[API] isnan isinf isfinite support bigtensor (#72517)

wanghuancoder · web-flow · commit 7a7f5cb153cb · 2025-04-28T14:41:23.000+08:00
* isnan isinf isfinite support bigtensor

* refine
diff --git a/paddle/phi/kernels/impl/isfinite_kernel_impl.h b/paddle/phi/kernels/impl/isfinite_kernel_impl.h
@@ -78,8 +78,8 @@ struct IsfiniteFunctor<
                   const DenseTensor& in,
                   DenseTensor* output) {
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       out_data[i] = true;
     }
   }
@@ -95,8 +95,8 @@ struct IsfiniteFunctor<
                   DenseTensor* output) {
     auto* in_a = in.data<T>();
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       const T& a = in_a[i];
       out_data[i] = std::isfinite(a);
     }
@@ -113,8 +113,8 @@ struct IsfiniteFunctor<
                   DenseTensor* output) {
     auto* in_a = in.data<T>();
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       const T& a = in_a[i];
       out_data[i] = phi::dtype::isfinite(a);
     }
@@ -131,8 +131,8 @@ struct IsfiniteFunctor<
                   DenseTensor* output) {
     auto* in_a = in.data<T>();
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       const T& a = in_a[i];
       out_data[i] = std::isfinite(a.real) && std::isfinite(a.imag);
     }
@@ -157,8 +157,8 @@ struct IsnanFunctor<
                   const DenseTensor& in,
                   DenseTensor* output) {
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       out_data[i] = false;
     }
   }
@@ -174,8 +174,8 @@ struct IsnanFunctor<
                   DenseTensor* output) {
     auto* in_a = in.data<T>();
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       const T& a = in_a[i];
       out_data[i] = std::isnan(a);
     }
@@ -191,8 +191,8 @@ struct IsnanFunctor<phi::CPUContext,
                   DenseTensor* output) {
     auto* in_a = in.data<T>();
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       const T& a = in_a[i];
       out_data[i] = phi::dtype::isnan(a);
     }
@@ -209,8 +209,8 @@ struct IsnanFunctor<
                   DenseTensor* output) {
     auto* in_a = in.data<T>();
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       const T& a = in_a[i];
       out_data[i] = std::isnan(a.real) || std::isnan(a.imag);
     }
@@ -236,7 +236,7 @@ struct IsinfFunctor<
                   DenseTensor* output) {
     auto* out_data = ctx.template Alloc<bool>(output);
     auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    for (int64_t i = 0; i < num; i++) {
       out_data[i] = false;
     }
   }
@@ -252,8 +252,8 @@ struct IsinfFunctor<
                   DenseTensor* output) {
     auto* in_a = in.data<T>();
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       const T& a = in_a[i];
       out_data[i] = std::isinf(a);
     }
@@ -269,8 +269,8 @@ struct IsinfFunctor<phi::CPUContext,
                   DenseTensor* output) {
     auto* in_a = in.data<T>();
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       const T& a = in_a[i];
       out_data[i] = phi::dtype::isinf(a);
     }
@@ -287,8 +287,8 @@ struct IsinfFunctor<
                   DenseTensor* output) {
     auto* in_a = in.data<T>();
     auto* out_data = ctx.template Alloc<bool>(output);
-    auto num = in.numel();
-    for (int i = 0; i < num; i++) {
+    int64_t num = in.numel();
+    for (int64_t i = 0; i < num; i++) {
       const T& a = in_a[i];
       out_data[i] = std::isinf(a.real) || std::isinf(a.imag);
     }
@@ -297,117 +297,117 @@ struct IsinfFunctor<
 
 #if defined(__NVCC__) || defined(__HIPCC__)
 /* IsfiniteFunctor */
-template <typename T>
+template <typename T, typename IndexType>
 __global__ void IsfiniteCUDAKernel(
     const T* in_data,
-    int num,
+    IndexType num,
     bool* out_data,
     typename std::enable_if<std::is_floating_point<T>::value>::type* = 0) {
-  unsigned int idx = threadIdx.x + blockIdx.x * blockDim.x;
-  for (int i = idx; i < num; i += blockDim.x * gridDim.x) {
+  IndexType idx = threadIdx.x + blockIdx.x * blockDim.x;
+  for (IndexType i = idx; i < num; i += blockDim.x * gridDim.x) {
     const T& a = in_data[i];
     out_data[i] = isfinite(a);
   }
 }
 
