[PATCH] Ignore device-side asm constraint errors while compiling CUDA code for host and vice versa.

[Eli Bendersky]

================
Comment at: test/SemaCUDA/asm-constraints.cu:7
@@ +6,3 @@
+  short h;
+  // asm with PTX constraints. Some of them are PTX-specific.
+  __asm__("dont care" : "=h"(h): "f"(0.0), "d"(0.0), "h"(0), "r"(0), "l"(0));
----------------
Are there checks anywhere that asm diagnostics are validated within device? 

Perhaps a positive test would be helpful (make sure that an invalid constraint produces a diagnostic within a kernel)

http://reviews.llvm.org/D8392

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