issue/291/style: 根据实际情况将 cuda 改为 nvidia

Signed-off-by: YdrMaster <ydrml@hotmail.com>

src/infiniop/ops/conv/cuda/conv_cuda.cuh → src/infiniop/ops/conv/nvidia/conv_nvidia.cuh

@@ -3,6 +3,6 @@
 
 #include "../conv.h"
 
-DESCRIPTOR(cuda)
+DESCRIPTOR(nvidia)
 
 #endif // __GEMM_CUDA_CUH__

src/infiniop/ops/conv/operator.cc

@@ -6,7 +6,7 @@
 #include "cpu/conv_cpu.h"
 #endif
 #ifdef ENABLE_NVIDIA_API
-#include "cuda/conv_cuda.cuh"
+#include "nvidia/conv_nvidia.cuh"
 #endif
 
 __C __export infiniStatus_t infiniopCreateConvDescriptor(infiniopHandle_t handle,
@@ -37,7 +37,7 @@ __C __export infiniStatus_t infiniopCreateConvDescriptor(infiniopHandle_t handle
         CREATE(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        CREATE(INFINI_DEVICE_NVIDIA, cuda);
+        CREATE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
     default:
         return INFINI_STATUS_DEVICE_TYPE_NOT_SUPPORTED;
@@ -61,7 +61,7 @@ infiniopGetConvWorkspaceSize(
         GET(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        GET(INFINI_DEVICE_NVIDIA, cuda);
+        GET(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
     default:
         return INFINI_STATUS_DEVICE_TYPE_NOT_SUPPORTED;
@@ -93,7 +93,7 @@ __C infiniStatus_t infiniopConv(
         CALCULATE(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        CALCULATE(INFINI_DEVICE_NVIDIA, cuda);
+        CALCULATE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 
     default:
@@ -114,7 +114,7 @@ infiniopDestroyConvDescriptor(infiniopConvDescriptor_t desc) {
         DELETE(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        DELETE(INFINI_DEVICE_NVIDIA, cuda);
+        DELETE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
     default:
         return INFINI_STATUS_DEVICE_TYPE_NOT_SUPPORTED;

src/infiniop/ops/gemm/cuda/gemm_cuda.cu → src/infiniop/ops/gemm/nvidia/gemm_nvidia.cu

-#include "../../../devices/cuda/cuda_handle.cuh"
-#include "gemm_cuda.cuh"
+#include "../../../devices/nvidia/nvidia_handle.cuh"
+#include "gemm_nvidia.cuh"
 
-namespace op::gemm::cuda {
+namespace op::gemm::nvidia {
 
 struct Descriptor::Opaque {
-    std::shared_ptr<device::cuda::Handle::Internal> internal;
+    std::shared_ptr<device::nvidia::Handle::Internal> internal;
 };
 
 Descriptor::~Descriptor() {
@@ -17,7 +17,7 @@ infiniStatus_t Descriptor::create(
     infiniopTensorDescriptor_t c_desc,
     infiniopTensorDescriptor_t a_desc,
     infiniopTensorDescriptor_t b_desc) {
-    auto handle = reinterpret_cast<device::cuda::nvidia::Handle *>(handle_);
+    auto handle = reinterpret_cast<device::nvidia::Handle *>(handle_);
     auto dtype = c_desc->dtype();
 
     CHECK_DTYPE(dtype, INFINI_DTYPE_F16, INFINI_DTYPE_F32, INFINI_DTYPE_BF16);
@@ -121,4 +121,4 @@ infiniStatus_t Descriptor::calculate(
     return INFINI_STATUS_SUCCESS;
 }
 
-} // namespace op::gemm::cuda
+} // namespace op::gemm::nvidia

src/infiniop/ops/gemm/cuda/gemm_cuda.cuh → src/infiniop/ops/gemm/nvidia/gemm_nvidia.cuh

@@ -3,6 +3,6 @@
 
 #include "../gemm.h"
 
-DESCRIPTOR(cuda)
+DESCRIPTOR(nvidia)
 
