Import compiler-rt r132478.

74 files changed, 855 insertions, 149 deletions

diff --git a/lib/abi.h b/lib/abi.h
new file mode 100644
index 0000000..2534317
--- /dev/null
+++ b/lib/abi.h
@@ -0,0 +1,23 @@
+/* ===------ abi.h - configuration header for compiler-rt  -----------------===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is dual licensed under the MIT and the University of Illinois Open
+ * Source Licenses. 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.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+
+#if __ARM_EABI__
+# define ARM_EABI_FNALIAS(aeabi_name, name)         \
+  void __aeabi_##aeabi_name() __attribute__((alias("__" #name)));
+# define COMPILER_RT_ABI __attribute__((pcs("aapcs")))
+#else
+# define ARM_EABI_FNALIAS(aeabi_name, name)
+# define COMPILER_RT_ABI
+#endif
diff --git a/lib/absvdi2.c b/lib/absvdi2.c
index 2ba6687..9c5d4a2 100644
--- a/lib/absvdi2.c
+++ b/lib/absvdi2.c
@@ -11,6 +11,7 @@
  *
  *===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 #include <stdlib.h>
@@ -19,7 +20,7 @@
 
 /* Effects: aborts if abs(x) < 0 */
 
-di_int
+COMPILER_RT_ABI di_int
 __absvdi2(di_int a)
 {
     const int N = (int)(sizeof(di_int) * CHAR_BIT);
diff --git a/lib/absvsi2.c b/lib/absvsi2.c
index ec5cf69..80a1a78 100644
--- a/lib/absvsi2.c
+++ b/lib/absvsi2.c
@@ -10,7 +10,8 @@
  * This file implements __absvsi2 for the compiler_rt library.
  *
  * ===----------------------------------------------------------------------===
- */ 
+ */
+#include "abi.h"
 
 #include "int_lib.h"
 #include <stdlib.h>
@@ -19,7 +20,7 @@
 
 /* Effects: aborts if abs(x) < 0 */
 
-si_int
+COMPILER_RT_ABI si_int
 __absvsi2(si_int a)
 {
     const int N = (int)(sizeof(si_int) * CHAR_BIT);
diff --git a/lib/adddf3.c b/lib/adddf3.c
index c401356..3cc997b 100644
--- a/lib/adddf3.c
+++ b/lib/adddf3.c
@@ -1,4 +1,4 @@
-//===-- lib/adddf3.c - Double-precision addition and subtraction --*- C -*-===//
+//===-- lib/adddf3.c - Double-precision addition ------------------*- C -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
@@ -7,15 +7,20 @@
 //
 //===----------------------------------------------------------------------===//
 //
-// This file implements double-precision soft-float addition and subtraction
-// with the IEEE-754 default rounding (to nearest, ties to even).
+// This file implements double-precision soft-float addition with the IEEE-754
+// default rounding (to nearest, ties to even).
 //
 //===----------------------------------------------------------------------===//
 
+#include "abi.h"
+
 #define DOUBLE_PRECISION
 #include "fp_lib.h"
 
-fp_t __adddf3(fp_t a, fp_t b) {
+ARM_EABI_FNALIAS(dadd, adddf3);
+
+COMPILER_RT_ABI fp_t
+__adddf3(fp_t a, fp_t b) {
     
     rep_t aRep = toRep(a);
     rep_t bRep = toRep(b);
@@ -147,8 +152,3 @@ fp_t __adddf3(fp_t a, fp_t b) {
     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));
-}
diff --git a/lib/addsf3.c b/lib/addsf3.c
index c0e8d8e..20610ef 100644
--- a/lib/addsf3.c
+++ b/lib/addsf3.c
@@ -1,4 +1,4 @@
-//===-- lib/addsf3.c - Single-precision addition and subtraction --*- C -*-===//
+//===-- lib/addsf3.c - Single-precision addition ------------------*- C -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
@@ -7,14 +7,18 @@
 //
 //===----------------------------------------------------------------------===//
 //
-// This file implements single-precision soft-float addition and subtraction
-// with the IEEE-754 default rounding (to nearest, ties to even).
+// This file implements single-precision soft-float addition with the IEEE-754
+// default rounding (to nearest, ties to even).
 //
 //===----------------------------------------------------------------------===//
 
+#include "abi.h"
+
 #define SINGLE_PRECISION
 #include "fp_lib.h"
 
