feat(kernel): 实现 concat cpu kernel

Signed-off-by: YdrMaster <ydrml@hotmail.com>
InfiniTensor · Nov 11, 2023 · 203cc42 · 203cc42
1 parent 8ec8306
commit 203cc42
Showing 3 changed files with 16 additions and 16 deletions.
diff --git a/src/04kernel/src/kernels/concat/cpu_kernel.cc b/src/04kernel/src/kernels/concat/cpu_kernel.cc
@@ -25,18 +25,18 @@ namespace refactor::kernel {
     Routine K::lower(Resources &) const noexcept {
         using namespace runtime;
         return [info = this->info](Resources &, void const **inputs, void **outputs) {
-            // auto data = reinterpret_cast<uint8_t const *>(inputs[0]);
-            // std::for_each_n(std::execution::par_unseq,
-            //                 natural_t(0), info.blockCount,
-            //                 [=, &info](auto i) {
-            //                     auto offset = i * info.sum;
-            //                     for (auto j : range0_(info.segments.size())) {
-            //                         auto len = info.segments[j];
-            //                         auto out = reinterpret_cast<uint8_t *>(outputs[j]);
-            //                         std::memcpy(out + i * len, data + offset, len);
-            //                         offset += len;
-            //                     }
-            //                 });
+            auto dst = reinterpret_cast<uint8_t *>(outputs[0]);
+            std::for_each_n(std::execution::par_unseq,
+                            natural_t(0), info.blockCount,
+                            [=, &info](auto i) {
+                                auto offset = i * info.sum;
+                                for (auto j : range0_(info.segments.size())) {
+                                    auto len = info.segments[j];
+                                    auto src = reinterpret_cast<uint8_t const *>(inputs[j]);
+                                    std::memcpy(dst + offset, src + i * len, len);
+                                    offset += len;
+                                }
+                            });
         };
     }
 

diff --git a/src/04kernel/src/kernels/split/cpu_kernel.cc b/src/04kernel/src/kernels/split/cpu_kernel.cc
@@ -25,15 +25,15 @@ namespace refactor::kernel {
     Routine K::lower(Resources &) const noexcept {
         using namespace runtime;
         return [info = this->info](Resources &, void const **inputs, void **outputs) {
-            auto data = reinterpret_cast<uint8_t const *>(inputs[0]);
+            auto src = reinterpret_cast<uint8_t const *>(inputs[0]);
             std::for_each_n(std::execution::par_unseq,
                             natural_t(0), info.blockCount,
                             [=, &info](auto i) {
                                 auto offset = i * info.sum;
                                 for (auto j : range0_(info.segments.size())) {
                                     auto len = info.segments[j];
-                                    auto out = reinterpret_cast<uint8_t *>(outputs[j]);
-                                    std::memcpy(out + i * len, data + offset, len);
+                                    auto dst = reinterpret_cast<uint8_t *>(outputs[j]);
+                                    std::memcpy(dst + i * len, src + offset, len);
                                     offset += len;
                                 }
                             });

diff --git a/src/04kernel/src/kernels/split/cuda_kernel.cc b/src/04kernel/src/kernels/split/cuda_kernel.cc
@@ -21,7 +21,7 @@ namespace refactor::kernel {
         return typeId();
     }
     auto K::description() const noexcept -> std::string_view {
-        return "Performing split operation using CUDA";
+        return "Performing concat operation using CUDA";
     }
 
 }// namespace refactor::kernel