_tensors_a != num_tensors_b) {
     return false;
   }
-  for (size_t i = 0; i < num_tensors_a; i++) {
+  for (const auto i : c10::irange(num_tensors_a)) {
     if (!tensors_are_close(tensors_a[i], tensors_b[i], rtol, opt_atol)) {
       return false;
     }
@@ -245,7 +246,7 @@ template <typename T>
 std::ostream& print_data(std::ostream& os, const T* data, size_t numel) {
   // TODO(dbort): Make this smarter: show dimensions, listen to strides,
   // break up or truncate data when it's huge
-  for (auto i = 0; i < numel; i++) {
+  for (const auto i : c10::irange(numel)) {
     os << data[i];
     if (i < numel - 1) {
       os << ", ";
@@ -257,7 +258,7 @@ std::ostream& print_data(std::ostream& os, const T* data, size_t numel) {
 template <typename T>
 std::ostream&
 print_data(std::ostream& os, const etensor::complex<T>* data, size_t numel) {
-  for (auto i = 0; i < numel; i++) {
+  for (const auto i : c10::irange(numel)) {
     os << data[i].real_ << " + " << data[i].imag_ << "j";
     if (i < numel - 1) {
       os << ", ";
@@ -276,7 +277,7 @@ template <>
 std::ostream& print_data(std::ostream& os, const uint8_t* data, size_t numel) {
   // TODO(dbort): Make this smarter: show dimensions, listen to strides,
   // break up or truncate data when it's huge
-  for (auto i = 0; i < numel; i++) {
+  for (const auto i : c10::irange(numel)) {
     os << (uint64_t)data[i];
     if (i < numel - 1) {
       os << ", ";
@@ -292,7 +293,7 @@ std::ostream& print_data(std::ostream& os, const uint8_t* data, size_t numel) {
  */
 std::ostream& operator<<(std::ostream& os, const Tensor& t) {
   os << "ETensor(sizes={";
-  for (auto dim = 0; dim < t.dim(); dim++) {
+  for (const auto dim : c10::irange(t.dim())) {
     os << t.size(dim);
     if (dim < t.dim() - 1) {
       os << ", ";

runtime/core/exec_aten/testing_util/test/targets.bzl

@@ -25,5 +25,6 @@ def define_common_targets():
             preprocessor_flags = preprocessor_flags,
             deps = [
                 "//executorch/runtime/core/exec_aten/testing_util:tensor_util" + aten_suffix,
+                "//executorch/runtime/core/portable_type/c10/c10:c10",
             ],
         )

runtime/core/exec_aten/testing_util/test/tensor_factory_test.cpp

@@ -6,6 +6,7 @@
  * LICENSE file in the root directory of this source tree.
  */
 
+#include <c10/util/irange.h>
 #include <executorch/runtime/core/error.h>
 #include <executorch/runtime/core/exec_aten/testing_util/tensor_factory.h>
 #include <executorch/runtime/core/exec_aten/testing_util/tensor_util.h>
@@ -86,7 +87,7 @@ using torch::executor::TensorImpl;
       "Arrays are not equal size." #a1 " size:%zu," #a2 " size:%zu", \
       a1.size(),                                                     \
       a2.size());                                                    \
-  for (size_t i = 0; i < a1.size(); ++i) {                           \
+  for (const auto i : c10::irange(a1.size())) {                      \
     ET_CHECK_MSG(                                                    \
         a1[i] == a2[i],                                              \
         "Value mismatch at index %zu, " #a1                          \
@@ -784,7 +785,7 @@ void run_zeros_like_test(Tensor input) {
 
   // A Tensor created manually, that should be identical to `actual`.
   std::vector<int32_t> expected_data;
-  for (int i = 0; i < input.numel(); i++) {
+  for (const auto i : c10::irange(input.numel())) {
     expected_data.push_back(0);
   }
 #ifdef USE_ATEN_LIB
@@ -842,7 +843,7 @@ void run_ones_like_test(Tensor input) {
 
   // A Tensor created manually, that should be identical to `actual`.
   std::vector<int32_t> expected_data;
-  for (int i = 0; i < input.numel(); i++) {
+  for (const auto i : c10::irange(input.numel())) {
     expected_data.push_back(1);
   }
 #ifdef USE_ATEN_LIB

runtime/core/exec_aten/util/dim_order_util.h

@@ -13,6 +13,7 @@
 #include <cstdio>
 #include <cstring>
 
+#include <c10/util/irange.h>
 #include <executorch/runtime/core/error.h>
 #include <executorch/runtime/platform/assert.h>
 #include <executorch/runtime/platform/compiler.h>
@@ -23,7 +24,7 @@ namespace runtime {
 namespace {
 template <typename DimOrderType>
 bool validate_dim_order(const DimOrderType* dim_order, const size_t dims) {
-  for (size_t i = 0; i < dims; ++i) {
+  for (const auto i : c10::irange(dims)) {
     if (dim_order[i] >= static_cast<DimOrderType>(dims)) {
       return false;
     }
@@ -43,7 +44,7 @@ template <typename DimOrderType>
 inline bool is_contiguous_dim_order(
     const DimOrderType* dim_order,
     const size_t dims) {
-  for (size_t i = 0; i < dims; ++i) {
+  for (const auto i : c10::irange(dims)) {
     if (dim_order[i] != static_cast<DimOrderType>(i)) {
       return false;
     }

