Merge remote-tracking branch 'upstream/master' into nvptx

Makefile

@@ -109,14 +109,13 @@ else
     BUILD_VERSION:=$(GIT_REVISION)
 endif
 
-CXX=g++
+CXX=clang++
-CPP=cpp
 OPT=-O2
 CXXFLAGS=$(OPT) $(LLVM_CXXFLAGS) -I. -Iobjs/ -I$(CLANG_INCLUDE)  \
 	$(LLVM_VERSION_DEF) \
 	-Wall \
 	-DBUILD_DATE="\"$(BUILD_DATE)\"" -DBUILD_VERSION="\"$(BUILD_VERSION)\"" \
-	-Wno-sign-compare
+	-Wno-sign-compare -Wno-unused-function
 ifneq ($(LLVM_VERSION),LLVM_3_1)
 	CXXFLAGS+=-Werror
 endif
@@ -202,9 +201,14 @@ ispc: print_llvm_src dirs $(OBJS)
 	$(CXX) $(OPT) $(LDFLAGS) -o $@ $(OBJS) $(ISPC_LIBS)
 
 # Use clang as a default compiler, instead of gcc
+# This is default now.
 clang: ispc
 clang: CXX=clang++
 
+# Use gcc as a default compiler, instead of gcc
+gcc: ispc
+gcc: CXX=clang++
+
 # Build ispc with address sanitizer instrumentation using clang compiler
 # Note that this is not portable build
 asan: clang

alloy.py

@@ -70,7 +70,7 @@ def try_do_LLVM(text, command, from_validation):
         error("can't " + text, 1)
     print_debug("DONE.\n", from_validation, alloy_build)
 
-def build_LLVM(version_LLVM, revision, folder, tarball, debug, selfbuild, from_validation, force, make):
+def build_LLVM(version_LLVM, revision, folder, tarball, debug, selfbuild, extra, from_validation, force, make):
     print_debug("Building LLVM. Version: " + version_LLVM + ". ", from_validation, alloy_build)
     if revision != "":
         print_debug("Revision: " + revision + ".\n", from_validation, alloy_build)
@@ -121,6 +121,15 @@ def build_LLVM(version_LLVM, revision, folder, tarball, debug, selfbuild, from_v
         try_do_LLVM("load clang from http://llvm.org/svn/llvm-project/cfe/" + SVN_PATH + " ",
                     "svn co " + revision + " http://llvm.org/svn/llvm-project/cfe/" + SVN_PATH + " clang",
                     from_validation)
+        if extra == True:
+            os.chdir("./clang/tools")
+            try_do_LLVM("load extra clang extra tools ",
+                    "svn co " + revision + " http://llvm.org/svn/llvm-project/clang-tools-extra/" + SVN_PATH + " extra",
+                    from_validation)
+            os.chdir("../../../projects")
+            try_do_LLVM("load extra clang compiler-rt ",
+                    "svn co " + revision + " http://llvm.org/svn/llvm-project/compiler-rt/" + SVN_PATH + " compiler-rt",
+                    from_validation)
         os.chdir("../")
     else:
         tar = tarball.split(" ")
@@ -286,6 +295,8 @@ def run_special_tests():
    i = 5 
 
