Merge pull request #2 from halide/master

Merge from mainline

Makefile

@@ -252,7 +252,6 @@ SOURCE_FILES = \
   AddImageChecks.cpp \
   AddParameterChecks.cpp \
   AllocationBoundsInference.cpp \
-  BlockFlattening.cpp \
   BoundaryConditions.cpp \
   Bounds.cpp \
   BoundsInference.cpp \
@@ -293,6 +292,7 @@ SOURCE_FILES = \
   FuseGPUThreadLoops.cpp \
   Generator.cpp \
   Image.cpp \
+  ImageParam.cpp \
   InjectHostDevBufferCopies.cpp \
   InjectImageIntrinsics.cpp \
   InjectOpenGLIntrinsics.cpp \
@@ -318,8 +318,8 @@ SOURCE_FILES = \
   ModulusRemainder.cpp \
   Monotonic.cpp \
   ObjectInstanceRegistry.cpp \
+  OutputImageParam.cpp \
   ParallelRVar.cpp \
-  Param.cpp \
   Parameter.cpp \
   PartitionLoops.cpp \
   Pipeline.cpp \
@@ -340,6 +340,7 @@ SOURCE_FILES = \
   SkipStages.cpp \
   SlidingWindow.cpp \
   Solve.cpp \
+  StaticLibrary.cpp \
   StmtToHtml.cpp \
   StorageFlattening.cpp \
   StorageFolding.cpp \
@@ -375,7 +376,6 @@ HEADER_FILES = \
   AddParameterChecks.h \
   AllocationBoundsInference.h \
   Argument.h \
-  BlockFlattening.h \
   BoundaryConditions.h \
   Bounds.h \
   BoundsInference.h \
@@ -419,6 +419,7 @@ HEADER_FILES = \
   Generator.h \
   runtime/HalideRuntime.h \
   Image.h \
+  ImageParam.h \
   InjectHostDevBufferCopies.h \
   InjectImageIntrinsics.h \
   InjectOpenGLIntrinsics.h \
@@ -448,6 +449,7 @@ HEADER_FILES = \
   Monotonic.h \
   ObjectInstanceRegistry.h \
   Outputs.h \
+  OutputImageParam.h \
   ParallelRVar.h \
   Parameter.h \
   Param.h \
@@ -470,6 +472,7 @@ HEADER_FILES = \
   SkipStages.h \
   SlidingWindow.h \
   Solve.h \
+  StaticLibrary.h \
   StmtToHtml.h \
   StorageFlattening.h \
   StorageFolding.h \
@@ -743,7 +746,7 @@ test_generators:  \
 ALL_TESTS = test_internal test_correctness test_errors test_tutorials test_warnings test_generators test_renderscript
 
 # These targets perform timings of each test. For most tests this includes Halide JIT compile times, and run times.
-# For static and generator tests they time the compile time only. The times are recorded in CSV files.
+# For generator tests they time the compile time only. The times are recorded in CSV files.
 time_compilation_correctness: init_time_compilation_correctness $(CORRECTNESS_TESTS:$(ROOT_DIR)/test/correctness/%.cpp=time_compilation_test_%)
 time_compilation_performance: init_time_compilation_performance $(PERFORMANCE_TESTS:$(ROOT_DIR)/test/performance/%.cpp=time_compilation_performance_%)
 time_compilation_opengl: init_time_compilation_opengl $(OPENGL_TESTS:$(ROOT_DIR)/test/opengl/%.cpp=time_compilation_opengl_%)
@@ -767,7 +770,7 @@ build_tests: $(CORRECTNESS_TESTS:$(ROOT_DIR)/test/correctness/%.cpp=$(BIN_DIR)/c
 	$(GENERATOR_EXTERNAL_TESTS:$(ROOT_DIR)/test/generator/%_jittest.cpp=$(BIN_DIR)/generator_jit_%) \
 	$(RENDERSCRIPT_TESTS:$(ROOT_DIR)/test/renderscript/%.cpp=$(BIN_DIR)/renderscript_%)
 
-time_compilation_tests: time_compilation_correctness time_compilation_performance time_compilation_static time_compilation_generators
+time_compilation_tests: time_compilation_correctness time_compilation_performance time_compilation_generators
 
