wip

Tencent · Apr 8, 2024 · ef8a99a · ef8a99a
1 parent 8711dd4
commit ef8a99a
Show file tree

Hide file tree

Showing 20 changed files with 102 additions and 48 deletions.
diff --git a/src/layer/loongarch/convolution1d_loongarch.cpp b/src/layer/loongarch/convolution1d_loongarch.cpp
@@ -78,6 +78,9 @@ int Convolution1D_loongarch::create_pipeline(const Option& opt)
         }
     }
 
+    if (opt.lightmode)
+        weight_data.release();
+
     return 0;
 }
 
@@ -281,7 +284,7 @@ int Convolution1D_loongarch::forward(const Mat& bottom_blob, Mat& top_blob, cons
                         sum = bias_data[p];
                     }
 
-                    const float* kptr = (const float*)weight_data + kernel_w * h * p;
+                    const float* kptr = weight_data_packed.channel(p);
 
                     for (int q = 0; q < h; q++)
                     {

diff --git a/src/layer/loongarch/convolution_loongarch.cpp b/src/layer/loongarch/convolution_loongarch.cpp
@@ -225,7 +225,8 @@ int Convolution_loongarch::create_pipeline(const Option& opt)
         }
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -789,7 +790,8 @@ int Convolution_loongarch::create_pipeline_int8_loongarch(const Option& opt)
         scale_in_data[p] = scale_in;
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/loongarch/convolutiondepthwise_loongarch.cpp b/src/layer/loongarch/convolutiondepthwise_loongarch.cpp
@@ -83,15 +83,17 @@ int ConvolutionDepthWise_loongarch::create_pipeline(const Option& opt)
             weight_data_tm = weight_data;
         }
 
-        weight_data.release();
+        if (opt.lightmode)
+            weight_data.release();
 
         return 0;
     }
 
     // group convolution
     create_group_ops(opt);
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -600,15 +602,17 @@ int ConvolutionDepthWise_loongarch::create_pipeline_int8_loongarch(const Option&
             weight_data_tm = weight_data;
         }
 
-        weight_data.release();
+        if (opt.lightmode)
+            weight_data.release();
 
         return 0;
     }
 
     // group convolution
     create_group_ops(opt);
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/loongarch/deconvolution_loongarch.cpp b/src/layer/loongarch/deconvolution_loongarch.cpp
@@ -126,7 +126,8 @@ int Deconvolution_loongarch::create_pipeline(const Option& opt)
     {
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/loongarch/deconvolutiondepthwise_loongarch.cpp b/src/layer/loongarch/deconvolutiondepthwise_loongarch.cpp
@@ -82,15 +82,17 @@ int DeconvolutionDepthWise_loongarch::create_pipeline(const Option& opt)
             weight_data_tm = weight_data_transposed;
         }
 
-        weight_data.release();
+        if (opt.lightmode)
+            weight_data.release();
 
         return 0;
     }
 
     // group convolution
     create_group_ops(opt);
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/loongarch/innerproduct_loongarch.cpp b/src/layer/loongarch/innerproduct_loongarch.cpp
@@ -99,7 +99,8 @@ int InnerProduct_loongarch::create_pipeline(const Option& opt)
         weight_data_tm = weight_data;
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -652,7 +653,8 @@ int InnerProduct_loongarch::create_pipeline_fp16s(const Option& opt)
         ncnn::cast_float32_to_float16(weight_data_r2, weight_data_tm, opt);
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -1140,7 +1142,8 @@ int InnerProduct_loongarch::create_pipeline_int8_loongarch(const Option& opt)
         scale_in_data[p] = scale_in;
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/mips/convolution1d_mips.cpp b/src/layer/mips/convolution1d_mips.cpp
@@ -78,6 +78,9 @@ int Convolution1D_mips::create_pipeline(const Option& opt)
         }
     }
 
+    if (opt.lightmode)
+        weight_data.release();
+
     return 0;
 }
 
@@ -281,7 +284,7 @@ int Convolution1D_mips::forward(const Mat& bottom_blob, Mat& top_blob, const Opt
                         sum = bias_data[p];
                     }
 
-                    const float* kptr = (const float*)weight_data + kernel_w * h * p;
+                    const float* kptr = weight_data_packed.channel(p);
 
                     for (int q = 0; q < h; q++)
                     {

diff --git a/src/layer/mips/convolution_mips.cpp b/src/layer/mips/convolution_mips.cpp
@@ -225,7 +225,8 @@ int Convolution_mips::create_pipeline(const Option& opt)
         }
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -789,7 +790,8 @@ int Convolution_mips::create_pipeline_int8_mips(const Option& opt)
         scale_in_data[p] = scale_in;
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/mips/convolutiondepthwise_mips.cpp b/src/layer/mips/convolutiondepthwise_mips.cpp
@@ -83,15 +83,17 @@ int ConvolutionDepthWise_mips::create_pipeline(const Option& opt)
             weight_data_tm = weight_data;
         }
 