 #endif // __GEMM_CUDA_CUH__

src/infiniop/ops/gemm/operator.cc

@@ -6,7 +6,7 @@
 #include "cpu/gemm_cpu.h"
 #endif
 #ifdef ENABLE_NVIDIA_API
-#include "cuda/gemm_cuda.cuh"
+#include "nvidia/gemm_nvidia.cuh"
 #endif
 #ifdef ENABLE_CAMBRICON_API
 #include "bang/gemm_bang.h"
@@ -46,7 +46,7 @@ __C infiniStatus_t infiniopCreateGemmDescriptor(
         CREATE(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        CREATE(INFINI_DEVICE_NVIDIA, cuda);
+        CREATE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_CAMBRICON_API
         CREATE(INFINI_DEVICE_CAMBRICON, bang);
@@ -88,7 +88,7 @@ infiniopGetGemmWorkspaceSize(
         GET(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        GET(INFINI_DEVICE_NVIDIA, cuda);
+        GET(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_CAMBRICON_API
         GET(INFINI_DEVICE_CAMBRICON, bang);
@@ -137,7 +137,7 @@ __C infiniStatus_t infiniopGemm(
         CALCULATE(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        CALCULATE(INFINI_DEVICE_NVIDIA, cuda);
+        CALCULATE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_CAMBRICON_API
         CALCULATE(INFINI_DEVICE_CAMBRICON, bang);
@@ -176,7 +176,7 @@ infiniopDestroyGemmDescriptor(infiniopGemmDescriptor_t desc) {
         DELETE(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        DELETE(INFINI_DEVICE_NVIDIA, cuda);
+        DELETE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_CAMBRICON_API
         DELETE(INFINI_DEVICE_CAMBRICON, bang);

src/infiniop/ops/mul/cpu/mul_cpu.h

@@ -3,7 +3,7 @@
 
 #include "../../../elementwise/cpu/elementwise_cpu.h"
 
-ELEMENTWISE_DESCRIPTOR(mul, cpu, cpu)
+ELEMENTWISE_DESCRIPTOR(mul, cpu)
 
 namespace op::mul::cpu {
 typedef struct MulOp {

src/infiniop/ops/mul/metax/mul_metax.h

@@ -3,6 +3,6 @@
 
 #include "../../../elementwise/metax/elementwise_metax_api.h"
 
-ELEMENTWISE_DESCRIPTOR(mul, metax, metax)
+ELEMENTWISE_DESCRIPTOR(mul, metax)
 
 #endif // __MUL_METAX_API_H__

src/infiniop/ops/mul/nvidia/mul_nvidia.cu

-#include "../../../elementwise/cuda/elementwise_cuda.cuh"
+#include "../../../elementwise/nvidia/elementwise_nvidia.cuh"
 
 #include "../cuda/kernel.cuh"
 #include "mul_nvidia.cuh"
@@ -13,7 +13,7 @@ infiniStatus_t Descriptor::create(
     infiniopTensorDescriptor_t out_desc,
     std::vector<infiniopTensorDescriptor_t> input_desc_vec) {
 
-    auto handle = reinterpret_cast<device::cuda::Handle *>(handle_);
+    auto handle = reinterpret_cast<device::nvidia::Handle *>(handle_);
     auto dtype = out_desc->dtype();
 
     const auto &a_desc = input_desc_vec.at(0);

src/infiniop/ops/mul/nvidia/mul_nvidia.cuh

 #ifndef __MUL_CUDA_API_H__
 #define __MUL_CUDA_API_H__
 
-#include "../../../elementwise/cuda/elementwise_cuda_api.cuh"
+#include "../../../elementwise/nvidia/elementwise_nvidia_api.cuh"
 
-ELEMENTWISE_DESCRIPTOR(mul, nvidia, cuda)
+ELEMENTWISE_DESCRIPTOR(mul, nvidia)
 
 #endif // __MUL_CUDA_API_H__

src/infiniop/ops/random_sample/cuda/random_sample_kernel.cuh → src/infiniop/ops/random_sample/nvidia/random_sample_kernel.cuh