-template <typename T>
+template <typename T, typename IndexType>
 __global__ void IsfiniteCUDAKernel(
     const T* in_data,
-    int num,
+    IndexType num,
     bool* out_data,
     typename std::enable_if<std::is_integral<T>::value>::type* = 0) {
-  unsigned int idx = threadIdx.x + blockIdx.x * blockDim.x;
-  for (int i = idx; i < num; i += blockDim.x * gridDim.x) {
+  IndexType idx = threadIdx.x + blockIdx.x * blockDim.x;
+  for (IndexType i = idx; i < num; i += blockDim.x * gridDim.x) {
     out_data[i] = true;
   }
 }
 
-template <typename T>
+template <typename T, typename IndexType>
 __global__ void IsfiniteCUDAKernel(
     const T* in_data,
-    int num,
+    IndexType num,
     bool* out_data,
     typename std::enable_if<is_complex64_or_complex128<T>::value>::type* = 0) {
-  unsigned int idx = threadIdx.x + blockIdx.x * blockDim.x;
-  for (int i = idx; i < num; i += blockDim.x * gridDim.x) {
+  IndexType idx = threadIdx.x + blockIdx.x * blockDim.x;
+  for (IndexType i = idx; i < num; i += blockDim.x * gridDim.x) {
     const T& a = in_data[i];
     out_data[i] = isfinite(a.real) && isfinite(a.imag);
   }
 }
 
 /* IsnanFunctor */
-template <typename T>
+template <typename T, typename IndexType>
 __global__ void IsnanCUDAKernel(
     const T* in_data,
-    int num,
+    IndexType num,
     bool* out_data,
     typename std::enable_if<std::is_floating_point<T>::value>::type* = 0) {
-  unsigned int idx = threadIdx.x + blockIdx.x * blockDim.x;
-  for (int i = idx; i < num; i += blockDim.x * gridDim.x) {
+  IndexType idx = threadIdx.x + blockIdx.x * blockDim.x;
+  for (IndexType i = idx; i < num; i += blockDim.x * gridDim.x) {
     const T& a = in_data[i];
     out_data[i] = isnan(a);
   }
 }
 
-template <typename T>
+template <typename T, typename IndexType>
 __global__ void IsnanCUDAKernel(
     const T* in_data,
-    int num,
+    IndexType num,
     bool* out_data,
     typename std::enable_if<std::is_integral<T>::value>::type* = 0) {
-  unsigned int idx = threadIdx.x + blockIdx.x * blockDim.x;
-  for (int i = idx; i < num; i += blockDim.x * gridDim.x) {
+  IndexType idx = threadIdx.x + blockIdx.x * blockDim.x;
+  for (IndexType i = idx; i < num; i += blockDim.x * gridDim.x) {
     out_data[i] = false;
   }
 }
 
-template <typename T>
+template <typename T, typename IndexType>
 __global__ void IsnanCUDAKernel(
     const T* in_data,
-    int num,
+    IndexType num,
     bool* out_data,
     typename std::enable_if<is_complex64_or_complex128<T>::value>::type* = 0) {
-  unsigned int idx = threadIdx.x + blockIdx.x * blockDim.x;
-  for (int i = idx; i < num; i += blockDim.x * gridDim.x) {
+  IndexType idx = threadIdx.x + blockIdx.x * blockDim.x;
+  for (IndexType i = idx; i < num; i += blockDim.x * gridDim.x) {
     const T& a = in_data[i];
     out_data[i] = isnan(a.real) || isnan(a.imag);
   }
 }
 
 /* IsinfFunctor */
-template <typename T>
+template <typename T, typename IndexType>
 __global__ void IsinfCUDAKernel(
     const T* in_data,
-    int num,
+    IndexType num,
     bool* out_data,
     typename std::enable_if<std::is_floating_point<T>::value>::type* = 0) {
-  unsigned int idx = threadIdx.x + blockIdx.x * blockDim.x;
-  for (int i = idx; i < num; i += blockDim.x * gridDim.x) {
+  IndexType idx = threadIdx.x + blockIdx.x * blockDim.x;
+  for (IndexType i = idx; i < num; i += blockDim.x * gridDim.x) {
     const T& a = in_data[i];
     out_data[i] = isinf(a);
   }
 }
 
-template <typename T>
+template <typename T, typename IndexType>
 __global__ void IsinfCUDAKernel(
     const T* in_data,
-    int num,
+    IndexType num,
     bool* out_data,
     typename std::enable_if<std::is_integral<T>::value>::type* = 0) {
-  unsigned int idx = threadIdx.x + blockIdx.x * blockDim.x;
-  for (int i = idx; i < num; i += blockDim.x * gridDim.x) {
+  IndexType idx = threadIdx.x + blockIdx.x * blockDim.x;
+  for (IndexType i = idx; i < num; i += blockDim.x * gridDim.x) {
     out_data[i] = false;
   }
 }
 
