[CORE] Fixed several issues that arose in the development of the

torch-blocksparse package:

* Now using warp shuffle in reductions when possible
* Various bugfixes in layout inference
* Added INFINITY, exponential and select
* Better error messages for unimplemented constructs

CMakeLists.txt

@@ -9,6 +9,7 @@ option(BUILD_PYTHON_MODULE "Build Python Triton bindings" OFF)
 
 # LLVM
 find_package(LLVM REQUIRED)
+link_directories(${LLVM_LIBRARY_DIRS})
 include_directories(${LLVM_INCLUDE_DIRS})
 add_definitions(${LLVM_DEFINITIONS})
 
@@ -40,6 +41,5 @@ endif()
 # Triton
 file(GLOB_RECURSE LIBTRITON_SRC lib/*.cc)
 add_library(triton SHARED ${LIBTRITON_SRC} ${PYTHON_SRC})
-link_directories(${LLVM_LIBRARY_DIRS})
 target_link_libraries(triton ${LLVM_LIBRARIES})