[cherry-pick] [ROCM] bugfix for bilinear_interp_v2_grad (PaddlePaddl…

…e#36160) PaddlePaddle#36161 ATT, cherry-pick PaddlePaddle#36160
gsq7474741 · Sep 28, 2021 · c576169 · c576169
1 parent cea0bc2
commit c576169
Showing 1 changed file with 14 additions and 2 deletions.
diff --git a/paddle/fluid/operators/interpolate_v2_op.cu b/paddle/fluid/operators/interpolate_v2_op.cu
@@ -1198,7 +1198,12 @@ static void Interpolate2DCUDAFwd(const framework::ExecutionContext& ctx,
         input_data, in_h, in_w, n, in_chw, output_data, out_h, out_w, n,
         out_chw, c, ratio_h, ratio_w, align_corners, align_mode, data_layout);
   } else if ("bicubic" == interp_method) {
-    KeBicubicInterpFw<T><<<config.block_per_grid, 512, 0,
+#ifdef __HIPCC__
+    constexpr int thread_per_block = 256;
+#else
+    constexpr int thread_per_block = 512;
+#endif
+    KeBicubicInterpFw<T><<<config.block_per_grid, thread_per_block, 0,
                            ctx.cuda_device_context().stream()>>>(
         input_data, in_h, in_w, n, in_chw, output_data, out_h, out_w, n,
         out_chw, c, ratio_h, ratio_w, align_corners, data_layout);
@@ -1606,9 +1611,11 @@ static void Interpolate2DCUDABwd(const framework::ExecutionContext& ctx,
     const T align_type_value = (align_mode == 0 && !align_corners) ? 0.5f : 0;
     bool is_nchw = (data_layout == DataLayout::kNCHW) ? true : false;
     bool optimize_flag = false;
+#ifndef __HIPCC__
     optimize_flag = (in_h < (out_h >> 6) && in_w < (out_w >> 6))
                         ? true
                         : ((in_h == 1 && in_w == 1) ? true : false);
+#endif
 
     if (optimize_flag & is_nchw) {
       KeBilinearInterpBwShareMemory<
@@ -1623,7 +1630,12 @@ static void Interpolate2DCUDABwd(const framework::ExecutionContext& ctx,
           ratio_h, ratio_w, align_type_value, is_nchw);
     }
   } else if ("bicubic" == interp_method) {
-    KeBicubicInterpBw<T><<<config.block_per_grid, 512, 0,
+#ifdef __HIPCC__
+    constexpr int thread_per_block = 256;
+#else
+    constexpr int thread_per_block = 512;
+#endif
+    KeBicubicInterpBw<T><<<config.block_per_grid, thread_per_block, 0,
                            ctx.cuda_device_context().stream()>>>(
         input_grad_data, in_h, in_w, n, in_chw, output_grad_data, out_h, out_w,
         n, out_chw, c, ratio_h, ratio_w, align_corners, data_layout);