 def validation_run(only, only_targets, reference_branch, number, notify, update, make):
+    if os.environ["ISPC_HOME"] != os.getcwd():
+        error("you ISPC_HOME and your current pass are different!\n", 2)
     os.chdir(os.environ["ISPC_HOME"])
     os.environ["PATH"] = os.environ["ISPC_HOME"] + ":" + os.environ["PATH"]
     if options.notify != "":
@@ -387,7 +398,7 @@ def validation_run(only, only_targets, reference_branch, number, notify, update,
         gen_archs = ["x86-64"]
         need_LLVM = check_LLVM(LLVM)
         for i in range(0,len(need_LLVM)):
-            build_LLVM(need_LLVM[i], "", "", "", False, False, True, False, make)
+            build_LLVM(need_LLVM[i], "", "", "", False, False, False, True, False, make)
 # begin validation run for stabitily
         common.remove_if_exists(stability.in_file)
         R = [[[],[]],[[],[]],[[],[]],[[],[]]]
@@ -465,7 +476,7 @@ def validation_run(only, only_targets, reference_branch, number, notify, update,
 # prepare LLVM 3.3 as newest LLVM
         need_LLVM = check_LLVM(["3.3"])
         if len(need_LLVM) != 0:
-            build_LLVM(need_LLVM[i], "", "", "", False, False, True, False, make)
+            build_LLVM(need_LLVM[i], "", "", "", False, False, False, True, False, make)
 # prepare reference point. build both test and reference compilers
         try_do_LLVM("apply git", "git branch", True)
         temp4 = take_lines("git branch", "all")
@@ -552,7 +563,7 @@ def Main():
     try:
         if options.build_llvm:
             build_LLVM(options.version, options.revision, options.folder, options.tarball,
-                    options.debug, options.selfbuild, False, options.force, make)
+                    options.debug, options.selfbuild, options.extra, False, options.force, make)
         if options.validation_run:
             validation_run(options.only, options.only_targets, options.branch,
                     options.number_for_performance, options.notify, options.update, make)
@@ -628,6 +639,8 @@ llvm_group.add_option('--selfbuild', dest='selfbuild',
     help='make selfbuild of LLVM and clang', default=False, action="store_true")
 llvm_group.add_option('--force', dest='force',
     help='rebuild LLVM', default=False, action='store_true')
+llvm_group.add_option('--extra', dest='extra',
+    help='load extra clang tools', default=False, action='store_true')
 parser.add_option_group(llvm_group)
 # options for activity "validation run"
 run_group = OptionGroup(parser, "Options for validation run",

builtins/util.m4

@@ -3834,9 +3834,9 @@ domixed:
   %first = call i64 @llvm.cttz.i64(i64 %mm)
   %first32 = trunc i64 %first to i32
   %baseval = extractelement <$1 x $2> %v, i32 %first32
-  %basev1 = bitcast $2 %baseval to <1 x $2>
+  %basev1 = insertelement <$1 x $2> undef, $2 %baseval, i32 0
   ; get a vector that is that value smeared across all elements
-  %basesmear = shufflevector <1 x $2> %basev1, <1 x $2> undef,
+  %basesmear = shufflevector <$1 x $2> %basev1, <$1 x $2> undef,
         <$1 x i32> < forloop(i, 0, eval($1-2), `i32 0, ') i32 0 >
 
   ; now to a blend of that vector with the original vector, such that the

cbackend.cpp

@@ -3066,7 +3066,7 @@ void CWriter::visitReturnInst(llvm::ReturnInst &I) {
   // Don't output a void return if this is the last basic block in the function
   if (I.getNumOperands() == 0 &&
       &*--I.getParent()->getParent()->end() == I.getParent() &&
-      !I.getParent()->size() == 1) {
+      (!I.getParent()->size()) == 1) {
     return;
   }
 

examples/aobench_instrumented/Makefile

@@ -1,5 +1,5 @@
 
-CXX=g++ -m64
+CXX=clang++ -m64
 CXXFLAGS=-Iobjs/ -g3 -Wall
 ISPC=ispc
 ISPCFLAGS=-O2 --instrument --arch=x86-64 --target=sse2

examples/common.mk

@@ -3,14 +3,14 @@ TASK_CXX=../tasksys.cpp
 TASK_LIB=-lpthread
 TASK_OBJ=objs/tasksys.o
 
-CXX=g++
+CXX=clang++
-CXXFLAGS=-Iobjs/ -O2
+CXXFLAGS+=-Iobjs/ -O2
-CC=gcc
+CC=clang
-CCFLAGS=-Iobjs/ -O2
+CCFLAGS+=-Iobjs/ -O2
 
 LIBS=-lm $(TASK_LIB) -lstdc++
 ISPC=ispc
-ISPC_FLAGS=-O2
+ISPC_FLAGS+=-O2
 ISPC_HEADER=objs/$(ISPC_SRC:.ispc=_ispc.h)
 