-#include "../../../devices/cuda/cuda_kernel_common.cuh"
+#include "../../../devices/nvidia/nvidia_kernel_common.cuh"
 #include "infinicore.h"
 #include <cub/device/device_radix_sort.cuh>
 #include <cub/device/device_reduce.cuh>
 #include <cub/device/device_scan.cuh>
 
-namespace op::random_sample::cuda {
+namespace op::random_sample::nvidia {
 
 // ↓↓↓ 重新封装 cub api，减少模板参数，方便调用
 
@@ -256,4 +256,4 @@ struct Algo {
     }
 };
 
-} // namespace op::random_sample::cuda
+} // namespace op::random_sample::nvidia

src/infiniop/ops/random_sample/cuda/random_sample_cuda.cu → src/infiniop/ops/random_sample/nvidia/random_sample_nvidia.cu

-#include "../../../devices/cuda/cuda_handle.cuh"
+#include "../../../devices/nvidia/nvidia_handle.cuh"
 #include "../info.h"
-#include "random_sample_cuda.cuh"
 #include "random_sample_kernel.cuh"
+#include "random_sample_nvidia.cuh"
 
-namespace op::random_sample::cuda {
+namespace op::random_sample::nvidia {
 
 struct Descriptor::Opaque {
-    std::shared_ptr<device::cuda::Handle::Internal> internal;
+    std::shared_ptr<device::nvidia::Handle::Internal> internal;
 };
 
 Descriptor::~Descriptor() {
@@ -18,7 +18,7 @@ infiniStatus_t Descriptor::create(
     Descriptor **desc_ptr,
     infiniopTensorDescriptor_t result_desc,
     infiniopTensorDescriptor_t probs_desc) {
-    auto handle = reinterpret_cast<device::cuda::Handle *>(handle_);
+    auto handle = reinterpret_cast<device::nvidia::Handle *>(handle_);
 
     auto result = RandomSampleInfo::create(result_desc, probs_desc);
     CHECK_RESULT(result);
@@ -99,4 +99,4 @@ infiniStatus_t Descriptor::calculate(
     return INFINI_STATUS_SUCCESS;
 }
 
-} // namespace op::random_sample::cuda
+} // namespace op::random_sample::nvidia

src/infiniop/ops/random_sample/cuda/random_sample_cuda.cuh → src/infiniop/ops/random_sample/nvidia/random_sample_nvidia.cuh

@@ -3,6 +3,6 @@
 
 #include "../random_sample.h"
 
-DESCRIPTOR(cuda)
+DESCRIPTOR(nvidia)
 
 #endif // __RANDOM_SAMPLE_CUDA_CUH__

src/infiniop/ops/random_sample/operator.cc

@@ -6,7 +6,7 @@
 #include "cpu/random_sample_cpu.h"
 #endif
 #ifdef ENABLE_NVIDIA_API
-#include "cuda/random_sample_cuda.cuh"
+#include "nvidia/random_sample_nvidia.cuh"
 #endif
 #ifdef ENABLE_METAX_API
 #include "metax/random_sample_metax.h"
@@ -36,7 +36,7 @@ infiniopCreateRandomSampleDescriptor(
         CREATE(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        CREATE(INFINI_DEVICE_NVIDIA, cuda);
+        CREATE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_METAX_API
         CREATE(INFINI_DEVICE_METAX, metax);
@@ -69,7 +69,7 @@ __C infiniStatus_t infiniopGetRandomSampleWorkspaceSize(
         GET(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        GET(INFINI_DEVICE_NVIDIA, cuda);
+        GET(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_METAX_API
         GET(INFINI_DEVICE_METAX, metax);
@@ -112,7 +112,7 @@ __C infiniStatus_t infiniopRandomSample(
         CALCULATE(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        CALCULATE(INFINI_DEVICE_NVIDIA, cuda);
+        CALCULATE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_METAX_API
         CALCULATE(INFINI_DEVICE_METAX, metax);
@@ -142,7 +142,7 @@ __C infiniStatus_t infiniopDestroyRandomSampleDescriptor(
         DELETE(INFINI_DEVICE_CPU, cpu);
 #endif
 #ifdef ENABLE_NVIDIA_API
-        DELETE(INFINI_DEVICE_NVIDIA, cuda);
+        DELETE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_METAX_API
         DELETE(INFINI_DEVICE_METAX, metax);

src/infiniop/ops/rearrange/cuda/rearrange_kernel.cuh → src/infiniop/ops/rearrange/nvidia/rearrange_kernel.cuh

