Import libcompiler_rt into HEAD and add Makefiles.

Obtained from:	user/ed/compiler-rt

contrib/compiler-rt/BlocksRuntime/Block.h

@@ -0,0 +1,59 @@
+/*
+ * Block.h
+ *
+ * Copyright 2008-2010 Apple, Inc. Permission is hereby granted, free of charge,
+ * to any person obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without restriction,
+ * including without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to permit
+ * persons to whom the Software is furnished to do so, subject to the following
+ * conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef _BLOCK_H_
+#define _BLOCK_H_
+
+#if !defined(BLOCK_EXPORT)
+#   if defined(__cplusplus)
+#       define BLOCK_EXPORT extern "C" 
+#   else
+#       define BLOCK_EXPORT extern
+#   endif
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+/* Create a heap based copy of a Block or simply add a reference to an existing one.
+ * This must be paired with Block_release to recover memory, even when running
+ * under Objective-C Garbage Collection.
+ */
+BLOCK_EXPORT void *_Block_copy(const void *aBlock);
+
+/* Lose the reference, and if heap based and last reference, recover the memory. */
+BLOCK_EXPORT void _Block_release(const void *aBlock);
+
+#if defined(__cplusplus)
+}
+#endif
+
+/* Type correct macros. */
+
+#define Block_copy(...) ((__typeof(__VA_ARGS__))_Block_copy((const void *)(__VA_ARGS__)))
+#define Block_release(...) _Block_release((const void *)(__VA_ARGS__))
+
+
+#endif

contrib/compiler-rt/BlocksRuntime/Block_private.h

@@ -0,0 +1,179 @@
+/*
+ * Block_private.h
+ *
+ * Copyright 2008-2010 Apple, Inc. Permission is hereby granted, free of charge,
+ * to any person obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without restriction,
+ * including without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to permit
+ * persons to whom the Software is furnished to do so, subject to the following
+ * conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef _BLOCK_PRIVATE_H_
+#define _BLOCK_PRIVATE_H_
+
+#if !defined(BLOCK_EXPORT)
+#   if defined(__cplusplus)
+#       define BLOCK_EXPORT extern "C" 
+#   else
+#       define BLOCK_EXPORT extern
+#   endif
+#endif
+
+#ifndef _MSC_VER
+#include <stdbool.h>
+#else
+/* MSVC doesn't have <stdbool.h>. Compensate. */
+typedef char bool;
+#define true (bool)1
+#define false (bool)0
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+
+enum {
+    BLOCK_REFCOUNT_MASK =     (0xffff),
+    BLOCK_NEEDS_FREE =        (1 << 24),
+    BLOCK_HAS_COPY_DISPOSE =  (1 << 25),
+    BLOCK_HAS_CTOR =          (1 << 26), /* Helpers have C++ code. */
+    BLOCK_IS_GC =             (1 << 27),
+    BLOCK_IS_GLOBAL =         (1 << 28),
+    BLOCK_HAS_DESCRIPTOR =    (1 << 29)
+};
+
+
+/* Revised new layout. */
+struct Block_descriptor {
+    unsigned long int reserved;
+    unsigned long int size;
+    void (*copy)(void *dst, void *src);
+    void (*dispose)(void *);
+};
+
+
+struct Block_layout {
+    void *isa;
+    int flags;
+    int reserved; 
+    void (*invoke)(void *, ...);
+    struct Block_descriptor *descriptor;
+    /* Imported variables. */
+};
+
+
+struct Block_byref {
+    void *isa;
+    struct Block_byref *forwarding;
+    int flags; /* refcount; */
+    int size;
+    void (*byref_keep)(struct Block_byref *dst, struct Block_byref *src);
+    void (*byref_destroy)(struct Block_byref *);
+    /* long shared[0]; */
+};
+
+
+struct Block_byref_header {
+    void *isa;
+    struct Block_byref *forwarding;
+    int flags;
+    int size;
+};
+
+
+/* Runtime support functions used by compiler when generating copy/dispose helpers. */
+
+enum {
+    /* See function implementation for a more complete description of these fields and combinations */
+    BLOCK_FIELD_IS_OBJECT   =  3,  /* id, NSObject, __attribute__((NSObject)), block, ... */
+    BLOCK_FIELD_IS_BLOCK    =  7,  /* a block variable */
+    BLOCK_FIELD_IS_BYREF    =  8,  /* the on stack structure holding the __block variable */
+    BLOCK_FIELD_IS_WEAK     = 16,  /* declared __weak, only used in byref copy helpers */
+    BLOCK_BYREF_CALLER      = 128  /* called from __block (byref) copy/dispose support routines. */
+};
+
+/* Runtime entry point called by compiler when assigning objects inside copy helper routines */
+BLOCK_EXPORT void _Block_object_assign(void *destAddr, const void *object, const int flags);
+    /* BLOCK_FIELD_IS_BYREF is only used from within block copy helpers */
+
+
+/* runtime entry point called by the compiler when disposing of objects inside dispose helper routine */
+BLOCK_EXPORT void _Block_object_dispose(const void *object, const int flags);
+
+
+
+/* Other support functions */
+
+/* Runtime entry to get total size of a closure */
+BLOCK_EXPORT unsigned long int Block_size(void *block_basic);
+
+
+
+/* the raw data space for runtime classes for blocks */
+/* class+meta used for stack, malloc, and collectable based blocks */
+BLOCK_EXPORT void * _NSConcreteStackBlock[32];
+BLOCK_EXPORT void * _NSConcreteMallocBlock[32];
+BLOCK_EXPORT void * _NSConcreteAutoBlock[32];
+BLOCK_EXPORT void * _NSConcreteFinalizingBlock[32];
+BLOCK_EXPORT void * _NSConcreteGlobalBlock[32];
+BLOCK_EXPORT void * _NSConcreteWeakBlockVariable[32];
+
+
+/* the intercept routines that must be used under GC */
+BLOCK_EXPORT void _Block_use_GC( void *(*alloc)(const unsigned long, const bool isOne, const bool isObject),
+                                  void (*setHasRefcount)(const void *, const bool),
+                                  void (*gc_assign_strong)(void *, void **),
+                                  void (*gc_assign_weak)(const void *, void *),
+                                  void (*gc_memmove)(void *, void *, unsigned long));
+
+/* earlier version, now simply transitional */
+BLOCK_EXPORT void _Block_use_GC5( void *(*alloc)(const unsigned long, const bool isOne, const bool isObject),
+                                  void (*setHasRefcount)(const void *, const bool),
+                                  void (*gc_assign_strong)(void *, void **),
+                                  void (*gc_assign_weak)(const void *, void *));
+
+BLOCK_EXPORT void _Block_use_RR( void (*retain)(const void *),
+                                 void (*release)(const void *));
+
+/* make a collectable GC heap based Block.  Not useful under non-GC. */
+BLOCK_EXPORT void *_Block_copy_collectable(const void *aBlock);
+
+/* thread-unsafe diagnostic */
+BLOCK_EXPORT const char *_Block_dump(const void *block);
+
+
+/* Obsolete */
+
+/* first layout */
+struct Block_basic {
+    void *isa;
+    int Block_flags;  /* int32_t */
+    int Block_size;  /* XXX should be packed into Block_flags */
+    void (*Block_invoke)(void *);
+    void (*Block_copy)(void *dst, void *src);  /* iff BLOCK_HAS_COPY_DISPOSE */
+    void (*Block_dispose)(void *);             /* iff BLOCK_HAS_COPY_DISPOSE */
+    /* long params[0];  // where const imports, __block storage references, etc. get laid down */
+};
+
+
+#if defined(__cplusplus)
+}
+#endif
+
+
+#endif /* _BLOCK_PRIVATE_H_ */

contrib/compiler-rt/BlocksRuntime/data.c

@@ -0,0 +1,41 @@
+/*
+ * data.c
+ *
+ * Copyright 2008-2010 Apple, Inc. Permission is hereby granted, free of charge,
+ * to any person obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without restriction,
+ * including without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to permit
+ * persons to whom the Software is furnished to do so, subject to the following
+ * conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+/********************
+NSBlock support
+
+We allocate space and export a symbol to be used as the Class for the on-stack and malloc'ed copies until ObjC arrives on the scene.  These data areas are set up by Foundation to link in as real classes post facto.
+
+We keep these in a separate file so that we can include the runtime code in test subprojects but not include the data so that compiled code that sees the data in libSystem doesn't get confused by a second copy.  Somehow these don't get unified in a common block.
+**********************/
+
+void * _NSConcreteStackBlock[32] = { 0 };
+void * _NSConcreteMallocBlock[32] = { 0 };
+void * _NSConcreteAutoBlock[32] = { 0 };
+void * _NSConcreteFinalizingBlock[32] = { 0 };
+void * _NSConcreteGlobalBlock[32] = { 0 };
+void * _NSConcreteWeakBlockVariable[32] = { 0 };
+
+void _Block_copy_error(void) {
+}

contrib/compiler-rt/BlocksRuntime/runtime.c

@@ -0,0 +1,700 @@
+/*
+ * runtime.c
+ *
+ * Copyright 2008-2010 Apple, Inc. Permission is hereby granted, free of charge,
+ * to any person obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without restriction,
+ * including without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to permit
+ * persons to whom the Software is furnished to do so, subject to the following
+ * conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#include "Block_private.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdint.h>
+
+#include "config.h"
+
+#ifdef HAVE_AVAILABILITY_MACROS_H
+#include <AvailabilityMacros.h>
+#endif /* HAVE_AVAILABILITY_MACROS_H */
+
+#ifdef HAVE_TARGET_CONDITIONALS_H
+#include <TargetConditionals.h>
+#endif /* HAVE_TARGET_CONDITIONALS_H */
+
+#if defined(HAVE_OSATOMIC_COMPARE_AND_SWAP_INT) && defined(HAVE_OSATOMIC_COMPARE_AND_SWAP_LONG)
+
+#ifdef HAVE_LIBKERN_OSATOMIC_H
+#include <libkern/OSAtomic.h>
+#endif /* HAVE_LIBKERN_OSATOMIC_H */
+
+#elif defined(__WIN32__) || defined(_WIN32)
+#define _CRT_SECURE_NO_WARNINGS 1
+#include <windows.h>
+
+static __inline bool OSAtomicCompareAndSwapLong(long oldl, long newl, long volatile *dst) {
+    /* fixme barrier is overkill -- see objc-os.h */
+    long original = InterlockedCompareExchange(dst, newl, oldl);
+    return (original == oldl);
+}
+
+static __inline bool OSAtomicCompareAndSwapInt(int oldi, int newi, int volatile *dst) {
+    /* fixme barrier is overkill -- see objc-os.h */
+    int original = InterlockedCompareExchange(dst, newi, oldi);
+    return (original == oldi);
+}
+
+/*
+ * Check to see if the GCC atomic built-ins are available.  If we're on
+ * a 64-bit system, make sure we have an 8-byte atomic function
+ * available.
+ *
+ */
+
+#elif defined(HAVE_SYNC_BOOL_COMPARE_AND_SWAP_INT) && defined(HAVE_SYNC_BOOL_COMPARE_AND_SWAP_LONG)
+
+static __inline bool OSAtomicCompareAndSwapLong(long oldl, long newl, long volatile *dst) {
+  return __sync_bool_compare_and_swap(dst, oldl, newl);
+}
+
+static __inline bool OSAtomicCompareAndSwapInt(int oldi, int newi, int volatile *dst) {
+  return __sync_bool_compare_and_swap(dst, oldi, newi);
+}
+
+#else
+#error unknown atomic compare-and-swap primitive
+#endif /* HAVE_OSATOMIC_COMPARE_AND_SWAP_INT && HAVE_OSATOMIC_COMPARE_AND_SWAP_LONG */
+
+
+/*
+ * Globals:
+ */
+
+static void *_Block_copy_class = _NSConcreteMallocBlock;
+static void *_Block_copy_finalizing_class = _NSConcreteMallocBlock;
+static int _Block_copy_flag = BLOCK_NEEDS_FREE;
+static int _Byref_flag_initial_value = BLOCK_NEEDS_FREE | 2;
+
+static const int WANTS_ONE = (1 << 16);
+
+static bool isGC = false;
+
+/*
+ * Internal Utilities:
+ */
+
+#if 0
+static unsigned long int latching_incr_long(unsigned long int *where) {
+    while (1) {
+        unsigned long int old_value = *(volatile unsigned long int *)where;
+        if ((old_value & BLOCK_REFCOUNT_MASK) == BLOCK_REFCOUNT_MASK) {
+            return BLOCK_REFCOUNT_MASK;
+        }
+        if (OSAtomicCompareAndSwapLong(old_value, old_value+1, (volatile long int *)where)) {
+            return old_value+1;
+        }
+    }
+}
+#endif /* if 0 */
+
+static int latching_incr_int(int *where) {
+    while (1) {
+        int old_value = *(volatile int *)where;
+        if ((old_value & BLOCK_REFCOUNT_MASK) == BLOCK_REFCOUNT_MASK) {
+            return BLOCK_REFCOUNT_MASK;
+        }
+        if (OSAtomicCompareAndSwapInt(old_value, old_value+1, (volatile int *)where)) {
+            return old_value+1;
+        }
+    }
+}
+
+#if 0
+static int latching_decr_long(unsigned long int *where) {
+    while (1) {
+        unsigned long int old_value = *(volatile int *)where;
+        if ((old_value & BLOCK_REFCOUNT_MASK) == BLOCK_REFCOUNT_MASK) {
+            return BLOCK_REFCOUNT_MASK;
+        }
+        if ((old_value & BLOCK_REFCOUNT_MASK) == 0) {
+            return 0;
+        }
+        if (OSAtomicCompareAndSwapLong(old_value, old_value-1, (volatile long int *)where)) {
+            return old_value-1;
+        }
+    }
+}
+#endif /* if 0 */
+
+static int latching_decr_int(int *where) {
+    while (1) {
+        int old_value = *(volatile int *)where;
+        if ((old_value & BLOCK_REFCOUNT_MASK) == BLOCK_REFCOUNT_MASK) {
+            return BLOCK_REFCOUNT_MASK;
+        }
+        if ((old_value & BLOCK_REFCOUNT_MASK) == 0) {
+            return 0;
+        }
+        if (OSAtomicCompareAndSwapInt(old_value, old_value-1, (volatile int *)where)) {
+            return old_value-1;
+        }
+    }
+}
+
+
+/*
+ * GC support stub routines:
+ */
+#if 0
+#pragma mark GC Support Routines
+#endif /* if 0 */
+
+
+static void *_Block_alloc_default(const unsigned long size, const bool initialCountIsOne, const bool isObject) {
+    return malloc(size);
+}
+
+static void _Block_assign_default(void *value, void **destptr) {
+    *destptr = value;
+}
+
+static void _Block_setHasRefcount_default(const void *ptr, const bool hasRefcount) {
+}
+
+static void _Block_do_nothing(const void *aBlock) { }
+
+static void _Block_retain_object_default(const void *ptr) {
+    if (!ptr) return;
+}
+
+static void _Block_release_object_default(const void *ptr) {
+    if (!ptr) return;
+}
+
+static void _Block_assign_weak_default(const void *ptr, void *dest) {
+    *(void **)dest = (void *)ptr;
+}
+
+static void _Block_memmove_default(void *dst, void *src, unsigned long size) {
+    memmove(dst, src, (size_t)size);
+}
+
+static void _Block_memmove_gc_broken(void *dest, void *src, unsigned long size) {
+    void **destp = (void **)dest;
+    void **srcp = (void **)src;
+    while (size) {
+        _Block_assign_default(*srcp, destp);
+        destp++;
+        srcp++;
+        size -= sizeof(void *);
+    }
+}
+
+/*
+ * GC support callout functions - initially set to stub routines:
+ */
+
+static void *(*_Block_allocator)(const unsigned long, const bool isOne, const bool isObject) = _Block_alloc_default;
+static void (*_Block_deallocator)(const void *) = (void (*)(const void *))free;
+static void (*_Block_assign)(void *value, void **destptr) = _Block_assign_default;
+static void (*_Block_setHasRefcount)(const void *ptr, const bool hasRefcount) = _Block_setHasRefcount_default;
+static void (*_Block_retain_object)(const void *ptr) = _Block_retain_object_default;
+static void (*_Block_release_object)(const void *ptr) = _Block_release_object_default;
+static void (*_Block_assign_weak)(const void *dest, void *ptr) = _Block_assign_weak_default;
+static void (*_Block_memmove)(void *dest, void *src, unsigned long size) = _Block_memmove_default;
+
+
+/*
+ * GC support SPI functions - called from ObjC runtime and CoreFoundation:
+ */
+
+/* Public SPI
+ * Called from objc-auto to turn on GC.
+ * version 3, 4 arg, but changed 1st arg
+ */
+void _Block_use_GC( void *(*alloc)(const unsigned long, const bool isOne, const bool isObject),
+                    void (*setHasRefcount)(const void *, const bool),
+                    void (*gc_assign)(void *, void **),
+                    void (*gc_assign_weak)(const void *, void *),
+                    void (*gc_memmove)(void *, void *, unsigned long)) {
+
+    isGC = true;
+    _Block_allocator = alloc;
+    _Block_deallocator = _Block_do_nothing;
+    _Block_assign = gc_assign;
+    _Block_copy_flag = BLOCK_IS_GC;
+    _Block_copy_class = _NSConcreteAutoBlock;
+    /* blocks with ctors & dtors need to have the dtor run from a class with a finalizer */
+    _Block_copy_finalizing_class = _NSConcreteFinalizingBlock;
+    _Block_setHasRefcount = setHasRefcount;
+    _Byref_flag_initial_value = BLOCK_IS_GC;   // no refcount
+    _Block_retain_object = _Block_do_nothing;
+    _Block_release_object = _Block_do_nothing;
+    _Block_assign_weak = gc_assign_weak;
+    _Block_memmove = gc_memmove;
+}
+
+/* transitional */
+void _Block_use_GC5( void *(*alloc)(const unsigned long, const bool isOne, const bool isObject),
+                    void (*setHasRefcount)(const void *, const bool),
+                    void (*gc_assign)(void *, void **),
+                    void (*gc_assign_weak)(const void *, void *)) {
+    /* until objc calls _Block_use_GC it will call us; supply a broken internal memmove implementation until then */
+    _Block_use_GC(alloc, setHasRefcount, gc_assign, gc_assign_weak, _Block_memmove_gc_broken);
+}
+
+ 
+/*
+ * Called from objc-auto to alternatively turn on retain/release.
+ * Prior to this the only "object" support we can provide is for those
+ * super special objects that live in libSystem, namely dispatch queues.
+ * Blocks and Block_byrefs have their own special entry points.
+ *
+ */
+void _Block_use_RR( void (*retain)(const void *),
+                    void (*release)(const void *)) {
+    _Block_retain_object = retain;
+    _Block_release_object = release;
+}
+
+/*
+ * Internal Support routines for copying:
+ */
+
+#if 0
+#pragma mark Copy/Release support
+#endif /* if 0 */
+
+/* Copy, or bump refcount, of a block.  If really copying, call the copy helper if present. */
+static void *_Block_copy_internal(const void *arg, const int flags) {
+    struct Block_layout *aBlock;
+    const bool wantsOne = (WANTS_ONE & flags) == WANTS_ONE;
+
+    //printf("_Block_copy_internal(%p, %x)\n", arg, flags);	
+    if (!arg) return NULL;
+    
+    
+    // The following would be better done as a switch statement
+    aBlock = (struct Block_layout *)arg;
+    if (aBlock->flags & BLOCK_NEEDS_FREE) {
+        // latches on high
+        latching_incr_int(&aBlock->flags);
+        return aBlock;
+    }
+    else if (aBlock->flags & BLOCK_IS_GC) {
+        // GC refcounting is expensive so do most refcounting here.
+        if (wantsOne && ((latching_incr_int(&aBlock->flags) & BLOCK_REFCOUNT_MASK) == 1)) {
+            // Tell collector to hang on this - it will bump the GC refcount version
+            _Block_setHasRefcount(aBlock, true);
+        }
+        return aBlock;
+    }
+    else if (aBlock->flags & BLOCK_IS_GLOBAL) {
+        return aBlock;
+    }
+
+    // Its a stack block.  Make a copy.
+    if (!isGC) {
+        struct Block_layout *result = malloc(aBlock->descriptor->size);
+        if (!result) return (void *)0;
+        memmove(result, aBlock, aBlock->descriptor->size); // bitcopy first
+        // reset refcount
+        result->flags &= ~(BLOCK_REFCOUNT_MASK);    // XXX not needed
+        result->flags |= BLOCK_NEEDS_FREE | 1;
+        result->isa = _NSConcreteMallocBlock;
+        if (result->flags & BLOCK_HAS_COPY_DISPOSE) {
+            //printf("calling block copy helper %p(%p, %p)...\n", aBlock->descriptor->copy, result, aBlock);
+            (*aBlock->descriptor->copy)(result, aBlock); // do fixup
+        }
+        return result;
+    }
+    else {
+        // Under GC want allocation with refcount 1 so we ask for "true" if wantsOne
+        // This allows the copy helper routines to make non-refcounted block copies under GC
+        unsigned long int flags = aBlock->flags;
+        bool hasCTOR = (flags & BLOCK_HAS_CTOR) != 0;
+        struct Block_layout *result = _Block_allocator(aBlock->descriptor->size, wantsOne, hasCTOR);
+        if (!result) return (void *)0;
+        memmove(result, aBlock, aBlock->descriptor->size); // bitcopy first
+        // reset refcount
+        // if we copy a malloc block to a GC block then we need to clear NEEDS_FREE.
+        flags &= ~(BLOCK_NEEDS_FREE|BLOCK_REFCOUNT_MASK);   // XXX not needed
+        if (wantsOne)
+            flags |= BLOCK_IS_GC | 1;
+        else
+            flags |= BLOCK_IS_GC;
+        result->flags = flags;
+        if (flags & BLOCK_HAS_COPY_DISPOSE) {
+            //printf("calling block copy helper...\n");
+            (*aBlock->descriptor->copy)(result, aBlock); // do fixup
+        }
+        if (hasCTOR) {
+            result->isa = _NSConcreteFinalizingBlock;
+        }
+        else {
+            result->isa = _NSConcreteAutoBlock;
+        }
+        return result;
+    }
+}
+
+
+/*
+ * Runtime entry points for maintaining the sharing knowledge of byref data blocks.
+ *
+ * A closure has been copied and its fixup routine is asking us to fix up the reference to the shared byref data
+ * Closures that aren't copied must still work, so everyone always accesses variables after dereferencing the forwarding ptr.
+ * We ask if the byref pointer that we know about has already been copied to the heap, and if so, increment it.
+ * Otherwise we need to copy it and update the stack forwarding pointer
+ * XXX We need to account for weak/nonretained read-write barriers.
+ */
+
+static void _Block_byref_assign_copy(void *dest, const void *arg, const int flags) {
+    struct Block_byref **destp = (struct Block_byref **)dest;
+    struct Block_byref *src = (struct Block_byref *)arg;
+        
+    //printf("_Block_byref_assign_copy called, byref destp %p, src %p, flags %x\n", destp, src, flags);
+    //printf("src dump: %s\n", _Block_byref_dump(src));
+    if (src->forwarding->flags & BLOCK_IS_GC) {
+        ;   // don't need to do any more work
+    }
+    else if ((src->forwarding->flags & BLOCK_REFCOUNT_MASK) == 0) {
+        //printf("making copy\n");
+        // src points to stack
+        bool isWeak = ((flags & (BLOCK_FIELD_IS_BYREF|BLOCK_FIELD_IS_WEAK)) == (BLOCK_FIELD_IS_BYREF|BLOCK_FIELD_IS_WEAK));
+        // if its weak ask for an object (only matters under GC)
+        struct Block_byref *copy = (struct Block_byref *)_Block_allocator(src->size, false, isWeak);
+        copy->flags = src->flags | _Byref_flag_initial_value; // non-GC one for caller, one for stack
+        copy->forwarding = copy; // patch heap copy to point to itself (skip write-barrier)
+        src->forwarding = copy;  // patch stack to point to heap copy
+        copy->size = src->size;
+        if (isWeak) {
+            copy->isa = &_NSConcreteWeakBlockVariable;  // mark isa field so it gets weak scanning
+        }
+        if (src->flags & BLOCK_HAS_COPY_DISPOSE) {
+            // Trust copy helper to copy everything of interest
+            // If more than one field shows up in a byref block this is wrong XXX
+            copy->byref_keep = src->byref_keep;
+            copy->byref_destroy = src->byref_destroy;
+            (*src->byref_keep)(copy, src);
+        }
+        else {
+            // just bits.  Blast 'em using _Block_memmove in case they're __strong
+            _Block_memmove(
+                (void *)&copy->byref_keep,
+                (void *)&src->byref_keep,
+                src->size - sizeof(struct Block_byref_header));
+        }
+    }
+    // already copied to heap
+    else if ((src->forwarding->flags & BLOCK_NEEDS_FREE) == BLOCK_NEEDS_FREE) {
+        latching_incr_int(&src->forwarding->flags);
+    }
+    // assign byref data block pointer into new Block
+    _Block_assign(src->forwarding, (void **)destp);
+}
+
+// Old compiler SPI
+static void _Block_byref_release(const void *arg) {
+    struct Block_byref *shared_struct = (struct Block_byref *)arg;
+    int refcount;
+
+    // dereference the forwarding pointer since the compiler isn't doing this anymore (ever?)
+    shared_struct = shared_struct->forwarding;
+    
+    //printf("_Block_byref_release %p called, flags are %x\n", shared_struct, shared_struct->flags);
+    // To support C++ destructors under GC we arrange for there to be a finalizer for this
+    // by using an isa that directs the code to a finalizer that calls the byref_destroy method.
+    if ((shared_struct->flags & BLOCK_NEEDS_FREE) == 0) {
+        return; // stack or GC or global
+    }
+    refcount = shared_struct->flags & BLOCK_REFCOUNT_MASK;
+    if (refcount <= 0) {
+        printf("_Block_byref_release: Block byref data structure at %p underflowed\n", arg);
+    }
+    else if ((latching_decr_int(&shared_struct->flags) & BLOCK_REFCOUNT_MASK) == 0) {
+        //printf("disposing of heap based byref block\n");
+        if (shared_struct->flags & BLOCK_HAS_COPY_DISPOSE) {
+            //printf("calling out to helper\n");
+            (*shared_struct->byref_destroy)(shared_struct);
+        }
+        _Block_deallocator((struct Block_layout *)shared_struct);
+    }
+}
+
+
+/*
+ *
+ * API supporting SPI
+ * _Block_copy, _Block_release, and (old) _Block_destroy
+ *
+ */
+
+#if 0
+#pragma mark SPI/API
+#endif /* if 0 */
+
+void *_Block_copy(const void *arg) {
+    return _Block_copy_internal(arg, WANTS_ONE);
+}
+
+
+// API entry point to release a copied Block
+void _Block_release(void *arg) {
+    struct Block_layout *aBlock = (struct Block_layout *)arg;
+    int32_t newCount;
+    if (!aBlock) return;
+    newCount = latching_decr_int(&aBlock->flags) & BLOCK_REFCOUNT_MASK;
+    if (newCount > 0) return;
+    // Hit zero
+    if (aBlock->flags & BLOCK_IS_GC) {
+        // Tell GC we no longer have our own refcounts.  GC will decr its refcount
+        // and unless someone has done a CFRetain or marked it uncollectable it will
+        // now be subject to GC reclamation.
+        _Block_setHasRefcount(aBlock, false);
+    }
+    else if (aBlock->flags & BLOCK_NEEDS_FREE) {
+        if (aBlock->flags & BLOCK_HAS_COPY_DISPOSE)(*aBlock->descriptor->dispose)(aBlock);
+        _Block_deallocator(aBlock);
+    }
+    else if (aBlock->flags & BLOCK_IS_GLOBAL) {
+        ;
+    }
+    else {
+        printf("Block_release called upon a stack Block: %p, ignored\n", (void *)aBlock);
+    }
+}
+
+
+
+// Old Compiler SPI point to release a copied Block used by the compiler in dispose helpers
+static void _Block_destroy(const void *arg) {
+    struct Block_layout *aBlock;
+    if (!arg) return;
+    aBlock = (struct Block_layout *)arg;
+    if (aBlock->flags & BLOCK_IS_GC) {
+        // assert(aBlock->Block_flags & BLOCK_HAS_CTOR);
+        return; // ignore, we are being called because of a DTOR
+    }
+    _Block_release(aBlock);
+}
+
+
+
+/*
+ *
+ * SPI used by other layers
+ *
+ */
+
+// SPI, also internal.  Called from NSAutoBlock only under GC
+void *_Block_copy_collectable(const void *aBlock) {
+    return _Block_copy_internal(aBlock, 0);
+}
+
+
+// SPI
+unsigned long int Block_size(void *arg) {
+    return ((struct Block_layout *)arg)->descriptor->size;
+}
+
+
+#if 0
+#pragma mark Compiler SPI entry points
+#endif /* if 0 */
+
+    
+/*******************************************************
+
+Entry points used by the compiler - the real API!
+
+
+A Block can reference four different kinds of things that require help when the Block is copied to the heap.
+1) C++ stack based objects
+2) References to Objective-C objects
+3) Other Blocks
+4) __block variables
+
+In these cases helper functions are synthesized by the compiler for use in Block_copy and Block_release, called the copy and dispose helpers.  The copy helper emits a call to the C++ const copy constructor for C++ stack based objects and for the rest calls into the runtime support function _Block_object_assign.  The dispose helper has a call to the C++ destructor for case 1 and a call into _Block_object_dispose for the rest.
+
+The flags parameter of _Block_object_assign and _Block_object_dispose is set to
+	* BLOCK_FIELD_IS_OBJECT (3), for the case of an Objective-C Object,
+	* BLOCK_FIELD_IS_BLOCK (7), for the case of another Block, and
+	* BLOCK_FIELD_IS_BYREF (8), for the case of a __block variable.
+If the __block variable is marked weak the compiler also or's in BLOCK_FIELD_IS_WEAK (16).
+
+So the Block copy/dispose helpers should only ever generate the four flag values of 3, 7, 8, and 24.
+
+When  a __block variable is either a C++ object, an Objective-C object, or another Block then the compiler also generates copy/dispose helper functions.  Similarly to the Block copy helper, the "__block" copy helper (formerly and still a.k.a. "byref" copy helper) will do a C++ copy constructor (not a const one though!) and the dispose helper will do the destructor.  And similarly the helpers will call into the same two support functions with the same values for objects and Blocks with the additional BLOCK_BYREF_CALLER (128) bit of information supplied.
+
+So the __block copy/dispose helpers will generate flag values of 3 or 7 for objects and Blocks respectively, with BLOCK_FIELD_IS_WEAK (16) or'ed as appropriate and always 128 or'd in, for the following set of possibilities:
+	__block id                   128+3
+        __weak block id              128+3+16
+	__block (^Block)             128+7
+	__weak __block (^Block)      128+7+16
+        
+The implementation of the two routines would be improved by switch statements enumerating the eight cases.
+
+********************************************************/
+
+/*
+ * When Blocks or Block_byrefs hold objects then their copy routine helpers use this entry point
+ * to do the assignment.
+ */
+void _Block_object_assign(void *destAddr, const void *object, const int flags) {
+    //printf("_Block_object_assign(*%p, %p, %x)\n", destAddr, object, flags);
+    if ((flags & BLOCK_BYREF_CALLER) == BLOCK_BYREF_CALLER) {
+        if ((flags & BLOCK_FIELD_IS_WEAK) == BLOCK_FIELD_IS_WEAK) {
+            _Block_assign_weak(object, destAddr);
+        }
+        else {
+            // do *not* retain or *copy* __block variables whatever they are
+            _Block_assign((void *)object, destAddr);
+        }
+    }
+    else if ((flags & BLOCK_FIELD_IS_BYREF) == BLOCK_FIELD_IS_BYREF)  {
+        // copying a __block reference from the stack Block to the heap
+        // flags will indicate if it holds a __weak reference and needs a special isa
+        _Block_byref_assign_copy(destAddr, object, flags);
+    }
+    // (this test must be before next one)
+    else if ((flags & BLOCK_FIELD_IS_BLOCK) == BLOCK_FIELD_IS_BLOCK) {
+        // copying a Block declared variable from the stack Block to the heap
+        _Block_assign(_Block_copy_internal(object, flags), destAddr);
+    }
+    // (this test must be after previous one)
+    else if ((flags & BLOCK_FIELD_IS_OBJECT) == BLOCK_FIELD_IS_OBJECT) {
+        //printf("retaining object at %p\n", object);
+        _Block_retain_object(object);
+        //printf("done retaining object at %p\n", object);
+        _Block_assign((void *)object, destAddr);
+    }
+}
+
+// When Blocks or Block_byrefs hold objects their destroy helper routines call this entry point
+// to help dispose of the contents
+// Used initially only for __attribute__((NSObject)) marked pointers.
+void _Block_object_dispose(const void *object, const int flags) {
+    //printf("_Block_object_dispose(%p, %x)\n", object, flags);
+    if (flags & BLOCK_FIELD_IS_BYREF)  {
+        // get rid of the __block data structure held in a Block
+        _Block_byref_release(object);
+    }
+    else if ((flags & (BLOCK_FIELD_IS_BLOCK|BLOCK_BYREF_CALLER)) == BLOCK_FIELD_IS_BLOCK) {
+        // get rid of a referenced Block held by this Block
+        // (ignore __block Block variables, compiler doesn't need to call us)
+        _Block_destroy(object);
+    }
+    else if ((flags & (BLOCK_FIELD_IS_WEAK|BLOCK_FIELD_IS_BLOCK|BLOCK_BYREF_CALLER)) == BLOCK_FIELD_IS_OBJECT) {
+        // get rid of a referenced object held by this Block
+        // (ignore __block object variables, compiler doesn't need to call us)
+        _Block_release_object(object);
+    }
+}
+
+
+/*
+ * Debugging support:
+ */
+#if 0
+#pragma mark Debugging
+#endif /* if 0 */
+
+
+const char *_Block_dump(const void *block) {
+    struct Block_layout *closure = (struct Block_layout *)block;
+    static char buffer[512];
+    char *cp = buffer;
+    if (closure == NULL) {
+        sprintf(cp, "NULL passed to _Block_dump\n");
+        return buffer;
+    }
+    if (! (closure->flags & BLOCK_HAS_DESCRIPTOR)) {
+        printf("Block compiled by obsolete compiler, please recompile source for this Block\n");
+        exit(1);
+    }
+    cp += sprintf(cp, "^%p (new layout) =\n", (void *)closure);
+    if (closure->isa == NULL) {
+        cp += sprintf(cp, "isa: NULL\n");
+    }
+    else if (closure->isa == _NSConcreteStackBlock) {
+        cp += sprintf(cp, "isa: stack Block\n");
+    }
+    else if (closure->isa == _NSConcreteMallocBlock) {
+        cp += sprintf(cp, "isa: malloc heap Block\n");
+    }
+    else if (closure->isa == _NSConcreteAutoBlock) {
+        cp += sprintf(cp, "isa: GC heap Block\n");
+    }
+    else if (closure->isa == _NSConcreteGlobalBlock) {
+        cp += sprintf(cp, "isa: global Block\n");
+    }
+    else if (closure->isa == _NSConcreteFinalizingBlock) {
+        cp += sprintf(cp, "isa: finalizing Block\n");
+    }
+    else {
+        cp += sprintf(cp, "isa?: %p\n", (void *)closure->isa);
+    }
+    cp += sprintf(cp, "flags:");
+    if (closure->flags & BLOCK_HAS_DESCRIPTOR) {
+        cp += sprintf(cp, " HASDESCRIPTOR");
+    }
+    if (closure->flags & BLOCK_NEEDS_FREE) {
+        cp += sprintf(cp, " FREEME");
+    }
+    if (closure->flags & BLOCK_IS_GC) {
+        cp += sprintf(cp, " ISGC");
+    }
+    if (closure->flags & BLOCK_HAS_COPY_DISPOSE) {
+        cp += sprintf(cp, " HASHELP");
+    }
+    if (closure->flags & BLOCK_HAS_CTOR) {
+        cp += sprintf(cp, " HASCTOR");
+    }
+    cp += sprintf(cp, "\nrefcount: %u\n", closure->flags & BLOCK_REFCOUNT_MASK);
+    cp += sprintf(cp, "invoke: %p\n", (void *)(uintptr_t)closure->invoke);
+    {
+        struct Block_descriptor *dp = closure->descriptor;
+        cp += sprintf(cp, "descriptor: %p\n", (void *)dp);
+        cp += sprintf(cp, "descriptor->reserved: %lu\n", dp->reserved);
+        cp += sprintf(cp, "descriptor->size: %lu\n", dp->size);
+
+        if (closure->flags & BLOCK_HAS_COPY_DISPOSE) {
+            cp += sprintf(cp, "descriptor->copy helper: %p\n", (void *)(uintptr_t)dp->copy);
+            cp += sprintf(cp, "descriptor->dispose helper: %p\n", (void *)(uintptr_t)dp->dispose);
+        }
+    }
+    return buffer;
+}
+
+
+const char *_Block_byref_dump(struct Block_byref *src) {
+    static char buffer[256];
+    char *cp = buffer;
+    cp += sprintf(cp, "byref data block %p contents:\n", (void *)src);
+    cp += sprintf(cp, "  forwarding: %p\n", (void *)src->forwarding);
+    cp += sprintf(cp, "  flags: 0x%x\n", src->flags);
+    cp += sprintf(cp, "  size: %d\n", src->size);
+    if (src->flags & BLOCK_HAS_COPY_DISPOSE) {
+        cp += sprintf(cp, "  copy helper: %p\n", (void *)(uintptr_t)src->byref_keep);
+        cp += sprintf(cp, "  dispose helper: %p\n", (void *)(uintptr_t)src->byref_destroy);
+    }
+    return buffer;
+}
+