 ARCH:=$(shell uname -m | sed -e s/x86_64/x86/ -e s/i686/x86/ -e s/arm.*/arm/ -e s/sa110/arm/)

examples/simple/Makefile

@@ -1,5 +1,5 @@
 
-CXX=g++ -m64
+CXX=clang++ -m64
 CXXFLAGS=-Iobjs/ -O3 -Wall
 ISPC=ispc
 ISPCFLAGS=-O2 --arch=x86-64 --target=sse2

fail_db.txt

@@ -8,42 +8,17 @@
 % 
 ./tests/masked-scatter-vector.ispc runfail  x86-64     sse2-i32x4   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/atomics-13.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-10.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-11.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-13.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-5.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-6.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/atomics-13.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-10.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-11.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-13.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-5.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-6.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/funcptr-null-4.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/funcptr-null-5.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/funcptr-null-6.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/atomics-13.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-10.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-11.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-13.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-5.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-6.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/funcptr-null-4.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/funcptr-null-5.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/funcptr-null-6.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/atomics-13.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-10.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-11.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-13.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-5.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/reduce-equal-6.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/avg-down-int8.ispc compfail     x86    avx1-i32x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/avg-up-int8.ispc compfail     x86    avx1-i32x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/avg-down-int8.ispc compfail  x86-64    avx1-i32x16   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/avg-up-int8.ispc compfail  x86-64    avx1-i32x16   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/atomics-4.ispc runfail  x86-64      generic-4   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/atomics-6.ispc runfail  x86-64      generic-4   Linux LLVM 3.3     g++4.4 -O2 *
-./tests/atomics-swap.ispc runfail  x86-64      generic-4   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/atomics-varyingptr-2.ispc runfail  x86-64      generic-4   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/atomics-varyingptr-4.ispc runfail  x86-64      generic-4   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/avg-down-uint16.ispc runfail  x86-64      generic-4   Linux LLVM 3.3     g++4.4 -O2 *
@@ -158,10 +133,6 @@
 ./tests/funcptr-null-4.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.4     g++4.4 -O2 *
 ./tests/funcptr-null-5.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.4     g++4.4 -O2 *
 ./tests/funcptr-null-6.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.4     g++4.4 -O2 *
-./tests/avg-down-int8.ispc compfail     x86    avx1-i32x16   Linux LLVM 3.4     g++4.4 -O2 *
-./tests/avg-up-int8.ispc compfail     x86    avx1-i32x16   Linux LLVM 3.4     g++4.4 -O2 *
-./tests/avg-down-int8.ispc compfail  x86-64    avx1-i32x16   Linux LLVM 3.4     g++4.4 -O2 *
-./tests/avg-up-int8.ispc compfail  x86-64    avx1-i32x16   Linux LLVM 3.4     g++4.4 -O2 *
 ./tests/array-gather-ifs.ispc runfail  x86-64      generic-4   Linux LLVM 3.4     g++4.4 -O2 *
 ./tests/array-gather-multi-unif.ispc runfail  x86-64      generic-4   Linux LLVM 3.4     g++4.4 -O2 *
 ./tests/array-gather-unif.ispc runfail  x86-64      generic-4   Linux LLVM 3.4     g++4.4 -O2 *
@@ -330,42 +301,17 @@
 ./tests/reduce-equal.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4     g++4.4 -O2 *
 ./tests/test-141.ispc runfail     x86    avx2-i32x16   Linux LLVM 3.4     g++4.4 -O2 *
 ./tests/test-141.ispc runfail  x86-64    avx2-i32x16   Linux LLVM 3.4     g++4.4 -O2 *
-./tests/atomics-swap.ispc runfail  x86-64      generic-4   Linux LLVM 3.4     g++4.4 -O2 *
 ./tests/masked-scatter-struct.ispc runfail  x86-64      generic-4   Linux LLVM 3.3     g++4.4 -O2 *
 ./tests/atomics-13.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-10.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-11.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-13.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-5.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-6.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/atomics-13.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-10.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-11.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-13.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-5.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-6.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/funcptr-null-4.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/funcptr-null-5.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/funcptr-null-6.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/atomics-13.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-10.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-11.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-13.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-5.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-6.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/funcptr-null-4.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/funcptr-null-5.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/funcptr-null-6.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/atomics-13.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-10.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-11.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-13.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-5.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/reduce-equal-6.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/avg-down-int8.ispc compfail     x86    avx1-i32x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/avg-up-int8.ispc compfail     x86    avx1-i32x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/avg-down-int8.ispc compfail  x86-64    avx1-i32x16   Linux LLVM 3.3     g++4.7 -O2 *
