diff --git a/mindspore/ccsrc/device/gpu/gpu_kernel_runtime.cc b/mindspore/ccsrc/device/gpu/gpu_kernel_runtime.cc index 387977d34503d0e63ebc3dd01c67e373e0408cd2..dc4487ccee1501e021ec61079b766053a89f1ace 100644 --- a/mindspore/ccsrc/device/gpu/gpu_kernel_runtime.cc +++ b/mindspore/ccsrc/device/gpu/gpu_kernel_runtime.cc @@ -199,7 +199,7 @@ bool GPUKernelRuntime::LaunchKernelDynamic(const session::KernelGraph *graph) { AddressPtrList kernel_workspaces; AddressPtrList kernel_outputs; AllocKernelDynamicRes(*kernel_mod, kernel, &kernel_inputs, &kernel_workspaces, &kernel_outputs); - if (!kernel_mod->Launch(kernel_inputs, kernel_workspaces, kernel_outputs, reinterpret_cast(stream_))) { + if (!kernel_mod->Launch(kernel_inputs, kernel_workspaces, kernel_outputs, stream_)) { MS_LOG(ERROR) << "Launch kernel failed."; return false; } diff --git a/mindspore/ccsrc/device/kernel_runtime.cc b/mindspore/ccsrc/device/kernel_runtime.cc index 42f56af8d527d3388c0356082d3524192fc3712d..0f5f282fd1b95da9190733de04300c6bd05a02ba 100644 --- a/mindspore/ccsrc/device/kernel_runtime.cc +++ b/mindspore/ccsrc/device/kernel_runtime.cc @@ -664,8 +664,7 @@ bool KernelRuntime::LaunchKernelMod(const session::KernelGraph &graph) { struct timeval start_time, end_time; (void)gettimeofday(&start_time, nullptr); #endif - auto ret = - kernel_mod->Launch(kernel_inputs, kernel_workspaces, kernel_outputs, reinterpret_cast(stream_)); + auto ret = kernel_mod->Launch(kernel_inputs, kernel_workspaces, kernel_outputs, stream_); if (!ret) { MS_LOG(ERROR) << "Launch kernel failed."; return false; diff --git a/mindspore/ccsrc/kernel/aicpu/aicpu_kernel_mod.cc b/mindspore/ccsrc/kernel/aicpu/aicpu_kernel_mod.cc index 5e07cd0ad7672fd56b3be3d4f95454743fc01388..7875baaf0ec8413894665f4e54a28047d2c7d0e9 100644 --- a/mindspore/ccsrc/kernel/aicpu/aicpu_kernel_mod.cc +++ b/mindspore/ccsrc/kernel/aicpu/aicpu_kernel_mod.cc @@ -103,14 +103,13 @@ void AicpuOpKernelMod::CreateCpuKernelInfo(const std::vector &inputs } bool AicpuOpKernelMod::Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) { - if (stream_ptr == 0) { + const std::vector &outputs, void *stream_ptr) { + if (stream_ptr == nullptr) { MS_LOG(ERROR) << "stream_ptr should not be nullptr."; return false; } CreateCpuKernelInfo(inputs, outputs); - auto *stream = reinterpret_cast(stream_ptr); if (node_name_ == "TopK") { node_name_ = "TopKV2"; } @@ -119,7 +118,7 @@ bool AicpuOpKernelMod::Launch(const std::vector &inputs, const std:: if (rtCpuKernelLaunch(reinterpret_cast(node_so_.c_str()), reinterpret_cast(node_name_.c_str()), 1, reinterpret_cast(args_.data()), static_cast(args_.length()), nullptr, - stream) != RT_ERROR_NONE) { + stream_ptr) != RT_ERROR_NONE) { MS_LOG(ERROR) << "Aicpu op launch failed!"; return false; diff --git a/mindspore/ccsrc/kernel/aicpu/aicpu_kernel_mod.h b/mindspore/ccsrc/kernel/aicpu/aicpu_kernel_mod.h index f7469c5e724cbbcdbbedd7a62ef136708fa7f871..3ee9bd2a15f9da96b66dc411f1130da32559a9e1 100644 --- a/mindspore/ccsrc/kernel/aicpu/aicpu_kernel_mod.h +++ b/mindspore/ccsrc/kernel/aicpu/aicpu_kernel_mod.h @@ -27,7 +27,7 @@ class AicpuOpKernelMod : public AscendKernelMod { AicpuOpKernelMod(); ~AicpuOpKernelMod() override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; diff --git a/mindspore/ccsrc/kernel/akg/gpu/akg_gpu_kernel_mod.cc b/mindspore/ccsrc/kernel/akg/gpu/akg_gpu_kernel_mod.cc index 7ce434b2d35638fda9105643654a42f5125df2f1..64590cd9b82655ff81c63ecd7e1474e0262cbb93 100644 --- a/mindspore/ccsrc/kernel/akg/gpu/akg_gpu_kernel_mod.cc +++ b/mindspore/ccsrc/kernel/akg/gpu/akg_gpu_kernel_mod.cc @@ -82,7 +82,7 @@ const std::vector &GpuKernelMod::GetOutputSizeList() const { return outp const std::vector &GpuKernelMod::GetWorkspaceSizeList() const { return workspace_size_list_; } bool GpuKernelMod::Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { if (stream_ptr == 0) { MS_LOG(ERROR) << "stream_ptr should not be nullptr."; return false; diff --git a/mindspore/ccsrc/kernel/akg/gpu/akg_gpu_kernel_mod.h b/mindspore/ccsrc/kernel/akg/gpu/akg_gpu_kernel_mod.h index 4ef903f1f32ff3d75ee512a8ed4d914a35bc34df..df9cb069f79a21bcf85378ce9008345d08137d8a 100644 --- a/mindspore/ccsrc/kernel/akg/gpu/akg_gpu_kernel_mod.h +++ b/mindspore/ccsrc/kernel/akg/gpu/akg_gpu_kernel_mod.h @@ -64,7 +64,7 @@ class GpuKernelMod : public KernelMod { const std::vector &GetOutputSizeList() const override; const std::vector &GetWorkspaceSizeList() const override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; static GpuKernelManagerPtr kernelmanager_; diff --git a/mindspore/ccsrc/kernel/cpu/cpu_kernel.h b/mindspore/ccsrc/kernel/cpu/cpu_kernel.h index 378f36ac5b547a3cedd6d0b4c2ea264b3e57c3b0..edd133bf13ada46c5e1c33c8449e3b75a2b1269f 100644 --- a/mindspore/ccsrc/kernel/cpu/cpu_kernel.h +++ b/mindspore/ccsrc/kernel/cpu/cpu_kernel.h @@ -56,7 +56,7 @@ class CPUKernel : public kernel::KernelMod { void Init(const CNodePtr &kernel_node); virtual void InitKernel(const CNodePtr &kernel_node) = 0; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t /*stream_ptr*/) override { + const std::vector &outputs, void * /*stream_ptr*/) override { return Launch(inputs, workspace, outputs); }; virtual bool Launch(const std::vector &inputs, const std::vector &workspace, diff --git a/mindspore/ccsrc/kernel/gpu/arrays/argmax_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/arrays/argmax_gpu_kernel.h index ccd691ae91e8847da6d4fefe778efb49be961581..3df70d0960de65ba7b8fd591239fd7af12b496ec 100644 --- a/mindspore/ccsrc/kernel/gpu/arrays/argmax_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/arrays/argmax_gpu_kernel.h @@ -35,7 +35,7 @@ class ArgmaxGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *input = GetDeviceAddress(inputs, 0); int *output = GetDeviceAddress(outputs, 0); CalArgmax(input, SizeToInt(batch_size_), SizeToInt(channel_size_), axis_, output, diff --git a/mindspore/ccsrc/kernel/gpu/arrays/array_reduce_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/arrays/array_reduce_gpu_kernel.h index c8410c419d37ded28663ddb9f835a276a44ad3a5..224a3da8ad00abddd5a36556e5da6d8b02710e2c 100644 --- a/mindspore/ccsrc/kernel/gpu/arrays/array_reduce_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/arrays/array_reduce_gpu_kernel.h @@ -55,7 +55,7 @@ class ArrayReduceGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/arrays/concatv2_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/arrays/concatv2_gpu_kernel.h index 5dabb3045c258d242ca4715c5a5b7c918c55ad5f..a91c50ce699aafa4695343db3651b222ffa61f59 100644 --- a/mindspore/ccsrc/kernel/gpu/arrays/concatv2_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/arrays/concatv2_gpu_kernel.h @@ -34,7 +34,7 @@ class ConcatV2GpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (inputs.size() == 2) { T *input_0 = GetDeviceAddress(inputs, 0); T *input_1 = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/arrays/gather_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/arrays/gather_gpu_kernel.h index c4424df59c603a2031f095fd947b7abe09ae21a4..72a05b09155a4c208a423da527bb3a872d09755e 100644 --- a/mindspore/ccsrc/kernel/gpu/arrays/gather_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/arrays/gather_gpu_kernel.h @@ -35,7 +35,7 @@ class GatherGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(workspace); T *input_addr = GetDeviceAddress(inputs, 0); S *indices_addr = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/arrays/one_hot_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/arrays/one_hot_gpu_kernel.h index d8059869f29349cd681662bf78e1e5bc4205f808..c8b64e7243036fbc3d0523ffe7f72365e077ffb9 100644 --- a/mindspore/ccsrc/kernel/gpu/arrays/one_hot_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/arrays/one_hot_gpu_kernel.h @@ -34,7 +34,7 @@ class OneHotGpuFwdKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(workspace); const S *indices = GetDeviceAddress(inputs, 0); const T *on_value = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/arrays/select_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/arrays/select_gpu_kernel.h index ba0bea4dee912034bcbd671d4718276bdcaf9f4b..f1b6c5853a17ee85cea16e16d4669ff2e686008d 100644 --- a/mindspore/ccsrc/kernel/gpu/arrays/select_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/arrays/select_gpu_kernel.h @@ -34,7 +34,7 @@ class SelectGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { bool *input_cond = GetDeviceAddress(inputs, 0); T *input_x = GetDeviceAddress(inputs, 1); T *input_y = GetDeviceAddress(inputs, 2); diff --git a/mindspore/ccsrc/kernel/gpu/arrays/slice_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/arrays/slice_gpu_kernel.h index f71ec23d2eac12193167cd3ad1939a13ce8f3f4e..eb829f73c6b8086a5e7cbf0882e11667a71e7de6 