+ARM_EABI_FNALIAS(fadd, addsf3);
+
 fp_t __addsf3(fp_t a, fp_t b) {
 
     rep_t aRep = toRep(a);
@@ -147,18 +151,3 @@ fp_t __addsf3(fp_t a, fp_t b) {
     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));
-}
-
-
-
-
-
-
-
-
-
-
diff --git a/lib/addvdi3.c b/lib/addvdi3.c
index ded023e..51ad397 100644
--- a/lib/addvdi3.c
+++ b/lib/addvdi3.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 #include <stdlib.h>
@@ -19,7 +20,7 @@
 
 /* Effects: aborts if a + b overflows */
 
-di_int
+COMPILER_RT_ABI di_int
 __addvdi3(di_int a, di_int b)
 {
     di_int s = a + b;
diff --git a/lib/addvsi3.c b/lib/addvsi3.c
index 7bdb39b..c18f7bd 100644
--- a/lib/addvsi3.c
+++ b/lib/addvsi3.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 #include <stdlib.h>
@@ -19,7 +20,7 @@
 
 /* Effects: aborts if a + b overflows */
 
-si_int
+COMPILER_RT_ABI si_int
 __addvsi3(si_int a, si_int b)
 {
     si_int s = a + b;
diff --git a/lib/apple_versioning.c b/lib/apple_versioning.c
index bb6302f..e838d72 100644
--- a/lib/apple_versioning.c
+++ b/lib/apple_versioning.c
@@ -143,6 +143,147 @@ NOT_HERE_BEFORE_10_6(__gcc_qsub)
 NOT_HERE_BEFORE_10_6(__trampoline_setup)
 #endif /* __ppc__ */
 