-./tests/avg-up-int8.ispc compfail  x86-64    avx1-i32x16   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/avg-down-int8.ispc compfail     x86  avx1.1-i32x16   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/avg-up-int8.ispc compfail     x86  avx1.1-i32x16   Linux LLVM 3.3     g++4.7 -O2 *
 ./tests/avg-down-int8.ispc compfail  x86-64  avx1.1-i32x16   Linux LLVM 3.3     g++4.7 -O2 *
@@ -425,10 +371,6 @@
 ./tests/funcptr-null-4.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.4     g++4.7 -O2 *
 ./tests/funcptr-null-5.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.4     g++4.7 -O2 *
 ./tests/funcptr-null-6.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.4     g++4.7 -O2 *
-./tests/avg-down-int8.ispc compfail     x86    avx1-i32x16   Linux LLVM 3.4     g++4.7 -O2 *
-./tests/avg-up-int8.ispc compfail     x86    avx1-i32x16   Linux LLVM 3.4     g++4.7 -O2 *
-./tests/avg-down-int8.ispc compfail  x86-64    avx1-i32x16   Linux LLVM 3.4     g++4.7 -O2 *
-./tests/avg-up-int8.ispc compfail  x86-64    avx1-i32x16   Linux LLVM 3.4     g++4.7 -O2 *
 ./tests/avg-down-int8.ispc compfail     x86  avx1.1-i32x16   Linux LLVM 3.4     g++4.7 -O2 *
 ./tests/avg-up-int8.ispc compfail     x86  avx1.1-i32x16   Linux LLVM 3.4     g++4.7 -O2 *
 ./tests/avg-down-int8.ispc compfail  x86-64  avx1.1-i32x16   Linux LLVM 3.4     g++4.7 -O2 *
@@ -949,3 +891,137 @@
 .\tests\uint64-min-1.ispc runfail     x86     avx1-i64x4 Windows LLVM 3.3         cl -O2 *
 .\tests\uint64-min.ispc runfail     x86     avx1-i64x4 Windows LLVM 3.3         cl -O2 *
 .\tests\reduce-min-uint64.ispc runfail     x86     avx1-i64x4 Windows LLVM 3.4         cl -O2 *
+./tests/atomics-13.ispc compfail     x86     sse4-i16x8   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/atomics-13.ispc compfail  x86-64     sse4-i16x8   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-4.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-5.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-6.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/atomics-13.ispc compfail     x86     sse4-i8x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-4.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-5.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-6.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/atomics-13.ispc compfail  x86-64     sse4-i8x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/ptr-assign-lhs-math-1.ispc compfail  x86-64      generic-4   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/short-vec-8.ispc compfail  x86-64      generic-4   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/test-141.ispc runfail  x86-64     generic-16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/test-143.ispc runfail  x86-64     generic-16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/ptr-assign-lhs-math-1.ispc compfail  x86-64     generic-16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/avg-down-int8.ispc compfail     x86  avx1.1-i32x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/avg-up-int8.ispc compfail     x86  avx1.1-i32x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/avg-down-int8.ispc compfail  x86-64  avx1.1-i32x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/avg-up-int8.ispc compfail  x86-64  avx1.1-i32x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/test-141.ispc runfail     x86    avx2-i32x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/test-141.ispc runfail  x86-64    avx2-i32x16   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/reduce-equal-4.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-4.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/funcptr-null-5.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/funcptr-null-6.ispc runfail     x86     sse4-i8x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/funcptr-null-4.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/funcptr-null-5.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/funcptr-null-6.ispc runfail  x86-64     sse4-i8x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-down-int8.ispc compfail     x86    avx1-i32x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-up-int8.ispc compfail     x86    avx1-i32x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-down-int8.ispc compfail  x86-64    avx1-i32x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-up-int8.ispc compfail  x86-64    avx1-i32x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/ptr-assign-lhs-math-1.ispc compfail  x86-64      generic-4   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/short-vec-8.ispc compfail  x86-64      generic-4   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/test-141.ispc runfail  x86-64     generic-16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/test-143.ispc runfail  x86-64     generic-16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/ptr-assign-lhs-math-1.ispc compfail  x86-64     generic-16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-down-int8.ispc compfail     x86  avx1.1-i32x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-up-int8.ispc compfail     x86  avx1.1-i32x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-down-int8.ispc compfail  x86-64  avx1.1-i32x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-up-int8.ispc compfail  x86-64  avx1.1-i32x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/atomics-varyingptr-2.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/atomics-varyingptr-3.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/atomics-varyingptr-4.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-11.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-12.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-13.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-4.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-5.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-6.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-7.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-8.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-swap.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-varyingptr-2.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-varyingptr-3.