100644 --- a/mindspore/ccsrc/kernel/gpu/arrays/slice_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/arrays/slice_gpu_kernel.h @@ -34,7 +34,7 @@ class SliceGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *input = GetDeviceAddress(inputs, 0); T *output = GetDeviceAddress(outputs, 0); if (is_strided_slice_) { diff --git a/mindspore/ccsrc/kernel/gpu/arrays/slice_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/arrays/slice_grad_gpu_kernel.h index 80eef23112b1684ec30a830fda8804290a629bea..bf24272d93dd6fadfb8de8c9c98ffd6def1bef53 100644 --- a/mindspore/ccsrc/kernel/gpu/arrays/slice_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/arrays/slice_grad_gpu_kernel.h @@ -34,7 +34,7 @@ class SliceGradGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *dy = GetDeviceAddress(inputs, 0); T *dx = GetDeviceAddress(outputs, 0); FillDeviceArray(outputs[0]->size / sizeof(T), dx, 0.f, reinterpret_cast(stream_ptr)); diff --git a/mindspore/ccsrc/kernel/gpu/arrays/transpose_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/arrays/transpose_gpu_kernel.h index 1c9cf925ea18721cebd37881a86432d54713a06f..61be9b68fea61cfd06b7f1e0510131aed667a9cc 100644 --- a/mindspore/ccsrc/kernel/gpu/arrays/transpose_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/arrays/transpose_gpu_kernel.h @@ -34,7 +34,7 @@ class TransposeGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *input = GetDeviceAddress(inputs, 0); T *output = GetDeviceAddress(outputs, 0); int *input_shape = GetDeviceAddress(workspace, 0); diff --git a/mindspore/ccsrc/kernel/gpu/arrays/unsorted_segment_sum_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/arrays/unsorted_segment_sum_gpu_kernel.h index 24c1f09097f84ff04a3e92f6bc8df256b64ccfb4..a20375ee299eb140ea35b6a57daf8a2b9280a4b8 100644 --- a/mindspore/ccsrc/kernel/gpu/arrays/unsorted_segment_sum_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/arrays/unsorted_segment_sum_gpu_kernel.h @@ -35,7 +35,7 @@ class UnsortedSegmentSumGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *input_addr = GetDeviceAddress(inputs, 0); S *indices_addr = GetDeviceAddress(inputs, 1); T *output_addr = GetDeviceAddress(outputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/control/recv_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/control/recv_gpu_kernel.h index 206eac5bd96b653f08650eb95ae36397585edb62..12b4eed1326f2cbe0ee4b5b2fe2d16206cbf7259 100644 --- a/mindspore/ccsrc/kernel/gpu/control/recv_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/control/recv_gpu_kernel.h @@ -33,7 +33,7 @@ class RecvGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &, const std::vector &, const std::vector &, - uintptr_t) override { + void *) override { CHECK_CUDA_RET_WITH_EXCEPT(cudaStreamWaitEvent(wait_stream_, wait_event_, 0), "Waiting cuda event failed."); return true; } diff --git a/mindspore/ccsrc/kernel/gpu/control/send_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/control/send_gpu_kernel.h index 156ec4160dcdd214c2e8a42cf26b92d9849ea0a1..a26e41aa1efdd6b99c1872195aa645bc136a9816 100644 --- a/mindspore/ccsrc/kernel/gpu/control/send_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/control/send_gpu_kernel.h @@ -33,7 +33,7 @@ class SendGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &, const std::vector &, const std::vector &, - uintptr_t) override { + void *) override { CHECK_CUDA_RET_WITH_EXCEPT(cudaEventRecord(record_event_, record_stream_), "Recording cuda event failed."); return true; } diff --git a/mindspore/ccsrc/kernel/gpu/data/dataset_init_kernel.cc b/mindspore/ccsrc/kernel/gpu/data/dataset_init_kernel.cc index d87a6cc72613c3ade818102353bce3f015d00aac..777310cebcae451ccbc2a44ec2ee17704bfca643 100644 --- a/mindspore/ccsrc/kernel/gpu/data/dataset_init_kernel.cc +++ b/mindspore/ccsrc/kernel/gpu/data/dataset_init_kernel.cc @@ -53,7 +53,7 @@ bool DatasetInitKernel::Init(const CNodePtr &kernel_node) { void DatasetInitKernel::InitSizeLists() { return; } bool DatasetInitKernel::Launch(const std::vector &, const std::vector &, - const std::vector &, uintptr_t) { + const std::vector &, void *) { void *addr = nullptr; size_t len = total_bytes_ * buffer_q_capacity_; diff --git a/mindspore/ccsrc/kernel/gpu/data/dataset_init_kernel.h b/mindspore/ccsrc/kernel/gpu/data/dataset_init_kernel.h index f7ffb419956219d42d835cf42673e4c272ac1269..318049f4ad5957e3861bf07f1c61c8e9094fc17d 100644 --- a/mindspore/ccsrc/kernel/gpu/data/dataset_init_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/data/dataset_init_kernel.h @@ -33,7 +33,7 @@ class DatasetInitKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override; const std::vector &GetWorkspaceSizeList() const override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; bool Init(const CNodePtr &kernel_node) override; protected: diff --git a/mindspore/ccsrc/kernel/gpu/data/dataset_iterator_kernel.cc b/mindspore/ccsrc/kernel/gpu/data/dataset_iterator_kernel.cc index f8ee13401893e6c381559e37d150096498b701b9..69e35585cf2186b13eb2f7c6e4ba5b4ec3b7e243 100644 --- a/mindspore/ccsrc/kernel/gpu/data/dataset_iterator_kernel.cc +++ b/mindspore/ccsrc/kernel/gpu/data/dataset_iterator_kernel.cc @@ -64,7 +64,7 @@ bool DatasetIteratorKernel::Init(const CNodePtr &kernel_node) { void DatasetIteratorKernel::InitSizeLists() { return; } bool DatasetIteratorKernel::Launch(const std::vector &, const std::vector &, - const std::vector &outputs, uintptr_t) { + const std::vector &outputs, void *) { void *addr = nullptr; size_t len = 0; diff --git a/mindspore/ccsrc/kernel/gpu/data/dataset_iterator_kernel.h b/mindspore/ccsrc/kernel/gpu/data/dataset_iterator_kernel.h index d3231cab3cfc520c47aeaaf6d235ec4e5c9a711e..cdd7a47e7ba81692c901dd1c4f41ede133d22a9e 100644 --- a/mindspore/ccsrc/kernel/gpu/data/dataset_iterator_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/data/dataset_iterator_kernel.h @@ -33,7 +33,7 @@ class DatasetIteratorKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override; const std::vector &GetWorkspaceSizeList() const override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; bool Init(const CNodePtr &kernel_node) override; protected: diff --git a/mindspore/ccsrc/kernel/gpu/math/addn_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/math/addn_gpu_kernel.h index 485d3b4f721ad7328ba88b21cdc101bc12b0eaf1..0b276027613adb364d095cae044e7aa750505c69 100644 --- a/mindspore/ccsrc/kernel/gpu/math/addn_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/math/addn_gpu_kernel.h @@ -43,7 +43,7 @@ class AddNGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t) override { + const std::vector &outputs, void *) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/math/assign_add_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/math/assign_add_gpu_kernel.h index 191bd0b1c8b1d356621ae38a4e2bbe001159ee8c..db69fd7be64af61ab242bd5c5f51dcd2371bae2e 100644 --- a/mindspore/ccsrc/kernel/gpu/math/assign_add_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/math/assign_add_gpu_kernel.h @@ -35,7 +35,7 @@ class AssignAddGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/math/bias_add_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/math/bias_add_gpu_kernel.h index 03192a36a3d2ae155b097b93dc8bb87bd91e4341..d7ed8e1072f4b3087f4b3495a6a9b06ef54f555e 100644 --- a/mindspore/ccsrc/kernel/gpu/math/bias_add_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/math/bias_add_gpu_kernel.h @@ -42,7 +42,7 @@ class BiasAddGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(workspace); VARIABLE_NOT_USED(stream_ptr); T *x_addr = GetDeviceAddress(inputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/math/broadcast_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/math/broadcast_gpu_kernel.h index 1b597cea70fbe280fadcc036e6d646775655cdf6..c652d9aae4efbe8cd3854ece88dee864d2d0b7ae 100644 --- a/mindspore/ccsrc/kernel/gpu/math/broadcast_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/math/broadcast_gpu_kernel.h @@ -39,7 +39,7 @@ class BroadcastOpGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *lhs = GetDeviceAddress(inputs, 0); T *rhs = GetDeviceAddress(inputs, 1); S *output = GetDeviceAddress(outputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/math/broadcast_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/math/broadcast_grad_gpu_kernel.h index 2c15e688293ba6c02965f0b13e923eefca75a08d..3e1f91b5b702b04185c301ce775fa694ab6cac97 100644 --- a/mindspore/ccsrc/kernel/gpu/math/broadcast_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/math/broadcast_grad_gpu_kernel.h @@ -39,7 +39,7 @@ class BroadcastOpGradGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *x1 = GetDeviceAddress(inputs, 0); T *x2 = GetDeviceAddress(inputs, 1); T *dy = GetDeviceAddress(inputs, 2); diff --git a/mindspore/ccsrc/kernel/gpu/math/equalcount_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/math/equalcount_gpu_kernel.h index 9e98d6be0f47ecedaab320a3d2b281dfc322f5ba..7d3f74970fe77e001ff6b27fb175bbbb2ee37416 100644 --- a/mindspore/ccsrc/kernel/gpu/math/equalcount_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/math/equalcount_gpu_kernel.h @@ -35,7 +35,7 @@ class EqualCountGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(workspace); T *input1 = GetDeviceAddress(inputs, 0); T *input2 = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/math/float_status_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/math/float_status_gpu_kernel.h index bdd93d5d54d005916ce8f08748e6c8571d92b458..1aa9b18684fefaaa86d548e8e0c59c13e264da66 100644 --- a/mindspore/ccsrc/kernel/gpu/math/float_status_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/math/float_status_gpu_kernel.h @@ -40,7 +40,7 @@ class FloatStatusGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *input = GetDeviceAddress(inputs, 0); switch (kernel_name_) { diff --git a/mindspore/ccsrc/kernel/gpu/math/matmul_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/math/matmul_gpu_kernel.h index 2dc164b4573aa9e7312c34c1c8079d40d3ef8b4f..765fb3d7d4b6180b98c30a4eeb1992db8c7f04dd 100644 --- a/mindspore/ccsrc/kernel/gpu/math/matmul_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/math/matmul_gpu_kernel.h @@ -48,7 +48,7 @@ class MatMulGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(workspace); VARIABLE_NOT_USED(stream_ptr); auto input1_addr = GetDeviceAddress(inputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/math/tensoradd_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/math/tensoradd_gpu_kernel.h index 52480b8c70d6d0d798e8dcfe8faebe2835f3594b..67c6a34f3fe16b54541f13ea4746a12a2a04eeb6 100644 --- a/mindspore/ccsrc/kernel/gpu/math/tensoradd_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/math/tensoradd_gpu_kernel.h @@ -43,7 +43,7 @@ class TensorAddGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t) { + const std::vector &outputs, void *) { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/math/unary_op_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/math/unary_op_gpu_kernel.h index d4c8a9b29cd6824577aa72a6ec8bb2415130c9ea..4503b805f6b6e77592c6d82fbe30e9319fe3a672 100644 --- a/mindspore/ccsrc/kernel/gpu/math/unary_op_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/math/unary_op_gpu_kernel.h @@ -62,7 +62,7 @@ class UnaryOpGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(workspace); T *input_addr = GetDeviceAddress(inputs, 0); T *output_addr = GetDeviceAddress(outputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/nccl/nccl_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nccl/nccl_gpu_kernel.h index 4ea332784dd8ab27ca70b050cb7ef28c7289e556..b5ab46a67d116f3a598ec56e5fdb92a22cf98990 100644 --- a/mindspore/ccsrc/kernel/gpu/nccl/nccl_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nccl/nccl_gpu_kernel.h @@ -60,7 +60,7 @@ class NcclGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *input_addr = GetDeviceAddress(inputs, 0); T *output_addr = GetDeviceAddress(outputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/nn/bias_add_grad_gpu_kenel.h b/mindspore/ccsrc/kernel/gpu/nn/bias_add_grad_gpu_kenel.h index 5c7153a172b2da0e6bc0be51584a709fc867a58c..c93a0506495ce2ab68754503a5c1312d303e6b53 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/bias_add_grad_gpu_kenel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/bias_add_grad_gpu_kenel.h @@ -45,7 +45,7 @@ class BiasAddGradGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *dy_addr = GetDeviceAddress(inputs, 0); T *db_addr = GetDeviceAddress(outputs, 0); T *indices_addr = GetDeviceAddress(workspace, 0); diff --git a/mindspore/ccsrc/kernel/gpu/nn/conv2d_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/conv2d_gpu_kernel.h index 6e218dddef83269631bff0e9dc40a3fbec67d9fe..7bb6aa2a6d2d19215c854aeb9782bf127f741d1f 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/conv2d_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/conv2d_gpu_kernel.h @@ -60,7 +60,7 @@ class Conv2dGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/conv2d_grad_filter_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/conv2d_grad_filter_gpu_kernel.h index faae0ae7952edbc9fb65a166d816837a325d908b..b126b542dd2f15b3e8c89e6d80f0b3e6881a4762 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/conv2d_grad_filter_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/conv2d_grad_filter_gpu_kernel.h @@ -61,7 +61,7 @@ class ConvGradFilterGpuBkwKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/conv2d_grad_input_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/conv2d_grad_input_gpu_kernel.h index 325da74486779eb21cbf8d00c46e07a8beb18cd4..f7f371067ff7abb3ca335538d934836e63c98020 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/conv2d_grad_input_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/conv2d_grad_input_gpu_kernel.h @@ -61,7 +61,7 @@ class ConvGradInputGpuBkwKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/dropout_gpu_kernel.cc b/mindspore/ccsrc/kernel/gpu/nn/dropout_gpu_kernel.cc index 937f38137fb30e9f2d1456d0577f02421efd15d8..0d2a6be9c8e1d7d24b0685029c736a46226be383 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/dropout_gpu_kernel.cc +++ b/mindspore/ccsrc/kernel/gpu/nn/dropout_gpu_kernel.cc @@ -76,7 +76,7 @@ void DropoutGpuFwdKernel::InitSizeLists() { } bool DropoutGpuFwdKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/dropout_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/dropout_gpu_kernel.h index 2b0d84a40c29e0054a8863afc1de07770a798aaa..accff17429e3a5a8ef630edb8973fb9c2bfa057d 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/dropout_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/dropout_gpu_kernel.h @@ -37,7 +37,7 @@ class DropoutGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; bool Init(const CNodePtr &kernel_node) override; diff --git a/mindspore/ccsrc/kernel/gpu/nn/dropout_grad_kernel.cc b/mindspore/ccsrc/kernel/gpu/nn/dropout_grad_kernel.cc index 42c3d279c40a5911ad9a2d82ac8e9c45b619090d..44f603f02d483678703796c1a83db3442ea9f16f 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/dropout_grad_kernel.cc +++ b/mindspore/ccsrc/kernel/gpu/nn/dropout_grad_kernel.cc @@ -75,7 +75,7 @@ void DropoutGradGpuFwdKernel::InitSizeLists() { } bool DropoutGradGpuFwdKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/dropout_grad_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/dropout_grad_kernel.h index b59b5d26704e1c2f81bbba1ff64708e41df6b4da..79d4117b58b9a0785d4ecd15888da1280c61836f 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/dropout_grad_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/dropout_grad_kernel.h @@ -32,7 +32,7 @@ class DropoutGradGpuFwdKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override; const std::vector &GetWorkspaceSizeList() const override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; bool Init(const CNodePtr &kernel_node) override; protected: diff --git a/mindspore/ccsrc/kernel/gpu/nn/flatten_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/flatten_gpu_kernel.h index 975dbd0082755b9d7822961584d5ad7462bd5f37..3b0ad8c94654e93b16f3bd4f6b70cafbdb59ae03 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/flatten_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/flatten_gpu_kernel.h @@ -35,7 +35,7 @@ class FlattenGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *input = GetDeviceAddress(inputs, 0); T *output = GetDeviceAddress(outputs, 0); cudaError_t ret = diff --git a/mindspore/ccsrc/kernel/gpu/nn/flatten_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/flatten_grad_gpu_kernel.h index bdae6404dcda4101ec107ae17d2f4cade17f0340..0748dc77db1b2f14bb50dadac8a4751d21c11da7 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/flatten_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/flatten_grad_gpu_kernel.h @@ -35,7 +35,7 @@ class FlattenGardGpuBkwKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(workspace); T *input = GetDeviceAddress(inputs, 0); T *output = GetDeviceAddress(outputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/nn/fused_batch_norm_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/fused_batch_norm_gpu_kernel.h index 3cdf4805404c254d4d477442966c7fba39bcb933..c08b341e78b1392fb01f0784914251ce67cbde0c 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/fused_batch_norm_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/fused_batch_norm_gpu_kernel.h @@ -49,7 +49,7 @@ class FusedBatchNormGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(workspace); VARIABLE_NOT_USED(stream_ptr); if (is_null_input_) { diff --git a/mindspore/ccsrc/kernel/gpu/nn/fused_batchnorm_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/fused_batchnorm_grad_gpu_kernel.h index 07372ad22d4fd960bb9c84a4f4f30be36e40cfc6..153b0286b34d6fd2667eab9863bd21f8a0117141 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/fused_batchnorm_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/fused_batchnorm_grad_gpu_kernel.h @@ -47,7 +47,7 @@ class FusedBatchNormGradGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(workspace); VARIABLE_NOT_USED(stream_ptr); if (is_null_input_) { diff --git a/mindspore/ccsrc/kernel/gpu/nn/gelu_grad_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/gelu_grad_kernel.h index 7ce6d4d491fba905234141c77941c72bd11692f7..64153490129adef227541b6451f3b65099b0805c 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/gelu_grad_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/gelu_grad_kernel.h @@ -35,7 +35,7 @@ class GeLUGpuGradKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *dy_addr = GetDeviceAddress(inputs, 0); T *x_addr = GetDeviceAddress(inputs, 1); T *dx_addr = GetDeviceAddress(outputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/nn/gelu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/gelu_kernel.h index f0dd37dec49dd9b4ec10bc3f64bcbf76ce0731f2..60968d109b1da59beb2ffedc6fd73a20b262ecc2 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/gelu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/gelu_kernel.h @@ -35,7 +35,7 @@ class GeluGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *input_addr = GetDeviceAddress(inputs, 0); T *output_addr = GetDeviceAddress(outputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/nn/layer_norm_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/layer_norm_gpu_kernel.h index 29a609d95ce3fc11cbb3efb72c26397c5f9762a5..d5ec3ff8f2514438c7bf0595679844f918edad3d 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/layer_norm_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/layer_norm_gpu_kernel.h @@ -35,7 +35,7 @@ class LayerNormGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { auto x = GetDeviceAddress(inputs, 0); auto gamma = GetDeviceAddress(inputs, 1); auto beta = GetDeviceAddress(inputs, 2); diff --git a/mindspore/ccsrc/kernel/gpu/nn/layer_norm_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/layer_norm_grad_gpu_kernel.h index 72cfc23266f28a71dcbefc550325182925c023e7..83bdedb9b35119ac2a7472a25733d9540ce5328f 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/layer_norm_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/layer_norm_grad_gpu_kernel.h @@ -35,7 +35,7 @@ class LayerNormGradGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { auto x = GetDeviceAddress(inputs, 0); auto dy = GetDeviceAddress(inputs, 1); auto var = GetDeviceAddress(inputs, 2); diff --git a/mindspore/ccsrc/kernel/gpu/nn/lstm_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/lstm_gpu_kernel.h index c3e839b9c534913c6e8e2013326e9094b85e612d..01247f0ed6259a1fa8768c9826fe700048c54963 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/lstm_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/lstm_gpu_kernel.h @@ -59,7 +59,7 @@ class LstmGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(stream_ptr); auto x_addr = GetDeviceAddress(inputs, 0); auto hx_addr = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/nn/lstm_grad_data_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/lstm_grad_data_gpu_kernel.h index b12fa3bea529529e52a36b44dd3045ea8beb4843..5591b0c8174fbaaeca6d10b27acba061bfe8d4eb 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/lstm_grad_data_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/lstm_grad_data_gpu_kernel.h @@ -61,7 +61,7 @@ class LstmGradDataGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(stream_ptr); auto y_addr = GetDeviceAddress(inputs, 0); auto dy_addr = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/nn/lstm_grad_weight_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/lstm_grad_weight_gpu_kernel.h index e081b9d0709d73c8f3154ec40697b17dac81a0f1..dd6aae9a00c7cc839f53850dbef38c9896574b62 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/lstm_grad_weight_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/lstm_grad_weight_gpu_kernel.h @@ -54,7 +54,7 @@ class LstmGradWeightGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { VARIABLE_NOT_USED(stream_ptr); auto x_addr = GetDeviceAddress(inputs, 0); auto hx_addr = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/nn/momentum_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/momentum_gpu_kernel.h index 2d8afb5780fdc150d674df8d95bd4137bb969215..8452c177dbea4a8338e1e2e45a9b5adcc5cf801a 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/momentum_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/momentum_gpu_kernel.h @@ -34,7 +34,7 @@ class MomentumGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, const std::vector &, - uintptr_t stream_ptr) override { + void *stream_ptr) override { T *variable = GetDeviceAddress(inputs, 0); T *accumulation = GetDeviceAddress(inputs, 1); T *learning_rate = GetDeviceAddress(inputs, 2); diff --git a/mindspore/ccsrc/kernel/gpu/nn/pooling_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/pooling_gpu_kernel.h index 2446c2295019b95d4791d4c06b4e7ad88c0d903f..faff4537756ab39e3074dd4a00aefc5a32aeb4bc 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/pooling_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/pooling_gpu_kernel.h @@ -59,7 +59,7 @@ class PoolingGpuFwdKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/pooling_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/pooling_grad_gpu_kernel.h index 535f96bbbf48d8d88c0690278168fbc01287dd13..df3454c5819aa38b2271eff6854757281f784d1e 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/pooling_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/pooling_grad_gpu_kernel.h @@ -61,7 +61,7 @@ class PoolingGradGpuFwdKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/relu_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/relu_gpu_kernel.h index d88efd3c7a59106b73868027ec942e0631fa04e3..4cebc4583134bb7f9a40d59b8c95216fb19f4c53 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/relu_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/relu_gpu_kernel.h @@ -43,7 +43,7 @@ class ReLUGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t) override { + const std::vector &outputs, void *) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/relu_grad_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/relu_grad_kernel.h index e93dc31f80da0ac5c41ef7ce4617fa4656b6aaa4..ccc037f6e780dc2e00e4a614f9be1d81a898b84d 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/relu_grad_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/relu_grad_kernel.h @@ -41,7 +41,7 @@ class ReluGradGpuFwdKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t) override { + const std::vector &outputs, void *) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/rmsprop_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/rmsprop_gpu_kernel.h index d1ca53110b858e75cb083389c07c19ee903dfb4d..7eaedfba520df4d4b9c4d8320e274cd2521664ef 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/rmsprop_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/rmsprop_gpu_kernel.h @@ -35,7 +35,7 @@ class RMSPropGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream) override { + const std::vector &outputs, void *stream) override { if (!use_center_) { T *variable = GetDeviceAddress(inputs, 0); T *mean_square = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/nn/softmax_cross_entropy_with_logits_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/softmax_cross_entropy_with_logits_gpu_kernel.h index 4d50d4753d4f2fe9ffd0eff0d1062beb7f1336c0..6840f0a1ebdcaec436ef342455b0318f0f973512 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/softmax_cross_entropy_with_logits_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/softmax_cross_entropy_with_logits_gpu_kernel.h @@ -52,7 +52,7 @@ class SoftmaxCrossEntropyWithLogitsGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/softmax_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/softmax_gpu_kernel.h index cc381d7a71265aa3e410a9ce7b07b46cf10b50c3..060bc57d56d2a6f87c8572651ba1817eb53e4280 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/softmax_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/softmax_gpu_kernel.h @@ -52,7 +52,7 @@ class SoftmaxGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/softmax_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/softmax_grad_gpu_kernel.h index a0356c3bc4fc862a68241cf4e63232f6aa0c76c4..003b55c0ed1bff726ae9d6616463dbed282250fc 