contrib/compiler-rt/CREDITS.TXT

@@ -0,0 +1,21 @@
+This file is a partial list of people who have contributed to the LLVM/CompilerRT
+project.  If you have contributed a patch or made some other contribution to
+LLVM/CompilerRT, please submit a patch to this file to add yourself, and it will be
+done!
+
+The list is sorted by surname and formatted to allow easy grepping and
+beautification by scripts.  The fields are: name (N), email (E), web-address
+(W), PGP key ID and fingerprint (P), description (D), and snail-mail address
+(S).
+
+N: Craig van Vliet
+E: cvanvliet@auroraux.org
+W: http://www.auroraux.org
+D: Code style and Readability fixes.
+
+N: Edward O'Callaghan
+E: eocallaghan@auroraux.org
+W: http://www.auroraux.org
+D: CMake'ify Compiler-RT build system
+D: Maintain Solaris & AuroraUX ports of Compiler-RT
+

contrib/compiler-rt/LICENSE.TXT

@@ -0,0 +1,63 @@
+==============================================================================
+LLVM Release License
+==============================================================================
+University of Illinois/NCSA
+Open Source License
+
+Copyright (c) 2003-2009 University of Illinois at Urbana-Champaign.
+All rights reserved.
+
+Developed by:
+
+    LLVM Team
+
+    University of Illinois at Urbana-Champaign
+
+    http://llvm.org
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal with
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+of the Software, and to permit persons to whom the Software is furnished to do
+so, subject to the following conditions:
+
+    * Redistributions of source code must retain the above copyright notice,
+      this list of conditions and the following disclaimers.
+
+    * Redistributions in binary form must reproduce the above copyright notice,
+      this list of conditions and the following disclaimers in the
+      documentation and/or other materials provided with the distribution.
+
+    * Neither the names of the LLVM Team, University of Illinois at
+      Urbana-Champaign, nor the names of its contributors may be used to
+      endorse or promote products derived from this Software without specific
+      prior written permission.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
+FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS WITH THE
+SOFTWARE.
+
+==============================================================================
+Copyrights and Licenses for Third Party Software Distributed with LLVM:
+==============================================================================
+The LLVM software contains code written by third parties.  Such software will
+have its own individual LICENSE.TXT file in the directory in which it appears.
+This file will describe the copyrights, license, and restrictions which apply
+to that code.
+
+The disclaimer of warranty in the University of Illinois Open Source License
+applies to all code in the LLVM Distribution, and nothing in any of the
+other licenses gives permission to use the names of the LLVM Team or the
+University of Illinois to endorse or promote products derived from this
+Software.
+
+The following pieces of software have additional or alternate copyrights,
+licenses, and/or restrictions:
+
+Program             Directory
+-------             ---------