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-varyingptr-4.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/memset-varying.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-1.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-12.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-13.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-2.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-3.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-4.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-5.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-6.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-7.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal.ispc compfail  x86-64     avx2-i32x8   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/test-141.ispc runfail     x86    avx2-i32x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/test-141.ispc runfail  x86-64    avx2-i32x16   Linux LLVM 3.4 clang++3.3 -O2 *
+./tests/atomics-13.ispc compfail     x86     sse4-i16x8     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/atomics-13.ispc compfail  x86-64     sse4-i16x8     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-4.ispc runfail     x86     sse4-i8x16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-5.ispc runfail     x86     sse4-i8x16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-6.ispc runfail     x86     sse4-i8x16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/atomics-13.ispc compfail     x86     sse4-i8x16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-4.ispc runfail  x86-64     sse4-i8x16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-5.ispc runfail  x86-64     sse4-i8x16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-6.ispc runfail  x86-64     sse4-i8x16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/atomics-13.ispc compfail  x86-64     sse4-i8x16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/ptr-assign-lhs-math-1.ispc compfail  x86-64      generic-4     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/short-vec-8.ispc compfail  x86-64      generic-4     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/test-141.ispc runfail  x86-64     generic-16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/test-143.ispc runfail  x86-64     generic-16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/ptr-assign-lhs-math-1.ispc compfail  x86-64     generic-16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/test-141.ispc runfail     x86    avx2-i32x16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/test-141.ispc runfail  x86-64    avx2-i32x16     Mac LLVM 3.3 clang++3.3 -O2 *
+./tests/funcptr-null-4.ispc runfail     x86     sse4-i8x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/funcptr-null-5.ispc runfail     x86     sse4-i8x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/funcptr-null-6.ispc runfail     x86     sse4-i8x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/funcptr-null-4.ispc runfail  x86-64     sse4-i8x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/funcptr-null-5.ispc runfail  x86-64     sse4-i8x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/funcptr-null-6.ispc runfail  x86-64     sse4-i8x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-down-int8.ispc compfail     x86    avx1-i32x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-up-int8.ispc compfail     x86    avx1-i32x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-down-int8.ispc compfail  x86-64    avx1-i32x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-up-int8.ispc compfail  x86-64    avx1-i32x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-down-int8.ispc compfail     x86  avx1.1-i32x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-up-int8.ispc compfail     x86  avx1.1-i32x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-down-int8.ispc compfail  x86-64  avx1.1-i32x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/avg-up-int8.ispc compfail  x86-64  avx1.1-i32x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/ptr-assign-lhs-math-1.ispc compfail  x86-64      generic-4     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/short-vec-8.ispc compfail  x86-64      generic-4     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/test-141.ispc runfail  x86-64     generic-16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/test-143.ispc runfail  x86-64     generic-16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/ptr-assign-lhs-math-1.ispc compfail  x86-64     generic-16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/double-3.ispc runfail     x86     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/atomics-varyingptr-2.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/atomics-varyingptr-3.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/atomics-varyingptr-4.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-11.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-12.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-13.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-4.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-5.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-6.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-7.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-8.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-swap.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-varyingptr-2.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-varyingptr-3.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/local-atomics-varyingptr-4.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/memset-varying.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-1.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-12.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-13.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-2.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-3.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-4.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-5.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-6.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal-7.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/reduce-equal.ispc compfail  x86-64     avx2-i32x8     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/test-141.ispc runfail     x86    avx2-i32x16     Mac LLVM 3.4 clang++3.3 -O2 *
+./tests/test-141.ispc runfail  x86-64    avx2-i32x16     Mac LLVM 3.4 clang++3.3 -O2 *

