[PATCH] [cuda] Include GPU binary into host object file and generate init/deinit code.

Eric Christopher echristo at gmail.com
Wed May 6 11:22:34 PDT 2015


Can you make all of the newly added function names lower case please? (Fixing the existing names in the CUDA stuff is preapproved).

Some more inline comments/nits.

-eric


================
Comment at: lib/CodeGen/CGCUDANV.cpp:23
@@ -22,2 +22,3 @@
 #include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/Verifier.h"
 #include <vector>
----------------
Unused include?

================
Comment at: lib/CodeGen/CGCUDANV.cpp:240
@@ +239,3 @@
+
+  for (const std::string &GpuBinaryFileName :
+       CGM.getCodeGenOpts().CudaGpuBinaryFileNames) {
----------------
Comment this loop please?

================
Comment at: test/CodeGenCUDA/device-stub.cu:5
@@ -4,1 +4,3 @@
 
+// Test for presence of elements we construct for constructor and destructor:
+// * constant unnamed string with the kernel name
----------------
Reword this in some way? It's a bit oddly phrased :)

================
Comment at: test/CodeGenCUDA/device-stub.cu:29
@@ +28,3 @@
+
+// Test that be build correct kernel launch sequence.
+// CHECK: define{{.*}}hostfunc
----------------
"Test that we"

http://reviews.llvm.org/D9507

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/






More information about the cfe-commits mailing list