Upgrade to LLVM 15 (exaloop#56)

* Upgrade to LLVM 15 (WIP)

* Call `setPresplitCoroutine()` on coro LLVM funcs

* Use new pass manager

* Update deps

* Update docs

* Fix exceptions on M1

* Add libunwind

* Use Orc and JITLink for "codon run"

* JITLink integration

* Fix callback

* Fix strbuf, fix GC root registration

* Fix test init

* Fix llvm function

* Fix pickle, float atomics

* Add TargetLibraryAnalysis

* Use new LLVM pass manager in GPU codegen

* Fix libunwind linking

* Fix libunwind

* Fix GPU passes

* Don't link libunwind explicitly

* Bump version

* Update plugins for new pass manager

* Fix bash error

* Fix GPU GV extraction

* Move simd module to experimental folder

* Update file read

* Add benchmark programs

* Add dynamic tuple

* Fix parser tuple slicing bug

* Make DynamicTuple interoperable with Tuple

* Fix DynamicTuple GPU interop

* Dockerize builds

* Simplify CMake

* Revert "Simplify CMake"

This reverts commit 08d2920.

Co-authored-by: Ibrahim Numanagić <[email protected]>

.github/workflows/ci.yml

@@ -196,6 +196,7 @@ jobs:
           cp build/codon codon-deploy/bin/
           cp build/libcodon*.${LIBEXT} codon-deploy/lib/codon/
           cp build/libomp.${LIBEXT} codon-deploy/lib/codon/
+          if [ -e build/libunwind.${LIBEXT} ]; then cp build/libunwind.${LIBEXT} codon-deploy/lib/codon/; fi
           cp -r build/include codon-deploy/
           cp -r stdlib codon-deploy/lib/codon/
           cp -r extra/python/dist/*.whl codon-deploy/

CMakeLists.txt

@@ -1,7 +1,7 @@
 cmake_minimum_required(VERSION 3.14)
 project(
   Codon
-  VERSION "0.14.0"
+  VERSION "0.15.0"
   HOMEPAGE_URL "https://github.com/exaloop/codon"
   DESCRIPTION "high-performance, extensible Python compiler")
 configure_file("${PROJECT_SOURCE_DIR}/cmake/config.h.in"
@@ -28,6 +28,11 @@ endif()
 set(CMAKE_CXX_FLAGS_RELEASE "-O3")
 include_directories(.)
 
+set(APPLE_ARM OFF)
+if (APPLE AND CMAKE_HOST_SYSTEM_PROCESSOR STREQUAL "arm64")
+  set(APPLE_ARM ON)
+endif()
+
 set(THREADS_PREFER_PTHREAD_FLAG ON)
 find_package(Threads REQUIRED)
 find_package(LLVM REQUIRED)
@@ -64,6 +69,9 @@ set(CODONRT_FILES codon/runtime/lib.h codon/runtime/lib.cpp
                   codon/runtime/gpu.cpp)
 add_library(codonrt SHARED ${CODONRT_FILES})
 add_dependencies(codonrt zlibstatic gc backtrace bz2 liblzma re2)
+if(APPLE AND APPLE_ARM)
+  add_dependencies(codonrt unwind_shared)
+endif()
 target_include_directories(codonrt PRIVATE ${backtrace_SOURCE_DIR}
                                            ${re2_SOURCE_DIR}
                                            "${gc_SOURCE_DIR}/include" runtime)
@@ -102,6 +110,13 @@ add_custom_command(
   POST_BUILD
   COMMAND ${CMAKE_COMMAND} -E copy_if_different $<TARGET_FILE:omp>
           ${CMAKE_BINARY_DIR})
+if(APPLE AND APPLE_ARM)
+  add_custom_command(
+    TARGET codonrt
+    POST_BUILD
+    COMMAND ${CMAKE_COMMAND} -E copy_if_different $<TARGET_FILE:unwind_shared>
+            ${CMAKE_BINARY_DIR})
+endif()
 