 # Make an empty generator for generating runtimes.
 $(BIN_DIR)/runtime.generator: $(ROOT_DIR)/tools/GenGen.cpp $(BIN_DIR)/libHalide.$(SHARED_EXT)
@@ -776,7 +779,7 @@ $(BIN_DIR)/runtime.generator: $(ROOT_DIR)/tools/GenGen.cpp $(BIN_DIR)/libHalide.
 # Generate a standalone runtime for a given target string
 $(RUNTIMES_DIR)/runtime_%.o: $(BIN_DIR)/runtime.generator
 	@mkdir -p $(RUNTIMES_DIR)
-	$(LD_PATH_SETUP) $(CURDIR)/$< -r $(notdir $@) -o $(CURDIR)/$(RUNTIMES_DIR) target=$*
+	$(LD_PATH_SETUP) $(CURDIR)/$< -r $(basename $(notdir $@)) -o $(CURDIR)/$(RUNTIMES_DIR) target=$*
 
 $(BIN_DIR)/test_internal: $(ROOT_DIR)/test/internal.cpp $(BIN_DIR)/libHalide.$(SHARED_EXT)
 	$(CXX) $(CXX_FLAGS)  $< -I$(SRC_DIR) -L$(BIN_DIR) -lHalide $(TEST_LDFLAGS) -lpthread $(LIBDL) -lz -o $@
@@ -801,11 +804,6 @@ $(BIN_DIR)/opengl_%: $(ROOT_DIR)/test/opengl/%.cpp $(BIN_DIR)/libHalide.$(SHARED
 $(BIN_DIR)/renderscript_%: $(ROOT_DIR)/test/renderscript/%.cpp $(BIN_DIR)/libHalide.$(SHARED_EXT) $(INCLUDE_DIR)/Halide.h
 	$(CXX) $(TEST_CXX_FLAGS) $(OPTIMIZE) $< -I$(INCLUDE_DIR) -I$(SRC_DIR) -L$(BIN_DIR) -lHalide $(TEST_LDFLAGS) -lpthread $(LIBDL) -lz -o $@
 
-$(TMP_DIR)/static/%/%.o: $(BIN_DIR)/static_%_generate
-	@-mkdir -p $(TMP_DIR)/static/$*
-	cd $(TMP_DIR)/static/$*; $(LD_PATH_SETUP) $(CURDIR)/$<
-	@-echo
-
 # TODO(srj): this doesn't auto-delete, why not?
 .INTERMEDIATE: $(FILTERS_DIR)/%.generator
 
@@ -1041,9 +1039,6 @@ time_compilation_opengl_%: $(BIN_DIR)/opengl_%
 time_compilation_renderscript_%: $(BIN_DIR)/renderscript_%
 	$(TIME_COMPILATION) compile_times_renderscript.csv make -f $(THIS_MAKEFILE) $(@:time_compilation_renderscript_%=renderscript_%)
 
-time_compilation_static_%: $(BIN_DIR)/static_%_generate
-	$(TIME_COMPILATION) compile_times_static.csv make -f $(THIS_MAKEFILE) $(@:time_compilation_static_%=$(TMP_DIR)/static/%/%.o)
-
 time_compilation_generator_%: $(FILTERS_DIR)/%.generator
 	$(TIME_COMPILATION) compile_times_generator.csv make -f $(THIS_MAKEFILE) $(@:time_compilation_generator_%=$(FILTERS_DIR)/%.o)
 

apps/glsl/Makefile

@@ -1,20 +1,13 @@
 include ../support/Makefile.inc
 
-TOP := $(abspath $(dir $(lastword $(MAKEFILE_LIST)))/../..)
-.PHONY: all $(TOP)
-all:
-HALIDE_LIB := $(TOP)/$(LIB_HALIDE)
-$(HALIDE_LIB): $(TOP)
-	$(MAKE) -C $(TOP) $(LIB_HALIDE)
-
 CXXFLAGS += -g -O0
 
