fix math function arch mismatch for older GPU

Author: kexinzhao

paddle/fluid/operators/math/math_function_test.cu

@@ -14,6 +14,8 @@
 #include "gtest/gtest.h"
 #include "paddle/fluid/operators/math/math_function.h"
 
+#include <iostream>
+
 void fill_fp16_data(paddle::platform::float16* in_ptr, size_t size,
                     const std::vector<float>& data) {
   PADDLE_ENFORCE_EQ(size, data.size());
@@ -22,6 +24,15 @@ void fill_fp16_data(paddle::platform::float16* in_ptr, size_t size,
   }
 }
 
+bool is_fp16_supported(int device_id) {
+  cudaDeviceProp device_prop;
+  cudaDeviceProperties(&device_prop, device_id);
+  PADDLE_ENFORCE_EQ(cudaGetLastError(), cudaSuccess);
+  int compute_capability = device_prop.major * 10 + device_prop.minor;
+  std::cout << "compute_capability is " << compute_capability << std::endl;
+  return compute_capability >= 53;
+}
+
 TEST(math_function, notrans_mul_trans_fp32) {
   using namespace paddle::framework;
   using namespace paddle::platform;
@@ -62,6 +73,10 @@ TEST(math_function, notrans_mul_trans_fp16) {
   using namespace paddle::framework;
   using namespace paddle::platform;
 
+  if (!is_fp16_supported(0)) {
+    return;
+  }
+
   Tensor input1;
   Tensor input1_gpu;
   Tensor input2_gpu;
@@ -139,6 +154,10 @@ TEST(math_function, trans_mul_notrans_fp16) {
   using namespace paddle::framework;
   using namespace paddle::platform;
 
+  if (!is_fp16_supported(0)) {
+    return;
+  }
+
   Tensor input1;
   Tensor input1_gpu;
   Tensor input2_gpu;
@@ -237,6 +256,10 @@ TEST(math_function, gemm_notrans_cublas_fp16) {
   using namespace paddle::framework;
   using namespace paddle::platform;
 
+  if (!is_fp16_supported(0)) {
+    return;
+  }
+
   Tensor input1;
   Tensor input2;
   Tensor input3;
@@ -344,6 +367,10 @@ TEST(math_function, gemm_trans_cublas_fp16) {
   using namespace paddle::framework;
   using namespace paddle::platform;
 
+  if (!is_fp16_supported(0)) {
+    return;
+  }
+
   Tensor input1;
   Tensor input2;
   Tensor input3;