 # Codon compiler library
 include_directories(${LLVM_INCLUDE_DIRS})
@@ -148,13 +163,6 @@ set(CODON_HPPFILES
     codon/sir/flow.h
     codon/sir/func.h
     codon/sir/instr.h
-    codon/sir/llvm/coro/CoroCleanup.h
-    codon/sir/llvm/coro/CoroEarly.h
-    codon/sir/llvm/coro/CoroElide.h
-    codon/sir/llvm/coro/CoroInstr.h
-    codon/sir/llvm/coro/CoroInternal.h
-    codon/sir/llvm/coro/CoroSplit.h
-    codon/sir/llvm/coro/Coroutines.h
     codon/sir/llvm/gpu.h
     codon/sir/llvm/llvisitor.h
     codon/sir/llvm/llvm.h
@@ -298,12 +306,6 @@ set(CODON_CPPFILES
     codon/sir/flow.cpp
     codon/sir/func.cpp
     codon/sir/instr.cpp
-    codon/sir/llvm/coro/CoroCleanup.cpp
-    codon/sir/llvm/coro/CoroEarly.cpp
-    codon/sir/llvm/coro/CoroElide.cpp
-    codon/sir/llvm/coro/CoroFrame.cpp
-    codon/sir/llvm/coro/CoroSplit.cpp
-    codon/sir/llvm/coro/Coroutines.cpp
     codon/sir/llvm/gpu.cpp
     codon/sir/llvm/llvisitor.cpp
     codon/sir/llvm/optimize.cpp
@@ -426,6 +428,20 @@ add_custom_target(
     "${CMAKE_BINARY_DIR}/lib/codon")
 add_dependencies(libs codonrt codonc)
 
+if(APPLE AND APPLE_ARM)
+  add_custom_target(
+    libs_apple ALL
+    COMMENT "Collecting Apple-specific libraries"
+    BYPRODUCTS "${CMAKE_BINARY_DIR}/lib"
+    VERBATIM
+    COMMAND ${CMAKE_COMMAND} -E make_directory "${CMAKE_BINARY_DIR}/lib/codon"
+    COMMAND
+      ${CMAKE_COMMAND} -E copy
+      "${CMAKE_BINARY_DIR}/libunwind${CMAKE_SHARED_LIBRARY_SUFFIX}"
+      "${CMAKE_BINARY_DIR}/lib/codon")
+  add_dependencies(libs_apple codonrt)
+endif()
+
 # Codon command-line tool
 add_executable(codon codon/app/main.cpp)
 target_link_libraries(codon PUBLIC ${STATIC_LIBCPP} codonc Threads::Threads)

bench/float/float.py

@@ -0,0 +1,55 @@
+from math import sin, cos, sqrt
+from time import time
+
+POINTS = 10000000
+
+
+class Point:
+    x: float
+    y: float
+    z: float
+
+    def __init__(self, i):
+        self.x = x = sin(i)
+        self.y = cos(i) * 3
+        self.z = (x * x) / 2
+
+    def __repr__(self):
+        return f"<Point: x={self.x}, y={self.y}, z={self.z}>"
+
+    def normalize(self):
+        x = self.x
+        y = self.y
+        z = self.z
+        norm = sqrt(x * x + y * y + z * z)
+        self.x /= norm
+        self.y /= norm
+        self.z /= norm
+
+    def maximize(self, other):
+        self.x = self.x if self.x > other.x else other.x
+        self.y = self.y if self.y > other.y else other.y
+        self.z = self.z if self.z > other.z else other.z
+        return self
+
+
+def maximize(points):
+    next = points[0]
+    for p in points[1:]:
+        next = next.maximize(p)
+    return next
+
+
+def benchmark(n):
+    points = [None] * n
+    for i in range(n):
+        points[i] = Point(i)
+    for p in points:
+        p.normalize()
+    return maximize(points)
+
+
+t0 = time()
+print(benchmark(POINTS))
+t1 = time()
+print(t1 - t0)