 #ifndef __REARRANGE_CUDA_KERNEL_H__
 #define __REARRANGE_CUDA_KERNEL_H__
 
-#include "../../../devices/cuda/cuda_common.cuh"
+#include "../../../devices/nvidia/nvidia_common.cuh"
 
 #define ARRAY_TYPE_STRIDE ptrdiff_t
 #define ARRAY_TYPE_SIZE size_t

src/infiniop/ops/rearrange/cuda/rearrange_cuda.cu → src/infiniop/ops/rearrange/nvidia/rearrange_nvidia.cu

-#include "../../../devices/cuda/cuda_common.cuh"
-#include "../../../devices/cuda/cuda_kernel_common.cuh"
+#include "../../../devices/nvidia/nvidia_common.cuh"
+#include "../../../devices/nvidia/nvidia_kernel_common.cuh"
 #include "../../../tensor.h"
-#include "rearrange_cuda.cuh"
 #include "rearrange_kernel.cuh"
+#include "rearrange_nvidia.cuh"
 #include <algorithm>
 #include <cmath>
 #include <memory>
 #include <stdint.h>
 #include <vector>
 
-namespace op::rearrange::cuda {
+namespace op::rearrange::nvidia {
 
 struct Descriptor::Opaque {
-    std::shared_ptr<device::cuda::Handle::Internal> internal;
+    std::shared_ptr<device::nvidia::Handle::Internal> internal;
 };
 
 Descriptor::~Descriptor() {
@@ -49,7 +49,7 @@ infiniStatus_t Descriptor::create(
 
     *desc_ptr = new Descriptor(
         std::move(*meta),
-        new Opaque{reinterpret_cast<device::cuda::Handle *>(handle)->internal()},
+        new Opaque{reinterpret_cast<device::nvidia::Handle *>(handle)->internal()},
         handle->device, handle->device_id);
     return INFINI_STATUS_SUCCESS;
 }
@@ -482,4 +482,4 @@ infiniStatus_t Descriptor::calculate(
     return status;
 }
 
-} // namespace op::rearrange::cuda
+} // namespace op::rearrange::nvidia

src/infiniop/ops/rearrange/cuda/rearrange_cuda.cuh → src/infiniop/ops/rearrange/nvidia/rearrange_nvidia.cuh

@@ -3,6 +3,6 @@
 
 #include "../rearrange.h"
 
-DESCRIPTOR(cuda)
+DESCRIPTOR(nvidia)
 
 #endif // __REARRANGE_CUDA_H__

src/infiniop/ops/rearrange/operator.cc

@@ -10,7 +10,7 @@
 #endif
 
 #ifdef ENABLE_NVIDIA_API
-#include "cuda/rearrange_cuda.cuh"
+#include "nvidia/rearrange_nvidia.cuh"
 #endif
 #ifdef ENABLE_METAX_API
 #include "metax/rearrange_metax.h"
@@ -40,7 +40,7 @@ __C infiniStatus_t infiniopCreateRearrangeDescriptor(
 #endif
 
 #ifdef ENABLE_NVIDIA_API
-        CREATE(INFINI_DEVICE_NVIDIA, cuda);
+        CREATE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_METAX_API
         CREATE(INFINI_DEVICE_METAX, metax);
@@ -73,7 +73,7 @@ __C infiniStatus_t infiniopRearrange(
 #endif
 
 #ifdef ENABLE_NVIDIA_API
-        CALCULATE(INFINI_DEVICE_NVIDIA, cuda);
+        CALCULATE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_METAX_API
         CALCULATE(INFINI_DEVICE_METAX, metax);
@@ -104,7 +104,7 @@ __C infiniStatus_t infiniopDestroyRearrangeDescriptor(
 #endif
 