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/softmax_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/softmax_grad_gpu_kernel.h @@ -51,7 +51,7 @@ class SoftmaxGradGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/sparse_softmax_cross_entropy_with_logits_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/sparse_softmax_cross_entropy_with_logits_gpu_kernel.h index d232f7a131dffb4094e6e7c98c09205dfc2f2ad4..0749172cc680975c663a0fe40fea49a502888946 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/sparse_softmax_cross_entropy_with_logits_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/sparse_softmax_cross_entropy_with_logits_gpu_kernel.h @@ -52,7 +52,7 @@ class SparseSoftmaxCrossEntropyWithLogitsGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/nn/tanh_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/tanh_gpu_kernel.h index 29fb8cab4865bc5af30f04bdc1519983f2a6acbe..7060ad1792eb0bdd115036a121c30172c0d73fd7 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/tanh_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/tanh_gpu_kernel.h @@ -37,7 +37,7 @@ class TanhGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { auto x_addr = GetDeviceAddress(inputs, 0); auto y_addr = GetDeviceAddress(outputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/nn/tanh_grad_kernel.h b/mindspore/ccsrc/kernel/gpu/nn/tanh_grad_kernel.h index 524dbe372b2b508e430b59584d47af5d23db2994..b5b52d0acf587a0236ee0168ebe10a079c1f02c9 100644 --- a/mindspore/ccsrc/kernel/gpu/nn/tanh_grad_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/nn/tanh_grad_kernel.h @@ -37,7 +37,7 @@ class TanhGradKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { auto y_addr = GetDeviceAddress(inputs, 0); auto dy_addr = GetDeviceAddress(inputs, 1); auto dx_addr = GetDeviceAddress(outputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/other/assign_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/other/assign_gpu_kernel.h index 1c1cde4fd4825bbdd51ba056672766fbe46e168d..b41d583a4302dd44fb218685c449f9ea2c00b98c 100644 --- a/mindspore/ccsrc/kernel/gpu/other/assign_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/other/assign_gpu_kernel.h @@ -33,7 +33,7 @@ class AssignGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { T *var = GetDeviceAddress(inputs, 0); T *value = GetDeviceAddress(inputs, 1); T *output = GetDeviceAddress(outputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold2_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold2_gpu_kernel.h index c1804a5b93b7bbe4cd4fd02e9dbba586da5bed4d..5d2dee3ec73c52ef3f73ef1312b3708a49e4a46e 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold2_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold2_gpu_kernel.h @@ -45,7 +45,7 @@ class BatchNormFold2GpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold2_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold2_grad_gpu_kernel.h index 38adda718ce6e5a31bc64fd1d1b4d3d6cd567e17..28a4cf6cd6ad5efe9877857d34c70c6a169d1d4a 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold2_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold2_grad_gpu_kernel.h @@ -45,7 +45,7 @@ class BatchNormFold2GradGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { if (is_null_input_) { return true; } diff --git a/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold_gpu_kernel.h index a5a8a10dc01baa027f1998e373d5fc1c18109837..a90e9b47d79229a3c46e06c9f68a14b969a5dd74 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold_gpu_kernel.h @@ -53,7 +53,7 @@ class BatchNormFoldGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { (void)workspace; auto x = GetDeviceAddress(inputs, 0); auto mean = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold_grad_gpu_kernel.h index cc420781da4454adfa6d2138e05a4826322a094e..8cbe5b69279d30dd05442370b7268fa548b4b8d5 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/quant/batchnorm_fold_grad_gpu_kernel.h @@ -47,7 +47,7 @@ class BatchNormFoldGradGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { (void)workspace; // 'd_batch_mean', 'd_batch_std', 'x', 'batch_mean', 'batch_std', 'current_step' T *d_batch_mean = GetDeviceAddress(inputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/quant/correction_mul_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/quant/correction_mul_gpu_kernel.h index eeab872ab367b4b48c0678c74f9aca51cd55b6a7..38a9532ef517f1eba95cb5a5a4e3b38aefb70494 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/correction_mul_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/quant/correction_mul_gpu_kernel.h @@ -34,7 +34,7 @@ class CorrectionMulGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override { return output_size_list_; } const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { auto *weight = GetDeviceAddress(inputs, 0); auto *gamma = GetDeviceAddress(inputs, 1); auto *running_std = GetDeviceAddress(inputs, 2); diff --git a/mindspore/ccsrc/kernel/gpu/quant/correction_mul_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/quant/correction_mul_grad_gpu_kernel.h index 29aeb3be139e7e16a839b37100027e5a714552eb..3feffa586b2a2255f3f4af6bb67e561014cf89db 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/correction_mul_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/quant/correction_mul_grad_gpu_kernel.h @@ -35,7 +35,7 @@ class CorrectionMulGradGpuKernel : public GpuKernel { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override { + const std::vector &outputs, void *stream_ptr) override { auto *d_out = GetDeviceAddress(inputs, 0); auto *weight = GetDeviceAddress(inputs, 1); auto *gamma = GetDeviceAddress(inputs, 2); diff --git a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_gpu_kernel.cc b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_gpu_kernel.cc index ee1cb0d01218159a68a0f3dbdb91d5ea1d1673ea..ade7c32da05ac2fa8367ec7b90b522b0a2d5282c 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_gpu_kernel.cc +++ b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_gpu_kernel.cc @@ -114,7 +114,7 @@ void FakeQuantGpuKernel::InitSizeLists() { } bool FakeQuantGpuKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { float *output = GetDeviceAddress(outputs, 0); float *input = GetDeviceAddress(inputs, 0); float *input_min = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_gpu_kernel.h index b14268ed62883c7886ec008fd481f99aeccf0ffb..5a594c615f44c6342aea541ea571fb02d4cd554a 100755 --- a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_gpu_kernel.h @@ -32,7 +32,7 @@ class FakeQuantGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override; const std::vector &GetWorkspaceSizeList() const override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; bool Init(const CNodePtr &kernel) override; protected: diff --git a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_grad_gpu_kernel.cc b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_grad_gpu_kernel.cc index 239e55b5b07bb3e43f423d079d1e2b8e745f8b53..7b7e3f1737dbd6c5efec4209a5015b05552dcc93 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_grad_gpu_kernel.cc +++ b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_grad_gpu_kernel.cc @@ -92,7 +92,7 @@ void FakeQuantGradGpuKernel::InitSizeLists() { } bool FakeQuantGradGpuKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { float *output = GetDeviceAddress(outputs, 0); float *gradient = GetDeviceAddress(inputs, 0); float *input = GetDeviceAddress(inputs, 1); diff --git a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_grad_gpu_kernel.h index cd0f9a4680025e27e9b0cc7ae0750e0669b87645..04c505d2bd96a55e7ee4c42ad09db43020f4c952 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_grad_gpu_kernel.h @@ -32,7 +32,7 @@ class FakeQuantGradGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override; const std::vector &GetWorkspaceSizeList() const override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; bool Init(const CNodePtr &kernel_node) override; protected: diff --git a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_gpu_kernel.cc b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_gpu_kernel.cc index c452bb5dd10d31f7f88816b98999e3778e2900ca..083bf7f011e8b580f0ab23e5679afd4e8b4a3bfc 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_gpu_kernel.cc +++ b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_gpu_kernel.cc @@ -118,7 +118,7 @@ void FakeQuantPerChannelGpuKernel::InitSizeLists() { void FakeQuantPerChannelGpuKernel::CalFakeQuantizeForTraining(float *input, float *output, float *input_min, float *input_max, float *d_nudge_min, float *d_nudge_max, - float *d_scale, uintptr_t stream_ptr) { + float *d_scale, void *stream_ptr) { // calculate the input min and max according by the