contrib/compiler-rt/README.txt

@@ -0,0 +1,330 @@
+Compiler-RT
+================================
+
+This directory and its subdirectories contain source code for the compiler
+support routines.
+
+Compiler-RT is open source software. You may freely distribute it under the
+terms of the license agreement found in LICENSE.txt.
+
+================================
+
+This is a replacement library for libgcc.  Each function is contained
+in its own file.  Each function has a corresponding unit test under
+test/Unit.
+
+A rudimentary script to test each file is in the file called
+test/Unit/test.
+
+Here is the specification for this library:
+
+http://gcc.gnu.org/onlinedocs/gccint/Libgcc.html#Libgcc
+
+Here is a synopsis of the contents of this library:
+
+typedef      int si_int;
+typedef unsigned su_int;
+
+typedef          long long di_int;
+typedef unsigned long long du_int;
+
+// Integral bit manipulation
+
+di_int __ashldi3(di_int a, si_int b);      // a << b
+ti_int __ashlti3(ti_int a, si_int b);      // a << b
+
+di_int __ashrdi3(di_int a, si_int b);      // a >> b  arithmetic (sign fill)
+ti_int __ashrti3(ti_int a, si_int b);      // a >> b  arithmetic (sign fill)
+di_int __lshrdi3(di_int a, si_int b);      // a >> b  logical    (zero fill)
+ti_int __lshrti3(ti_int a, si_int b);      // a >> b  logical    (zero fill)
+
+si_int __clzsi2(si_int a);  // count leading zeros
+si_int __clzdi2(di_int a);  // count leading zeros
+si_int __clzti2(ti_int a);  // count leading zeros
+si_int __ctzsi2(si_int a);  // count trailing zeros
+si_int __ctzdi2(di_int a);  // count trailing zeros
+si_int __ctzti2(ti_int a);  // count trailing zeros
+
+si_int __ffsdi2(di_int a);  // find least significant 1 bit
+si_int __ffsti2(ti_int a);  // find least significant 1 bit
+
+si_int __paritysi2(si_int a);  // bit parity
+si_int __paritydi2(di_int a);  // bit parity
+si_int __parityti2(ti_int a);  // bit parity
+
+si_int __popcountsi2(si_int a);  // bit population
+si_int __popcountdi2(di_int a);  // bit population
+si_int __popcountti2(ti_int a);  // bit population
+
+uint32_t __bswapsi2(uint32_t a);   // a byteswapped, arm only
+uint64_t __bswapdi2(uint64_t a);   // a byteswapped, arm only
+
+// Integral arithmetic
+
+di_int __negdi2    (di_int a);                         // -a
+ti_int __negti2    (ti_int a);                         // -a
+di_int __muldi3    (di_int a, di_int b);               // a * b
+ti_int __multi3    (ti_int a, ti_int b);               // a * b
+si_int __divsi3    (si_int a, si_int b);               // a / b   signed
+di_int __divdi3    (di_int a, di_int b);               // a / b   signed
+ti_int __divti3    (ti_int a, ti_int b);               // a / b   signed
+su_int __udivsi3   (su_int n, su_int d);               // a / b   unsigned
+du_int __udivdi3   (du_int a, du_int b);               // a / b   unsigned
+tu_int __udivti3   (tu_int a, tu_int b);               // a / b   unsigned
+si_int __modsi3    (si_int a, si_int b);               // a % b   signed
+di_int __moddi3    (di_int a, di_int b);               // a % b   signed
+ti_int __modti3    (ti_int a, ti_int b);               // a % b   signed
+su_int __umodsi3   (su_int a, su_int b);               // a % b   unsigned
+du_int __umoddi3   (du_int a, du_int b);               // a % b   unsigned
+tu_int __umodti3   (tu_int a, tu_int b);               // a % b   unsigned
+du_int __udivmoddi4(du_int a, du_int b, du_int* rem);  // a / b, *rem = a % b
+tu_int __udivmodti4(tu_int a, tu_int b, tu_int* rem);  // a / b, *rem = a % b
+
+//  Integral arithmetic with trapping overflow
+
+si_int __absvsi2(si_int a);           // abs(a)
+di_int __absvdi2(di_int a);           // abs(a)
+ti_int __absvti2(ti_int a);           // abs(a)
+
+si_int __negvsi2(si_int a);           // -a
+di_int __negvdi2(di_int a);           // -a
+ti_int __negvti2(ti_int a);           // -a
+
+si_int __addvsi3(si_int a, si_int b);  // a + b
+di_int __addvdi3(di_int a, di_int b);  // a + b
+ti_int __addvti3(ti_int a, ti_int b);  // a + b
+
+si_int __subvsi3(si_int a, si_int b);  // a - b
+di_int __subvdi3(di_int a, di_int b);  // a - b
+ti_int __subvti3(ti_int a, ti_int b);  // a - b
+
+si_int __mulvsi3(si_int a, si_int b);  // a * b
+di_int __mulvdi3(di_int a, di_int b);  // a * b
+ti_int __mulvti3(ti_int a, ti_int b);  // a * b
+
+//  Integral comparison: a  < b -> 0
+//                       a == b -> 1
+//                       a  > b -> 2
+
+si_int __cmpdi2 (di_int a, di_int b);
+si_int __cmpti2 (ti_int a, ti_int b);
+si_int __ucmpdi2(du_int a, du_int b);
+si_int __ucmpti2(tu_int a, tu_int b);
+
+//  Integral / floating point conversion
+
+di_int __fixsfdi(      float a);
+di_int __fixdfdi(     double a);
+di_int __fixxfdi(long double a);
+
+ti_int __fixsfti(      float a);
+ti_int __fixdfti(     double a);
+ti_int __fixxfti(long double a);
+uint64_t __fixtfdi(long double input);  // ppc only, doesn't match documentation
+
+su_int __fixunssfsi(      float a);
+su_int __fixunsdfsi(     double a);
+su_int __fixunsxfsi(long double a);
+
+du_int __fixunssfdi(      float a);
+du_int __fixunsdfdi(     double a);
+du_int __fixunsxfdi(long double a);
+
+tu_int __fixunssfti(      float a);
+tu_int __fixunsdfti(     double a);
+tu_int __fixunsxfti(long double a);
+uint64_t __fixunstfdi(long double input);  // ppc only
+
+float       __floatdisf(di_int a);
+double      __floatdidf(di_int a);
+long double __floatdixf(di_int a);
+long double __floatditf(int64_t a);        // ppc only
+
+float       __floattisf(ti_int a);
+double      __floattidf(ti_int a);
+long double __floattixf(ti_int a);
+
+float       __floatundisf(du_int a);
+double      __floatundidf(du_int a);
+long double __floatundixf(du_int a);
+long double __floatunditf(uint64_t a);     // ppc only
+
+float       __floatuntisf(tu_int a);
+double      __floatuntidf(tu_int a);
+long double __floatuntixf(tu_int a);
+
+//  Floating point raised to integer power
+
+float       __powisf2(      float a, si_int b);  // a ^ b
+double      __powidf2(     double a, si_int b);  // a ^ b
+long double __powixf2(long double a, si_int b);  // a ^ b
+long double __powitf2(long double a, si_int b);  // ppc only, a ^ b
+
+//  Complex arithmetic
+
+//  (a + ib) * (c + id)
+
+      float _Complex __mulsc3( float a,  float b,  float c,  float d);
+     double _Complex __muldc3(double a, double b, double c, double d);
+long double _Complex __mulxc3(long double a, long double b,
+                              long double c, long double d);
+long double _Complex __multc3(long double a, long double b,
+                              long double c, long double d); // ppc only
+
+//  (a + ib) / (c + id)
+
+      float _Complex __divsc3( float a,  float b,  float c,  float d);
+     double _Complex __divdc3(double a, double b, double c, double d);
+long double _Complex __divxc3(long double a, long double b,
+                              long double c, long double d);
+long double _Complex __divtc3(long double a, long double b,
+                              long double c, long double d);  // ppc only
+
+
+//         Runtime support
+
+// __clear_cache() is used to tell process that new instructions have been
+// written to an address range.  Necessary on processors that do not have
+// a unified instuction and data cache.
+void __clear_cache(void* start, void* end);
+
+// __enable_execute_stack() is used with nested functions when a trampoline
+// function is written onto the stack and that page range needs to be made
+// executable.
+void __enable_execute_stack(void* addr);
+
+// __gcc_personality_v0() is normally only called by the system unwinder.
+// C code (as opposed to C++) normally does not need a personality function
+// because there are no catch clauses or destructors to be run.  But there
+// is a C language extension __attribute__((cleanup(func))) which marks local
+// variables as needing the cleanup function "func" to be run when the
+// variable goes out of scope.  That includes when an exception is thrown,
+// so a personality handler is needed.  
+_Unwind_Reason_Code __gcc_personality_v0(int version, _Unwind_Action actions,
+         uint64_t exceptionClass, struct _Unwind_Exception* exceptionObject,
+         _Unwind_Context_t context);
+
+// for use with some implementations of assert() in <assert.h>
+void __eprintf(const char* format, const char* assertion_expression,
+				const char* line, const char* file);
+				
+
+
+//   Power PC specific functions
+
+// There is no C interface to the saveFP/restFP functions.  They are helper
+// functions called by the prolog and epilog of functions that need to save
+// a number of non-volatile float point registers.  
+saveFP
+restFP
+
+// PowerPC has a standard template for trampoline functions.  This function
+// generates a custom trampoline function with the specific realFunc
+// and localsPtr values.
+void __trampoline_setup(uint32_t* trampOnStack, int trampSizeAllocated, 
+                                const void* realFunc, void* localsPtr);
+
+// adds two 128-bit double-double precision values ( x + y )
+long double __gcc_qadd(long double x, long double y);  
+
+// subtracts two 128-bit double-double precision values ( x - y )
+long double __gcc_qsub(long double x, long double y); 
+
+// multiples two 128-bit double-double precision values ( x * y )
+long double __gcc_qmul(long double x, long double y);  
+
+// divides two 128-bit double-double precision values ( x / y )
+long double __gcc_qdiv(long double a, long double b);  
+
+
+//    ARM specific functions
+
+// There is no C interface to the switch* functions.  These helper functions
+// are only needed by Thumb1 code for efficient switch table generation.
+switch16
+switch32
+switch8
+switchu8
+
+// There is no C interface to the *_vfp_d8_d15_regs functions.  There are
+// called in the prolog and epilog of Thumb1 functions.  When the C++ ABI use
+// SJLJ for exceptions, each function with a catch clause or destuctors needs
+// to save and restore all registers in it prolog and epliog.  But there is 
+// no way to access vector and high float registers from thumb1 code, so the 
+// compiler must add call outs to these helper functions in the prolog and 
+// epilog.
+restore_vfp_d8_d15_regs
+save_vfp_d8_d15_regs
+
+
+// Note: long ago ARM processors did not have floating point hardware support.
+// Floating point was done in software and floating point parameters were 
+// passed in integer registers.  When hardware support was added for floating
+// point, new *vfp functions were added to do the same operations but with 
+// floating point parameters in floating point registers.
+
+// Undocumented functions
+
+float  __addsf3vfp(float a, float b);   // Appears to return a + b
+double __adddf3vfp(double a, double b); // Appears to return a + b
+float  __divsf3vfp(float a, float b);   // Appears to return a / b
+double __divdf3vfp(double a, double b); // Appears to return a / b
+int    __eqsf2vfp(float a, float b);    // Appears to return  one
+                                        //     iff a == b and neither is NaN.
+int    __eqdf2vfp(double a, double b);  // Appears to return  one
+                                        //     iff a == b and neither is NaN.
+double __extendsfdf2vfp(float a);       // Appears to convert from
+                                        //     float to double.
+int    __fixdfsivfp(double a);          // Appears to convert from
+                                        //     double to int.
+int    __fixsfsivfp(float a);           // Appears to convert from
+                                        //     float to int.
+unsigned int __fixunssfsivfp(float a);  // Appears to convert from
+                                        //     float to unsigned int.
+unsigned int __fixunsdfsivfp(double a); // Appears to convert from
+                                        //     double to unsigned int.
+double __floatsidfvfp(int a);           // Appears to convert from
+                                        //     int to double.
+float __floatsisfvfp(int a);            // Appears to convert from
+                                        //     int to float.
+double __floatunssidfvfp(unsigned int a); // Appears to convert from
+                                        //     unisgned int to double.
+float __floatunssisfvfp(unsigned int a); // Appears to convert from
+                                        //     unisgned int to float.
+int __gedf2vfp(double a, double b);     // Appears to return __gedf2
+                                        //     (a >= b)
+int __gesf2vfp(float a, float b);       // Appears to return __gesf2
+                                        //     (a >= b)
+int __gtdf2vfp(double a, double b);     // Appears to return __gtdf2
+                                        //     (a > b)
+int __gtsf2vfp(float a, float b);       // Appears to return __gtsf2
+                                        //     (a > b)
+int __ledf2vfp(double a, double b);     // Appears to return __ledf2
+                                        //     (a <= b)
+int __lesf2vfp(float a, float b);       // Appears to return __lesf2
+                                        //     (a <= b)
+int __ltdf2vfp(double a, double b);     // Appears to return __ltdf2
+                                        //     (a < b)
+int __ltsf2vfp(float a, float b);       // Appears to return __ltsf2
+                                        //     (a < b)
+double __muldf3vfp(double a, double b); // Appears to return a * b
+float __mulsf3vfp(float a, float b);    // Appears to return a * b
+int __nedf2vfp(double a, double b);     // Appears to return __nedf2
+                                        //     (a != b)
+double __negdf2vfp(double a);           // Appears to return -a
+float __negsf2vfp(float a);             // Appears to return -a
+float __negsf2vfp(float a);             // Appears to return -a
+double __subdf3vfp(double a, double b); // Appears to return a - b
+float __subsf3vfp(float a, float b);    // Appears to return a - b
+float __truncdfsf2vfp(double a);        // Appears to convert from
+                                        //     double to float.
+int __unorddf2vfp(double a, double b);  // Appears to return __unorddf2
+int __unordsf2vfp(float a, float b);    // Appears to return __unordsf2
+
+
+Preconditions are listed for each function at the definition when there are any.
+Any preconditions reflect the specification at
+http://gcc.gnu.org/onlinedocs/gccint/Libgcc.html#Libgcc.
+
+Assumptions are listed in "int_lib.h", and in individual files.  Where possible
+assumptions are checked at compile time.

contrib/compiler-rt/lib/absvdi2.c

@@ -0,0 +1,30 @@
+/*===-- absvdi2.c - Implement __absvdi2 -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ *===----------------------------------------------------------------------===
+ *
+ * This file implements __absvdi2 for the compiler_rt library.
+ *
+ *===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+#include <stdlib.h>
+
+/* Returns: absolute value */
+
+/* Effects: aborts if abs(x) < 0 */
+
+di_int
+__absvdi2(di_int a)
+{
+    const int N = (int)(sizeof(di_int) * CHAR_BIT);
+    if (a == ((di_int)1 << (N-1)))
+        compilerrt_abort();
+    const di_int t = a >> (N - 1);
+    return (a ^ t) - t;
+}

contrib/compiler-rt/lib/absvsi2.c

@@ -0,0 +1,30 @@
+/* ===-- absvsi2.c - Implement __absvsi2 -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __absvsi2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */ 
+
+#include "int_lib.h"
+#include <stdlib.h>
+
+/* Returns: absolute value */
+
+/* Effects: aborts if abs(x) < 0 */
+
+si_int
+__absvsi2(si_int a)
+{
+    const int N = (int)(sizeof(si_int) * CHAR_BIT);
+    if (a == (1 << (N-1)))
+        compilerrt_abort();
+    const si_int t = a >> (N - 1);
+    return (a ^ t) - t;
+}

contrib/compiler-rt/lib/absvti2.c

@@ -0,0 +1,34 @@
+/* ===-- absvti2.c - Implement __absvdi2 -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __absvti2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+#include <stdlib.h>
+
+/* Returns: absolute value */
+
+/* Effects: aborts if abs(x) < 0 */
+
+ti_int
+__absvti2(ti_int a)
+{
+    const int N = (int)(sizeof(ti_int) * CHAR_BIT);
+    if (a == ((ti_int)1 << (N-1)))
+        compilerrt_abort();
+    const ti_int s = a >> (N - 1);
+    return (a ^ s) - s;
+}
+
+#endif

contrib/compiler-rt/lib/adddf3.c

@@ -0,0 +1,154 @@
+//===-- lib/adddf3.c - Double-precision addition and subtraction --*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements double-precision soft-float addition and subtraction
+// with the IEEE-754 default rounding (to nearest, ties to even).
+//
+//===----------------------------------------------------------------------===//
+
+#define DOUBLE_PRECISION
+#include "fp_lib.h"
+
+fp_t __adddf3(fp_t a, fp_t b) {
+    
+    rep_t aRep = toRep(a);
+    rep_t bRep = toRep(b);
+    const rep_t aAbs = aRep & absMask;
+    const rep_t bAbs = bRep & absMask;
+    
+    // Detect if a or b is zero, infinity, or NaN.
+    if (aAbs - 1U >= infRep - 1U || bAbs - 1U >= infRep - 1U) {
+        
+        // NaN + anything = qNaN
+        if (aAbs > infRep) return fromRep(toRep(a) | quietBit);
+        // anything + NaN = qNaN
+        if (bAbs > infRep) return fromRep(toRep(b) | quietBit);
+        
+        if (aAbs == infRep) {
+            // +/-infinity + -/+infinity = qNaN
+            if ((toRep(a) ^ toRep(b)) == signBit) return fromRep(qnanRep);
+            // +/-infinity + anything remaining = +/- infinity
+            else return a;
+        }
+        
+        // anything remaining + +/-infinity = +/-infinity
+        if (bAbs == infRep) return b;
+        
+        // zero + anything = anything
+        if (!aAbs) {
+            // but we need to get the sign right for zero + zero
+            if (!bAbs) return fromRep(toRep(a) & toRep(b));
+            else return b;
+        }
+        
+        // anything + zero = anything
+        if (!bAbs) return a;
+    }
+    
+    // Swap a and b if necessary so that a has the larger absolute value.
+    if (bAbs > aAbs) {
+        const rep_t temp = aRep;
+        aRep = bRep;
+        bRep = temp;
+    }
+    
+    // Extract the exponent and significand from the (possibly swapped) a and b.
+    int aExponent = aRep >> significandBits & maxExponent;
+    int bExponent = bRep >> significandBits & maxExponent;
+    rep_t aSignificand = aRep & significandMask;
+    rep_t bSignificand = bRep & significandMask;
+    
+    // Normalize any denormals, and adjust the exponent accordingly.
+    if (aExponent == 0) aExponent = normalize(&aSignificand);
+    if (bExponent == 0) bExponent = normalize(&bSignificand);
+    
+    // The sign of the result is the sign of the larger operand, a.  If they
+    // have opposite signs, we are performing a subtraction; otherwise addition.
+    const rep_t resultSign = aRep & signBit;
+    const bool subtraction = (aRep ^ bRep) & signBit;
+    
+    // Shift the significands to give us round, guard and sticky, and or in the
+    // implicit significand bit.  (If we fell through from the denormal path it
+    // was already set by normalize( ), but setting it twice won't hurt
+    // anything.)
+    aSignificand = (aSignificand | implicitBit) << 3;
+    bSignificand = (bSignificand | implicitBit) << 3;
+    
+    // Shift the significand of b by the difference in exponents, with a sticky
+    // bottom bit to get rounding correct.
+    const int align = aExponent - bExponent;
+    if (align) {
+        if (align < typeWidth) {
+            const bool sticky = bSignificand << (typeWidth - align);
+            bSignificand = bSignificand >> align | sticky;
+        } else {
+            bSignificand = 1; // sticky; b is known to be non-zero.
+        }
+    }
+    
+    if (subtraction) {
+        aSignificand -= bSignificand;
+        
+        // If a == -b, return +zero.
+        if (aSignificand == 0) return fromRep(0);
+        
+        // If partial cancellation occured, we need to left-shift the result
+        // and adjust the exponent:
+        if (aSignificand < implicitBit << 3) {
+            const int shift = rep_clz(aSignificand) - rep_clz(implicitBit << 3);
+            aSignificand <<= shift;
+            aExponent -= shift;
+        }
+    }
+    
+    else /* addition */ {
+        aSignificand += bSignificand;
+        
+        // If the addition carried up, we need to right-shift the result and
+        // adjust the exponent:
+        if (aSignificand & implicitBit << 4) {
+            const bool sticky = aSignificand & 1;
+            aSignificand = aSignificand >> 1 | sticky;
+            aExponent += 1;
+        }
+    }
+    
+    // If we have overflowed the type, return +/- infinity:
+    if (aExponent >= maxExponent) return fromRep(infRep | resultSign);
+    
+    if (aExponent <= 0) {
+        // Result is denormal before rounding; the exponent is zero and we
+        // need to shift the significand.
+        const int shift = 1 - aExponent;
+        const bool sticky = aSignificand << (typeWidth - shift);
+        aSignificand = aSignificand >> shift | sticky;
+        aExponent = 0;
+    }
+    
+    // Low three bits are round, guard, and sticky.
+    const int roundGuardSticky = aSignificand & 0x7;
+    
+    // Shift the significand into place, and mask off the implicit bit.
+    rep_t result = aSignificand >> 3 & significandMask;
+    
+    // Insert the exponent and sign.
+    result |= (rep_t)aExponent << significandBits;
+    result |= resultSign;
+    
+    // Final rounding.  The result may overflow to infinity, but that is the
+    // correct result in that case.
+    if (roundGuardSticky > 0x4) result++;
+    if (roundGuardSticky == 0x4) result += result & 1;
+    return fromRep(result);
+}
+
+// Subtraction; flip the sign bit of b and add.
+fp_t __subdf3(fp_t a, fp_t b) {
+    return __adddf3(a, fromRep(toRep(b) ^ signBit));
+}

contrib/compiler-rt/lib/addsf3.c

@@ -0,0 +1,164 @@
+//===-- lib/addsf3.c - Single-precision addition and subtraction --*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements single-precision soft-float addition and subtraction
+// with the IEEE-754 default rounding (to nearest, ties to even).
+//
+//===----------------------------------------------------------------------===//
+
+#define SINGLE_PRECISION
+#include "fp_lib.h"
+
+fp_t __addsf3(fp_t a, fp_t b) {
+
+    rep_t aRep = toRep(a);
+    rep_t bRep = toRep(b);
+    const rep_t aAbs = aRep & absMask;
+    const rep_t bAbs = bRep & absMask;
+    
+    // Detect if a or b is zero, infinity, or NaN.
+    if (aAbs - 1U >= infRep - 1U || bAbs - 1U >= infRep - 1U) {
+        
+        // NaN + anything = qNaN
+        if (aAbs > infRep) return fromRep(toRep(a) | quietBit);
+        // anything + NaN = qNaN
+        if (bAbs > infRep) return fromRep(toRep(b) | quietBit);
+        
+        if (aAbs == infRep) {
+            // +/-infinity + -/+infinity = qNaN
+            if ((toRep(a) ^ toRep(b)) == signBit) return fromRep(qnanRep);
+            // +/-infinity + anything remaining = +/- infinity
+            else return a;
+        }
+        
+        // anything remaining + +/-infinity = +/-infinity
+        if (bAbs == infRep) return b;
+        
+        // zero + anything = anything
+        if (!aAbs) {
+            // but we need to get the sign right for zero + zero
+            if (!bAbs) return fromRep(toRep(a) & toRep(b));
+            else return b;
+        }
+        
+        // anything + zero = anything
+        if (!bAbs) return a;
+    }
+    
+    // Swap a and b if necessary so that a has the larger absolute value.
+    if (bAbs > aAbs) {
+        const rep_t temp = aRep;
+        aRep = bRep;
+        bRep = temp;
+    }
+    
+    // Extract the exponent and significand from the (possibly swapped) a and b.
+    int aExponent = aRep >> significandBits & maxExponent;
+    int bExponent = bRep >> significandBits & maxExponent;
+    rep_t aSignificand = aRep & significandMask;
+    rep_t bSignificand = bRep & significandMask;
+    
+    // Normalize any denormals, and adjust the exponent accordingly.
+    if (aExponent == 0) aExponent = normalize(&aSignificand);
+    if (bExponent == 0) bExponent = normalize(&bSignificand);
+    
+    // The sign of the result is the sign of the larger operand, a.  If they
+    // have opposite signs, we are performing a subtraction; otherwise addition.
+    const rep_t resultSign = aRep & signBit;
+    const bool subtraction = (aRep ^ bRep) & signBit;
+    
+    // Shift the significands to give us round, guard and sticky, and or in the
+    // implicit significand bit.  (If we fell through from the denormal path it
+    // was already set by normalize( ), but setting it twice won't hurt
+    // anything.)
+    aSignificand = (aSignificand | implicitBit) << 3;
+    bSignificand = (bSignificand | implicitBit) << 3;
+    
+    // Shift the significand of b by the difference in exponents, with a sticky
+    // bottom bit to get rounding correct.
+    const int align = aExponent - bExponent;
+    if (align) {
+        if (align < typeWidth) {
+            const bool sticky = bSignificand << (typeWidth - align);
+            bSignificand = bSignificand >> align | sticky;
+        } else {
+            bSignificand = 1; // sticky; b is known to be non-zero.
+        }
+    }
+    
+    if (subtraction) {
+        aSignificand -= bSignificand;
+        
+        // If a == -b, return +zero.
+        if (aSignificand == 0) return fromRep(0);
+        
+        // If partial cancellation occured, we need to left-shift the result
+        // and adjust the exponent:
+        if (aSignificand < implicitBit << 3) {
+            const int shift = rep_clz(aSignificand) - rep_clz(implicitBit << 3);
+            aSignificand <<= shift;
+            aExponent -= shift;
+        }
+    }
+    
+    else /* addition */ {
+        aSignificand += bSignificand;
+        
+        // If the addition carried up, we need to right-shift the result and
+        // adjust the exponent:
+        if (aSignificand & implicitBit << 4) {
+            const bool sticky = aSignificand & 1;
+            aSignificand = aSignificand >> 1 | sticky;
+            aExponent += 1;
+        }
+    }
+    
+    // If we have overflowed the type, return +/- infinity:
+    if (aExponent >= maxExponent) return fromRep(infRep | resultSign);
+    
+    if (aExponent <= 0) {
+        // Result is denormal before rounding; the exponent is zero and we
+        // need to shift the significand.
+        const int shift = 1 - aExponent;
+        const bool sticky = aSignificand << (typeWidth - shift);
+        aSignificand = aSignificand >> shift | sticky;
+        aExponent = 0;
+    }
+    
+    // Low three bits are round, guard, and sticky.
+    const int roundGuardSticky = aSignificand & 0x7;
+    
+    // Shift the significand into place, and mask off the implicit bit.
+    rep_t result = aSignificand >> 3 & significandMask;
+    
+    // Insert the exponent and sign.
+    result |= (rep_t)aExponent << significandBits;
+    result |= resultSign;
+    
+    // Final rounding.  The result may overflow to infinity, but that is the
+    // correct result in that case.
+    if (roundGuardSticky > 0x4) result++;
+    if (roundGuardSticky == 0x4) result += result & 1;
+    return fromRep(result);
+}
+
+// Subtraction; flip the sign bit of b and add.
+fp_t __subsf3(fp_t a, fp_t b) {
+    return __addsf3(a, fromRep(toRep(b) ^ signBit));
+}
+
+
+
+
+
+
+
+
+
+

contrib/compiler-rt/lib/addvdi3.c

@@ -0,0 +1,37 @@
+/* ===-- addvdi3.c - Implement __addvdi3 -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __addvdi3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+#include <stdlib.h>
+
+/* Returns: a + b */
+
+/* Effects: aborts if a + b overflows */
+
+di_int
+__addvdi3(di_int a, di_int b)
+{
+    di_int s = a + b;
+    if (b >= 0)
+    {
+        if (s < a)
+            compilerrt_abort();
+    }
+    else
+    {
+        if (s >= a)
+            compilerrt_abort();
+    }
+    return s;
+}

contrib/compiler-rt/lib/addvsi3.c

@@ -0,0 +1,37 @@
+/* ===-- addvsi3.c - Implement __addvsi3 -----------------------------------===
+ *
+ *                    The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __addvsi3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+#include <stdlib.h>
+
+/* Returns: a + b */
+
+/* Effects: aborts if a + b overflows */
+
+si_int
+__addvsi3(si_int a, si_int b)
+{
+    si_int s = a + b;
+    if (b >= 0)
+    {
+        if (s < a)
+            compilerrt_abort();
+    }
+    else
+    {
+        if (s >= a)
+            compilerrt_abort();
+    }
+    return s;
+}

contrib/compiler-rt/lib/addvti3.c

@@ -0,0 +1,41 @@
+/* ===-- addvti3.c - Implement __addvti3 -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __addvti3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+#include <stdlib.h>
+
+/* Returns: a + b */
+
+/* Effects: aborts if a + b overflows */
+
+ti_int
+__addvti3(ti_int a, ti_int b)
+{
+    ti_int s = a + b;
+    if (b >= 0)
+    {
+        if (s < a)
+            compilerrt_abort();
+    }
+    else
+    {
+        if (s >= a)
+            compilerrt_abort();
+    }
+    return s;
+}
+
+#endif

