commit 746a163c9d6d150eb35e1b7a5ec6b3512fe993c9
author soumith <soumith@fb.com> Thu Nov 06 09:06:53 2014 -0800
committer soumith <soumith@fb.com> Thu Nov 06 09:29:26 2014 -0800
tree 6067ed8c9752c338834c27e3d38ee5c1bc73d040
parent ff84e38d233578971a5806f575a1018de9ad8f50

fixing inconsistent API in addcmul, addcdiv

THCTensorMath.cu

diff --git a/THCTensorMath.cu b/THCTensorMath.cu
index 9fabe7e..915707d 100644
--- a/THCTensorMath.cu
+++ b/THCTensorMath.cu
@@ -189,8 +189,13 @@
 }
 
 
-void THCudaTensor_addcmul(THCudaTensor *self_, float value, THCudaTensor *src1, THCudaTensor *src2)
+void THCudaTensor_addcmul(THCudaTensor *self_, THCudaTensor *t, float value, THCudaTensor *src1, THCudaTensor *src2)
 {
+  if(self_ != t)
+  {
+    THCudaTensor_resizeAs(self_, t);
+    THCudaTensor_copy(self_, t);
+  }
   THCudaTensor_resizeAs(self_, src1);
   THArgCheck(THCudaTensor_nElement(src1) == THCudaTensor_nElement(src2), 3, "size do not match");
   {
@@ -225,8 +230,14 @@
 }
 
 
-void THCudaTensor_addcdiv(THCudaTensor *self_, float value, THCudaTensor *src1, THCudaTensor *src2)
+void THCudaTensor_addcdiv(THCudaTensor *self_, THCudaTensor *t, float value, THCudaTensor *src1, THCudaTensor *src2)
 {
+  if(self_ != t)
+  {
+    THCudaTensor_resizeAs(self_, t);
+    THCudaTensor_copy(self_, t);
+  }
+
   THCudaTensor_resizeAs(self_, src1);
   THArgCheck(THCudaTensor_nElement(src1) == THCudaTensor_nElement(src2), 3, "size do not match");
   {

THCTensorMath.h

diff --git a/THCTensorMath.h b/THCTensorMath.h
index 915e269..7bd76d7 100644
--- a/THCTensorMath.h
+++ b/THCTensorMath.h
@@ -16,8 +16,8 @@
 THC_API void THCudaTensor_cmul(THCudaTensor *self, THCudaTensor *src1, THCudaTensor *src2);
 THC_API void THCudaTensor_cdiv(THCudaTensor *self, THCudaTensor *src1, THCudaTensor *src2);
 
-THC_API void THCudaTensor_addcmul(THCudaTensor *self, float value, THCudaTensor *src1, THCudaTensor *src2);
-THC_API void THCudaTensor_addcdiv(THCudaTensor *self, float value, THCudaTensor *src1, THCudaTensor *src2);
+THC_API void THCudaTensor_addcmul(THCudaTensor *self, THCudaTensor* t, float value, THCudaTensor *src1, THCudaTensor *src2);
+THC_API void THCudaTensor_addcdiv(THCudaTensor *self, THCudaTensor* t, float value, THCudaTensor *src1, THCudaTensor *src2);
 
 THC_API float THCudaTensor_dot(THCudaTensor *self, THCudaTensor *src);