ispc.cpp

@@ -735,7 +735,7 @@ const char *
 Target::SupportedTargets() {
     return "nvptx64, "
 #ifdef ISPC_ARM_ENABLED
-        "neon-i8x16, neon-16x8, neon-32x4, "
+        "neon-i8x16, neon-i16x8, neon-i32x4, "
 #endif
         "sse2-i32x4, sse2-i32x8, "
         "sse4-i32x4, sse4-i32x8, sse4-i16x8, sse4-i8x16, "

llvm_patches/3_3_0001-Fix-PR16807.patch

@@ -0,0 +1,78 @@
+From b9c47f44691cb9a648b9fa1ae373f0defe53c757 Mon Sep 17 00:00:00 2001
+From: Michael Liao <michael.hliao@gmail.com>
+Date: Thu, 10 Oct 2013 16:47:00 -0700
+Subject: [PATCH] Fix PR16807
+
+- Lower signed division by constant powers-of-2 to target-independent
+  DAG operators instead of target-dependent ones to support them on
+  targets where vector types are legal but shift operators on that types
+  are illegal, e.g. on AVX, PSRAW is only available on <8 x i16> though
+  <16 x i16> is a legal type.
+---
+ lib/Target/X86/X86ISelLowering.cpp | 22 ++++++++++++++++------
+ test/CodeGen/X86/pr16807.ll        | 18 ++++++++++++++++++
+ 2 files changed, 34 insertions(+), 6 deletions(-)
+ create mode 100644 test/CodeGen/X86/pr16807.ll
+
+diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
+index bd5ad4e..518bb90 100644
+--- lib/Target/X86/X86ISelLowering.cpp
++++ lib/Target/X86/X86ISelLowering.cpp
+@@ -12462,14 +12462,24 @@ static SDValue LowerSDIV(SDValue Op, SelectionDAG &DAG) {
+       (SplatValue.isPowerOf2() || (-SplatValue).isPowerOf2())) {
+     unsigned lg2 = SplatValue.countTrailingZeros();
+     // Splat the sign bit.
+-    SDValue Sz = DAG.getConstant(EltTy.getSizeInBits()-1, MVT::i32);
+-    SDValue SGN = getTargetVShiftNode(X86ISD::VSRAI, dl, VT, N0, Sz, DAG);
++    SmallVector<SDValue, 16> Sz(NumElts,
++                                DAG.getConstant(EltTy.getSizeInBits() - 1,
++                                                EltTy));
++    SDValue SGN = DAG.getNode(ISD::SRA, dl, VT, N0,
++                              DAG.getNode(ISD::BUILD_VECTOR, dl, VT, &Sz[0],
++                                          NumElts));
+     // Add (N0 < 0) ? abs2 - 1 : 0;
+-    SDValue Amt = DAG.getConstant(EltTy.getSizeInBits() - lg2, MVT::i32);
+-    SDValue SRL = getTargetVShiftNode(X86ISD::VSRLI, dl, VT, SGN, Amt, DAG);
++    SmallVector<SDValue, 16> Amt(NumElts,
++                                 DAG.getConstant(EltTy.getSizeInBits() - lg2,
++                                                 EltTy));
++    SDValue SRL = DAG.getNode(ISD::SRL, dl, VT, SGN,
++                              DAG.getNode(ISD::BUILD_VECTOR, dl, VT, &Amt[0],
++                                          NumElts));
+     SDValue ADD = DAG.getNode(ISD::ADD, dl, VT, N0, SRL);
+-    SDValue Lg2Amt = DAG.getConstant(lg2, MVT::i32);
+-    SDValue SRA = getTargetVShiftNode(X86ISD::VSRAI, dl, VT, ADD, Lg2Amt, DAG);
++    SmallVector<SDValue, 16> Lg2Amt(NumElts, DAG.getConstant(lg2, EltTy));
++    SDValue SRA = DAG.getNode(ISD::SRA, dl, VT, ADD,
++                              DAG.getNode(ISD::BUILD_VECTOR, dl, VT, &Lg2Amt[0],
++                                          NumElts));
+ 
+     // If we're dividing by a positive value, we're done.  Otherwise, we must
+     // negate the result.
+diff --git a/test/CodeGen/X86/pr16807.ll b/test/CodeGen/X86/pr16807.ll
+new file mode 100644
+index 0000000..6d55d99
+--- /dev/null
++++ test/CodeGen/X86/pr16807.ll
+@@ -0,0 +1,18 @@
++; RUN: llc < %s -mtriple=x86_64-linux-gnu -mcpu=core-avx-i | FileCheck %s
++
++define <16 x i16> @f_fu(<16 x i16> %bf) {
++allocas:
++  %avg.i.i = sdiv <16 x i16> %bf, <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4>
++  ret <16 x i16> %avg.i.i
++}
++
++; CHECK: f_fu
++; CHECK: psraw
++; CHECK: psrlw
++; CHECK: paddw
++; CHECK: psraw
++; CHECK: psraw
++; CHECK: psrlw
++; CHECK: paddw
++; CHECK: psraw
++; CHECK: ret
+-- 
+1.8.1.2
+