contrib/compiler-rt/lib/apple_versioning.c

@@ -0,0 +1,150 @@
+/* ===-- apple_versioning.c - Adds versioning symbols for ld ---------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+
+
+#if __APPLE__
+  #if __arm__
+    #define NOT_HERE_BEFORE_10_6(sym) 
+  #elif __ppc__
+    #define NOT_HERE_BEFORE_10_6(sym) \
+        extern const char sym##_tmp3 __asm("$ld$hide$os10.3$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp3 = 0; \
+         extern const char sym##_tmp4 __asm("$ld$hide$os10.4$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp4 = 0; \
+        extern const char sym##_tmp5 __asm("$ld$hide$os10.5$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp5 = 0; 
+  #else
+    #define NOT_HERE_BEFORE_10_6(sym) \
+         extern const char sym##_tmp4 __asm("$ld$hide$os10.4$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp4 = 0; \
+        extern const char sym##_tmp5 __asm("$ld$hide$os10.5$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp5 = 0; 
+  #endif /* __ppc__ */
+
+
+/* Symbols in libSystem.dylib in 10.6 and later, 
+ *  but are in libgcc_s.dylib in earlier versions
+ */
+
+NOT_HERE_BEFORE_10_6(__absvdi2)
+NOT_HERE_BEFORE_10_6(__absvsi2)
+NOT_HERE_BEFORE_10_6(__absvti2)
+NOT_HERE_BEFORE_10_6(__addvdi3)
+NOT_HERE_BEFORE_10_6(__addvsi3)
+NOT_HERE_BEFORE_10_6(__addvti3)
+NOT_HERE_BEFORE_10_6(__ashldi3)
+NOT_HERE_BEFORE_10_6(__ashlti3)
+NOT_HERE_BEFORE_10_6(__ashrdi3)
+NOT_HERE_BEFORE_10_6(__ashrti3)
+NOT_HERE_BEFORE_10_6(__clear_cache)
+NOT_HERE_BEFORE_10_6(__clzdi2)
+NOT_HERE_BEFORE_10_6(__clzsi2)
+NOT_HERE_BEFORE_10_6(__clzti2)
+NOT_HERE_BEFORE_10_6(__cmpdi2)
+NOT_HERE_BEFORE_10_6(__cmpti2)
+NOT_HERE_BEFORE_10_6(__ctzdi2)
+NOT_HERE_BEFORE_10_6(__ctzsi2)
+NOT_HERE_BEFORE_10_6(__ctzti2)
+NOT_HERE_BEFORE_10_6(__divdc3)
+NOT_HERE_BEFORE_10_6(__divdi3)
+NOT_HERE_BEFORE_10_6(__divsc3)
+NOT_HERE_BEFORE_10_6(__divtc3)
+NOT_HERE_BEFORE_10_6(__divti3)
+NOT_HERE_BEFORE_10_6(__divxc3)
+NOT_HERE_BEFORE_10_6(__enable_execute_stack)
+NOT_HERE_BEFORE_10_6(__ffsdi2)
+NOT_HERE_BEFORE_10_6(__ffsti2)
+NOT_HERE_BEFORE_10_6(__fixdfdi)
+NOT_HERE_BEFORE_10_6(__fixdfti)
+NOT_HERE_BEFORE_10_6(__fixsfdi)
+NOT_HERE_BEFORE_10_6(__fixsfti)
+NOT_HERE_BEFORE_10_6(__fixtfdi)
+NOT_HERE_BEFORE_10_6(__fixunsdfdi)
+NOT_HERE_BEFORE_10_6(__fixunsdfsi)
+NOT_HERE_BEFORE_10_6(__fixunsdfti)
+NOT_HERE_BEFORE_10_6(__fixunssfdi)
+NOT_HERE_BEFORE_10_6(__fixunssfsi)
+NOT_HERE_BEFORE_10_6(__fixunssfti)
+NOT_HERE_BEFORE_10_6(__fixunstfdi)
+NOT_HERE_BEFORE_10_6(__fixunsxfdi)
+NOT_HERE_BEFORE_10_6(__fixunsxfsi)
+NOT_HERE_BEFORE_10_6(__fixunsxfti)
+NOT_HERE_BEFORE_10_6(__fixxfdi)
+NOT_HERE_BEFORE_10_6(__fixxfti)
+NOT_HERE_BEFORE_10_6(__floatdidf)
+NOT_HERE_BEFORE_10_6(__floatdisf)
+NOT_HERE_BEFORE_10_6(__floatditf)
+NOT_HERE_BEFORE_10_6(__floatdixf)
+NOT_HERE_BEFORE_10_6(__floattidf)
+NOT_HERE_BEFORE_10_6(__floattisf)
+NOT_HERE_BEFORE_10_6(__floattixf)
+NOT_HERE_BEFORE_10_6(__floatundidf)
+NOT_HERE_BEFORE_10_6(__floatundisf)
+NOT_HERE_BEFORE_10_6(__floatunditf)
+NOT_HERE_BEFORE_10_6(__floatundixf)
+NOT_HERE_BEFORE_10_6(__floatuntidf)
+NOT_HERE_BEFORE_10_6(__floatuntisf)
+NOT_HERE_BEFORE_10_6(__floatuntixf)
+NOT_HERE_BEFORE_10_6(__gcc_personality_v0)
+NOT_HERE_BEFORE_10_6(__lshrdi3)
+NOT_HERE_BEFORE_10_6(__lshrti3)
+NOT_HERE_BEFORE_10_6(__moddi3)
+NOT_HERE_BEFORE_10_6(__modti3)
+NOT_HERE_BEFORE_10_6(__muldc3)
+NOT_HERE_BEFORE_10_6(__muldi3)
+NOT_HERE_BEFORE_10_6(__mulsc3)
+NOT_HERE_BEFORE_10_6(__multc3)
+NOT_HERE_BEFORE_10_6(__multi3)
+NOT_HERE_BEFORE_10_6(__mulvdi3)
+NOT_HERE_BEFORE_10_6(__mulvsi3)
+NOT_HERE_BEFORE_10_6(__mulvti3)
+NOT_HERE_BEFORE_10_6(__mulxc3)
+NOT_HERE_BEFORE_10_6(__negdi2)
+NOT_HERE_BEFORE_10_6(__negti2)
+NOT_HERE_BEFORE_10_6(__negvdi2)
+NOT_HERE_BEFORE_10_6(__negvsi2)
+NOT_HERE_BEFORE_10_6(__negvti2)
+NOT_HERE_BEFORE_10_6(__paritydi2)
+NOT_HERE_BEFORE_10_6(__paritysi2)
+NOT_HERE_BEFORE_10_6(__parityti2)
+NOT_HERE_BEFORE_10_6(__popcountdi2)
+NOT_HERE_BEFORE_10_6(__popcountsi2)
+NOT_HERE_BEFORE_10_6(__popcountti2)
+NOT_HERE_BEFORE_10_6(__powidf2)
+NOT_HERE_BEFORE_10_6(__powisf2)
+NOT_HERE_BEFORE_10_6(__powitf2)
+NOT_HERE_BEFORE_10_6(__powixf2)
+NOT_HERE_BEFORE_10_6(__subvdi3)
+NOT_HERE_BEFORE_10_6(__subvsi3)
+NOT_HERE_BEFORE_10_6(__subvti3)
+NOT_HERE_BEFORE_10_6(__ucmpdi2)
+NOT_HERE_BEFORE_10_6(__ucmpti2)
+NOT_HERE_BEFORE_10_6(__udivdi3)
+NOT_HERE_BEFORE_10_6(__udivmoddi4)
+NOT_HERE_BEFORE_10_6(__udivmodti4)
+NOT_HERE_BEFORE_10_6(__udivti3)
+NOT_HERE_BEFORE_10_6(__umoddi3)
+NOT_HERE_BEFORE_10_6(__umodti3)
+
+
+#if __ppc__
+NOT_HERE_BEFORE_10_6(__gcc_qadd)
+NOT_HERE_BEFORE_10_6(__gcc_qdiv)
+NOT_HERE_BEFORE_10_6(__gcc_qmul)
+NOT_HERE_BEFORE_10_6(__gcc_qsub)
+NOT_HERE_BEFORE_10_6(__trampoline_setup)
+#endif /* __ppc__ */
+
+#else /* !__APPLE__ */
+
+extern int avoid_empty_file;
+
+#endif /* !__APPLE__*/

contrib/compiler-rt/lib/arm/adddf3vfp.S

@@ -0,0 +1,24 @@
+//===-- adddf3vfp.S - Implement adddf3vfp ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// double __adddf3vfp(double a, double b) { return a + b; }
+//
+// Adds two double precision floating point numbers using the Darwin
+// calling convention where double arguments are passsed in GPR pairs
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__adddf3vfp)
+	fmdrr	d6, r0, r1		// move first param from r0/r1 pair into d6
+	fmdrr	d7, r2, r3		// move second param from r2/r3 pair into d7
+	faddd	d6, d6, d7		
+	fmrrd	r0, r1, d6		// move result back to r0/r1 pair
+	bx	lr

contrib/compiler-rt/lib/arm/addsf3vfp.S

@@ -0,0 +1,24 @@
+//===-- addsf3vfp.S - Implement addsf3vfp ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern float __addsf3vfp(float a, float b);
+//
+// Adds two single precision floating point numbers using the Darwin
+// calling convention where single arguments are passsed in GPRs
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__addsf3vfp)
+	fmsr	s14, r0		// move first param from r0 into float register
+	fmsr	s15, r1		// move second param from r1 into float register
+	fadds	s14, s14, s15
+	fmrs	r0, s14		// move result back to r0
+	bx	lr

contrib/compiler-rt/lib/arm/bswapdi2.S

@@ -0,0 +1,36 @@
+//===------- bswapdi2 - Implement bswapdi2 --------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern uint64_t __bswapdi2(uint64_t);
+//
+// Reverse all the bytes in a 64-bit integer.
+//
+.align 2
+DEFINE_COMPILERRT_FUNCTION(__bswapdi2)
+#if __ARM_ARCH_5TEJ__ || __ARM_ARCH_4T__
+    // before armv6 does not have "rev" instruction
+    // r2 = rev(r0)
+    eor r2, r0, r0, ror #16
+    bic r2, r2, #0xff0000
+    mov r2, r2, lsr #8
+    eor r2, r2, r0, ror #8
+    // r0 = rev(r1)
+    eor r0, r1, r1, ror #16
+    bic r0, r0, #0xff0000
+    mov r0, r0, lsr #8
+    eor r0, r0, r1, ror #8
+#else
+    rev r2, r0  // r2 = rev(r0)
+    rev r0, r1  // r0 = rev(r1)
+#endif
+    mov r1, r2  // r1 = r2 = rev(r0)
+    bx  lr

contrib/compiler-rt/lib/arm/bswapsi2.S

@@ -0,0 +1,28 @@
+//===------- bswapsi2 - Implement bswapsi2 --------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern uint32_t __bswapsi2(uint32_t);
+//
+// Reverse all the bytes in a 32-bit integer.
+//
+.align 2
+DEFINE_COMPILERRT_FUNCTION(__bswapsi2)
+#if __ARM_ARCH_5TEJ__ || __ARM_ARCH_4T__
+    // before armv6 does not have "rev" instruction
+ 	eor	r1, r0, r0, ror #16
+ 	bic	r1, r1, #0xff0000
+ 	mov	r1, r1, lsr #8
+ 	eor	r0, r1, r0, ror #8
+#else
+    rev r0, r0
+#endif
+    bx  lr

contrib/compiler-rt/lib/arm/comparesf2.S

@@ -0,0 +1,130 @@
+//===-- comparesf2.S - Implement single-precision soft-float comparisons --===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the following soft-fp_t comparison routines:
+//
+//   __eqsf2   __gesf2   __unordsf2
+//   __lesf2   __gtsf2
+//   __ltsf2
+//   __nesf2
+//
+// The semantics of the routines grouped in each column are identical, so there
+// is a single implementation for each, with multiple names.
+//
+// The routines behave as follows:
+//
+//   __lesf2(a,b) returns -1 if a < b
+//                         0 if a == b
+//                         1 if a > b
+//                         1 if either a or b is NaN
+//
+//   __gesf2(a,b) returns -1 if a < b
+//                         0 if a == b
+//                         1 if a > b
+//                        -1 if either a or b is NaN
+//
+//   __unordsf2(a,b) returns 0 if both a and b are numbers
+//                           1 if either a or b is NaN
+//
+// Note that __lesf2( ) and __gesf2( ) are identical except in their handling of
+// NaN values.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+.syntax unified
+
+.align 2
+DEFINE_COMPILERRT_FUNCTION(__eqsf2)
+DEFINE_COMPILERRT_FUNCTION(__lesf2)
+DEFINE_COMPILERRT_FUNCTION(__ltsf2)
+DEFINE_COMPILERRT_FUNCTION(__nesf2)
+    // Make copies of a and b with the sign bit shifted off the top.  These will
+    // be used to detect zeros and NaNs.
+    mov     r2,         r0, lsl #1
+    mov     r3,         r1, lsl #1
+    
+    // We do the comparison in three stages (ignoring NaN values for the time
+    // being).  First, we orr the absolute values of a and b; this sets the Z
+    // flag if both a and b are zero (of either sign).  The shift of r3 doesn't
+    // effect this at all, but it *does* make sure that the C flag is clear for
+    // the subsequent operations.
+    orrs    r12,    r2, r3, lsr #1
+    
+    // Next, we check if a and b have the same or different signs.  If they have
+    // opposite signs, this eor will set the N flag.
+    eorsne  r12,    r0, r1
+    
+    // If a and b are equal (either both zeros or bit identical; again, we're
+    // ignoring NaNs for now), this subtract will zero out r0.  If they have the
+    // same sign, the flags are updated as they would be for a comparison of the
+    // absolute values of a and b.
+    subspl  r0,     r2, r3
+    
+    // If a is smaller in magnitude than b and both have the same sign, place
+    // the negation of the sign of b in r0.  Thus, if both are negative and
+    // a > b, this sets r0 to 0; if both are positive and a < b, this sets
+    // r0 to -1.
+    //
+    // This is also done if a and b have opposite signs and are not both zero,
+    // because in that case the subtract was not performed and the C flag is
+    // still clear from the shift argument in orrs; if a is positive and b
+    // negative, this places 0 in r0; if a is negative and b positive, -1 is
+    // placed in r0.
+    mvnlo   r0,         r1, asr #31
+
+    // If a is greater in magnitude than b and both have the same sign, place
+    // the sign of b in r0.  Thus, if both are negative and a < b, -1 is placed
+    // in r0, which is the desired result.  Conversely, if both are positive
+    // and a > b, zero is placed in r0.
+    movhi   r0,         r1, asr #31
+    
+    // If you've been keeping track, at this point r0 contains -1 if a < b and
+    // 0 if a >= b.  All that remains to be done is to set it to 1 if a > b.
+    // If a == b, then the Z flag is set, so we can get the correct final value
+    // into r0 by simply or'ing with 1 if Z is clear.
+	orrne	r0,     r0, #1
+    
+    // Finally, we need to deal with NaNs.  If either argument is NaN, replace
+    // the value in r0 with 1.
+    cmp     r2,         #0xff000000
+    cmpls   r3,         #0xff000000
+    movhi   r0,         #1
+    bx      lr
+    
+.align 2
+DEFINE_COMPILERRT_FUNCTION(__gesf2)
+DEFINE_COMPILERRT_FUNCTION(__gtsf2)
+    // Identical to the preceeding except in that we return -1 for NaN values.
+    // Given that the two paths share so much code, one might be tempted to 
+    // unify them; however, the extra code needed to do so makes the code size
+    // to performance tradeoff very hard to justify for such small functions.
+    mov     r2,         r0, lsl #1
+    mov     r3,         r1, lsl #1
+    orrs    r12,    r2, r3, lsr #1
+    eorsne  r12,    r0, r1
+    subspl  r0,     r2, r3
+    mvnlo   r0,         r1, asr #31
+    movhi   r0,         r1, asr #31
+	orrne	r0,     r0, #1
+    cmp     r2,         #0xff000000
+    cmpls   r3,         #0xff000000
+    movhi   r0,         #-1
+    bx      lr
+    
+.align 2
+DEFINE_COMPILERRT_FUNCTION(__unordsf2)
+    // Return 1 for NaN values, 0 otherwise.
+    mov     r2,         r0, lsl #1
+    mov     r3,         r1, lsl #1
+    mov     r0,         #0
+    cmp     r2,         #0xff000000
+    cmpls   r3,         #0xff000000
+    movhi   r0,         #1
+    bx      lr

contrib/compiler-rt/lib/arm/divdf3vfp.S

@@ -0,0 +1,24 @@
+//===-- divdf3vfp.S - Implement divdf3vfp ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __divdf3vfp(double a, double b);
+//
+// Divides two double precision floating point numbers using the Darwin
+// calling convention where double arguments are passsed in GPR pairs
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__divdf3vfp)
+	fmdrr	d6, r0, r1		// move first param from r0/r1 pair into d6
+	fmdrr	d7, r2, r3		// move second param from r2/r3 pair into d7
+	fdivd	d5, d6, d7		
+	fmrrd	r0, r1, d5		// move result back to r0/r1 pair
+	bx	lr

contrib/compiler-rt/lib/arm/divsf3vfp.S

@@ -0,0 +1,24 @@
+//===-- divsf3vfp.S - Implement divsf3vfp ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern float __divsf3vfp(float a, float b);
+//
+// Divides two single precision floating point numbers using the Darwin
+// calling convention where single arguments are passsed like 32-bit ints.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__divsf3vfp)
+	fmsr	s14, r0		// move first param from r0 into float register
+	fmsr	s15, r1		// move second param from r1 into float register
+	fdivs	s13, s14, s15
+	fmrs	r0, s13		// move result back to r0
+	bx	lr

contrib/compiler-rt/lib/arm/eqdf2vfp.S

@@ -0,0 +1,27 @@
+//===-- eqdf2vfp.S - Implement eqdf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __eqdf2vfp(double a, double b);
+//
+// Returns one iff a == b and neither is NaN.
+// Uses Darwin calling convention where double precision arguments are passsed 
+// like in GPR pairs.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__eqdf2vfp)
+	fmdrr	d6, r0, r1	// load r0/r1 pair in double register
+	fmdrr	d7, r2, r3	// load r2/r3 pair in double register
+	fcmpd	d6, d7		
+	fmstat
+	moveq	r0, #1		// set result register to 1 if equal
+	movne	r0, #0
+	bx	lr

contrib/compiler-rt/lib/arm/eqsf2vfp.S

@@ -0,0 +1,28 @@
+//===-- eqsf2vfp.S - Implement eqsf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __eqsf2vfp(float a, float b);
+//
+// Returns one iff a == b and neither is NaN.
+// Uses Darwin calling convention where single precision arguments are passsed 
+// like 32-bit ints
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__eqsf2vfp)
+	fmsr	s14, r0     // move from GPR 0 to float register
+	fmsr	s15, r1	    // move from GPR 1 to float register
+	fcmps	s14, s15
+	fmstat
+	moveq	r0, #1      // set result register to 1 if equal
+	movne	r0, #0
+	bx	lr
+

contrib/compiler-rt/lib/arm/extendsfdf2vfp.S

@@ -0,0 +1,24 @@
+//===-- extendsfdf2vfp.S - Implement extendsfdf2vfp -----------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __extendsfdf2vfp(float a);
+//
+// Converts single precision float to double precision result.
+// Uses Darwin calling convention where a single precision parameter is 
+// passed in a GPR and a double precision result is returned in R0/R1 pair.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__extendsfdf2vfp)
+	fmsr	s15, r0      // load float register from R0
+	fcvtds	d7, s15      // convert single to double
+	fmrrd	r0, r1, d7   // return result in r0/r1 pair
+	bx	lr

contrib/compiler-rt/lib/arm/fixdfsivfp.S

@@ -0,0 +1,24 @@
+//===-- fixdfsivfp.S - Implement fixdfsivfp -----------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __fixdfsivfp(double a);
+//
+// Converts double precision float to a 32-bit int rounding towards zero.
+// Uses Darwin calling convention where a double precision parameter is 
+// passed in GPR register pair.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__fixdfsivfp)
+	fmdrr	d7, r0, r1    // load double register from R0/R1
+	ftosizd	s15, d7       // convert double to 32-bit int into s15
+	fmrs	r0, s15	      // move s15 to result register
+	bx	lr

contrib/compiler-rt/lib/arm/fixsfsivfp.S

@@ -0,0 +1,24 @@
+//===-- fixsfsivfp.S - Implement fixsfsivfp -----------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __fixsfsivfp(float a);
+//
+// Converts single precision float to a 32-bit int rounding towards zero.
+// Uses Darwin calling convention where a single precision parameter is 
+// passed in a GPR..
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__fixsfsivfp)
+	fmsr	s15, r0      // load float register from R0
+	ftosizs	s15, s15     // convert single to 32-bit int into s15
+	fmrs	r0, s15	     // move s15 to result register
+	bx	lr

contrib/compiler-rt/lib/arm/fixunsdfsivfp.S

@@ -0,0 +1,25 @@
+//===-- fixunsdfsivfp.S - Implement fixunsdfsivfp -------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern unsigned int __fixunsdfsivfp(double a);
+//
+// Converts double precision float to a 32-bit unsigned int rounding towards 
+// zero. All negative values become zero.
+// Uses Darwin calling convention where a double precision parameter is 
+// passed in GPR register pair.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__fixunsdfsivfp)
+	fmdrr	d7, r0, r1    // load double register from R0/R1
+	ftouizd	s15, d7       // convert double to 32-bit int into s15
+	fmrs	r0, s15	      // move s15 to result register
+	bx	lr

contrib/compiler-rt/lib/arm/fixunssfsivfp.S

@@ -0,0 +1,25 @@
+//===-- fixunssfsivfp.S - Implement fixunssfsivfp -------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern unsigned int __fixunssfsivfp(float a);
+//
+// Converts single precision float to a 32-bit unsigned int rounding towards 
+// zero. All negative values become zero.
+// Uses Darwin calling convention where a single precision parameter is 
+// passed in a GPR..
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__fixunssfsivfp)
+	fmsr	s15, r0      // load float register from R0
+	ftouizs	s15, s15     // convert single to 32-bit unsigned into s15
+	fmrs	r0, s15	     // move s15 to result register
+	bx	lr

contrib/compiler-rt/lib/arm/floatsidfvfp.S

@@ -0,0 +1,24 @@
+//===-- floatsidfvfp.S - Implement floatsidfvfp ---------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __floatsidfvfp(int a);
+//
+// Converts a 32-bit int to a double precision float.
+// Uses Darwin calling convention where a double precision result is 
+// return in GPR register pair.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__floatsidfvfp)
+	fmsr	s15, r0		   // move int to float register s15
+	fsitod	d7, s15        // convert 32-bit int in s15 to double in d7
+	fmrrd	r0, r1, d7     // move d7 to result register pair r0/r1
+	bx	lr

contrib/compiler-rt/lib/arm/floatsisfvfp.S

@@ -0,0 +1,24 @@
+//===-- floatsisfvfp.S - Implement floatsisfvfp ---------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern float __floatsisfvfp(int a);
+//
+// Converts single precision float to a 32-bit int rounding towards zero.
+// Uses Darwin calling convention where a single precision result is 
+// return in a GPR..
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__floatsisfvfp)
+	fmsr	s15, r0	     // move int to float register s15
+	fsitos	s15, s15     // convert 32-bit int in s15 to float in s15
+	fmrs	r0, s15      // move s15 to result register
+	bx	lr

contrib/compiler-rt/lib/arm/floatunssidfvfp.S

@@ -0,0 +1,24 @@
+//===-- floatunssidfvfp.S - Implement floatunssidfvfp ---------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __floatunssidfvfp(unsigned int a);
+//
+// Converts a 32-bit int to a double precision float.
+// Uses Darwin calling convention where a double precision result is 
+// return in GPR register pair.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__floatunssidfvfp)
+	fmsr	s15, r0		   // move int to float register s15
+	fuitod	d7, s15        // convert 32-bit int in s15 to double in d7
+	fmrrd	r0, r1, d7     // move d7 to result register pair r0/r1
+	bx	lr