parameter ema and ema_decay. CalMinMaxPerChannel(input, input_min, input_max, input_size_ / sizeof(float), channel_out_, ema_decay_, ema_, reinterpret_cast(stream_ptr)); @@ -139,7 +139,7 @@ void FakeQuantPerChannelGpuKernel::CalFakeQuantizeForTraining(float *input, floa void FakeQuantPerChannelGpuKernel::CalFakeQuantizeForInfer(float *input, float *output, float *input_min, float *input_max, float *d_nudge_min, float *d_nudge_max, - float *d_scale, uintptr_t stream_ptr) { + float *d_scale, void *stream_ptr) { // real launch CalNudgePerChannel(input_min, input_max, quant_min_, quant_max_, d_nudge_min, d_nudge_max, d_scale, channel_out_, reinterpret_cast(stream_ptr)); @@ -149,7 +149,7 @@ void FakeQuantPerChannelGpuKernel::CalFakeQuantizeForInfer(float *input, float * bool FakeQuantPerChannelGpuKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { (void)workspace; float *output = GetDeviceAddress(outputs, 0); float *input = GetDeviceAddress(inputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_gpu_kernel.h index 8a1bb7293aafd3b1653c8c3dcec639922566154d..bea1a7421fc5c574f15d6e9ae7823f56d69b7cb2 100755 --- a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_gpu_kernel.h @@ -32,7 +32,7 @@ class FakeQuantPerChannelGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override; const std::vector &GetWorkspaceSizeList() const override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; bool Init(const CNodePtr &kernel) override; protected: @@ -40,9 +40,9 @@ class FakeQuantPerChannelGpuKernel : public GpuKernel { private: void CalFakeQuantizeForTraining(float *input, float *output, float *input_min, float *input_max, float *d_nudge_min, - float *d_nudge_max, float *d_scale, uintptr_t stream_ptr); + float *d_nudge_max, float *d_scale, void *stream_ptr); void CalFakeQuantizeForInfer(float *input, float *output, float *input_min, float *input_max, float *d_nudge_min, - float *d_nudge_max, float *d_scale, uintptr_t stream_ptr); + float *d_nudge_max, float *d_scale, void *stream_ptr); size_t input_size_; size_t min_size_; diff --git a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_grad_gpu_kernel.cc b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_grad_gpu_kernel.cc index f995f81190899ac3cae6dc713e2c221be4e0203d..88c976285cc8ab4b54a19e7ec74ae6d56f51ed0f 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_grad_gpu_kernel.cc +++ b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_grad_gpu_kernel.cc @@ -104,7 +104,7 @@ void FakeQuantPerChannelGradGpuKernel::InitSizeLists() { bool FakeQuantPerChannelGradGpuKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { (void)workspace; float *output = GetDeviceAddress(outputs, 0); float *gradient = GetDeviceAddress(inputs, 0); diff --git a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_grad_gpu_kernel.h b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_grad_gpu_kernel.h index c210f4cc810d080511f79f7eaf9992f41c521bb8..fe760d85d24fe42fede6b9926f73f71fd51abd94 100644 --- a/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_grad_gpu_kernel.h +++ b/mindspore/ccsrc/kernel/gpu/quant/fake_quant_per_channel_grad_gpu_kernel.h @@ -32,7 +32,7 @@ class FakeQuantPerChannelGradGpuKernel : public GpuKernel { const std::vector &GetOutputSizeList() const override; const std::vector &GetWorkspaceSizeList() const override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; bool Init(const CNodePtr &kernel_node) override; protected: diff --git a/mindspore/ccsrc/kernel/hccl/hcom_all_broadcast.cc b/mindspore/ccsrc/kernel/hccl/hcom_all_broadcast.cc index 3cc57fe6d88d2958a90da16fd961b5c6582bfdcc..dba692606c5affa42daf4a5015099f32b3b5e451 100644 --- a/mindspore/ccsrc/kernel/hccl/hcom_all_broadcast.cc +++ b/mindspore/ccsrc/kernel/hccl/hcom_all_broadcast.cc @@ -24,17 +24,17 @@ namespace mindspore { namespace kernel { -bool HcomAllBroadCastKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { +bool HcomAllBroadCastKernel::Launch(const std::vector &inputs, + const std::vector & /*workspace*/, + const std::vector & /*outputs*/, void *stream_ptr) { auto context_ptr = MsContext::GetInstance(); MS_EXCEPTION_IF_NULL(context_ptr); if (context_ptr->enable_task_sink()) { return true; } const char *tag = "Hccl-BroadCast"; - auto stream = reinterpret_cast(stream_ptr); hcclResult_t ret = - hcom_broadcast(tag, inputs[0]->addr, hccl_count_, hccl_data_type_list_[0], root_id_, nullptr, stream); + hcom_broadcast(tag, inputs[0]->addr, hccl_count_, hccl_data_type_list_[0], root_id_, nullptr, stream_ptr); if (ret != HCCL_SUCCESS) { MS_LOG(ERROR) << "HcomBroadcastOp : hcom_broadcast fail, return: " << static_cast(ret); return false; diff --git a/mindspore/ccsrc/kernel/hccl/hcom_all_broadcast.h b/mindspore/ccsrc/kernel/hccl/hcom_all_broadcast.h index d7d02a945144a2fa1099d3c15614601aa28511c3..ca8eba91af3be9ffa9c17afdc27e3b4a1a1a4e9d 100644 --- a/mindspore/ccsrc/kernel/hccl/hcom_all_broadcast.h +++ b/mindspore/ccsrc/kernel/hccl/hcom_all_broadcast.h @@ -31,7 +31,7 @@ class HcomAllBroadCastKernel : public HcclKernel { /* Inherit from kernelmod */ bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; private: }; diff --git a/mindspore/ccsrc/kernel/hccl/hcom_all_gather.cc b/mindspore/ccsrc/kernel/hccl/hcom_all_gather.cc index fde1e3bb12deca12b72b826d9122649ff3982ae4..67cd1001e3bfc91ae74427def3da2ca2f15ab365 100644 --- a/mindspore/ccsrc/kernel/hccl/hcom_all_gather.cc +++ b/mindspore/ccsrc/kernel/hccl/hcom_all_gather.cc @@ -24,17 +24,16 @@ namespace mindspore { namespace kernel { -bool HcomAllGatherKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { +bool HcomAllGatherKernel::Launch(const std::vector &inputs, const std::vector & /*workspace*/, + const std::vector &outputs, void *stream_ptr) { auto context_ptr = MsContext::GetInstance(); MS_EXCEPTION_IF_NULL(context_ptr); if (context_ptr->enable_task_sink()) { return true; } const char *tag = "Hccl-AllGather"; - auto stream = reinterpret_cast(stream_ptr); hcclResult_t ret = - hcom_all_gather(tag, inputs[0]->addr, outputs[0]->addr, hccl_count_, hccl_data_type_list_[0], nullptr, stream); + hcom_all_gather(tag, inputs[0]->addr, outputs[0]->addr, hccl_count_, hccl_data_type_list_[0], nullptr, stream_ptr); if (ret != HCCL_SUCCESS) { MS_LOG(ERROR) << "HcomAllGatherKernelOp : hcom_all_gather fail, return: " << static_cast(ret); return false; diff --git a/mindspore/ccsrc/kernel/hccl/hcom_all_gather.h b/mindspore/ccsrc/kernel/hccl/hcom_all_gather.h index f29b5cc0f695e3a45509ff782fdea3f8205243ff..5de2c513cff5b6dbf14a79b541863038705a847e 100644 --- a/mindspore/ccsrc/kernel/hccl/hcom_all_gather.h +++ b/mindspore/ccsrc/kernel/hccl/hcom_all_gather.h @@ -31,7 +31,7 @@ class HcomAllGatherKernel : public HcclKernel { /* Inherit from kernelmod */ bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; private: }; diff --git a/mindspore/ccsrc/kernel/hccl/hcom_all_reduce.cc b/mindspore/ccsrc/kernel/hccl/hcom_all_reduce.cc index a0d96683c22723838151d64394b5cb3ae378d352..2bf9823e5d1fc254a08d75c3a1327bd64fcd42f8 100644 --- a/mindspore/ccsrc/kernel/hccl/hcom_all_reduce.cc +++ b/mindspore/ccsrc/kernel/hccl/hcom_all_reduce.cc @@ -24,17 +24,16 @@ namespace mindspore { namespace kernel { -bool HcomAllReduceKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { +bool HcomAllReduceKernel::Launch(const std::vector &inputs, const std::vector & /*workspace*/, + const std::vector &outputs, void *stream_ptr) { auto context_ptr = MsContext::GetInstance(); MS_EXCEPTION_IF_NULL(context_ptr); if (context_ptr->enable_task_sink()) { return true; } const char *tag = "Hccl-AllReduce"; - auto stream = reinterpret_cast(stream_ptr); hcclResult_t ret = hcom_all_reduce(tag, inputs[0]->addr, outputs[0]->addr, hccl_count_, hccl_data_type_list_[0], - op_type_, nullptr, stream); + op_type_, nullptr, stream_ptr); if (ret != HCCL_SUCCESS) { MS_LOG(ERROR) << "HcomAllReduceKernelOp : hcom_all_reduce fail, return: " << static_cast(ret); return false; diff --git a/mindspore/ccsrc/kernel/hccl/hcom_all_reduce.h b/mindspore/ccsrc/kernel/hccl/hcom_all_reduce.h index 0a3bdb32844996d2677e8bc61f67562fb1bbc8b2..939abd9de7f93b3bdbf2ea5baf6b8510e95e0dc9 100644 --- a/mindspore/ccsrc/kernel/hccl/hcom_all_reduce.h +++ b/mindspore/ccsrc/kernel/hccl/hcom_all_reduce.h @@ -30,7 +30,7 @@ class HcomAllReduceKernel : public HcclKernel { /* Inherit from kernelmod */ bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; private: }; diff --git a/mindspore/ccsrc/kernel/hccl/hcom_all_reduce_scatter.cc b/mindspore/ccsrc/kernel/hccl/hcom_all_reduce_scatter.cc index 36341ed6a7561cb9536026e9ac6eedf71a40fdbc..05217108d907e12299565517dff350e5816d47b8 100644 --- a/mindspore/ccsrc/kernel/hccl/hcom_all_reduce_scatter.cc +++ b/mindspore/ccsrc/kernel/hccl/hcom_all_reduce_scatter.cc @@ -24,17 +24,17 @@ namespace mindspore { namespace kernel { -bool HcomAllReduceScatterKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { +bool HcomAllReduceScatterKernel::Launch(const std::vector &inputs, + const std::vector & /*workspace*/, + const std::vector &outputs, void *stream_ptr) { auto context_ptr = MsContext::GetInstance(); MS_EXCEPTION_IF_NULL(context_ptr); if (context_ptr->enable_task_sink()) { return true; } const char *tag = "Hccl-ReduceScatter"; - auto stream = reinterpret_cast(stream_ptr); hcclResult_t ret = hcom_reduce_scatter(tag, inputs[0]->addr, outputs[0]->addr, hccl_count_, hccl_data_type_list_[0], - op_type_, nullptr, stream); + op_type_, nullptr, stream_ptr); if (ret != HCCL_SUCCESS) { MS_LOG(ERROR) << "HcomReduceScatterOp : hcom_reduce_scatter fail, return: " << static_cast(ret); return false; diff --git a/mindspore/ccsrc/kernel/hccl/hcom_all_reduce_scatter.h b/mindspore/ccsrc/kernel/hccl/hcom_all_reduce_scatter.h index 4c4f821d36b631e876ad251e366ab8213c7805e5..c734b517c6725e4c3c9f294505436e28835fbf6c 100644 --- a/mindspore/ccsrc/kernel/hccl/hcom_all_reduce_scatter.h +++ b/mindspore/ccsrc/kernel/hccl/hcom_all_reduce_scatter.h @@ -31,7 +31,7 @@ class HcomAllReduceScatterKernel : public HcclKernel { /* Inherit from kernelmod */ bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; private: }; diff --git a/mindspore/ccsrc/kernel/kernel.h b/mindspore/ccsrc/kernel/kernel.h index 4217b56625e1cb2f686543da2335ab2e69bff008..684e6cf160d88cd702ab8ac3e4474a5c38c18229 100644 --- a/mindspore/ccsrc/kernel/kernel.h +++ b/mindspore/ccsrc/kernel/kernel.h @@ -124,7 +124,7 @@ class KernelMod { virtual const std::vector &GetOutputSizeList() const = 0; virtual const std::vector &GetWorkspaceSizeList() const = 0; virtual bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) = 0; + const std::vector &outputs, void *stream_ptr) = 0; virtual std::vector GenParameters() { return {}; } virtual ~KernelMod() = default; diff --git a/mindspore/ccsrc/kernel/rts/assign.cc b/mindspore/ccsrc/kernel/rts/assign.cc index b7212c06f67a51b4c0b254a985ed5577c96a57d4..7f214b6e6f6dd0df93549926ec8ff74186007049 100644 --- a/mindspore/ccsrc/kernel/rts/assign.cc +++ b/mindspore/ccsrc/kernel/rts/assign.cc @@ -30,10 +30,8 @@ AssignKernel::AssignKernel() {} AssignKernel::~AssignKernel() {} -bool AssignKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { - auto stream = reinterpret_cast(stream_ptr); - +bool AssignKernel::Launch(const std::vector &inputs, const std::vector & /*workspace*/, + const std::vector & /*outputs*/, void *stream_ptr) { if (inputs.size() != 2) { MS_LOG(ERROR) << "inputs size is not two"; return false; @@ -44,7 +42,7 @@ bool AssignKernel::Launch(const std::vector &inputs, const std::vect return true; } rtError_t status = rtMemcpyAsync(inputs[0]->addr, inputs[0]->size, inputs[1]->addr, inputs[1]->size, - RT_MEMCPY_DEVICE_TO_DEVICE, stream); + RT_MEMCPY_DEVICE_TO_DEVICE, stream_ptr); if (status != RT_ERROR_NONE) { MS_LOG(ERROR) << "Assign op rtMemcpyAsync failed!"; return false; diff --git a/mindspore/ccsrc/kernel/rts/assign.h b/mindspore/ccsrc/kernel/rts/assign.h index c4c6014e8a23a37b00fc2aa2751a9bc01759f827..0e7e52d48fe9642c082e1f4295ce8ff8ca6570c5 100644 --- a/mindspore/ccsrc/kernel/rts/assign.h +++ b/mindspore/ccsrc/kernel/rts/assign.h @@ -29,7 +29,7 @@ class AssignKernel : public RtKernel { ~AssignKernel() override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; }; diff --git a/mindspore/ccsrc/kernel/rts/label_goto.cc b/mindspore/ccsrc/kernel/rts/label_goto.cc index b60361bc607e1b8d67ec5ac372fd64bd58fbe3cc..7bcf42a210d3170535ce99be07ecc8085ef5a228 100644 --- a/mindspore/ccsrc/kernel/rts/label_goto.cc +++ b/mindspore/ccsrc/kernel/rts/label_goto.cc @@ -45,8 +45,8 @@ bool LabelGotoKernel::Init(const AnfNodePtr &anf_node) { return true; } -bool LabelGotoKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { +bool LabelGotoKernel::Launch(const std::vector & /*inputs*/, const std::vector & /*workspace*/, + const std::vector & /*outputs*/, void * /*stream_ptr*/) { MS_LOG(INFO) << "LabelGotoKernel launch"; return true; } diff --git a/mindspore/ccsrc/kernel/rts/label_goto.h b/mindspore/ccsrc/kernel/rts/label_goto.h index f4356391e675106157a5442c5ffe5062c04c3a5b..efccc12d6f5262a23a6bc8cda9cca63c47445706 100644 --- a/mindspore/ccsrc/kernel/rts/label_goto.h +++ b/mindspore/ccsrc/kernel/rts/label_goto.h @@ -32,7 +32,7 @@ class LabelGotoKernel : public RtKernel { bool Init(const AnfNodePtr &anf_node) override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; diff --git a/mindspore/ccsrc/kernel/rts/label_set.cc b/mindspore/ccsrc/kernel/rts/label_set.cc index 0686cb021960f93fc315de96daa903e273926cee..5aedd012dc60fc151f275fbdbde98ac44263d9e1 100644 --- a/mindspore/ccsrc/kernel/rts/label_set.cc +++ b/mindspore/ccsrc/kernel/rts/label_set.cc @@ -45,8 +45,8 @@ bool LabelSetKernel::Init(const AnfNodePtr &anf_node) { return true; } -bool LabelSetKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { +bool LabelSetKernel::Launch(const std::vector & /*inputs*/, const std::vector & /*workspace*/, + const std::vector & /*outputs*/, void * /*stream_ptr*/) { MS_LOG(INFO) << "LabelSetKernel launch"; return true; } diff --git a/mindspore/ccsrc/kernel/rts/label_set.h b/mindspore/ccsrc/kernel/rts/label_set.h index 12f0d5b8069c01b5c7d5cc2357376ba07cfa696b..d05d81f89859038523f637f7685a84307b9eff7c 100644 --- a/mindspore/ccsrc/kernel/rts/label_set.h +++ b/mindspore/ccsrc/kernel/rts/label_set.h @@ -32,7 +32,7 @@ class LabelSetKernel : public RtKernel { bool Init(const AnfNodePtr &anf_node) override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; diff --git a/mindspore/ccsrc/kernel/rts/label_switch.cc b/mindspore/ccsrc/kernel/rts/label_switch.cc index f428996410dffc1a31c2c612382002869b0b6128..6647ac7eb693d06febb49af825fcfe03b9f48f16 100644 --- a/mindspore/ccsrc/kernel/rts/label_switch.cc +++ b/mindspore/ccsrc/kernel/rts/label_switch.cc @@ -53,8 +53,9 @@ bool LabelSwitchKernel::Init(const AnfNodePtr &anf_node) { return true; } -bool LabelSwitchKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { +bool LabelSwitchKernel::Launch(const std::vector & /*inputs*/, + const std::vector & /*workspace*/, + const std::vector & /*outputs*/, void * /*stream_ptr*/) { MS_LOG(INFO) << "LabelSwitchKernel launch"; return true; } diff --git a/mindspore/ccsrc/kernel/rts/label_switch.h b/mindspore/ccsrc/kernel/rts/label_switch.h index 372dc0cd7a6bd1bc223cc219bcd15f96fd8639f5..0accd26afbf638b33a7e47a9fda7b6f68b68cd8e 100644 --- a/mindspore/ccsrc/kernel/rts/label_switch.h +++ b/mindspore/ccsrc/kernel/rts/label_switch.h @@ -32,7 +32,7 @@ class LabelSwitchKernel : public RtKernel { bool Init(const AnfNodePtr &anf_node) override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; diff --git a/mindspore/ccsrc/kernel/rts/memcpy_async.cc b/mindspore/ccsrc/kernel/rts/memcpy_async.cc index 08b040cd151c962c02ab7deaa242a04ee34e2990..3d5a7c88abc4cc64382c2d0e725dbbd121b713f6 100644 --- a/mindspore/ccsrc/kernel/rts/memcpy_async.cc +++ b/mindspore/ccsrc/kernel/rts/memcpy_async.cc @@ -34,9 +34,7 @@ MemCpyAsyncKernel::MemCpyAsyncKernel() {} MemCpyAsyncKernel::~MemCpyAsyncKernel() {} bool MemCpyAsyncKernel::Launch(const std::vector &inputs, const std::vector & /*workspace*/, - const std::vector &outputs, uintptr_t stream_ptr) { - auto stream = reinterpret_cast(stream_ptr); - + const std::vector &outputs, void *stream_ptr) { if (inputs.size() != 1) { MS_LOG(ERROR) << "inputs size is not one"; return false; @@ -51,7 +49,7 @@ bool MemCpyAsyncKernel::Launch(const std::vector &inputs, const std: return true; } rtError_t status = rtMemcpyAsync(outputs[0]->addr, outputs[0]->size, inputs[0]->addr, inputs[0]->size, - RT_MEMCPY_DEVICE_TO_DEVICE, stream); + RT_MEMCPY_DEVICE_TO_DEVICE, stream_ptr); if (status != RT_ERROR_NONE) { MS_LOG(ERROR) << "MemCpyAsync op rtMemcpyAsync failed!"; return false; diff --git a/mindspore/ccsrc/kernel/rts/memcpy_async.h b/mindspore/ccsrc/kernel/rts/memcpy_async.h index b8e39374f86fb624c8be90827f232a6c3b050b0f..94bbf1ca1cb0c650f57a3e7c14824f064880f567 100644 --- a/mindspore/ccsrc/kernel/rts/memcpy_async.h +++ b/mindspore/ccsrc/kernel/rts/memcpy_async.h @@ -31,7 +31,7 @@ class MemCpyAsyncKernel : public RtKernel { bool Init(const AnfNodePtr &anf_node) override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; diff --git a/mindspore/ccsrc/kernel/rts/profiling_kernel_mod.cc b/mindspore/ccsrc/kernel/rts/profiling_kernel_mod.cc index 002c2523746fe1a3938385972ffd28ab8c860a25..ff005f399bd7b1e7c1d16c66f898d51fd9ad1121 100644 --- a/mindspore/ccsrc/kernel/rts/profiling_kernel_mod.cc +++ b/mindspore/ccsrc/kernel/rts/profiling_kernel_mod.cc @@ -50,10 +50,9 @@ bool ProfilingKernelMod::Init(const AnfNodePtr &anf_node) { return true; } -bool ProfilingKernelMod::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { - MS_LOG(INFO) << "gen task inputs size:" << inputs.size() << ", workspace size:" << workspace.size() - << ", outputs size:" << outputs.size() << ", stream_ptr:" << stream_ptr; +bool ProfilingKernelMod::Launch(const std::vector & /*inputs*/, + const std::vector & /*workspace*/, + const std::vector & /*outputs*/, void * /*stream_ptr*/) { return true; } diff --git a/mindspore/ccsrc/kernel/rts/profiling_kernel_mod.h b/mindspore/ccsrc/kernel/rts/profiling_kernel_mod.h index b9b4a1c05cb322c85b8342c931e282fa98222729..f77f3b5c6735f6adb82c011aa7c3421f0d408099 100644 --- a/mindspore/ccsrc/kernel/rts/profiling_kernel_mod.h +++ b/mindspore/ccsrc/kernel/rts/profiling_kernel_mod.h @@ -24,7 +24,7 @@ class ProfilingKernelMod : public RtKernel { ProfilingKernelMod() = default; ~ProfilingKernelMod() override = default; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; bool Init(const AnfNodePtr &anf_node) override; diff --git a/mindspore/ccsrc/kernel/rts/recv.cc b/mindspore/ccsrc/kernel/rts/recv.cc index af921083b5aa08a2c397fb6f4190bdfa4689e957..b68380dac8e9252012a490f9d3891c03985abba1 100644 --- a/mindspore/ccsrc/kernel/rts/recv.cc +++ b/mindspore/ccsrc/kernel/rts/recv.cc @@ -43,10 +43,9 @@ bool RecvKernel::Init(const AnfNodePtr &anf_node) { } bool RecvKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { rtEvent_t stream_event{}; - auto stream = reinterpret_cast(stream_ptr); - auto status = rtStreamWaitEvent(stream, stream_event); + auto status = rtStreamWaitEvent(stream_ptr, stream_event); if (status != RT_ERROR_NONE) { MS_LOG(ERROR) << "Recv rtStreamWaitEvent failed!"; return false; diff --git a/mindspore/ccsrc/kernel/rts/recv.h b/mindspore/ccsrc/kernel/rts/recv.h index 3a32fb98126ccba867d9e3e495334be21ab2f34c..68f0b69cc5c833dddfab7e23a3d1228f79308a26 100644 --- a/mindspore/ccsrc/kernel/rts/recv.h +++ b/mindspore/ccsrc/kernel/rts/recv.h @@ -31,7 +31,7 @@ class RecvKernel : public RtKernel { bool Init(const AnfNodePtr &anf_node) override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; diff --git a/mindspore/ccsrc/kernel/rts/send.cc b/mindspore/ccsrc/kernel/rts/send.cc index abb93f5af4178f9b628283e165883552c84f0abb..ebcb53069eb5408aa33c27d06bd881cac72c6a41 100644 --- a/mindspore/ccsrc/kernel/rts/send.cc +++ b/mindspore/ccsrc/kernel/rts/send.cc @@ -40,10 +40,9 @@ bool SendKernel::Init(const AnfNodePtr &anf_node) { } bool SendKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { rtEvent_t event{}; - auto stream = reinterpret_cast(stream_ptr); - rtError_t status = rtEventRecord(event, stream); + rtError_t status = rtEventRecord(event, stream_ptr); if (status != RT_ERROR_NONE) { MS_LOG(ERROR) << "Send op rtEventRecord failed!"; return false; diff --git a/mindspore/ccsrc/kernel/rts/send.h b/mindspore/ccsrc/kernel/rts/send.h index bbf571b6ab7a3be5e2fea477906ce234f95f59a5..5c5b7cf09e683044e4d108816463cbe3da3d465e 100644 --- a/mindspore/ccsrc/kernel/rts/send.h +++ b/mindspore/ccsrc/kernel/rts/send.h @@ -29,7 +29,7 @@ class SendKernel : public RtKernel { ~SendKernel() override; bool Init(const AnfNodePtr &anf_node) override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; diff --git a/mindspore/ccsrc/kernel/rts/stream_active.cc b/mindspore/ccsrc/kernel/rts/stream_active.cc index ce6d94e4a572a69e61f8a4f089f89ebb759b7a31..3666dd670f7ae62d162189d7a8438a67d5bb7316 100644 --- a/mindspore/ccsrc/kernel/rts/stream_active.cc +++ b/mindspore/ccsrc/kernel/rts/stream_active.cc @@ -41,9 +41,8 @@ bool StreamActiveKernel::Init(const AnfNodePtr &anf_node) { } bool StreamActiveKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { MS_LOG(INFO) << "Stream active op launch start"; - auto stream = reinterpret_cast(stream_ptr); if (active_streams_index_.empty()) { MS_LOG(ERROR) << "activeStreamList_ is empty!"; @@ -54,7 +53,7 @@ bool StreamActiveKernel::Launch(const std::vector &inputs, const std rtError_t status; for (auto index : active_streams_index_) { act_stream = kernel::TaskStream::GetInstance()->gen_stream_list()[index]; - status = rtStreamActive(act_stream, stream); + status = rtStreamActive(act_stream, stream_ptr); if (status != RT_ERROR_NONE) { MS_LOG(ERROR) << "Stream active failed!"; return false; diff --git a/mindspore/ccsrc/kernel/rts/stream_active.h b/mindspore/ccsrc/kernel/rts/stream_active.h index 0955de8cef179adaf250f18e72b528edfaec245b..68c422e7c29495a286cc1728dc3633b7698e5357 100644 --- a/mindspore/ccsrc/kernel/rts/stream_active.h +++ b/mindspore/ccsrc/kernel/rts/stream_active.h @@ -31,7 +31,7 @@ class StreamActiveKernel : public RtKernel { bool Init(const AnfNodePtr &anf_node) override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; diff --git a/mindspore/ccsrc/kernel/rts/stream_switch.cc b/mindspore/ccsrc/kernel/rts/stream_switch.cc index 4fc8f8ef59b75882166b806399a0fefce1ad086d..9dfb3e8de04fe195b93cf435c0e1828e7d53f973 100644 --- a/mindspore/ccsrc/kernel/rts/stream_switch.cc +++ b/mindspore/ccsrc/kernel/rts/stream_switch.cc @@ -51,7 +51,7 @@ bool StreamSwitchKernel::Init(const AnfNodePtr &anf_node) { } bool StreamSwitchKernel::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { + const std::vector &outputs, void *stream_ptr) { MS_LOG(INFO) << "stream switch op launch start"; if (inputs.size() != 2) { MS_LOG(ERROR) << "Stream switch inputs size is " << inputs.size() << ", only support 2"; @@ -59,9 +59,8 @@ bool StreamSwitchKernel::Launch(const std::vector &inputs, const std void *loop_cnt = inputs[0]->addr; void *ites_per_loop = inputs[1]->addr; - auto stream = reinterpret_cast(stream_ptr); rtStream_t true_stream_ = kernel::TaskStream::GetInstance()->gen_stream_list()[true_stream_index_]; - rtError_t status = rtStreamSwitchEx(loop_cnt, cond_, ites_per_loop, true_stream_, stream, data_type_); + rtError_t status = rtStreamSwitchEx(loop_cnt, cond_, ites_per_loop, true_stream_, stream_ptr, data_type_); if (status != RT_ERROR_NONE) { MS_LOG(ERROR) << "Stream switch failed!"; return false; diff --git a/mindspore/ccsrc/kernel/rts/stream_switch.h b/mindspore/ccsrc/kernel/rts/stream_switch.h index 3cc09b749454fdf69fc0e8819d91f78d775b9c00..4e927f30599869d8b2cc6bd0d68ce4c937a2d415 100644 --- a/mindspore/ccsrc/kernel/rts/stream_switch.h +++ b/mindspore/ccsrc/kernel/rts/stream_switch.h @@ -32,7 +32,7 @@ class StreamSwitchKernel : public RtKernel { bool Init(const AnfNodePtr &anf_node) override; bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspace, const std::vector &outputs, uint32_t stream_id) override; diff --git a/mindspore/ccsrc/kernel/tbe/tbe_kernel_mod.cc b/mindspore/ccsrc/kernel/tbe/tbe_kernel_mod.cc index 6a44ea2cc145837249b9d1499ff5014475481956..0f377940daf6df24562b972935f664b8d63b2201 100644 --- a/mindspore/ccsrc/kernel/tbe/tbe_kernel_mod.cc +++ b/mindspore/ccsrc/kernel/tbe/tbe_kernel_mod.cc @@ -26,8 +26,8 @@ using TbeTaskInfoPtr = std::shared_ptr; using tbe::KernelManager; bool TbeKernelMod::Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) { - if (stream_ptr == 0) { + const std::vector &outputs, void *stream_ptr) { + if (stream_ptr == nullptr) { MS_LOG(ERROR) << "stream_ptr should not be nullptr."; return false; } @@ -55,10 +55,9 @@ bool TbeKernelMod::Launch(const std::vector &inpu [](const AddressPtr &addr) -> void * { return addr->addr; }); } rtL2Ctrl_t *l2ctrl = nullptr; - auto *stream = reinterpret_cast(stream_ptr); const void *stubFunc = reinterpret_cast(func_stub); auto argsSize = static_cast(UlongToUint(sizeof(void *)) * runtimeargs.size()); - if (RT_ERROR_NONE != rtKernelLaunch(stubFunc, blockdim, runtimeargs.data(), argsSize, l2ctrl, stream)) { + if (RT_ERROR_NONE != rtKernelLaunch(stubFunc, blockdim, runtimeargs.data(), argsSize, l2ctrl, stream_ptr)) { MS_LOG(ERROR) << "Call runtime rtKernelLaunch error."; return false; } diff --git a/mindspore/ccsrc/kernel/tbe/tbe_kernel_mod.h b/mindspore/ccsrc/kernel/tbe/tbe_kernel_mod.h index d7af83e2c971c15be442a1d8af80b8d761361b96..e0e7ab46461436a50bb50ba58b9414cee7b2bca4 100644 --- a/mindspore/ccsrc/kernel/tbe/tbe_kernel_mod.h +++ b/mindspore/ccsrc/kernel/tbe/tbe_kernel_mod.h @@ -39,7 +39,7 @@ class TbeKernelMod : public AscendKernelMod { const std::vector &GetWorkspaceSizeList() const override { return workspace_size_list_; } bool Launch(const std::vector &inputs, const std::vector &workspace, - const std::vector &outputs, uintptr_t stream_ptr) override; + const std::vector &outputs, void *stream_ptr) override; std::vector GenTask(const std::vector &inputs, const std::vector &workspaces, const std::vector &outputs, uint32_t stream_id) override; std::vector GenParameters() override;