[GitHub] anirudh2290 commented on a change in pull request #10014: Fix crash with mx.nd.ones

2018-03-09 Thread GitBox 
anirudh2290 commented on a change in pull request #10014: Fix crash with 
mx.nd.ones
URL: https://github.com/apache/incubator-mxnet/pull/10014#discussion_r173614117
 
 

 ##
 File path: src/imperative/imperative_utils.h
 ##
 @@ -70,12 +70,14 @@ inline Context GetContext(const nnvm::NodeAttrs& attrs,
   if (ctx.dev_mask() != ctx.dev_type) {
 ctx = Context::Create(ctx.dev_mask(), ctx.dev_id);
   }
-#if !MXNET_USE_CUDA
   if (ctx.dev_mask() == gpu::kDevMask) {
+#if !MXNET_USE_CUDA
 LOG(INFO) << "GPU support is disabled. Compile MXNet with "
   << "USE_CUDA=1 to enable GPU support.";
-  }
+#else
+mshadow::SetDevice(ctx.dev_id);
 
 Review comment:
   @piiswrong what do you suggest ?


This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

[GitHub] anirudh2290 commented on a change in pull request #10014: Fix crash with mx.nd.ones

2018-03-09 Thread GitBox 
anirudh2290 commented on a change in pull request #10014: Fix crash with 
mx.nd.ones
URL: https://github.com/apache/incubator-mxnet/pull/10014#discussion_r173405771
 
 

 ##
 File path: src/imperative/imperative_utils.h
 ##
 @@ -70,12 +70,14 @@ inline Context GetContext(const nnvm::NodeAttrs& attrs,
   if (ctx.dev_mask() != ctx.dev_type) {
 ctx = Context::Create(ctx.dev_mask(), ctx.dev_id);
   }
-#if !MXNET_USE_CUDA
   if (ctx.dev_mask() == gpu::kDevMask) {
+#if !MXNET_USE_CUDA
 LOG(INFO) << "GPU support is disabled. Compile MXNet with "
   << "USE_CUDA=1 to enable GPU support.";
-  }
+#else
+mshadow::SetDevice(ctx.dev_id);
 
 Review comment:
   @piiswrong Where would the async error be raised first ? Currently only 
exception thrown inside ExecuteOprBlock(during operator exec) is handled. One 
of the arguments to ExecuteOprBlock is the runtime context and so the stream 
needs to be set based on device availability before calling ExecuteOprBlock. 
Therefore, I cannot push this to engine/handle it as async error in engine.


This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

[GitHub] anirudh2290 commented on a change in pull request #10014: Fix crash with mx.nd.ones

2018-03-09 Thread GitBox 
anirudh2290 commented on a change in pull request #10014: Fix crash with 
mx.nd.ones
URL: https://github.com/apache/incubator-mxnet/pull/10014#discussion_r173405771
 
 

 ##
 File path: src/imperative/imperative_utils.h
 ##
 @@ -70,12 +70,14 @@ inline Context GetContext(const nnvm::NodeAttrs& attrs,
   if (ctx.dev_mask() != ctx.dev_type) {
 ctx = Context::Create(ctx.dev_mask(), ctx.dev_id);
   }
-#if !MXNET_USE_CUDA
   if (ctx.dev_mask() == gpu::kDevMask) {
+#if !MXNET_USE_CUDA
 LOG(INFO) << "GPU support is disabled. Compile MXNet with "
   << "USE_CUDA=1 to enable GPU support.";
-  }
+#else
+mshadow::SetDevice(ctx.dev_id);
 
 Review comment:
   @piiswrong Where would the async error be raised first ? Currently only 
exception thrown inside ExecuteOprBlock(during operator exec) is handled. One 
of the arguments to ExecuteOprBlock is the runtime context and so the stream 
needs to be set based on device availability before calling ExecuteOprBlock. 
Therefore, I cannot add this when pushing to engine/handle it as async error in 
engine.


This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

[GitHub] anirudh2290 commented on a change in pull request #10014: Fix crash with mx.nd.ones

2018-03-06 Thread GitBox 
anirudh2290 commented on a change in pull request #10014: Fix crash with 
mx.nd.ones
URL: https://github.com/apache/incubator-mxnet/pull/10014#discussion_r172751325
 
 

 ##
 File path: src/imperative/imperative_utils.h
 ##
 @@ -70,12 +70,14 @@ inline Context GetContext(const nnvm::NodeAttrs& attrs,
   if (ctx.dev_mask() != ctx.dev_type) {
 ctx = Context::Create(ctx.dev_mask(), ctx.dev_id);
   }
-#if !MXNET_USE_CUDA
   if (ctx.dev_mask() == gpu::kDevMask) {
+#if !MXNET_USE_CUDA
 LOG(INFO) << "GPU support is disabled. Compile MXNet with "
   << "USE_CUDA=1 to enable GPU support.";
-  }
+#else
+mshadow::SetDevice(ctx.dev_id);
 
 Review comment:
   The call is intended to check if the device_id is available. The overhead 
should be pretty low as it is only setting the dev id corresponding to the 
thread and there is no synchronization taking place. The other option is to get 
the cuda device count and check if it the device falls in the range but that 
would have a similar overhead. 


This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services