contrib/compiler-rt/lib/arm/floatunssisfvfp.S

@@ -0,0 +1,24 @@
+//===-- floatunssisfvfp.S - Implement floatunssisfvfp ---------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern float __floatunssisfvfp(unsigned int a);
+//
+// Converts single precision float to a 32-bit int rounding towards zero.
+// Uses Darwin calling convention where a single precision result is 
+// return in a GPR..
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__floatunssisfvfp)
+	fmsr	s15, r0	     // move int to float register s15
+	fuitos 	s15, s15     // convert 32-bit int in s15 to float in s15
+	fmrs	r0, s15      // move s15 to result register
+	bx	lr

contrib/compiler-rt/lib/arm/gedf2vfp.S

@@ -0,0 +1,27 @@
+//===-- gedf2vfp.S - Implement gedf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __gedf2vfp(double a, double b);
+//
+// Returns one iff a >= b and neither is NaN.
+// Uses Darwin calling convention where double precision arguments are passsed 
+// like in GPR pairs.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__gedf2vfp)
+	fmdrr	d6, r0, r1	// load r0/r1 pair in double register
+	fmdrr	d7, r2, r3	// load r2/r3 pair in double register
+	fcmpd	d6, d7		
+	fmstat
+	movge	r0, #1      // set result register to 1 if greater than or equal
+	movlt	r0, #0
+	bx	lr

contrib/compiler-rt/lib/arm/gesf2vfp.S

@@ -0,0 +1,28 @@
+//===-- gesf2vfp.S - Implement gesf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __gesf2vfp(float a, float b);
+//
+// Returns one iff a >= b and neither is NaN.
+// Uses Darwin calling convention where single precision arguments are passsed 
+// like 32-bit ints
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__gesf2vfp)
+	fmsr	s14, r0	    // move from GPR 0 to float register
+	fmsr	s15, r1	    // move from GPR 1 to float register
+	fcmps	s14, s15
+	fmstat
+	movge	r0, #1      // set result register to 1 if greater than or equal
+	movlt	r0, #0
+	bx	lr
+

contrib/compiler-rt/lib/arm/gtdf2vfp.S

@@ -0,0 +1,27 @@
+//===-- gtdf2vfp.S - Implement gtdf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __gtdf2vfp(double a, double b);
+//
+// Returns one iff a > b and neither is NaN.
+// Uses Darwin calling convention where double precision arguments are passsed 
+// like in GPR pairs.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__gtdf2vfp)
+	fmdrr	d6, r0, r1	// load r0/r1 pair in double register
+	fmdrr	d7, r2, r3	// load r2/r3 pair in double register
+	fcmpd	d6, d7		
+	fmstat
+	movgt	r0, #1		// set result register to 1 if equal
+	movle	r0, #0
+	bx	lr

contrib/compiler-rt/lib/arm/gtsf2vfp.S

@@ -0,0 +1,28 @@
+//===-- gtsf2vfp.S - Implement gtsf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __gtsf2vfp(float a, float b);
+//
+// Returns one iff a > b and neither is NaN.
+// Uses Darwin calling convention where single precision arguments are passsed 
+// like 32-bit ints
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__gtsf2vfp)
+	fmsr	s14, r0		// move from GPR 0 to float register
+	fmsr	s15, r1		// move from GPR 1 to float register
+	fcmps	s14, s15
+	fmstat
+	movgt	r0, #1		// set result register to 1 if equal
+	movle	r0, #0
+	bx	lr
+

contrib/compiler-rt/lib/arm/ledf2vfp.S

@@ -0,0 +1,27 @@
+//===-- ledf2vfp.S - Implement ledf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __ledf2vfp(double a, double b);
+//
+// Returns one iff a <= b and neither is NaN.
+// Uses Darwin calling convention where double precision arguments are passsed 
+// like in GPR pairs.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__ledf2vfp)
+	fmdrr	d6, r0, r1	// load r0/r1 pair in double register
+	fmdrr	d7, r2, r3	// load r2/r3 pair in double register
+	fcmpd	d6, d7		
+	fmstat
+	movls	r0, #1		// set result register to 1 if equal
+	movhi	r0, #0
+	bx	lr

contrib/compiler-rt/lib/arm/lesf2vfp.S

@@ -0,0 +1,28 @@
+//===-- lesf2vfp.S - Implement lesf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __lesf2vfp(float a, float b);
+//
+// Returns one iff a <= b and neither is NaN.
+// Uses Darwin calling convention where single precision arguments are passsed 
+// like 32-bit ints
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__lesf2vfp)
+	fmsr	s14, r0     // move from GPR 0 to float register
+	fmsr	s15, r1     // move from GPR 1 to float register
+	fcmps	s14, s15
+	fmstat
+	movls	r0, #1      // set result register to 1 if equal
+	movhi	r0, #0
+	bx	lr
+

contrib/compiler-rt/lib/arm/ltdf2vfp.S

@@ -0,0 +1,27 @@
+//===-- ltdf2vfp.S - Implement ltdf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __ltdf2vfp(double a, double b);
+//
+// Returns one iff a < b and neither is NaN.
+// Uses Darwin calling convention where double precision arguments are passsed 
+// like in GPR pairs.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__ltdf2vfp)
+	fmdrr	d6, r0, r1	// load r0/r1 pair in double register
+	fmdrr	d7, r2, r3	// load r2/r3 pair in double register
+	fcmpd	d6, d7		
+	fmstat
+	movmi	r0, #1		// set result register to 1 if equal
+	movpl	r0, #0
+	bx	lr

contrib/compiler-rt/lib/arm/ltsf2vfp.S

@@ -0,0 +1,28 @@
+//===-- ltsf2vfp.S - Implement ltsf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __ltsf2vfp(float a, float b);
+//
+// Returns one iff a < b and neither is NaN.
+// Uses Darwin calling convention where single precision arguments are passsed 
+// like 32-bit ints
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__ltsf2vfp)
+	fmsr	s14, r0     // move from GPR 0 to float register
+	fmsr	s15, r1     // move from GPR 1 to float register
+	fcmps	s14, s15
+	fmstat
+	movmi	r0, #1      // set result register to 1 if equal
+	movpl	r0, #0
+	bx	lr
+

contrib/compiler-rt/lib/arm/modsi3.S

@@ -0,0 +1,36 @@
+//===-------- modsi3.S - Implement modsi3 ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int32_t __modsi3(int32_t a, int32_t b);
+//
+// Returns the remainder when dividing two 32-bit signed integers.
+// Conceptually, the function is: { return a - (a / b) * b; }
+// But if you write that in C, llvm compiles it to a call to __modsi3...
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__modsi3)
+	push	{r4, r5, r7, lr}
+	add	r7, sp, #8	// set stack frame
+	mov	r5, r0		// save a
+	mov	r4, r1		// save b
+	bl      ___divsi3	// compute a/b
+#if __ARM_ARCH_7A__
+	mls     r0, r4, r0, r5  // mulitple result * b and subtract from a
+#else	
+	// before armv7, does not have "mls" instruction
+	mul	r3, r0, r4	// multiple result * b
+	sub     r0, r5, r3      // a - result
+#endif
+	pop     {r4, r5, r7, pc}
+	
+
+

contrib/compiler-rt/lib/arm/muldf3vfp.S

@@ -0,0 +1,24 @@
+//===-- muldf3vfp.S - Implement muldf3vfp ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __muldf3vfp(double a, double b);
+//
+// Multiplies two double precision floating point numbers using the Darwin
+// calling convention where double arguments are passsed in GPR pairs
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__muldf3vfp)
+	fmdrr	d6, r0, r1		// move first param from r0/r1 pair into d6
+	fmdrr	d7, r2, r3		// move second param from r2/r3 pair into d7
+	fmuld	d6, d6, d7		
+	fmrrd	r0, r1, d6		// move result back to r0/r1 pair
+	bx	lr

contrib/compiler-rt/lib/arm/mulsf3vfp.S

@@ -0,0 +1,24 @@
+//===-- mulsf3vfp.S - Implement mulsf3vfp ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern float __mulsf3vfp(float a, float b);
+//
+// Multiplies two single precision floating point numbers using the Darwin
+// calling convention where single arguments are passsed like 32-bit ints.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__mulsf3vfp)
+	fmsr	s14, r0		// move first param from r0 into float register
+	fmsr	s15, r1		// move second param from r1 into float register
+	fmuls	s13, s14, s15
+	fmrs	r0, s13		// move result back to r0
+	bx	lr

contrib/compiler-rt/lib/arm/nedf2vfp.S

@@ -0,0 +1,27 @@
+//===-- nedf2vfp.S - Implement nedf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __nedf2vfp(double a, double b);
+//
+// Returns zero if a and b are unequal and neither is NaN.
+// Uses Darwin calling convention where double precision arguments are passsed 
+// like in GPR pairs.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__nedf2vfp)
+	fmdrr	d6, r0, r1	// load r0/r1 pair in double register
+	fmdrr	d7, r2, r3	// load r2/r3 pair in double register
+	fcmpd	d6, d7		
+	fmstat
+	movne	r0, #1		// set result register to 0 if unequal
+	moveq	r0, #0
+	bx	lr

contrib/compiler-rt/lib/arm/negdf2vfp.S

@@ -0,0 +1,21 @@
+//===-- negdf2vfp.S - Implement negdf2vfp ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __negdf2vfp(double a, double b);
+//
+// Returns the negation a double precision floating point numbers using the 
+// Darwin calling convention where double arguments are passsed in GPR pairs.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__negdf2vfp)
+	eor	r1, r1, #-2147483648	// flip sign bit on double in r0/r1 pair
+	bx	lr

contrib/compiler-rt/lib/arm/negsf2vfp.S

@@ -0,0 +1,21 @@
+//===-- negsf2vfp.S - Implement negsf2vfp ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern float __negsf2vfp(float a);
+//
+// Returns the negation of a single precision floating point numbers using the 
+// Darwin calling convention where single arguments are passsed like 32-bit ints
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__negsf2vfp)
+	eor	r0, r0, #-2147483648	// flip sign bit on float in r0
+	bx	lr

contrib/compiler-rt/lib/arm/nesf2vfp.S

@@ -0,0 +1,28 @@
+//===-- nesf2vfp.S - Implement nesf2vfp -----------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __nesf2vfp(float a, float b);
+//
+// Returns one iff a != b and neither is NaN.
+// Uses Darwin calling convention where single precision arguments are passsed 
+// like 32-bit ints
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__nesf2vfp)
+	fmsr	s14, r0	    // move from GPR 0 to float register
+	fmsr	s15, r1	    // move from GPR 1 to float register
+	fcmps	s14, s15
+	fmstat
+	movne	r0, #1      // set result register to 1 if unequal
+	moveq	r0, #0
+	bx	lr
+

contrib/compiler-rt/lib/arm/restore_vfp_d8_d15_regs.S

@@ -0,0 +1,37 @@
+//===-- save_restore_regs.S - Implement save/restore* ---------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// When compiling C++ functions that need to handle thrown exceptions the
+// compiler is required to save all registers and call __Unwind_SjLj_Register
+// in the function prolog.  But when compiling for thumb1, there are
+// no instructions to access the floating point registers, so the
+// compiler needs to add a call to the helper function _save_vfp_d8_d15_regs
+// written in ARM to save the float registers.  In the epilog, the compiler
+// must also add a call to __restore_vfp_d8_d15_regs to restore those registers.
+//
+
+	.text
+	.syntax unified
+
+//
+// Restore registers d8-d15 from stack
+//
+	.align 2
+DEFINE_COMPILERRT_PRIVATE_FUNCTION(__restore_vfp_d8_d15_regs)
+	vldmia	sp!, {d8-d15}           // pop registers d8-d15 off stack
+	bx      lr                      // return to prolog
+
+
+
+	// tell linker it can break up file at label boundaries
+	.subsections_via_symbols
+

contrib/compiler-rt/lib/arm/save_vfp_d8_d15_regs.S

@@ -0,0 +1,35 @@
+//===-- save_restore_regs.S - Implement save/restore* ---------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// When compiling C++ functions that need to handle thrown exceptions the
+// compiler is required to save all registers and call __Unwind_SjLj_Register
+// in the function prolog.  But when compiling for thumb1, there are
+// no instructions to access the floating point registers, so the
+// compiler needs to add a call to the helper function _save_vfp_d8_d15_regs
+// written in ARM to save the float registers.  In the epilog, the compiler
+// must also add a call to __restore_vfp_d8_d15_regs to restore those registers.
+//
+
+	.text
+	.syntax unified
+
+//
+// Save registers d8-d15 onto stack
+//
+	.align 2
+DEFINE_COMPILERRT_PRIVATE_FUNCTION(__save_vfp_d8_d15_regs)
+	vstmdb	sp!, {d8-d15}           // push registers d8-d15 onto stack
+	bx      lr                      // return to prolog
+
+	// tell linker it can break up file at label boundaries
+	.subsections_via_symbols
+

contrib/compiler-rt/lib/arm/subdf3vfp.S

@@ -0,0 +1,24 @@
+//===-- subdf3vfp.S - Implement subdf3vfp ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern double __subdf3vfp(double a, double b);
+//
+// Returns difference between two double precision floating point numbers using 
+// the Darwin calling convention where double arguments are passsed in GPR pairs
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__subdf3vfp)
+	fmdrr	d6, r0, r1		// move first param from r0/r1 pair into d6
+	fmdrr	d7, r2, r3		// move second param from r2/r3 pair into d7
+	fsubd	d6, d6, d7		
+	fmrrd	r0, r1, d6		// move result back to r0/r1 pair
+	bx	lr

contrib/compiler-rt/lib/arm/subsf3vfp.S

@@ -0,0 +1,25 @@
+//===-- subsf3vfp.S - Implement subsf3vfp ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern float __subsf3vfp(float a, float b);
+//
+// Returns the difference between two single precision floating point numbers 
+// using the Darwin calling convention where single arguments are passsed
+// like 32-bit ints.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__subsf3vfp)
+	fmsr	s14, r0		// move first param from r0 into float register
+	fmsr	s15, r1		// move second param from r1 into float register
+	fsubs	s14, s14, s15
+	fmrs	r0, s14		// move result back to r0
+	bx	lr

contrib/compiler-rt/lib/arm/switch16.S

@@ -0,0 +1,44 @@
+//===-- switch.S - Implement switch* --------------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// When compiling switch statements in thumb mode, the compiler
+// can use these __switch* helper functions  The compiler emits a blx to
+// the __switch* function followed by a table of displacements for each
+// case statement.  On entry, R0 is the index into the table. The __switch*
+// function uses the return address in lr to find the start of the table.
+// The first entry in the table is the count of the entries in the table.
+// It then uses R0 to index into the table and get the displacement of the
+// address to jump to.  If R0 is greater than the size of the table, it jumps
+// to the last entry in the table. Each displacement in the table is actually
+// the distance from lr to the label, thus making the tables PIC.
+
+
+	.text
+	.syntax unified
+
+//
+// The table contains signed 2-byte sized elements which are 1/2 the distance
+// from lr to the target label.
+//
+	.align 2
+DEFINE_COMPILERRT_PRIVATE_FUNCTION(__switch16)
+	ldrh    ip, [lr, #-1]           // get first 16-bit word in table
+	cmp     r0, ip                  // compare with index
+	add     r0, lr, r0, lsl #1      // compute address of element in table
+	ldrshcc r0, [r0, #1]            // load 16-bit element if r0 is in range
+	add     ip, lr, ip, lsl #1      // compute address of last element in table
+	ldrshhs r0, [ip, #1]            // load 16-bit element if r0 out of range
+	add     ip, lr, r0, lsl #1      // compute label = lr + element*2
+	bx      ip                      // jump to computed label
+
+	// tell linker it can break up file at label boundaries
+	.subsections_via_symbols

contrib/compiler-rt/lib/arm/switch32.S

@@ -0,0 +1,46 @@
+//===-- switch.S - Implement switch* --------------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// When compiling switch statements in thumb mode, the compiler
+// can use these __switch* helper functions  The compiler emits a blx to
+// the __switch* function followed by a table of displacements for each
+// case statement.  On entry, R0 is the index into the table. The __switch*
+// function uses the return address in lr to find the start of the table.
+// The first entry in the table is the count of the entries in the table.
+// It then uses R0 to index into the table and get the displacement of the
+// address to jump to.  If R0 is greater than the size of the table, it jumps
+// to the last entry in the table. Each displacement in the table is actually
+// the distance from lr to the label, thus making the tables PIC.
+
+
+	.text
+	.syntax unified
+
+//
+// The table contains signed 4-byte sized elements which are the distance
+// from lr to the target label.
+//
+	.align 2
+DEFINE_COMPILERRT_PRIVATE_FUNCTION(__switch32)
+	ldr     ip, [lr, #-1]            // get first 32-bit word in table
+	cmp     r0, ip                   // compare with index
+	add     r0, lr, r0, lsl #2       // compute address of element in table
+	ldrcc   r0, [r0, #3]             // load 32-bit element if r0 is in range
+	add     ip, lr, ip, lsl #2       // compute address of last element in table
+	ldrcs   r0, [ip, #3]             // load 32-bit element if r0 out of range
+	add     ip, lr, r0               // compute label = lr + element
+	bx      ip                       // jump to computed label
+
+
+	// tell linker it can break up file at label boundaries
+	.subsections_via_symbols
+

contrib/compiler-rt/lib/arm/switch8.S

@@ -0,0 +1,43 @@
+//===-- switch.S - Implement switch* --------------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// When compiling switch statements in thumb mode, the compiler
+// can use these __switch* helper functions  The compiler emits a blx to
+// the __switch* function followed by a table of displacements for each
+// case statement.  On entry, R0 is the index into the table. The __switch*
+// function uses the return address in lr to find the start of the table.
+// The first entry in the table is the count of the entries in the table.
+// It then uses R0 to index into the table and get the displacement of the
+// address to jump to.  If R0 is greater than the size of the table, it jumps
+// to the last entry in the table. Each displacement in the table is actually
+// the distance from lr to the label, thus making the tables PIC.
+
+
+	.text
+	.syntax unified
+
+//
+// The table contains signed byte sized elements which are 1/2 the distance
+// from lr to the target label.
+//
+	.align 2
+DEFINE_COMPILERRT_PRIVATE_FUNCTION(__switch8)
+	ldrb    ip, [lr, #-1]           // get first byte in table
+	cmp     r0, ip                  // signed compare with index
+	ldrsbcc r0, [lr, r0]            // get indexed byte out of table
+	ldrsbhs r0, [lr, ip]            // if out of range, use last entry in table
+	add     ip, lr, r0, lsl #1      // compute label = lr + element*2
+	bx      ip                      // jump to computed label
+
+	// tell linker it can break up file at label boundaries
+	.subsections_via_symbols
+

contrib/compiler-rt/lib/arm/switchu8.S

@@ -0,0 +1,43 @@
+//===-- switch.S - Implement switch* --------------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// When compiling switch statements in thumb mode, the compiler
+// can use these __switch* helper functions  The compiler emits a blx to
+// the __switch* function followed by a table of displacements for each
+// case statement.  On entry, R0 is the index into the table. The __switch*
+// function uses the return address in lr to find the start of the table.
+// The first entry in the table is the count of the entries in the table.
+// It then uses R0 to index into the table and get the displacement of the
+// address to jump to.  If R0 is greater than the size of the table, it jumps
+// to the last entry in the table. Each displacement in the table is actually
+// the distance from lr to the label, thus making the tables PIC.
+
+
+	.text
+	.syntax unified
+
+//
+// The table contains unsigned byte sized elements which are 1/2 the distance
+// from lr to the target label.
+//
+	.align 2
+DEFINE_COMPILERRT_PRIVATE_FUNCTION(__switchu8)
+	ldrb    ip, [lr, #-1]           // get first byte in table
+	cmp     r0, ip                  // compare with index
+	ldrbcc  r0, [lr, r0]            // get indexed byte out of table
+	ldrbhs  r0, [lr, ip]            // if out of range, use last entry in table
+	add     ip, lr, r0, lsl #1      // compute label = lr + element*2
+	bx      ip                      // jump to computed label
+
+	// tell linker it can break up file at label boundaries
+	.subsections_via_symbols
+

contrib/compiler-rt/lib/arm/sync_synchronize.S

@@ -0,0 +1,34 @@
+//===-- sync_synchronize - Implement memory barrier * ----------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// When compiling a use of the gcc built-in __sync_synchronize() in thumb1 mode
+// the compiler may emit a call to __sync_synchronize.  
+// On Darwin the implementation jumps to an OS supplied function named 
+// OSMemoryBarrier
+//
+
+	.text
+	.syntax unified
+
+#if __APPLE__
+
+	.align 2
+DEFINE_COMPILERRT_PRIVATE_FUNCTION(__sync_synchronize)
+	stmfd	sp!, {r7, lr}
+	add		r7, sp, #0
+	bl		_OSMemoryBarrier
+	ldmfd	sp!, {r7, pc}
+
+	// tell linker it can break up file at label boundaries
+	.subsections_via_symbols
+		
+#endif

contrib/compiler-rt/lib/arm/truncdfsf2vfp.S

@@ -0,0 +1,24 @@
+//===-- truncdfsf2vfp.S - Implement truncdfsf2vfp -------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern float __truncdfsf2vfp(double a);
+//
+// Converts double precision float to signle precision result.
+// Uses Darwin calling convention where a double precision parameter is 
+// passed in a R0/R1 pair and a signle precision result is returned in R0.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__truncdfsf2vfp)
+	fmdrr	d7, r0, r1   // load double from r0/r1 pair
+	fcvtsd	s15, d7      // convert double to single (trucate precision)
+	fmrs	r0, s15      // return result in r0
+	bx	lr

contrib/compiler-rt/lib/arm/unorddf2vfp.S

@@ -0,0 +1,27 @@
+//===-- unorddf2vfp.S - Implement unorddf2vfp ------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __unorddf2vfp(double a, double b);
+//
+// Returns one iff a or b is NaN
+// Uses Darwin calling convention where double precision arguments are passsed 
+// like in GPR pairs.
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__unorddf2vfp)
+	fmdrr	d6, r0, r1	// load r0/r1 pair in double register
+	fmdrr	d7, r2, r3	// load r2/r3 pair in double register
+	fcmpd	d6, d7		
+	fmstat
+	movvs	r0, #1      // set result register to 1 if "overflow" (any NaNs)
+	movvc	r0, #0
+	bx	lr

contrib/compiler-rt/lib/arm/unordsf2vfp.S

@@ -0,0 +1,28 @@
+//===-- unordsf2vfp.S - Implement unordsf2vfp -----------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int __unordsf2vfp(float a, float b);
+//
+// Returns one iff a or b is NaN
+// Uses Darwin calling convention where single precision arguments are passsed 
+// like 32-bit ints
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__unordsf2vfp)
+	fmsr	s14, r0     // move from GPR 0 to float register
+	fmsr	s15, r1	    // move from GPR 1 to float register
+	fcmps	s14, s15
+	fmstat
+	movvs	r0, #1      // set result register to 1 if "overflow" (any NaNs)
+	movvc	r0, #0
+	bx	lr
+