-        weight_data.release();
+        if (opt.lightmode)
+            weight_data.release();
 
         return 0;
     }
 
     // group convolution
     create_group_ops(opt);
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -600,15 +602,17 @@ int ConvolutionDepthWise_mips::create_pipeline_int8_mips(const Option& opt)
             weight_data_tm = weight_data;
         }
 
-        weight_data.release();
+        if (opt.lightmode)
+            weight_data.release();
 
         return 0;
     }
 
     // group convolution
     create_group_ops(opt);
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/mips/deconvolution_mips.cpp b/src/layer/mips/deconvolution_mips.cpp
@@ -126,7 +126,8 @@ int Deconvolution_mips::create_pipeline(const Option& opt)
     {
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/mips/deconvolutiondepthwise_mips.cpp b/src/layer/mips/deconvolutiondepthwise_mips.cpp
@@ -82,15 +82,17 @@ int DeconvolutionDepthWise_mips::create_pipeline(const Option& opt)
             weight_data_tm = weight_data_transposed;
         }
 
-        weight_data.release();
+        if (opt.lightmode)
+            weight_data.release();
 
         return 0;
     }
 
     // group convolution
     create_group_ops(opt);
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/mips/innerproduct_mips.cpp b/src/layer/mips/innerproduct_mips.cpp
@@ -99,7 +99,8 @@ int InnerProduct_mips::create_pipeline(const Option& opt)
         weight_data_tm = weight_data;
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -652,7 +653,8 @@ int InnerProduct_mips::create_pipeline_fp16s(const Option& opt)
         ncnn::cast_float32_to_float16(weight_data_r2, weight_data_tm, opt);
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -1140,7 +1142,8 @@ int InnerProduct_mips::create_pipeline_int8_mips(const Option& opt)
         scale_in_data[p] = scale_in;
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/riscv/convolution1d_riscv.cpp b/src/layer/riscv/convolution1d_riscv.cpp
@@ -95,6 +95,9 @@ int Convolution1D_riscv::create_pipeline(const Option& opt)
         }
     }
 
+    if (opt.lightmode)
+        weight_data.release();
+
     return 0;
 }
 
@@ -308,7 +311,7 @@ int Convolution1D_riscv::forward(const Mat& bottom_blob, Mat& top_blob, const Op
                         sum = bias_data[p];
                     }
 
-                    const float* kptr = (const float*)weight_data + kernel_w * h * p;
+                    const float* kptr = weight_data_packed.channel(p);
 
                     for (int q = 0; q < h; q++)
                     {
@@ -470,7 +473,8 @@ int Convolution1D_riscv::create_pipeline_fp16s(const Option& opt)
 
     ncnn::cast_float32_to_float16(bias_data, bias_data_fp16, opt);
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/riscv/convolution_riscv.cpp b/src/layer/riscv/convolution_riscv.cpp
@@ -237,7 +237,8 @@ int Convolution_riscv::create_pipeline(const Option& opt)
         }
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -834,7 +835,8 @@ int Convolution_riscv::create_pipeline_fp16s(const Option& opt)
         ncnn::cast_float32_to_float16(bias_data, bias_data_fp16, opt);
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/riscv/convolutiondepthwise_riscv.cpp b/src/layer/riscv/convolutiondepthwise_riscv.cpp
@@ -104,15 +104,17 @@ int ConvolutionDepthWise_riscv::create_pipeline(const Option& opt)
             weight_data_tm = weight_data;
         }
 
-        weight_data.release();
+        if (opt.lightmode)
+            weight_data.release();
 
         return 0;
     }
 
     // group convolution
     create_group_ops(opt);
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -682,15 +684,17 @@ int ConvolutionDepthWise_riscv::create_pipeline_fp16s(const Option& opt)
 
         ncnn::cast_float32_to_float16(bias_data, bias_data_fp16, opt);
 
-        weight_data.release();
+        if (opt.lightmode)
+            weight_data.release();
 
         return 0;
     }
 
     // group convolution
     create_group_ops(opt);
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }

diff --git a/src/layer/riscv/deconvolution_riscv.cpp b/src/layer/riscv/deconvolution_riscv.cpp
@@ -148,7 +148,8 @@ int Deconvolution_riscv::create_pipeline(const Option& opt)
     {
     }
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }
@@ -530,7 +531,8 @@ int Deconvolution_riscv::create_pipeline_fp16s(const Option& opt)
 
     ncnn::cast_float32_to_float16(bias_data, bias_data_fp16, opt);
 
-    weight_data.release();
+    if (opt.lightmode)
+        weight_data.release();
 
     return 0;
 }