 #ifdef ENABLE_NVIDIA_API
-        DELETE(INFINI_DEVICE_NVIDIA, cuda);
+        DELETE(INFINI_DEVICE_NVIDIA, nvidia);
 #endif
 #ifdef ENABLE_METAX_API
         DELETE(INFINI_DEVICE_METAX, metax);

src/infiniop/ops/relu/cpu/relu_cpu.h

@@ -5,7 +5,7 @@
 
 #include "../../../elementwise/cpu/elementwise_cpu.h"
 
-ELEMENTWISE_DESCRIPTOR(relu, cpu, cpu)
+ELEMENTWISE_DESCRIPTOR(relu, cpu)
 
 namespace op::relu::cpu {
 typedef struct ReluOp {

src/infiniop/ops/rms_norm/nvidia/rms_norm_nvidia.cu

-#include "../../../devices/cuda/cuda_common.cuh"
+#include "../../../devices/nvidia/nvidia_common.cuh"
 #include "rms_norm_nvidia.cuh"
 
-#include "../../../devices/cuda/cuda_kernel_common.cuh"
+#include "../../../devices/nvidia/nvidia_kernel_common.cuh"
 #include <cub/block/block_reduce.cuh>
 
 #include "../../../reduce/cuda/reduce.cuh"
@@ -23,7 +23,7 @@ INFINIOP_CUDA_KERNEL rmsnormKernel(
 namespace op::rms_norm::nvidia {
 
 struct Descriptor::Opaque {
-    std::shared_ptr<device::cuda::Handle::Internal> internal;
+    std::shared_ptr<device::nvidia::Handle::Internal> internal;
 };
 
 Descriptor::~Descriptor() {
@@ -47,7 +47,7 @@ infiniStatus_t Descriptor::create(
     }
 
     *desc_ptr = new Descriptor(
-        new Opaque{reinterpret_cast<device::cuda::Handle *>(handle)->internal()},
+        new Opaque{reinterpret_cast<device::nvidia::Handle *>(handle)->internal()},
         std::move(info),
         0,
         handle->device, handle->device_id);

src/infiniop/ops/rope/nvidia/rope_nvidia.cu

-#include "../../../devices/cuda/cuda_common.cuh"
+#include "../../../devices/nvidia/nvidia_common.cuh"
 #include "rope_nvidia.cuh"
 
-#include "../../../devices/cuda/cuda_kernel_common.cuh"
+#include "../../../devices/nvidia/nvidia_kernel_common.cuh"
 
 #include "../cuda/kernel.cuh"
 
@@ -28,7 +28,7 @@ INFINIOP_CUDA_KERNEL ropeThreadPerItemKernel(
 namespace op::rope::nvidia {
 
 struct Descriptor::Opaque {
-    std::shared_ptr<device::cuda::Handle::Internal> internal;
+    std::shared_ptr<device::nvidia::Handle::Internal> internal;
 };
 
 Descriptor::~Descriptor() {
@@ -44,7 +44,7 @@ infiniStatus_t Descriptor::create(
     infiniopTensorDescriptor_t sin_desc,
     infiniopTensorDescriptor_t cos_desc) {
 
-    auto handle = reinterpret_cast<device::cuda::Handle *>(handle_);
+    auto handle = reinterpret_cast<device::nvidia::Handle *>(handle_);
 
     auto info = RoPEInfo::createRoPEInfo(y_desc, x_desc, pos_desc, sin_desc, cos_desc);
     CHECK_RESULT(info);
@@ -53,7 +53,7 @@ infiniStatus_t Descriptor::create(
     *desc_ptr = new Descriptor(
         info.take(),
         0,
-        new Opaque{reinterpret_cast<device::cuda::Handle *>(handle)->internal()},
+        new Opaque{reinterpret_cast<device::nvidia::Handle *>(handle)->internal()},
         handle->device,
         handle->device_id);