contrib/compiler-rt/lib/ashldi3.c

@@ -0,0 +1,41 @@
+/* ====-- ashldi3.c - Implement __ashldi3 -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __ashldi3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: a << b */
+
+/* Precondition:  0 <= b < bits_in_dword */
+
+di_int
+__ashldi3(di_int a, si_int b)
+{
+    const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT);
+    dwords input;
+    dwords result;
+    input.all = a;
+    if (b & bits_in_word)  /* bits_in_word <= b < bits_in_dword */
+    {
+        result.s.low = 0;
+        result.s.high = input.s.low << (b - bits_in_word);
+    }
+    else  /* 0 <= b < bits_in_word */
+    {
+        if (b == 0)
+            return a;
+        result.s.low  = input.s.low << b;
+        result.s.high = (input.s.high << b) | (input.s.low >> (bits_in_word - b));
+    }
+    return result.all;
+}

contrib/compiler-rt/lib/ashlti3.c

@@ -0,0 +1,45 @@
+/* ===-- ashlti3.c - Implement __ashlti3 -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __ashlti3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+
+/* Returns: a << b */
+
+/* Precondition:  0 <= b < bits_in_tword */
+
+ti_int
+__ashlti3(ti_int a, si_int b)
+{
+    const int bits_in_dword = (int)(sizeof(di_int) * CHAR_BIT);
+    twords input;
+    twords result;
+    input.all = a;
+    if (b & bits_in_dword)  /* bits_in_dword <= b < bits_in_tword */
+    {
+        result.s.low = 0;
+        result.s.high = input.s.low << (b - bits_in_dword);
+    }
+    else  /* 0 <= b < bits_in_dword */
+    {
+        if (b == 0)
+            return a;
+        result.s.low  = input.s.low << b;
+        result.s.high = (input.s.high << b) | (input.s.low >> (bits_in_dword - b));
+    }
+    return result.all;
+}
+
+#endif /* __x86_64 */

contrib/compiler-rt/lib/ashrdi3.c

@@ -0,0 +1,42 @@
+/*===-- ashrdi3.c - Implement __ashrdi3 -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __ashrdi3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: arithmetic a >> b */
+
+/* Precondition:  0 <= b < bits_in_dword */
+
+di_int
+__ashrdi3(di_int a, si_int b)
+{
+    const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT);
+    dwords input;
+    dwords result;
+    input.all = a;
+    if (b & bits_in_word)  /* bits_in_word <= b < bits_in_dword */
+    {
+        /* result.s.high = input.s.high < 0 ? -1 : 0 */
+        result.s.high = input.s.high >> (bits_in_word - 1);
+        result.s.low = input.s.high >> (b - bits_in_word);
+    }
+    else  /* 0 <= b < bits_in_word */
+    {
+        if (b == 0)
+            return a;
+        result.s.high  = input.s.high >> b;
+        result.s.low = (input.s.high << (bits_in_word - b)) | (input.s.low >> b);
+    }
+    return result.all;
+}

contrib/compiler-rt/lib/ashrti3.c

@@ -0,0 +1,46 @@
+/* ===-- ashrti3.c - Implement __ashrti3 -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __ashrti3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+
+/* Returns: arithmetic a >> b */
+
+/* Precondition:  0 <= b < bits_in_tword */
+
+ti_int
+__ashrti3(ti_int a, si_int b)
+{
+    const int bits_in_dword = (int)(sizeof(di_int) * CHAR_BIT);
+    twords input;
+    twords result;
+    input.all = a;
+    if (b & bits_in_dword)  /* bits_in_dword <= b < bits_in_tword */
+    {
+        /* result.s.high = input.s.high < 0 ? -1 : 0 */
+        result.s.high = input.s.high >> (bits_in_dword - 1);
+        result.s.low = input.s.high >> (b - bits_in_dword);
+    }
+    else  /* 0 <= b < bits_in_dword */
+    {
+        if (b == 0)
+            return a;
+        result.s.high  = input.s.high >> b;
+        result.s.low = (input.s.high << (bits_in_dword - b)) | (input.s.low >> b);
+    }
+    return result.all;
+}
+
+#endif /* __x86_64 */

contrib/compiler-rt/lib/assembly.h

@@ -0,0 +1,54 @@
+/* ===-- assembly.h - compiler-rt assembler support macros -----------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file defines macros for use in compiler-rt assembler source.
+ * This file is not part of the interface of this library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#ifndef COMPILERRT_ASSEMBLY_H
+#define COMPILERRT_ASSEMBLY_H
+
+#if defined(__POWERPC__) || defined(__powerpc__) || defined(__ppc__)
+#define SEPARATOR @
+#else
+#define SEPARATOR ;
+#endif
+
+/* We can't use __USER_LABEL_PREFIX__ here, it isn't possible to concatenate the
+   *values* of two macros. This is quite brittle, though. */
+#if defined(__APPLE__)
+#define SYMBOL_NAME(name) _##name
+#else
+#define SYMBOL_NAME(name) name
+#endif
+
+#ifdef VISIBILITY_HIDDEN
+#define DEFINE_COMPILERRT_FUNCTION(name) \
+  .globl SYMBOL_NAME(name) SEPARATOR     \
+  .private_extern SYMBOL_NAME(name) SEPARATOR    \
+  SYMBOL_NAME(name):
+#else
+#define DEFINE_COMPILERRT_FUNCTION(name) \
+  .globl SYMBOL_NAME(name) SEPARATOR     \
+  SYMBOL_NAME(name):
+#endif
+
+#define DEFINE_COMPILERRT_PRIVATE_FUNCTION(name) \
+  .globl SYMBOL_NAME(name) SEPARATOR             \
+  .private_extern SYMBOL_NAME(name) SEPARATOR    \
+  SYMBOL_NAME(name):
+
+#define DEFINE_COMPILERRT_PRIVATE_FUNCTION_UNMANGLED(name) \
+  .globl name SEPARATOR             \
+  .private_extern name SEPARATOR    \
+  name:
+
+#endif /* COMPILERRT_ASSEMBLY_H */

contrib/compiler-rt/lib/clear_cache.c

@@ -0,0 +1,41 @@
+/* ===-- clear_cache.c - Implement __clear_cache ---------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+#include <stdlib.h>
+
+#if __APPLE__
+  #include <libkern/OSCacheControl.h>
+#endif
+
+/*
+ * The compiler generates calls to __clear_cache() when creating 
+ * trampoline functions on the stack for use with nested functions.
+ * It is expected to invalidate the instruction cache for the 
+ * specified range.
+ */
+
+void __clear_cache(void* start, void* end)
+{
+#if __i386__ || __x86_64__
+/*
+ * Intel processors have a unified instruction and data cache
+ * so there is nothing to do
+ */
+#else
+    #if __APPLE__
+        /* On Darwin, sys_icache_invalidate() provides this functionality */
+        sys_icache_invalidate(start, end-start);
+    #else
+        compilerrt_abort();
+    #endif
+#endif
+}
+

contrib/compiler-rt/lib/clzdi2.c

@@ -0,0 +1,29 @@
+/* ===-- clzdi2.c - Implement __clzdi2 -------------------------------------===
+ *
+ *      	       The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __clzdi2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: the number of leading 0-bits */
+
+/* Precondition: a != 0 */
+
+si_int
+__clzdi2(di_int a)
+{
+    dwords x;
+    x.all = a;
+    const si_int f = -(x.s.high == 0);
+    return __builtin_clz((x.s.high & ~f) | (x.s.low & f)) +
+           (f & ((si_int)(sizeof(si_int) * CHAR_BIT)));
+}

contrib/compiler-rt/lib/clzsi2.c

@@ -0,0 +1,53 @@
+/* ===-- clzsi2.c - Implement __clzsi2 -------------------------------------===
+ *
+ *      	       The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __clzsi2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: the number of leading 0-bits */
+
+/* Precondition: a != 0 */
+
+si_int
+__clzsi2(si_int a)
+{
+    su_int x = (su_int)a;
+    si_int t = ((x & 0xFFFF0000) == 0) << 4;  /* if (x is small) t = 16 else 0 */
+    x >>= 16 - t;      /* x = [0 - 0xFFFF] */
+    su_int r = t;       /* r = [0, 16] */
+    /* return r + clz(x) */
+    t = ((x & 0xFF00) == 0) << 3;
+    x >>= 8 - t;       /* x = [0 - 0xFF] */
+    r += t;            /* r = [0, 8, 16, 24] */
+    /* return r + clz(x) */
+    t = ((x & 0xF0) == 0) << 2;
+    x >>= 4 - t;       /* x = [0 - 0xF] */
+    r += t;            /* r = [0, 4, 8, 12, 16, 20, 24, 28] */
+    /* return r + clz(x) */
+    t = ((x & 0xC) == 0) << 1;
+    x >>= 2 - t;       /* x = [0 - 3] */
+    r += t;            /* r = [0 - 30] and is even */
+    /* return r + clz(x) */
+/*     switch (x)
+ *     {
+ *     case 0:
+ *         return r + 2;
+ *     case 1:
+ *         return r + 1;
+ *     case 2:
+ *     case 3:
+ *         return r;
+ *     }
+ */
+    return r + ((2 - x) & -((x & 2) == 0));
+}

contrib/compiler-rt/lib/clzti2.c

@@ -0,0 +1,33 @@
+/* ===-- clzti2.c - Implement __clzti2 -------------------------------------===
+ *
+ *      	       The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __clzti2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+
+/* Returns: the number of leading 0-bits */
+
+/* Precondition: a != 0 */
+
+si_int
+__clzti2(ti_int a)
+{
+    twords x;
+    x.all = a;
+    const di_int f = -(x.s.high == 0);
+    return __builtin_clzll((x.s.high & ~f) | (x.s.low & f)) +
+           ((si_int)f & ((si_int)(sizeof(di_int) * CHAR_BIT)));
+}
+
+#endif /* __x86_64 */

contrib/compiler-rt/lib/cmpdi2.c

@@ -0,0 +1,38 @@
+/* ===-- cmpdi2.c - Implement __cmpdi2 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ * 
+ * This file implements __cmpdi2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns:  if (a <  b) returns 0
+*           if (a == b) returns 1
+*           if (a >  b) returns 2
+*/
+
+si_int
+__cmpdi2(di_int a, di_int b)
+{
+    dwords x;
+    x.all = a;
+    dwords y;
+    y.all = b;
+    if (x.s.high < y.s.high)
+        return 0;
+    if (x.s.high > y.s.high)
+        return 2;
+    if (x.s.low < y.s.low)
+        return 0;
+    if (x.s.low > y.s.low)
+        return 2;
+    return 1;
+}

contrib/compiler-rt/lib/cmpti2.c

@@ -0,0 +1,42 @@
+/* ===-- cmpti2.c - Implement __cmpti2 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __cmpti2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+
+/* Returns:  if (a <  b) returns 0
+ *           if (a == b) returns 1
+ *           if (a >  b) returns 2
+ */
+
+si_int
+__cmpti2(ti_int a, ti_int b)
+{
+    twords x;
+    x.all = a;
+    twords y;
+    y.all = b;
+    if (x.s.high < y.s.high)
+        return 0;
+    if (x.s.high > y.s.high)
+        return 2;
+    if (x.s.low < y.s.low)
+        return 0;
+    if (x.s.low > y.s.low)
+        return 2;
+    return 1;
+}
+
+#endif

contrib/compiler-rt/lib/comparedf2.c

@@ -0,0 +1,132 @@
+//===-- lib/comparedf2.c - Double-precision comparisons -----------*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// // This file implements the following soft-float comparison routines:
+//
+//   __eqdf2   __gedf2   __unorddf2
+//   __ledf2   __gtdf2
+//   __ltdf2
+//   __nedf2
+//
+// The semantics of the routines grouped in each column are identical, so there
+// is a single implementation for each, and wrappers to provide the other names.
+//
+// The main routines behave as follows:
+//
+//   __ledf2(a,b) returns -1 if a < b
+//                         0 if a == b
+//                         1 if a > b
+//                         1 if either a or b is NaN
+//
+//   __gedf2(a,b) returns -1 if a < b
+//                         0 if a == b
+//                         1 if a > b
+//                        -1 if either a or b is NaN
+//
+//   __unorddf2(a,b) returns 0 if both a and b are numbers
+//                           1 if either a or b is NaN
+//
+// Note that __ledf2( ) and __gedf2( ) are identical except in their handling of
+// NaN values.
+//
+//===----------------------------------------------------------------------===//
+
+#define DOUBLE_PRECISION
+#include "fp_lib.h"
+
+enum LE_RESULT {
+    LE_LESS      = -1,
+    LE_EQUAL     =  0,
+    LE_GREATER   =  1,
+    LE_UNORDERED =  1
+};
+
+enum LE_RESULT __ledf2(fp_t a, fp_t b) {
+    
+    const srep_t aInt = toRep(a);
+    const srep_t bInt = toRep(b);
+    const rep_t aAbs = aInt & absMask;
+    const rep_t bAbs = bInt & absMask;
+    
+    // If either a or b is NaN, they are unordered.
+    if (aAbs > infRep || bAbs > infRep) return LE_UNORDERED;
+    
+    // If a and b are both zeros, they are equal.
+    if ((aAbs | bAbs) == 0) return LE_EQUAL;
+    
+    // If at least one of a and b is positive, we get the same result comparing
+    // a and b as signed integers as we would with a floating-point compare.
+    if ((aInt & bInt) >= 0) {
+        if (aInt < bInt) return LE_LESS;
+        else if (aInt == bInt) return LE_EQUAL;
+        else return LE_GREATER;
+    }
+    
+    // Otherwise, both are negative, so we need to flip the sense of the
+    // comparison to get the correct result.  (This assumes a twos- or ones-
+    // complement integer representation; if integers are represented in a
+    // sign-magnitude representation, then this flip is incorrect).
+    else {
+        if (aInt > bInt) return LE_LESS;
+        else if (aInt == bInt) return LE_EQUAL;
+        else return LE_GREATER;
+    }
+}
+
+enum GE_RESULT {
+    GE_LESS      = -1,
+    GE_EQUAL     =  0,
+    GE_GREATER   =  1,
+    GE_UNORDERED = -1   // Note: different from LE_UNORDERED
+};
+
+enum GE_RESULT __gedf2(fp_t a, fp_t b) {
+    
+    const srep_t aInt = toRep(a);
+    const srep_t bInt = toRep(b);
+    const rep_t aAbs = aInt & absMask;
+    const rep_t bAbs = bInt & absMask;
+    
+    if (aAbs > infRep || bAbs > infRep) return GE_UNORDERED;
+    if ((aAbs | bAbs) == 0) return GE_EQUAL;
+    if ((aInt & bInt) >= 0) {
+        if (aInt < bInt) return GE_LESS;
+        else if (aInt == bInt) return GE_EQUAL;
+        else return GE_GREATER;
+    } else {
+        if (aInt > bInt) return GE_LESS;
+        else if (aInt == bInt) return GE_EQUAL;
+        else return GE_GREATER;
+    }
+}
+
+int __unorddf2(fp_t a, fp_t b) {
+    const rep_t aAbs = toRep(a) & absMask;
+    const rep_t bAbs = toRep(b) & absMask;
+    return aAbs > infRep || bAbs > infRep;
+}
+
+// The following are alternative names for the preceeding routines.
+
+enum LE_RESULT __eqdf2(fp_t a, fp_t b) {
+    return __ledf2(a, b);
+}
+
+enum LE_RESULT __ltdf2(fp_t a, fp_t b) {
+    return __ledf2(a, b);
+}
+
+enum LE_RESULT __nedf2(fp_t a, fp_t b) {
+    return __ledf2(a, b);
+}
+
+enum GE_RESULT __gtdf2(fp_t a, fp_t b) {
+    return __gedf2(a, b);
+}
+

contrib/compiler-rt/lib/comparesf2.c

@@ -0,0 +1,131 @@
+//===-- lib/comparesf2.c - Single-precision comparisons -----------*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the following soft-fp_t comparison routines:
+//
+//   __eqsf2   __gesf2   __unordsf2
+//   __lesf2   __gtsf2
+//   __ltsf2
+//   __nesf2
+//
+// The semantics of the routines grouped in each column are identical, so there
+// is a single implementation for each, and wrappers to provide the other names.
+//
+// The main routines behave as follows:
+//
+//   __lesf2(a,b) returns -1 if a < b
+//                         0 if a == b
+//                         1 if a > b
+//                         1 if either a or b is NaN
+//
+//   __gesf2(a,b) returns -1 if a < b
+//                         0 if a == b
+//                         1 if a > b
+//                        -1 if either a or b is NaN
+//
+//   __unordsf2(a,b) returns 0 if both a and b are numbers
+//                           1 if either a or b is NaN
+//
+// Note that __lesf2( ) and __gesf2( ) are identical except in their handling of
+// NaN values.
+//
+//===----------------------------------------------------------------------===//
+
+#define SINGLE_PRECISION
+#include "fp_lib.h"
+
+enum LE_RESULT {
+    LE_LESS      = -1,
+    LE_EQUAL     =  0,
+    LE_GREATER   =  1,
+    LE_UNORDERED =  1
+};
+
+enum LE_RESULT __lesf2(fp_t a, fp_t b) {
+    
+    const srep_t aInt = toRep(a);
+    const srep_t bInt = toRep(b);
+    const rep_t aAbs = aInt & absMask;
+    const rep_t bAbs = bInt & absMask;
+    
+    // If either a or b is NaN, they are unordered.
+    if (aAbs > infRep || bAbs > infRep) return LE_UNORDERED;
+    
+    // If a and b are both zeros, they are equal.
+    if ((aAbs | bAbs) == 0) return LE_EQUAL;
+    
+    // If at least one of a and b is positive, we get the same result comparing
+    // a and b as signed integers as we would with a fp_ting-point compare.
+    if ((aInt & bInt) >= 0) {
+        if (aInt < bInt) return LE_LESS;
+        else if (aInt == bInt) return LE_EQUAL;
+        else return LE_GREATER;
+    }
+    
+    // Otherwise, both are negative, so we need to flip the sense of the
+    // comparison to get the correct result.  (This assumes a twos- or ones-
+    // complement integer representation; if integers are represented in a
+    // sign-magnitude representation, then this flip is incorrect).
+    else {
+        if (aInt > bInt) return LE_LESS;
+        else if (aInt == bInt) return LE_EQUAL;
+        else return LE_GREATER;
+    }
+}
+
+enum GE_RESULT {
+    GE_LESS      = -1,
+    GE_EQUAL     =  0,
+    GE_GREATER   =  1,
+    GE_UNORDERED = -1   // Note: different from LE_UNORDERED
+};
+
+enum GE_RESULT __gesf2(fp_t a, fp_t b) {
+    
+    const srep_t aInt = toRep(a);
+    const srep_t bInt = toRep(b);
+    const rep_t aAbs = aInt & absMask;
+    const rep_t bAbs = bInt & absMask;
+    
+    if (aAbs > infRep || bAbs > infRep) return GE_UNORDERED;
+    if ((aAbs | bAbs) == 0) return GE_EQUAL;
+    if ((aInt & bInt) >= 0) {
+        if (aInt < bInt) return GE_LESS;
+        else if (aInt == bInt) return GE_EQUAL;
+        else return GE_GREATER;
+    } else {
+        if (aInt > bInt) return GE_LESS;
+        else if (aInt == bInt) return GE_EQUAL;
+        else return GE_GREATER;
+    }
+}
+
+int __unordsf2(fp_t a, fp_t b) {
+    const rep_t aAbs = toRep(a) & absMask;
+    const rep_t bAbs = toRep(b) & absMask;
+    return aAbs > infRep || bAbs > infRep;
+}
+
+// The following are alternative names for the preceeding routines.
+
+enum LE_RESULT __eqsf2(fp_t a, fp_t b) {
+    return __lesf2(a, b);
+}
+
+enum LE_RESULT __ltsf2(fp_t a, fp_t b) {
+    return __lesf2(a, b);
+}
+
+enum LE_RESULT __nesf2(fp_t a, fp_t b) {
+    return __lesf2(a, b);
+}
+
+enum GE_RESULT __gtsf2(fp_t a, fp_t b) {
+    return __gesf2(a, b);
+}

contrib/compiler-rt/lib/ctzdi2.c

@@ -0,0 +1,29 @@
+/* ===-- ctzdi2.c - Implement __ctzdi2 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __ctzdi2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: the number of trailing 0-bits  */
+
+/* Precondition: a != 0 */
+
+si_int
+__ctzdi2(di_int a)
+{
+    dwords x;
+    x.all = a;
+    const si_int f = -(x.s.low == 0);
+    return __builtin_ctz((x.s.high & f) | (x.s.low & ~f)) +
+              (f & ((si_int)(sizeof(si_int) * CHAR_BIT)));
+}

contrib/compiler-rt/lib/ctzsi2.c

@@ -0,0 +1,57 @@
+/* ===-- ctzsi2.c - Implement __ctzsi2 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __ctzsi2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: the number of trailing 0-bits */
+
+/* Precondition: a != 0 */
+
+si_int
+__ctzsi2(si_int a)
+{
+    su_int x = (su_int)a;
+    si_int t = ((x & 0x0000FFFF) == 0) << 4;  /* if (x has no small bits) t = 16 else 0 */
+    x >>= t;           /* x = [0 - 0xFFFF] + higher garbage bits */
+    su_int r = t;       /* r = [0, 16]  */
+    /* return r + ctz(x) */
+    t = ((x & 0x00FF) == 0) << 3;
+    x >>= t;           /* x = [0 - 0xFF] + higher garbage bits */
+    r += t;            /* r = [0, 8, 16, 24] */
+    /* return r + ctz(x) */
+    t = ((x & 0x0F) == 0) << 2;
+    x >>= t;           /* x = [0 - 0xF] + higher garbage bits */
+    r += t;            /* r = [0, 4, 8, 12, 16, 20, 24, 28] */
+    /* return r + ctz(x) */
+    t = ((x & 0x3) == 0) << 1;
+    x >>= t;
+    x &= 3;            /* x = [0 - 3] */
+    r += t;            /* r = [0 - 30] and is even */
+    /* return r + ctz(x) */
+
+/*  The branch-less return statement below is equivalent
+ *  to the following switch statement:
+ *     switch (x)
+ *    {
+ *     case 0:
+ *         return r + 2;
+ *     case 2:
+ *         return r + 1;
+ *     case 1:
+ *     case 3:
+ *         return r;
+ *     }
+ */
+    return r + ((2 - (x >> 1)) & -((x & 1) == 0));
+}

contrib/compiler-rt/lib/ctzti2.c

@@ -0,0 +1,33 @@
+/* ===-- ctzti2.c - Implement __ctzti2 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __ctzti2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+
+/* Returns: the number of trailing 0-bits */
+
+/* Precondition: a != 0 */
+
+si_int
+__ctzti2(ti_int a)
+{
+    twords x;
+    x.all = a;
+    const di_int f = -(x.s.low == 0);
+    return __builtin_ctzll((x.s.high & f) | (x.s.low & ~f)) +
+              ((si_int)f & ((si_int)(sizeof(di_int) * CHAR_BIT)));
+}
+
+#endif

