[GENERAL] Merged v1.0alpha into master. Added features are:

- A100 support via mma.16816
- Thread swizzling for conflict-free shared memory accesses without
padding
- Complete overhaul of the LLVM code generation in
codegen/selection/generator.cc to remove overengineering
- Added debugging capabilities in the Python binding
- Compilation error for kernels that spill

lib/driver/dispatch.cc

@@ -93,6 +93,7 @@ namespace driver
 
 bool dispatch::cuinit(){
   if(cuda_==nullptr){
+    putenv((char*)"CUDA_CACHE_DISABLE=1");
     std::string libcuda = tools::getenv("TRITON_LIBCUDA");
     if(libcuda.empty())
       cuda_ = dlopen("libcuda.so", RTLD_LAZY);