 UNAME := $(shell uname)
 ifeq ($(UNAME), Linux)
 OPENGL_LIBS = -lX11 -lGL
 endif
 
-HALIDE_LIBS = $(HALIDE_LIB) -ldl $(LLVM_LDFLAGS)
+HALIDE_LIBS = $(LIB_HALIDE) -ldl $(LLVM_LDFLAGS)
 
 all: opengl_test
 
@@ -25,10 +18,10 @@ opengl_test: blur.o ycc.o opengl_test.o
 	$(CXX) $(CXXFLAGS) $^ -o $@ -ldl -lpthread -L$(TOP)/bin $(OPENGL_LIBS)
 
 blur.h blur.o: halide_blur
-	HL_TARGET=host-opengl-gpu_debug ./halide_blur
+	HL_TARGET=host-opengl-debug ./halide_blur
 
 ycc.h ycc.o: halide_ycc
-	HL_TARGET=host-opengl-gpu_debug ./halide_ycc
+	HL_TARGET=host-opengl-debug ./halide_ycc
 
 halide_blur: halide_blur.o $(HALIDE_LIB)
 	$(CXX) $(CXXFLAGS) $< -o $@ $(CXXFLAGS) $(HALIDE_LIBS)

apps/wavelet/Makefile

@@ -25,7 +25,7 @@ $(BUILD_DIR)/%.o $(BUILD_DIR)/%.h: $(BUILD_DIR)/%.generator
 $(BUILD_DIR)/runtime_$(HL_TARGET).o: $(BUILD_DIR)/haar_x.generator
 	@echo Compiling Halide runtime for target $(HL_TARGET)
 	@mkdir -p $(BUILD_DIR)
-	@$< -r runtime_$(HL_TARGET).o -o $(BUILD_DIR) target=$(HL_TARGET)
+	@$< -r runtime_$(HL_TARGET) -o $(BUILD_DIR) target=$(HL_TARGET)
 
 HL_MODULES = \
 	$(BUILD_DIR)/daubechies_x.o \

python_bindings/CMakeLists.txt

@@ -78,7 +78,8 @@ endif()
 
 file(GLOB SrcCpp
 python/*.cpp
-../src/Param.cpp)
+../src/OutputImageParam.cpp
+../src/ImageParam.cpp)
 
 add_library(halide SHARED ${SrcCpp})
 target_link_libraries(halide

python_bindings/python/BoundaryConditions.cpp

@@ -4,6 +4,7 @@
 #include <boost/python.hpp>
 
 #include "../../src/Lambda.h" // needed by BoundaryConditions.h
+#include "../../src/ImageParam.h"
 #include "../../src/BoundaryConditions.h"
 #include "../../src/Func.h"
 
@@ -142,7 +143,7 @@ void def_mirror_interior_for_image<void>()
 
 void defineBoundaryConditions()
 {
-    
+
 
     // Other variants of constant_exterior exist, but not yet implemented
 
@@ -244,4 +245,3 @@ void defineBoundaryConditions()
 
     return;
 }
-

python_bindings/python/Param.cpp

@@ -7,6 +7,8 @@
 #include "add_operators.h"
 
 #include "../../src/Param.h"
+#include "../../src/OutputImageParam.h"
+#include "../../src/ImageParam.h"
 #include "../../src/IROperator.h" // enables Param + Expr operations (which include is it ?)
 #include "Type.h"
 
@@ -124,7 +126,6 @@ void defineImageParam()
             ;
 
     p::implicitly_convertible<ImageParam, h::Argument>();
-    p::implicitly_convertible<ImageParam, h::Expr>();
 
     // "Using a param as the argument to an external stage treats it as an Expr"
     //p::implicitly_convertible<ImageParam, h::ExternFuncArgument>();

python_bindings/python/RDom.cpp

@@ -5,6 +5,7 @@
 #include "add_operators.h"
 
 #include "../../src/RDom.h"
+#include "../../src/ImageParam.h"
 #include "../../src/IROperator.h" // for operations with RVar
 