contrib/compiler-rt/lib/divdc3.c

@@ -0,0 +1,59 @@
+/* ===-- divdc3.c - Implement __divdc3 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __divdc3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+#include <math.h>
+#include <complex.h>
+
+/* Returns: the quotient of (a + ib) / (c + id) */
+
+double _Complex
+__divdc3(double __a, double __b, double __c, double __d)
+{
+    int __ilogbw = 0;
+    double __logbw = logb(fmax(fabs(__c), fabs(__d)));
+    if (isfinite(__logbw))
+    {
+        __ilogbw = (int)__logbw;
+        __c = scalbn(__c, -__ilogbw);
+        __d = scalbn(__d, -__ilogbw);
+    }
+    double __denom = __c * __c + __d * __d;
+    double _Complex z;
+    __real__ z = scalbn((__a * __c + __b * __d) / __denom, -__ilogbw);
+    __imag__ z = scalbn((__b * __c - __a * __d) / __denom, -__ilogbw);
+    if (isnan(__real__ z) && isnan(__imag__ z))
+    {
+        if ((__denom == 0.0) && (!isnan(__a) || !isnan(__b)))
+        {
+            __real__ z = copysign(INFINITY, __c) * __a;
+            __imag__ z = copysign(INFINITY, __c) * __b;
+        }
+        else if ((isinf(__a) || isinf(__b)) && isfinite(__c) && isfinite(__d))
+        {
+            __a = copysign(isinf(__a) ? 1.0 : 0.0, __a);
+            __b = copysign(isinf(__b) ? 1.0 : 0.0, __b);
+            __real__ z = INFINITY * (__a * __c + __b * __d);
+            __imag__ z = INFINITY * (__b * __c - __a * __d);
+        }
+        else if (isinf(__logbw) && __logbw > 0.0 && isfinite(__a) && isfinite(__b))
+        {
+            __c = copysign(isinf(__c) ? 1.0 : 0.0, __c);
+            __d = copysign(isinf(__d) ? 1.0 : 0.0, __d);
+            __real__ z = 0.0 * (__a * __c + __b * __d);
+            __imag__ z = 0.0 * (__b * __c - __a * __d);
+        }
+    }
+    return z;
+}

contrib/compiler-rt/lib/divdf3.c

@@ -0,0 +1,182 @@
+//===-- lib/divdf3.c - Double-precision division ------------------*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements double-precision soft-float division
+// with the IEEE-754 default rounding (to nearest, ties to even).
+//
+// For simplicity, this implementation currently flushes denormals to zero.
+// It should be a fairly straightforward exercise to implement gradual
+// underflow with correct rounding.
+//
+//===----------------------------------------------------------------------===//
+
+#define DOUBLE_PRECISION
+#include "fp_lib.h"
+
+fp_t __divdf3(fp_t a, fp_t b) {
+    
+    const unsigned int aExponent = toRep(a) >> significandBits & maxExponent;
+    const unsigned int bExponent = toRep(b) >> significandBits & maxExponent;
+    const rep_t quotientSign = (toRep(a) ^ toRep(b)) & signBit;
+    
+    rep_t aSignificand = toRep(a) & significandMask;
+    rep_t bSignificand = toRep(b) & significandMask;
+    int scale = 0;
+    
+    // Detect if a or b is zero, denormal, infinity, or NaN.
+    if (aExponent-1U >= maxExponent-1U || bExponent-1U >= maxExponent-1U) {
+        
+        const rep_t aAbs = toRep(a) & absMask;
+        const rep_t bAbs = toRep(b) & absMask;
+        
+        // NaN / anything = qNaN
+        if (aAbs > infRep) return fromRep(toRep(a) | quietBit);
+        // anything / NaN = qNaN
+        if (bAbs > infRep) return fromRep(toRep(b) | quietBit);
+        
+        if (aAbs == infRep) {
+            // infinity / infinity = NaN
+            if (bAbs == infRep) return fromRep(qnanRep);
+            // infinity / anything else = +/- infinity
+            else return fromRep(aAbs | quotientSign);
+        }
+        
+        // anything else / infinity = +/- 0
+        if (bAbs == infRep) return fromRep(quotientSign);
+        
+        if (!aAbs) {
+            // zero / zero = NaN
+            if (!bAbs) return fromRep(qnanRep);
+            // zero / anything else = +/- zero
+            else return fromRep(quotientSign);
+        }
+        // anything else / zero = +/- infinity
+        if (!bAbs) return fromRep(infRep | quotientSign);
+        
+        // one or both of a or b is denormal, the other (if applicable) is a
+        // normal number.  Renormalize one or both of a and b, and set scale to
+        // include the necessary exponent adjustment.
+        if (aAbs < implicitBit) scale += normalize(&aSignificand);
+        if (bAbs < implicitBit) scale -= normalize(&bSignificand);
+    }
+    
+    // Or in the implicit significand bit.  (If we fell through from the
+    // denormal path it was already set by normalize( ), but setting it twice
+    // won't hurt anything.)
+    aSignificand |= implicitBit;
+    bSignificand |= implicitBit;
+    int quotientExponent = aExponent - bExponent + scale;
+    
+    // Align the significand of b as a Q31 fixed-point number in the range
+    // [1, 2.0) and get a Q32 approximate reciprocal using a small minimax
+    // polynomial approximation: reciprocal = 3/4 + 1/sqrt(2) - b/2.  This
+    // is accurate to about 3.5 binary digits.
+    const uint32_t q31b = bSignificand >> 21;
+    uint32_t recip32 = UINT32_C(0x7504f333) - q31b;
+    
+    // Now refine the reciprocal estimate using a Newton-Raphson iteration:
+    //
+    //     x1 = x0 * (2 - x0 * b)
+    //
+    // This doubles the number of correct binary digits in the approximation
+    // with each iteration, so after three iterations, we have about 28 binary
+    // digits of accuracy.
+    uint32_t correction32;
+    correction32 = -((uint64_t)recip32 * q31b >> 32);
+    recip32 = (uint64_t)recip32 * correction32 >> 31;
+    correction32 = -((uint64_t)recip32 * q31b >> 32);
+    recip32 = (uint64_t)recip32 * correction32 >> 31;
+    correction32 = -((uint64_t)recip32 * q31b >> 32);
+    recip32 = (uint64_t)recip32 * correction32 >> 31;
+    
+    // recip32 might have overflowed to exactly zero in the preceeding
+    // computation if the high word of b is exactly 1.0.  This would sabotage
+    // the full-width final stage of the computation that follows, so we adjust
+    // recip32 downward by one bit.
+    recip32--;
+    
+    // We need to perform one more iteration to get us to 56 binary digits;
+    // The last iteration needs to happen with extra precision.
+    const uint32_t q63blo = bSignificand << 11;
+    uint64_t correction, reciprocal;
+    correction = -((uint64_t)recip32*q31b + ((uint64_t)recip32*q63blo >> 32));
+    uint32_t cHi = correction >> 32;
+    uint32_t cLo = correction;
+    reciprocal = (uint64_t)recip32*cHi + ((uint64_t)recip32*cLo >> 32);
+    
+    // We already adjusted the 32-bit estimate, now we need to adjust the final
+    // 64-bit reciprocal estimate downward to ensure that it is strictly smaller
+    // than the infinitely precise exact reciprocal.  Because the computation
+    // of the Newton-Raphson step is truncating at every step, this adjustment
+    // is small; most of the work is already done.
+    reciprocal -= 2;
+    
+    // The numerical reciprocal is accurate to within 2^-56, lies in the
+    // interval [0.5, 1.0), and is strictly smaller than the true reciprocal
+    // of b.  Multiplying a by this reciprocal thus gives a numerical q = a/b
+    // in Q53 with the following properties:
+    //
+    //    1. q < a/b
+    //    2. q is in the interval [0.5, 2.0)
+    //    3. the error in q is bounded away from 2^-53 (actually, we have a
+    //       couple of bits to spare, but this is all we need).
+    
+    // We need a 64 x 64 multiply high to compute q, which isn't a basic
+    // operation in C, so we need to be a little bit fussy.
+    rep_t quotient, quotientLo;
+    wideMultiply(aSignificand << 2, reciprocal, &quotient, &quotientLo);
+    
+    // Two cases: quotient is in [0.5, 1.0) or quotient is in [1.0, 2.0).
+    // In either case, we are going to compute a residual of the form
+    //
+    //     r = a - q*b
+    //
+    // We know from the construction of q that r satisfies:
+    //
+    //     0 <= r < ulp(q)*b
+    // 
+    // if r is greater than 1/2 ulp(q)*b, then q rounds up.  Otherwise, we
+    // already have the correct result.  The exact halfway case cannot occur.
+    // We also take this time to right shift quotient if it falls in the [1,2)
+    // range and adjust the exponent accordingly.
+    rep_t residual;
+    if (quotient < (implicitBit << 1)) {
+        residual = (aSignificand << 53) - quotient * bSignificand;
+        quotientExponent--;
+    } else {
+        quotient >>= 1;
+        residual = (aSignificand << 52) - quotient * bSignificand;
+    }
+    
+    const int writtenExponent = quotientExponent + exponentBias;
+    
+    if (writtenExponent >= maxExponent) {
+        // If we have overflowed the exponent, return infinity.
+        return fromRep(infRep | quotientSign);
+    }
+    
+    else if (writtenExponent < 1) {
+        // Flush denormals to zero.  In the future, it would be nice to add
+        // code to round them correctly.
+        return fromRep(quotientSign);
+    }
+    
+    else {
+        const bool round = (residual << 1) > bSignificand;
+        // Clear the implicit bit
+        rep_t absResult = quotient & significandMask;
+        // Insert the exponent
+        absResult |= (rep_t)writtenExponent << significandBits;
+        // Round
+        absResult += round;
+        // Insert the sign and return
+        const double result = fromRep(absResult | quotientSign);
+        return result;
+    }
+}

contrib/compiler-rt/lib/divdi3.c

@@ -0,0 +1,31 @@
+/* ===-- divdi3.c - Implement __divdi3 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __divdi3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+du_int __udivmoddi4(du_int a, du_int b, du_int* rem);
+
+/* Returns: a / b */
+
+di_int
+__divdi3(di_int a, di_int b)
+{
+    const int bits_in_dword_m1 = (int)(sizeof(di_int) * CHAR_BIT) - 1;
+    di_int s_a = a >> bits_in_dword_m1;           /* s_a = a < 0 ? -1 : 0 */
+    di_int s_b = b >> bits_in_dword_m1;           /* s_b = b < 0 ? -1 : 0 */
+    a = (a ^ s_a) - s_a;                         /* negate if s_a == -1 */
+    b = (b ^ s_b) - s_b;                         /* negate if s_b == -1 */
+    s_a ^= s_b;                                  /*sign of quotient */
+    return (__udivmoddi4(a, b, (du_int*)0) ^ s_a) - s_a;  /* negate if s_a == -1 */
+}

contrib/compiler-rt/lib/divsc3.c

@@ -0,0 +1,59 @@
+/*===-- divsc3.c - Implement __divsc3 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __divsc3 for the compiler_rt library.
+ *
+ *===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+#include <math.h>
+#include <complex.h>
+
+/* Returns: the quotient of (a + ib) / (c + id) */
+
+float _Complex
+__divsc3(float __a, float __b, float __c, float __d)
+{
+    int __ilogbw = 0;
+    float __logbw = logbf(fmaxf(fabsf(__c), fabsf(__d)));
+    if (isfinite(__logbw))
+    {
+        __ilogbw = (int)__logbw;
+        __c = scalbnf(__c, -__ilogbw);
+        __d = scalbnf(__d, -__ilogbw);
+    }
+    float __denom = __c * __c + __d * __d;
+    float _Complex z;
+    __real__ z = scalbnf((__a * __c + __b * __d) / __denom, -__ilogbw);
+    __imag__ z = scalbnf((__b * __c - __a * __d) / __denom, -__ilogbw);
+    if (isnan(__real__ z) && isnan(__imag__ z))
+    {
+        if ((__denom == 0) && (!isnan(__a) || !isnan(__b)))
+        {
+            __real__ z = copysignf(INFINITY, __c) * __a;
+            __imag__ z = copysignf(INFINITY, __c) * __b;
+        }
+        else if ((isinf(__a) || isinf(__b)) && isfinite(__c) && isfinite(__d))
+        {
+            __a = copysignf(isinf(__a) ? 1 : 0, __a);
+            __b = copysignf(isinf(__b) ? 1 : 0, __b);
+            __real__ z = INFINITY * (__a * __c + __b * __d);
+            __imag__ z = INFINITY * (__b * __c - __a * __d);
+        }
+        else if (isinf(__logbw) && __logbw > 0 && isfinite(__a) && isfinite(__b))
+        {
+            __c = copysignf(isinf(__c) ? 1 : 0, __c);
+            __d = copysignf(isinf(__d) ? 1 : 0, __d);
+            __real__ z = 0 * (__a * __c + __b * __d);
+            __imag__ z = 0 * (__b * __c - __a * __d);
+        }
+    }
+    return z;
+}

contrib/compiler-rt/lib/divsf3.c

@@ -0,0 +1,166 @@
+//===-- lib/divsf3.c - Single-precision division ------------------*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements single-precision soft-float division
+// with the IEEE-754 default rounding (to nearest, ties to even).
+//
+// For simplicity, this implementation currently flushes denormals to zero.
+// It should be a fairly straightforward exercise to implement gradual
+// underflow with correct rounding.
+//
+//===----------------------------------------------------------------------===//
+
+#define SINGLE_PRECISION
+#include "fp_lib.h"
+
+fp_t __divsf3(fp_t a, fp_t b) {
+    
+    const unsigned int aExponent = toRep(a) >> significandBits & maxExponent;
+    const unsigned int bExponent = toRep(b) >> significandBits & maxExponent;
+    const rep_t quotientSign = (toRep(a) ^ toRep(b)) & signBit;
+    
+    rep_t aSignificand = toRep(a) & significandMask;
+    rep_t bSignificand = toRep(b) & significandMask;
+    int scale = 0;
+    
+    // Detect if a or b is zero, denormal, infinity, or NaN.
+    if (aExponent-1U >= maxExponent-1U || bExponent-1U >= maxExponent-1U) {
+        
+        const rep_t aAbs = toRep(a) & absMask;
+        const rep_t bAbs = toRep(b) & absMask;
+        
+        // NaN / anything = qNaN
+        if (aAbs > infRep) return fromRep(toRep(a) | quietBit);
+        // anything / NaN = qNaN
+        if (bAbs > infRep) return fromRep(toRep(b) | quietBit);
+        
+        if (aAbs == infRep) {
+            // infinity / infinity = NaN
+            if (bAbs == infRep) return fromRep(qnanRep);
+            // infinity / anything else = +/- infinity
+            else return fromRep(aAbs | quotientSign);
+        }
+        
+        // anything else / infinity = +/- 0
+        if (bAbs == infRep) return fromRep(quotientSign);
+        
+        if (!aAbs) {
+            // zero / zero = NaN
+            if (!bAbs) return fromRep(qnanRep);
+            // zero / anything else = +/- zero
+            else return fromRep(quotientSign);
+        }
+        // anything else / zero = +/- infinity
+        if (!bAbs) return fromRep(infRep | quotientSign);
+        
+        // one or both of a or b is denormal, the other (if applicable) is a
+        // normal number.  Renormalize one or both of a and b, and set scale to
+        // include the necessary exponent adjustment.
+        if (aAbs < implicitBit) scale += normalize(&aSignificand);
+        if (bAbs < implicitBit) scale -= normalize(&bSignificand);
+    }
+    
+    // Or in the implicit significand bit.  (If we fell through from the
+    // denormal path it was already set by normalize( ), but setting it twice
+    // won't hurt anything.)
+    aSignificand |= implicitBit;
+    bSignificand |= implicitBit;
+    int quotientExponent = aExponent - bExponent + scale;
+    
+    // Align the significand of b as a Q31 fixed-point number in the range
+    // [1, 2.0) and get a Q32 approximate reciprocal using a small minimax
+    // polynomial approximation: reciprocal = 3/4 + 1/sqrt(2) - b/2.  This
+    // is accurate to about 3.5 binary digits.
+    uint32_t q31b = bSignificand << 8;
+    uint32_t reciprocal = UINT32_C(0x7504f333) - q31b;
+    
+    // Now refine the reciprocal estimate using a Newton-Raphson iteration:
+    //
+    //     x1 = x0 * (2 - x0 * b)
+    //
+    // This doubles the number of correct binary digits in the approximation
+    // with each iteration, so after three iterations, we have about 28 binary
+    // digits of accuracy.
+    uint32_t correction;
+    correction = -((uint64_t)reciprocal * q31b >> 32);
+    reciprocal = (uint64_t)reciprocal * correction >> 31;
+    correction = -((uint64_t)reciprocal * q31b >> 32);
+    reciprocal = (uint64_t)reciprocal * correction >> 31;
+    correction = -((uint64_t)reciprocal * q31b >> 32);
+    reciprocal = (uint64_t)reciprocal * correction >> 31;
+    
+    // Exhaustive testing shows that the error in reciprocal after three steps
+    // is in the interval [-0x1.f58108p-31, 0x1.d0e48cp-29], in line with our
+    // expectations.  We bump the reciprocal by a tiny value to force the error
+    // to be strictly positive (in the range [0x1.4fdfp-37,0x1.287246p-29], to
+    // be specific).  This also causes 1/1 to give a sensible approximation
+    // instead of zero (due to overflow).
+    reciprocal -= 2;
+    
+    // The numerical reciprocal is accurate to within 2^-28, lies in the
+    // interval [0x1.000000eep-1, 0x1.fffffffcp-1], and is strictly smaller
+    // than the true reciprocal of b.  Multiplying a by this reciprocal thus
+    // gives a numerical q = a/b in Q24 with the following properties:
+    //
+    //    1. q < a/b
+    //    2. q is in the interval [0x1.000000eep-1, 0x1.fffffffcp0)
+    //    3. the error in q is at most 2^-24 + 2^-27 -- the 2^24 term comes
+    //       from the fact that we truncate the product, and the 2^27 term
+    //       is the error in the reciprocal of b scaled by the maximum
+    //       possible value of a.  As a consequence of this error bound,
+    //       either q or nextafter(q) is the correctly rounded 
+    rep_t quotient = (uint64_t)reciprocal*(aSignificand << 1) >> 32;
+    
+    // Two cases: quotient is in [0.5, 1.0) or quotient is in [1.0, 2.0).
+    // In either case, we are going to compute a residual of the form
+    //
+    //     r = a - q*b
+    //
+    // We know from the construction of q that r satisfies:
+    //
+    //     0 <= r < ulp(q)*b
+    // 
+    // if r is greater than 1/2 ulp(q)*b, then q rounds up.  Otherwise, we
+    // already have the correct result.  The exact halfway case cannot occur.
+    // We also take this time to right shift quotient if it falls in the [1,2)
+    // range and adjust the exponent accordingly.
+    rep_t residual;
+    if (quotient < (implicitBit << 1)) {
+        residual = (aSignificand << 24) - quotient * bSignificand;
+        quotientExponent--;
+    } else {
+        quotient >>= 1;
+        residual = (aSignificand << 23) - quotient * bSignificand;
+    }
+
+    const int writtenExponent = quotientExponent + exponentBias;
+    
+    if (writtenExponent >= maxExponent) {
+        // If we have overflowed the exponent, return infinity.
+        return fromRep(infRep | quotientSign);
+    }
+    
+    else if (writtenExponent < 1) {
+        // Flush denormals to zero.  In the future, it would be nice to add
+        // code to round them correctly.
+        return fromRep(quotientSign);
+    }
+    
+    else {
+        const bool round = (residual << 1) > bSignificand;
+        // Clear the implicit bit
+        rep_t absResult = quotient & significandMask;
+        // Insert the exponent
+        absResult |= (rep_t)writtenExponent << significandBits;
+        // Round
+        absResult += round;
+        // Insert the sign and return
+        return fromRep(absResult | quotientSign);
+    }
+}

contrib/compiler-rt/lib/divsi3.c

@@ -0,0 +1,31 @@
+/* ===-- divsi3.c - Implement __divsi3 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __divsi3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+su_int __udivsi3(su_int n, su_int d);
+
+/* Returns: a / b */
+
+si_int
+__divsi3(si_int a, si_int b)
+{
+    const int bits_in_word_m1 = (int)(sizeof(si_int) * CHAR_BIT) - 1;
+    si_int s_a = a >> bits_in_word_m1;           /* s_a = a < 0 ? -1 : 0 */
+    si_int s_b = b >> bits_in_word_m1;           /* s_b = b < 0 ? -1 : 0 */
+    a = (a ^ s_a) - s_a;                         /* negate if s_a == -1 */
+    b = (b ^ s_b) - s_b;                         /* negate if s_b == -1 */
+    s_a ^= s_b;                                  /* sign of quotient */
+    return (__udivsi3(a, b) ^ s_a) - s_a;        /* negate if s_a == -1 */
+}

contrib/compiler-rt/lib/divti3.c

@@ -0,0 +1,35 @@
+/* ===-- divti3.c - Implement __divti3 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __divti3 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+
+tu_int __udivmodti4(tu_int a, tu_int b, tu_int* rem);
+
+/* Returns: a / b */
+
+ti_int
+__divti3(ti_int a, ti_int b)
+{
+    const int bits_in_tword_m1 = (int)(sizeof(ti_int) * CHAR_BIT) - 1;
+    ti_int s_a = a >> bits_in_tword_m1;           /* s_a = a < 0 ? -1 : 0 */
+    ti_int s_b = b >> bits_in_tword_m1;           /* s_b = b < 0 ? -1 : 0 */
+    a = (a ^ s_a) - s_a;                         /* negate if s_a == -1 */
+    b = (b ^ s_b) - s_b;                         /* negate if s_b == -1 */
+    s_a ^= s_b;                                  /* sign of quotient */
+    return (__udivmodti4(a, b, (tu_int*)0) ^ s_a) - s_a;  /* negate if s_a == -1 */
+}
+
+#endif

contrib/compiler-rt/lib/divxc3.c

@@ -0,0 +1,62 @@
+/* ===-- divxc3.c - Implement __divxc3 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __divxc3 for the compiler_rt library.
+ *
+ */
+
+#if !_ARCH_PPC
+
+#include "int_lib.h"
+#include <math.h>
+#include <complex.h>
+
+/* Returns: the quotient of (a + ib) / (c + id) */
+
+long double _Complex
+__divxc3(long double __a, long double __b, long double __c, long double __d)
+{
+    int __ilogbw = 0;
+    long double __logbw = logbl(fmaxl(fabsl(__c), fabsl(__d)));
+    if (isfinite(__logbw))
+    {
+        __ilogbw = (int)__logbw;
+        __c = scalbnl(__c, -__ilogbw);
+        __d = scalbnl(__d, -__ilogbw);
+    }
+    long double __denom = __c * __c + __d * __d;
+    long double _Complex z;
+    __real__ z = scalbnl((__a * __c + __b * __d) / __denom, -__ilogbw);
+    __imag__ z = scalbnl((__b * __c - __a * __d) / __denom, -__ilogbw);
+    if (isnan(__real__ z) && isnan(__imag__ z))
+    {
+        if ((__denom == 0) && (!isnan(__a) || !isnan(__b)))
+        {
+            __real__ z = copysignl(INFINITY, __c) * __a;
+            __imag__ z = copysignl(INFINITY, __c) * __b;
+        }
+        else if ((isinf(__a) || isinf(__b)) && isfinite(__c) && isfinite(__d))
+        {
+            __a = copysignl(isinf(__a) ? 1 : 0, __a);
+            __b = copysignl(isinf(__b) ? 1 : 0, __b);
+            __real__ z = INFINITY * (__a * __c + __b * __d);
+            __imag__ z = INFINITY * (__b * __c - __a * __d);
+        }
+        else if (isinf(__logbw) && __logbw > 0 && isfinite(__a) && isfinite(__b))
+        {
+            __c = copysignl(isinf(__c) ? 1 : 0, __c);
+            __d = copysignl(isinf(__d) ? 1 : 0, __d);
+            __real__ z = 0 * (__a * __c + __b * __d);
+            __imag__ z = 0 * (__b * __c - __a * __d);
+        }
+    }
+    return z;
+}
+
+#endif