runtime/core/exec_aten/util/targets.bzl

@@ -40,6 +40,7 @@ def define_common_targets():
             ],
             exported_deps = [
                 "//executorch/runtime/core:core",
+                "//executorch/runtime/core/portable_type/c10/c10:c10",
             ],
             visibility = [
                 "//executorch/...",
@@ -62,6 +63,7 @@ def define_common_targets():
             exported_deps = [
                 ":tensor_dimension_limit",
                 "//executorch/runtime/core:core",
+                "//executorch/runtime/core/portable_type/c10/c10:c10",
             ] + [
                 "//executorch/runtime/core/exec_aten:lib" + aten_suffix,
                 ":scalar_type_util" + aten_suffix,

runtime/core/exec_aten/util/tensor_util.h

@@ -17,6 +17,7 @@
 
 #include <limits>
 
+#include <c10/util/irange.h>
 #include <executorch/runtime/core/array_ref.h>
 #include <executorch/runtime/core/error.h>
 #include <executorch/runtime/core/exec_aten/exec_aten.h>
@@ -277,7 +278,7 @@
         a_strides = a__.strides();                                             \
     const ::executorch::aten::ArrayRef<executorch::aten::StridesType>          \
         b_strides = b__.strides();                                             \
-    for (size_t i = 0; i < a__.dim(); i++) {                                   \
+    for (const auto i : c10::irange(a__.dim())) {                              \
       ET_CHECK_MSG(                                                            \
           a_strides[i] == b_strides[i],                                        \
           "a.strides()[%zu] shall equal to b.strides()[%zu], "                 \
@@ -309,7 +310,7 @@
         b_strides = b__.strides();                                      \
     const ::executorch::aten::ArrayRef<executorch::aten::StridesType>   \
         c_strides = c__.strides();                                      \
-    for (size_t i = 0; i < a__.dim(); i++) {                            \
+    for (const auto i : c10::irange(a__.dim())) {                       \
       ET_CHECK_MSG(                                                     \
           a_strides[i] == b_strides[i] && b_strides[i] == c_strides[i], \
           "a_strides[%zu], b_strides[%zu] and c_strides[%zu] "          \
@@ -967,7 +968,7 @@ inline size_t coordinateToIndex(
     const executorch::aten::Tensor& tensor,
     const size_t* const coordinate) {
   size_t index = 0;
-  for (int d = 0; d < tensor.dim(); ++d) {
+  for (const auto d : c10::irange(tensor.dim())) {
     index += coordinate[d] * getTrailingDims(tensor, d);
   }
   return index;
@@ -999,7 +1000,7 @@ inline size_t coordinateToIndexWithTrailingDimsMemo(
     const size_t* const coordinate,
     const size_t trailing_dims_memo[kTensorDimensionLimit]) {
   size_t index = 0;
-  for (int d = 0; d < tensor.dim(); ++d) {
+  for (const auto d : c10::irange(tensor.dim())) {
     index += coordinate[d] * trailing_dims_memo[d];
   }
   return index;
@@ -1021,7 +1022,7 @@ inline void indexToCoordinate(
     size_t index,
     size_t* coordinate) {
   ET_CHECK(index < static_cast<size_t>(tensor.numel()));
-  for (auto i = 0; i < tensor.dim(); ++i) {
+  for (const auto i : c10::irange(tensor.dim())) {
     auto dim = tensor.dim() - 1 - i;
     size_t dim_size = tensor.size(dim);
     coordinate[dim] = index % dim_size;
@@ -1211,7 +1212,7 @@ ET_NODISCARD inline Error resize_tensor(
   std::array<executorch::aten::SizesType, kTensorDimensionLimit>
       new_sizes_casted{};
   size_t new_sizes_ndim = new_sizes.size();
-  for (size_t i = 0; i < new_sizes_ndim; ++i) {
+  for (const auto i : c10::irange(new_sizes_ndim)) {
     new_sizes_casted[i] =
         static_cast<executorch::aten::SizesType>(new_sizes[i]);
   }
@@ -1342,7 +1343,7 @@ inline size_t calculate_linear_index(
     const executorch::aten::StridesType* strides,
     const size_t ndim) {
   size_t index = 0;
-  for (size_t i = 0; i < ndim; i++) {
+  for (const auto i : c10::irange(ndim)) {
     index += coordinate[i] * strides[i];
   }
   return index;

runtime/core/exec_aten/util/tensor_util_aten.cpp

@@ -9,6 +9,7 @@
 #include <executorch/runtime/core/exec_aten/util/tensor_util.h>
 
 #include <ATen/Tensor.h> // @manual
+#include <c10/util/irange.h>
 #include <executorch/runtime/platform/assert.h>
 
 namespace executorch {
@@ -41,7 +42,7 @@ bool tensor_has_valid_dim_order(at::Tensor t) {
 
   if (!validate_dim_order(dim_order, t.dim())) {
     ET_LOG(Error, "Tensor dim order is not valid:");
-    for (size_t d = 0; d < t.dim(); ++d) {
+    for (const auto d : c10::irange(t.dim())) {
       ET_LOG(
           Error,
           "    dim_order(%zu): %zu",
@@ -66,7 +67,7 @@ inline bool tensor_is_default_or_channels_last_dim_order(at::Tensor t) {
     ET_LOG(
         Error,
         "Expected tensor to have default or channels last dim order, but got");
-    for (size_t d = 0; d < t.dim(); ++d) {
+    for (const auto d : c10::irange(t.dim())) {
       ET_LOG(
           Error,
           "    dim_order(%zu): %zu",
@@ -96,7 +97,7 @@ bool tensors_have_same_dim_order(
   bool all_channels_last =
       is_channels_last_dim_order(first_dim_order, tensor_list[0].dim());
 
-  for (size_t i = 1; i < tensor_list.size(); ++i) {
+  for (const auto i : c10::irange(1, tensor_list.size())) {
     ET_CHECK_OR_RETURN_FALSE(
         get_dim_order(tensor_list[i], other_dim_order, tensor_list[i].dim()) ==
             Error::Ok,