 #include <string>

src/CMakeLists.txt

@@ -203,7 +203,6 @@ set(HEADER_FILES
   AddParameterChecks.h
   AllocationBoundsInference.h
   Argument.h
-  BlockFlattening.h
   BoundaryConditions.h
   Bounds.h
   BoundsInference.h
@@ -252,6 +251,7 @@ set(HEADER_FILES
   IRPrinter.h
   IRVisitor.h
   Image.h
+  ImageParam.h
   InjectHostDevBufferCopies.h
   InjectImageIntrinsics.h
   InjectOpenGLIntrinsics.h
@@ -273,6 +273,7 @@ set(HEADER_FILES
   ModulusRemainder.h
   Monotonic.h
   ObjectInstanceRegistry.h
+  OutputImageParam.h
   Outputs.h
   ParallelRVar.h
   Param.h
@@ -296,6 +297,7 @@ set(HEADER_FILES
   SkipStages.h
   SlidingWindow.h
   Solve.h
+  StaticLibrary.h
   StmtToHtml.h
   StorageFlattening.h
   StorageFolding.h
@@ -350,7 +352,6 @@ add_library(Halide ${HALIDE_LIBRARY_TYPE}
   AddImageChecks.cpp
   AddParameterChecks.cpp
   AllocationBoundsInference.cpp
-  BlockFlattening.cpp
   BoundaryConditions.cpp
   Bounds.cpp
   BoundsInference.cpp
@@ -399,6 +400,7 @@ add_library(Halide ${HALIDE_LIBRARY_TYPE}
   IRPrinter.cpp
   IRVisitor.cpp
   Image.cpp
+  ImageParam.cpp
   InjectHostDevBufferCopies.cpp
   InjectImageIntrinsics.cpp
   InjectOpenGLIntrinsics.cpp
@@ -417,8 +419,8 @@ add_library(Halide ${HALIDE_LIBRARY_TYPE}
   ModulusRemainder.cpp
   Monotonic.cpp
   ObjectInstanceRegistry.cpp
+  OutputImageParam.cpp
   ParallelRVar.cpp
-  Param.cpp
   Parameter.cpp
   PartitionLoops.cpp
   Pipeline.cpp
@@ -439,6 +441,7 @@ add_library(Halide ${HALIDE_LIBRARY_TYPE}
   SkipStages.cpp
   SlidingWindow.cpp
   Solve.cpp
+  StaticLibrary.cpp
   StmtToHtml.cpp
   StorageFlattening.cpp
   StorageFolding.cpp

src/CodeGen_Internal.cpp

@@ -354,7 +354,7 @@ void clone_target_options(const llvm::Module &from, llvm::Module &to) {
     }
 }
 
-llvm::TargetMachine *get_target_machine(const llvm::Module &module) {
+std::unique_ptr<llvm::TargetMachine> make_target_machine(const llvm::Module &module) {
     std::string error_string;
 
     const llvm::Target *target = llvm::TargetRegistry::lookupTarget(module.getTargetTriple(), error_string);
@@ -369,12 +369,12 @@ llvm::TargetMachine *get_target_machine(const llvm::Module &module) {
     std::string mattrs = "";
     get_target_options(module, options, mcpu, mattrs);
 
-    return target->createTargetMachine(module.getTargetTriple(),
-                                       mcpu, mattrs,
-                                       options,
-                                       llvm::Reloc::PIC_,
-                                       llvm::CodeModel::Default,
-                                       llvm::CodeGenOpt::Aggressive);
+    return std::unique_ptr<llvm::TargetMachine>(target->createTargetMachine(module.getTargetTriple(),
+                                                mcpu, mattrs,
+                                                options,
+                                                llvm::Reloc::PIC_,
+                                                llvm::CodeModel::Default,
+                                                llvm::CodeGenOpt::Aggressive));
 }
 
 }

src/CodeGen_Internal.h

@@ -8,6 +8,8 @@
  * front-end-facing interface to CodeGen).
  */
 
+#include <memory>
+
 #include "IR.h"
 #include "IRVisitor.h"
 #include "LLVM_Headers.h"
@@ -62,7 +64,7 @@ void get_target_options(const llvm::Module &module, llvm::TargetOptions &options
 void clone_target_options(const llvm::Module &from, llvm::Module &to);
 
 /** Given an llvm::Module, get or create an llvm:TargetMachine */
-llvm::TargetMachine *get_target_machine(const llvm::Module &module);
+std::unique_ptr<llvm::TargetMachine> make_target_machine(const llvm::Module &module);
 
 }}