contrib/compiler-rt/lib/enable_execute_stack.c

@@ -0,0 +1,58 @@
+/* ===-- enable_execute_stack.c - Implement __enable_execute_stack ---------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include <stdint.h>
+#include <sys/mman.h>
+
+/* #include "config.h"
+ * FIXME: CMake - include when cmake system is ready.
+ * Remove #define HAVE_SYSCONF 1 line.
+ */
+#define HAVE_SYSCONF 1
+
+#ifndef __APPLE__
+#include <unistd.h>
+#endif /* __APPLE__ */
+
+#if __LP64__
+	#define TRAMPOLINE_SIZE 48
+#else
+	#define TRAMPOLINE_SIZE 40
+#endif
+
+/*
+ * The compiler generates calls to __enable_execute_stack() when creating 
+ * trampoline functions on the stack for use with nested functions.
+ * It is expected to mark the page(s) containing the address 
+ * and the next 48 bytes as executable.  Since the stack is normally rw-
+ * that means changing the protection on those page(s) to rwx. 
+ */
+
+void __enable_execute_stack(void* addr)
+{
+
+#if __APPLE__
+	/* On Darwin, pagesize is always 4096 bytes */
+	const uintptr_t pageSize = 4096;
+#elif !defined(HAVE_SYSCONF)
+#error "HAVE_SYSCONF not defined! See enable_execute_stack.c"
+#else
+        const uintptr_t pageSize = sysconf(_SC_PAGESIZE);
+#endif /* __APPLE__ */
+
+	const uintptr_t pageAlignMask = ~(pageSize-1);
+	uintptr_t p = (uintptr_t)addr;
+	unsigned char* startPage = (unsigned char*)(p & pageAlignMask);
+	unsigned char* endPage = (unsigned char*)((p+TRAMPOLINE_SIZE+pageSize) & pageAlignMask);
+	size_t length = endPage - startPage;
+	(void) mprotect((void *)startPage, length, PROT_READ | PROT_WRITE | PROT_EXEC);
+}
+
+

contrib/compiler-rt/lib/endianness.h

@@ -0,0 +1,94 @@
+/* ===-- endianness.h - configuration header for compiler-rt ---------------===
+ *
+ *		       The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file is a configuration header for compiler-rt.
+ * This file is not part of the interface of this library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#ifndef ENDIANNESS_H
+#define ENDIANNESS_H
+
+/*
+ * Known limitations:
+ *   Middle endian systems are not handled currently.
+ */
+
+#if defined(__SVR4) && defined(__sun)
+#include <sys/byteorder.h>
+
+#if _BYTE_ORDER == _BIG_ENDIAN
+#define _YUGA_LITTLE_ENDIAN 0
+#define _YUGA_BIG_ENDIAN    1
+#elif _BYTE_ORDER == _LITTLE_ENDIAN 
+#define _YUGA_LITTLE_ENDIAN 1
+#define _YUGA_BIG_ENDIAN    0
+#endif /* _BYTE_ORDER */
+
+#endif /* Solaris and AuroraUX. */
+
+/* .. */
+
+#if defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) || defined(__DragonFly__)
+#include <sys/endian.h>
+
+#if _BYTE_ORDER == _BIG_ENDIAN
+#define _YUGA_LITTLE_ENDIAN 0
+#define _YUGA_BIG_ENDIAN    1
+#elif _BYTE_ORDER == _LITTLE_ENDIAN
+#define _YUGA_LITTLE_ENDIAN 1
+#define _YUGA_BIG_ENDIAN    0
+#endif /* _BYTE_ORDER */
+
+#endif /* *BSD */
+
+/* .. */
+
+/* Mac OSX has __BIG_ENDIAN__ or __LITTLE_ENDIAN__ automatically set by the compiler (at least with GCC) */
+#if defined(__APPLE__) && defined(__MACH__) || defined(__ellcc__ )
+
+#ifdef __BIG_ENDIAN__
+#if __BIG_ENDIAN__
+#define _YUGA_LITTLE_ENDIAN 0
+#define _YUGA_BIG_ENDIAN    1
+#endif
+#endif /* __BIG_ENDIAN__ */
+
+#ifdef __LITTLE_ENDIAN__
+#if __LITTLE_ENDIAN__
+#define _YUGA_LITTLE_ENDIAN 1
+#define _YUGA_BIG_ENDIAN    0
+#endif
+#endif /* __LITTLE_ENDIAN__ */
+
+#endif /* Mac OSX */
+
+/* .. */
+
+#if defined(__linux__)
+#include <endian.h>
+
+#if __BYTE_ORDER == __BIG_ENDIAN
+#define _YUGA_LITTLE_ENDIAN 0
+#define _YUGA_BIG_ENDIAN    1
+#elif __BYTE_ORDER == __LITTLE_ENDIAN
+#define _YUGA_LITTLE_ENDIAN 1
+#define _YUGA_BIG_ENDIAN    0
+#endif /* __BYTE_ORDER */
+
+#endif /* GNU/Linux */
+
+/* . */
+
+#if !defined(_YUGA_LITTLE_ENDIAN) || !defined(_YUGA_BIG_ENDIAN)
+#error Unable to determine endian
+#endif /* Check we found an endianness correctly. */
+
+#endif /* ENDIANNESS_H */

contrib/compiler-rt/lib/eprintf.c

@@ -0,0 +1,33 @@
+/* ===---------- eprintf.c - Implements __eprintf --------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+
+
+#include "int_lib.h"
+#include <stdio.h>
+#include <stdlib.h>
+
+
+/*
+ * __eprintf() was used in an old version of <assert.h>.
+ * It can eventually go away, but it is needed when linking
+ * .o files built with the old <assert.h>.
+ *
+ * It should never be exported from a dylib, so it is marked
+ * visibility hidden.
+ */
+__attribute__((visibility("hidden")))
+void __eprintf(const char* format, const char* assertion_expression,
+				const char* line, const char* file)
+{
+	fprintf(stderr, format, assertion_expression, line, file);
+	fflush(stderr);
+	compilerrt_abort();
+}

contrib/compiler-rt/lib/extendsfdf2.c

@@ -0,0 +1,136 @@
+//===-- lib/extendsfdf2.c - single -> double conversion -----------*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements a fairly generic conversion from a narrower to a wider
+// IEEE-754 floating-point type.  The constants and types defined following the
+// includes below parameterize the conversion.
+//
+// This routine can be trivially adapted to support conversions from 
+// half-precision or to quad-precision. It does not support types that don't
+// use the usual IEEE-754 interchange formats; specifically, some work would be
+// needed to adapt it to (for example) the Intel 80-bit format or PowerPC
+// double-double format.
+//
+// Note please, however, that this implementation is only intended to support
+// *widening* operations; if you need to convert to a *narrower* floating-point
+// type (e.g. double -> float), then this routine will not do what you want it
+// to.
+//
+// It also requires that integer types at least as large as both formats
+// are available on the target platform; this may pose a problem when trying
+// to add support for quad on some 32-bit systems, for example.  You also may
+// run into trouble finding an appropriate CLZ function for wide source types;
+// you will likely need to roll your own on some platforms.
+//
+// Finally, the following assumptions are made:
+//
+// 1. floating-point types and integer types have the same endianness on the
+//    target platform
+//
+// 2. quiet NaNs, if supported, are indicated by the leading bit of the
+//    significand field being set
+//
+//===----------------------------------------------------------------------===//
+
+#include <stdint.h>
+#include <limits.h>
+
+typedef float src_t;
+typedef uint32_t src_rep_t;
+#define SRC_REP_C UINT32_C
+static const int srcSigBits = 23;
+#define src_rep_t_clz __builtin_clz
+
+typedef double dst_t;
+typedef uint64_t dst_rep_t;
+#define DST_REP_C UINT64_C
+static const int dstSigBits = 52;
+
+// End of specialization parameters.  Two helper routines for conversion to and
+// from the representation of floating-point data as integer values follow.
+
+static inline src_rep_t srcToRep(src_t x) {
+    const union { src_t f; src_rep_t i; } rep = {.f = x};
+    return rep.i;
+}
+
+static inline dst_t dstFromRep(dst_rep_t x) {
+    const union { dst_t f; dst_rep_t i; } rep = {.i = x};
+    return rep.f;
+}
+
+// End helper routines.  Conversion implementation follows.
+
+dst_t __extendsfdf2(src_t a) {
+    
+    // Various constants whose values follow from the type parameters.
+    // Any reasonable optimizer will fold and propagate all of these.
+    const int srcBits = sizeof(src_t)*CHAR_BIT;
+    const int srcExpBits = srcBits - srcSigBits - 1;
+    const int srcInfExp = (1 << srcExpBits) - 1;
+    const int srcExpBias = srcInfExp >> 1;
+    
+    const src_rep_t srcMinNormal = SRC_REP_C(1) << srcSigBits;
+    const src_rep_t srcInfinity = (src_rep_t)srcInfExp << srcSigBits;
+    const src_rep_t srcSignMask = SRC_REP_C(1) << (srcSigBits + srcExpBits);
+    const src_rep_t srcAbsMask = srcSignMask - 1;
+    const src_rep_t srcQNaN = SRC_REP_C(1) << (srcSigBits - 1);
+    const src_rep_t srcNaNCode = srcQNaN - 1;
+    
+    const int dstBits = sizeof(dst_t)*CHAR_BIT;
+    const int dstExpBits = dstBits - dstSigBits - 1;
+    const int dstInfExp = (1 << dstExpBits) - 1;
+    const int dstExpBias = dstInfExp >> 1;
+    
+    const dst_rep_t dstMinNormal = DST_REP_C(1) << dstSigBits;
+    
+    // Break a into a sign and representation of the absolute value
+    const src_rep_t aRep = srcToRep(a);
+    const src_rep_t aAbs = aRep & srcAbsMask;
+    const src_rep_t sign = aRep & srcSignMask;
+    dst_rep_t absResult;
+    
+    if (aAbs - srcMinNormal < srcInfinity - srcMinNormal) {
+        // a is a normal number.
+        // Extend to the destination type by shifting the significand and
+        // exponent into the proper position and rebiasing the exponent.
+        absResult = (dst_rep_t)aAbs << (dstSigBits - srcSigBits);
+        absResult += (dst_rep_t)(dstExpBias - srcExpBias) << dstSigBits;
+    }
+    
+    else if (aAbs >= srcInfinity) {
+        // a is NaN or infinity.
+        // Conjure the result by beginning with infinity, then setting the qNaN
+        // bit (if needed) and right-aligning the rest of the trailing NaN
+        // payload field.
+        absResult = (dst_rep_t)dstInfExp << dstSigBits;
+        absResult |= (dst_rep_t)(aAbs & srcQNaN) << (dstSigBits - srcSigBits);
+        absResult |= aAbs & srcNaNCode;
+    }
+    
+    else if (aAbs) {
+        // a is denormal.
+        // renormalize the significand and clear the leading bit, then insert
+        // the correct adjusted exponent in the destination type.
+        const int scale = src_rep_t_clz(aAbs) - src_rep_t_clz(srcMinNormal);
+        absResult = (dst_rep_t)aAbs << (dstSigBits - srcSigBits + scale);
+        absResult ^= dstMinNormal;
+        const int resultExponent = dstExpBias - srcExpBias - scale + 1;
+        absResult |= (dst_rep_t)resultExponent << dstSigBits;
+    }
+
+    else {
+        // a is zero.
+        absResult = 0;
+    }
+    
+    // Apply the signbit to (dst_t)abs(a).
+    const dst_rep_t result = absResult | (dst_rep_t)sign << (dstBits - srcBits);
+    return dstFromRep(result);
+}

contrib/compiler-rt/lib/ffsdi2.c

@@ -0,0 +1,33 @@
+/* ===-- ffsdi2.c - Implement __ffsdi2 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __ffsdi2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: the index of the least significant 1-bit in a, or
+ * the value zero if a is zero. The least significant bit is index one.
+ */
+
+si_int
+__ffsdi2(di_int a)
+{
+    dwords x;
+    x.all = a;
+    if (x.s.low == 0)
+    {
+        if (x.s.high == 0)
+            return 0;
+        return __builtin_ctz(x.s.high) + (1 + sizeof(si_int) * CHAR_BIT);
+    }
+    return __builtin_ctz(x.s.low) + 1;
+}

contrib/compiler-rt/lib/ffsti2.c

@@ -0,0 +1,37 @@
+/* ===-- ffsti2.c - Implement __ffsti2 -------------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __ffsti2 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+
+/* Returns: the index of the least significant 1-bit in a, or
+ * the value zero if a is zero. The least significant bit is index one.
+ */
+
+si_int
+__ffsti2(ti_int a)
+{
+    twords x;
+    x.all = a;
+    if (x.s.low == 0)
+    {
+        if (x.s.high == 0)
+            return 0;
+        return __builtin_ctzll(x.s.high) + (1 + sizeof(di_int) * CHAR_BIT);
+    }
+    return __builtin_ctzll(x.s.low) + 1;
+}
+
+#endif /* __x86_64 */

contrib/compiler-rt/lib/fixdfdi.c

@@ -0,0 +1,43 @@
+/* ===-- fixdfdi.c - Implement __fixdfdi -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __fixdfdi for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: convert a to a signed long long, rounding toward zero. */
+
+/* Assumption: double is a IEEE 64 bit floating point type 
+ *            su_int is a 32 bit integral type
+ *            value in double is representable in di_int (no range checking performed)
+ */
+
+/* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */
+
+di_int
+__fixdfdi(double a)
+{
+    double_bits fb;
+    fb.f = a;
+    int e = ((fb.u.s.high & 0x7FF00000) >> 20) - 1023;
+    if (e < 0)
+        return 0;
+    di_int s = (si_int)(fb.u.s.high & 0x80000000) >> 31;
+    dwords r;
+    r.s.high = (fb.u.s.high & 0x000FFFFF) | 0x00100000;
+    r.s.low = fb.u.s.low;
+    if (e > 52)
+        r.all <<= (e - 52);
+    else
+        r.all >>= (52 - e);
+    return (r.all ^ s) - s;
+} 

contrib/compiler-rt/lib/fixdfsi.c

@@ -0,0 +1,45 @@
+//===-- lib/fixdfsi.c - Double-precision -> integer conversion ----*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements double-precision to integer conversion for the
+// compiler-rt library.  No range checking is performed; the behavior of this
+// conversion is undefined for out of range values in the C standard.
+//
+//===----------------------------------------------------------------------===//
+
+#define DOUBLE_PRECISION
+#include "fp_lib.h"
+
+int __fixdfsi(fp_t a) {
+    
+    // Break a into sign, exponent, significand
+    const rep_t aRep = toRep(a);
+    const rep_t aAbs = aRep & absMask;
+    const int sign = aRep & signBit ? -1 : 1;
+    const int exponent = (aAbs >> significandBits) - exponentBias;
+    const rep_t significand = (aAbs & significandMask) | implicitBit;
+    
+    // If 0 < exponent < significandBits, right shift to get the result.
+    if ((unsigned int)exponent < significandBits) {
+        return sign * (significand >> (significandBits - exponent));
+    }
+    
+    // If exponent is negative, the result is zero.
+    else if (exponent < 0) {
+        return 0;
+    }
+    
+    // If significandBits < exponent, left shift to get the result.  This shift
+    // may end up being larger than the type width, which incurs undefined
+    // behavior, but the conversion itself is undefined in that case, so
+    // whatever the compiler decides to do is fine.
+    else {
+        return sign * (significand << (exponent - significandBits));
+    }
+}

contrib/compiler-rt/lib/fixdfti.c

@@ -0,0 +1,45 @@
+/* ===-- fixdfti.c - Implement __fixdfti -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __fixdfti for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+
+/* Returns: convert a to a signed long long, rounding toward zero. */
+
+/* Assumption: double is a IEEE 64 bit floating point type 
+ *             su_int is a 32 bit integral type
+ *             value in double is representable in ti_int (no range checking performed)
+ */
+
+/* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */
+
+ti_int
+__fixdfti(double a)
+{
+    double_bits fb;
+    fb.f = a;
+    int e = ((fb.u.s.high & 0x7FF00000) >> 20) - 1023;
+    if (e < 0)
+        return 0;
+    ti_int s = (si_int)(fb.u.s.high & 0x80000000) >> 31;
+    ti_int r = 0x0010000000000000uLL | (0x000FFFFFFFFFFFFFuLL & fb.u.all);
+    if (e > 52)
+        r <<= (e - 52);
+    else
+        r >>= (52 - e);
+    return (r ^ s) - s;
+}
+
+#endif

contrib/compiler-rt/lib/fixsfdi.c

@@ -0,0 +1,41 @@
+/* ===-- fixsfdi.c - Implement __fixsfdi -----------------------------------===
+ *
+ *                    The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __fixsfdi for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: convert a to a signed long long, rounding toward zero. */
+
+/* Assumption: float is a IEEE 32 bit floating point type 
+ *             su_int is a 32 bit integral type
+ *             value in float is representable in di_int (no range checking performed)
+ */
+
+/* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */
+
+di_int
+__fixsfdi(float a)
+{
+    float_bits fb;
+    fb.f = a;
+    int e = ((fb.u & 0x7F800000) >> 23) - 127;
+    if (e < 0)
+        return 0;
+    di_int s = (si_int)(fb.u & 0x80000000) >> 31;
+    di_int r = (fb.u & 0x007FFFFF) | 0x00800000;
+    if (e > 23)
+        r <<= (e - 23);
+    else
+        r >>= (23 - e);
+    return (r ^ s) - s;
+}

contrib/compiler-rt/lib/fixsfsi.c

@@ -0,0 +1,45 @@
+//===-- lib/fixsfsi.c - Single-precision -> integer conversion ----*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements single-precision to integer conversion for the
+// compiler-rt library.  No range checking is performed; the behavior of this
+// conversion is undefined for out of range values in the C standard.
+//
+//===----------------------------------------------------------------------===//
+
+#define SINGLE_PRECISION
+#include "fp_lib.h"
+
+int __fixsfsi(fp_t a) {
+    
+    // Break a into sign, exponent, significand
+    const rep_t aRep = toRep(a);
+    const rep_t aAbs = aRep & absMask;
+    const int sign = aRep & signBit ? -1 : 1;
+    const int exponent = (aAbs >> significandBits) - exponentBias;
+    const rep_t significand = (aAbs & significandMask) | implicitBit;
+    
+    // If 0 < exponent < significandBits, right shift to get the result.
+    if ((unsigned int)exponent < significandBits) {
+        return sign * (significand >> (significandBits - exponent));
+    }
+    
+    // If exponent is negative, the result is zero.
+    else if (exponent < 0) {
+        return 0;
+    }
+    
+    // If significandBits < exponent, left shift to get the result.  This shift
+    // may end up being larger than the type width, which incurs undefined
+    // behavior, but the conversion itself is undefined in that case, so
+    // whatever the compiler decides to do is fine.
+    else {
+        return sign * (significand << (exponent - significandBits));
+    }
+}

contrib/compiler-rt/lib/fixsfti.c

@@ -0,0 +1,45 @@
+/* ===-- fixsfti.c - Implement __fixsfti -----------------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __fixsfti for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+
+/* Returns: convert a to a signed long long, rounding toward zero. */
+
+/* Assumption: float is a IEEE 32 bit floating point type 
+ *             su_int is a 32 bit integral type
+ *             value in float is representable in ti_int (no range checking performed)
+ */
+
+/* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */
+
+ti_int
+__fixsfti(float a)
+{
+    float_bits fb;
+    fb.f = a;
+    int e = ((fb.u & 0x7F800000) >> 23) - 127;
+    if (e < 0)
+        return 0;
+    ti_int s = (si_int)(fb.u & 0x80000000) >> 31;
+    ti_int r = (fb.u & 0x007FFFFF) | 0x00800000;
+    if (e > 23)
+        r <<= (e - 23);
+    else
+        r >>= (23 - e);
+    return (r ^ s) - s;
+}
+
+#endif

contrib/compiler-rt/lib/fixunsdfdi.c

@@ -0,0 +1,45 @@
+/* ===-- fixunsdfdi.c - Implement __fixunsdfdi -----------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __fixunsdfdi for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: convert a to a unsigned long long, rounding toward zero.
+ *          Negative values all become zero.
+ */
+
+/* Assumption: double is a IEEE 64 bit floating point type 
+ *             du_int is a 64 bit integral type
+ *             value in double is representable in du_int or is negative 
+ *                 (no range checking performed)
+ */
+
+/* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */
+
+du_int
+__fixunsdfdi(double a)
+{
+    double_bits fb;
+    fb.f = a;
+    int e = ((fb.u.s.high & 0x7FF00000) >> 20) - 1023;
+    if (e < 0 || (fb.u.s.high & 0x80000000))
+        return 0;
+    udwords r;
+    r.s.high = (fb.u.s.high & 0x000FFFFF) | 0x00100000;
+    r.s.low = fb.u.s.low;
+    if (e > 52)
+        r.all <<= (e - 52);
+    else
+        r.all >>= (52 - e);
+    return r.all;
+}

contrib/compiler-rt/lib/fixunsdfsi.c

@@ -0,0 +1,42 @@
+/* ===-- fixunsdfsi.c - Implement __fixunsdfsi -----------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __fixunsdfsi for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#include "int_lib.h"
+
+/* Returns: convert a to a unsigned int, rounding toward zero.
+ *          Negative values all become zero.
+ */
+
+/* Assumption: double is a IEEE 64 bit floating point type 
+ *             su_int is a 32 bit integral type
+ *             value in double is representable in su_int or is negative 
+ *                 (no range checking performed)
+ */
+
+/* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */
+
+su_int
+__fixunsdfsi(double a)
+{
+    double_bits fb;
+    fb.f = a;
+    int e = ((fb.u.s.high & 0x7FF00000) >> 20) - 1023;
+    if (e < 0 || (fb.u.s.high & 0x80000000))
+        return 0;
+    return (
+                0x80000000u                      |
+                ((fb.u.s.high & 0x000FFFFF) << 11) |
+                (fb.u.s.low >> 21)
+           ) >> (31 - e);
+}

contrib/compiler-rt/lib/fixunsdfti.c

@@ -0,0 +1,47 @@
+/* ===-- fixunsdfti.c - Implement __fixunsdfti -----------------------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is distributed under the University of Illinois Open Source
+ * License. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __fixunsdfti for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __x86_64
+
+#include "int_lib.h"
+
+/* Returns: convert a to a unsigned long long, rounding toward zero.
+ *          Negative values all become zero.
+ */
+
+/* Assumption: double is a IEEE 64 bit floating point type 
+ *             tu_int is a 64 bit integral type
+ *             value in double is representable in tu_int or is negative 
+ *                 (no range checking performed)
+ */
+
+/* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */
+
+tu_int
+__fixunsdfti(double a)
+{
+    double_bits fb;
+    fb.f = a;
+    int e = ((fb.u.s.high & 0x7FF00000) >> 20) - 1023;
+    if (e < 0 || (fb.u.s.high & 0x80000000))
+        return 0;
+    tu_int r = 0x0010000000000000uLL | (fb.u.all & 0x000FFFFFFFFFFFFFuLL);
+    if (e > 52)
+        r <<= (e - 52);
+    else
+        r >>= (52 - e);
+    return r;
+}
+
+#endif