module.cpp

@@ -936,7 +936,7 @@ Module::AddExportedTypes(const std::vector<std::pair<const Type *,
 bool
 Module::writeOutput(OutputType outputType, const char *outFileName,
                     const char *includeFileName) {
-    if (diBuilder != NULL && outputType != Header) {
+    if (diBuilder != NULL && (outputType != Header && outputType != Deps)) {
         diBuilder->finalize();
 
         lStripUnusedDebugInfo(module);

perf.py

@@ -299,6 +299,22 @@ def perf(options1, args):
     if cpu_percent > 20:
         error("CPU Usage is very high.\nClose other applications.\n", 2)
 
+    # prepare build.log, perf_temp and perf.log files
+    global perf_log
+    if options.in_file:
+        perf_log = pwd + options.in_file
+        common.remove_if_exists(perf_log)
+    else:
+        perf_log = ""
+    global build_log
+    build_log = pwd + os.sep + "logs" + os.sep + "perf_build.log"
+    common.remove_if_exists(build_log)
+    if os.path.exists(pwd + os.sep + "logs") == False:
+        os.makedirs(pwd + os.sep + "logs")
+    global perf_temp
+    perf_temp = pwd + "perf_temp"
+
+
     global ispc_test
     global ispc_ref
     global ref_compiler
@@ -310,8 +326,8 @@ def perf(options1, args):
     ref_compiler_exists = False
     if is_windows == False:
         ispc_test = "ispc"
-        ref_compiler = "g++"
+        ref_compiler = "clang++"
-        refc_compiler = "gcc"
+        refc_compiler = "clang"
         if options.compiler != "":
             if options.compiler == "clang" or options.compiler == "clang++":
                 ref_compiler = "clang++"
@@ -319,21 +335,30 @@ def perf(options1, args):
             if options.compiler == "icc" or options.compiler == "icpc":
                 ref_compiler = "icpc"
                 refc_compiler = "icc"
+            if options.compiler == "gcc" or options.compiler == "g++":
+                ref_compiler = "g++"
+                refc_compiler = "gcc"
     else:
         ispc_test = "ispc.exe"
         ref_compiler = "cl.exe"
     ispc_ref = options.ref
     if options.ref != "":
         options.ref = True
-    for counter in PATH_dir:
+    if os.environ.get("ISPC_HOME") != None:
-        if os.path.exists(counter + os.sep + ispc_test):
+        if os.path.exists(os.environ["ISPC_HOME"] + os.sep + ispc_test):
             ispc_test_exists = True
+            ispc_test = os.environ["ISPC_HOME"] + os.sep + ispc_test
+    for counter in PATH_dir:
+        if ispc_test_exists == False:
+            if os.path.exists(counter + os.sep + ispc_test):
+                ispc_test_exists = True
+                ispc_test = counter + os.sep + ispc_test
         if os.path.exists(counter + os.sep + ref_compiler):
             ref_compiler_exists = True
         if os.path.exists(counter + os.sep + ispc_ref):
             ispc_ref_exists = True
     if not ispc_test_exists:
-        error("ISPC compiler not found.\nAdded path to ispc compiler to your PATH variable.\n", 1)
+        error("ISPC compiler not found.\nAdded path to ispc compiler to your PATH variable or ISPC_HOME variable\n", 1)
     if not ref_compiler_exists:
         error("C/C++ compiler %s not found.\nAdded path to %s compiler to your PATH variable.\n" % (ref_compiler, ref_compiler), 1)
     if options.ref:
@@ -355,26 +380,11 @@ def perf(options1, args):
         if f_lines[i][0] != "%":
             lines.append(f_lines[i])
     length = len(lines)
-
-    # prepare build.log, perf_temp and perf.log files
-    global perf_log
-    if options.in_file:
-        perf_log = pwd + options.in_file
-        common.remove_if_exists(perf_log)
-    else:
-        perf_log = ""
-    global build_log
-    build_log = pwd + os.sep + "logs" + os.sep + "perf_build.log"
-    common.remove_if_exists(build_log)
-    if os.path.exists(pwd + os.sep + "logs") == False:
-        os.makedirs(pwd + os.sep + "logs")
-
-    global perf_temp
-    perf_temp = pwd + "perf_temp"
     # end of preparations
  