-template <typename T>
+template <typename T, typename IndexType>
 __global__ void IsinfCUDAKernel(
     const T* in_data,
-    int num,
+    IndexType num,
     bool* out_data,
     typename std::enable_if<is_complex64_or_complex128<T>::value>::type* = 0) {
-  unsigned int idx = threadIdx.x + blockIdx.x * blockDim.x;
-  for (int i = idx; i < num; i += blockDim.x * gridDim.x) {
+  IndexType idx = threadIdx.x + blockIdx.x * blockDim.x;
+  for (IndexType i = idx; i < num; i += blockDim.x * gridDim.x) {
     const T& a = in_data[i];
     out_data[i] = isinf(a.real) || isinf(a.imag);
   }
@@ -418,14 +418,19 @@ struct IsfiniteFunctor<phi::GPUContext, T> {
   void operator()(const phi::GPUContext& dev_ctx,
                   const DenseTensor& in,
                   DenseTensor* output) {
-    int num = in.numel();
+    int64_t num = in.numel();
     const T* in_data = in.data<T>();
     bool* out_data = dev_ctx.template Alloc<bool>(output);
-    int block = 1024;
-    int grid = (block - 1 + num) / block;
+    int64_t block = 1024;
+    int64_t grid = (block - 1 + num) / block;
     grid = (grid > block) ? block : grid;
-    IsfiniteCUDAKernel<T>
-        <<<grid, block, 0, dev_ctx.stream()>>>(in_data, num, out_data);
+    if (num + block * grid + 1 > std::numeric_limits<unsigned int>::max()) {
+      IsfiniteCUDAKernel<T, int64_t>
+          <<<grid, block, 0, dev_ctx.stream()>>>(in_data, num, out_data);
+    } else {
+      IsfiniteCUDAKernel<T, unsigned int>
+          <<<grid, block, 0, dev_ctx.stream()>>>(in_data, num, out_data);
+    }
   }
 };
 
@@ -434,14 +439,19 @@ struct IsnanFunctor<phi::GPUContext, T> {
   void operator()(const phi::GPUContext& dev_ctx,
                   const DenseTensor& in,
                   DenseTensor* output) {
-    int num = in.numel();
+    int64_t num = in.numel();
     const T* in_data = in.data<T>();
     bool* out_data = dev_ctx.template Alloc<bool>(output);
-    int block = 1024;
-    int grid = (block - 1 + num) / block;
+    int64_t block = 1024;
+    int64_t grid = (block - 1 + num) / block;
     grid = (grid > block) ? block : grid;
-    IsnanCUDAKernel<T>
-        <<<grid, block, 0, dev_ctx.stream()>>>(in_data, num, out_data);
+    if (num + block * grid + 1 > std::numeric_limits<unsigned int>::max()) {
+      IsnanCUDAKernel<T, int64_t>
+          <<<grid, block, 0, dev_ctx.stream()>>>(in_data, num, out_data);
+    } else {
+      IsnanCUDAKernel<T, unsigned int>
+          <<<grid, block, 0, dev_ctx.stream()>>>(in_data, num, out_data);
+    }
   }
 };
 
@@ -450,14 +460,19 @@ struct IsinfFunctor<phi::GPUContext, T> {
   void operator()(const phi::GPUContext& dev_ctx,
                   const DenseTensor& in,
                   DenseTensor* output) {
-    int num = in.numel();
+    int64_t num = in.numel();
     const T* in_data = in.data<T>();
     bool* out_data = dev_ctx.template Alloc<bool>(output);
-    int block = 1024;
-    int grid = (block - 1 + num) / block;
+    int64_t block = 1024;
+    int64_t grid = (block - 1 + num) / block;
     grid = (grid > block) ? block : grid;
-    IsinfCUDAKernel<T>
-        <<<grid, block, 0, dev_ctx.stream()>>>(in_data, num, out_data);
+    if (num + block * grid + 1 > std::numeric_limits<unsigned int>::max()) {
+      IsinfCUDAKernel<T, int64_t>
+          <<<grid, block, 0, dev_ctx.stream()>>>(in_data, num, out_data);
+    } else {
+      IsinfCUDAKernel<T, unsigned int>
+          <<<grid, block, 0, dev_ctx.stream()>>>(in_data, num, out_data);
+    }
   }
 };
 #endif