     print_debug("Okey go go go!\n\n", s, perf_log)
-    
+    print_debug("Testing ispc: " + ispc_test + "\n", s, perf_log)
+ 
     #print compilers versions   
     common.print_version(ispc_test, ispc_ref, ref_compiler, False, perf_log, is_windows) 
 

run_tests.py

@@ -446,7 +446,7 @@ def verify():
     f = open(test_states, 'r')
     f_lines = f.readlines()
     f.close()
-    check = [["g++", "clang", "cl"],["-O0", "-O2"],["x86","x86-64"],
+    check = [["g++", "clang++", "cl"],["-O0", "-O2"],["x86","x86-64"],
              ["Linux","Windows","Mac"],["LLVM 3.1","LLVM 3.2","LLVM 3.3","LLVM head"],
              ["sse2-i32x4", "sse2-i32x8", "sse4-i32x4", "sse4-i32x8", "sse4-i16x8",
               "sse4-i8x16", "avx1-i32x8", "avx1-i32x16", "avx1-i64x4", "avx1.1-i32x8", "avx1.1-i32x16",
@@ -501,16 +501,26 @@ def run_tests(options1, args, print_version):
     # use relative path to not depend on host directory, which may possibly
     # have white spaces and unicode characters.
     global ispc_exe
+    ispc_exe = ""
     if not is_windows:
-        ispc_exe = "./ispc"
+        if os.environ.get("ISPC_HOME") != None:
+            if os.path.exists(os.environ["ISPC_HOME"] + os.sep + "ispc"):
+                ispc_exe = os.environ["ISPC_HOME"] + os.sep + "ispc"
+            else:
+                PATH_dir = string.split(os.getenv("PATH"), os.pathsep)
+                for counter in PATH_dir:
+                    if os.path.exists(counter + os.sep + "ispc"):
+                        ispc_exe = counter + os.sep + "ispc"
     else:
-        ispc_exe = ".\\Release\\ispc.exe"
+        if os.path.exists(".\\Release\\ispc.exe"):
- 
+            ispc_exe = ".\\Release\\ispc.exe"
+        else:
+            error("You don't have ispc.exe compiler in .\\Release.\n", 1)
     # checks the required ispc compiler otherwise prints an error message
-    if not os.path.exists(ispc_exe):
+    if ispc_exe == "":
-        error("missing ispc compiler: %s\n" % ispc_exe, 1)
+        error("ISPC compiler not found.\nAdded path to ispc compiler to your PATH variable or ISPC_HOME variable\n", 1)
+    print_debug("Testing ispc: " + ispc_exe + "\n", s, run_tests_log)
     ispc_exe += " " + options.ispc_flags
-    print_debug("ispc compiler: %s\n" % ispc_exe, s, run_tests_log)
 
     global is_generic_target 
     is_generic_target = (options.target.find("generic-") != -1 and
@@ -540,7 +550,7 @@ def run_tests(options1, args, print_version):
         if is_windows:
             options.compiler_exe = "cl.exe"
         else:
-            options.compiler_exe = "g++"
+            options.compiler_exe = "clang++"
  
     # checks the required compiler otherwise prints an error message
     PATH_dir = string.split(os.getenv("PATH"), os.pathsep)