+#if __arm__ && __DYNAMIC__
+   #define NOT_HERE_UNTIL_AFTER_4_3(sym) \
+        extern const char sym##_tmp1 __asm("$ld$hide$os3.0$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp1 = 0; \
+        extern const char sym##_tmp2 __asm("$ld$hide$os3.1$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp2 = 0; \
+        extern const char sym##_tmp3 __asm("$ld$hide$os3.2$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp3 = 0; \
+        extern const char sym##_tmp4 __asm("$ld$hide$os4.0$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp4 = 0; \
+        extern const char sym##_tmp5 __asm("$ld$hide$os4.1$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp5 = 0; \
+        extern const char sym##_tmp6 __asm("$ld$hide$os4.2$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp6 = 0; \
+        extern const char sym##_tmp7 __asm("$ld$hide$os4.3$_" #sym ); \
+            __attribute__((visibility("default"))) const char sym##_tmp7 = 0; 
+            
+NOT_HERE_UNTIL_AFTER_4_3(__absvdi2)
+NOT_HERE_UNTIL_AFTER_4_3(__absvsi2)
+NOT_HERE_UNTIL_AFTER_4_3(__adddf3)
+NOT_HERE_UNTIL_AFTER_4_3(__adddf3vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__addsf3)
+NOT_HERE_UNTIL_AFTER_4_3(__addsf3vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__addvdi3)
+NOT_HERE_UNTIL_AFTER_4_3(__addvsi3)
+NOT_HERE_UNTIL_AFTER_4_3(__ashldi3)
+NOT_HERE_UNTIL_AFTER_4_3(__ashrdi3)
+NOT_HERE_UNTIL_AFTER_4_3(__bswapdi2)
+NOT_HERE_UNTIL_AFTER_4_3(__bswapsi2)
+NOT_HERE_UNTIL_AFTER_4_3(__clzdi2)
+NOT_HERE_UNTIL_AFTER_4_3(__clzsi2)
+NOT_HERE_UNTIL_AFTER_4_3(__cmpdi2)
+NOT_HERE_UNTIL_AFTER_4_3(__ctzdi2)
+NOT_HERE_UNTIL_AFTER_4_3(__ctzsi2)
+NOT_HERE_UNTIL_AFTER_4_3(__divdc3)
+NOT_HERE_UNTIL_AFTER_4_3(__divdf3)
+NOT_HERE_UNTIL_AFTER_4_3(__divdf3vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__divdi3)
+NOT_HERE_UNTIL_AFTER_4_3(__divsc3)
+NOT_HERE_UNTIL_AFTER_4_3(__divsf3)
+NOT_HERE_UNTIL_AFTER_4_3(__divsf3vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__divsi3)
+NOT_HERE_UNTIL_AFTER_4_3(__eqdf2)
+NOT_HERE_UNTIL_AFTER_4_3(__eqdf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__eqsf2)
+NOT_HERE_UNTIL_AFTER_4_3(__eqsf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__extendsfdf2)
+NOT_HERE_UNTIL_AFTER_4_3(__extendsfdf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__ffsdi2)
+NOT_HERE_UNTIL_AFTER_4_3(__fixdfdi)
+NOT_HERE_UNTIL_AFTER_4_3(__fixdfsi)
+NOT_HERE_UNTIL_AFTER_4_3(__fixdfsivfp)
+NOT_HERE_UNTIL_AFTER_4_3(__fixsfdi)
+NOT_HERE_UNTIL_AFTER_4_3(__fixsfsi)
+NOT_HERE_UNTIL_AFTER_4_3(__fixsfsivfp)
+NOT_HERE_UNTIL_AFTER_4_3(__fixunsdfdi)
+NOT_HERE_UNTIL_AFTER_4_3(__fixunsdfsi)
+NOT_HERE_UNTIL_AFTER_4_3(__fixunsdfsivfp)
+NOT_HERE_UNTIL_AFTER_4_3(__fixunssfdi)
+NOT_HERE_UNTIL_AFTER_4_3(__fixunssfsi)
+NOT_HERE_UNTIL_AFTER_4_3(__fixunssfsivfp)
+NOT_HERE_UNTIL_AFTER_4_3(__floatdidf)
+NOT_HERE_UNTIL_AFTER_4_3(__floatdisf)
+NOT_HERE_UNTIL_AFTER_4_3(__floatsidf)
+NOT_HERE_UNTIL_AFTER_4_3(__floatsidfvfp)
+NOT_HERE_UNTIL_AFTER_4_3(__floatsisf)
+NOT_HERE_UNTIL_AFTER_4_3(__floatsisfvfp)
+NOT_HERE_UNTIL_AFTER_4_3(__floatundidf)
+NOT_HERE_UNTIL_AFTER_4_3(__floatundisf)
+NOT_HERE_UNTIL_AFTER_4_3(__floatunsidf)
+NOT_HERE_UNTIL_AFTER_4_3(__floatunsisf)
+NOT_HERE_UNTIL_AFTER_4_3(__floatunssidfvfp)
+NOT_HERE_UNTIL_AFTER_4_3(__floatunssisfvfp)
+NOT_HERE_UNTIL_AFTER_4_3(__gedf2)
+NOT_HERE_UNTIL_AFTER_4_3(__gedf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__gesf2)
+NOT_HERE_UNTIL_AFTER_4_3(__gesf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__gtdf2)
+NOT_HERE_UNTIL_AFTER_4_3(__gtdf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__gtsf2)
+NOT_HERE_UNTIL_AFTER_4_3(__gtsf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__ledf2)
+NOT_HERE_UNTIL_AFTER_4_3(__ledf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__lesf2)
+NOT_HERE_UNTIL_AFTER_4_3(__lesf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__lshrdi3)
+NOT_HERE_UNTIL_AFTER_4_3(__ltdf2)
+NOT_HERE_UNTIL_AFTER_4_3(__ltdf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__ltsf2)
+NOT_HERE_UNTIL_AFTER_4_3(__ltsf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__moddi3)
+NOT_HERE_UNTIL_AFTER_4_3(__modsi3)
+NOT_HERE_UNTIL_AFTER_4_3(__muldc3)
+NOT_HERE_UNTIL_AFTER_4_3(__muldf3)
+NOT_HERE_UNTIL_AFTER_4_3(__muldf3vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__muldi3)
+NOT_HERE_UNTIL_AFTER_4_3(__mulsc3)
+NOT_HERE_UNTIL_AFTER_4_3(__mulsf3)
+NOT_HERE_UNTIL_AFTER_4_3(__mulsf3vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__mulvdi3)
+NOT_HERE_UNTIL_AFTER_4_3(__mulvsi3)
+NOT_HERE_UNTIL_AFTER_4_3(__nedf2)
+NOT_HERE_UNTIL_AFTER_4_3(__nedf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__negdi2)
+NOT_HERE_UNTIL_AFTER_4_3(__negvdi2)
+NOT_HERE_UNTIL_AFTER_4_3(__negvsi2)
+NOT_HERE_UNTIL_AFTER_4_3(__nesf2)
+NOT_HERE_UNTIL_AFTER_4_3(__nesf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__paritydi2)
+NOT_HERE_UNTIL_AFTER_4_3(__paritysi2)
+NOT_HERE_UNTIL_AFTER_4_3(__popcountdi2)
+NOT_HERE_UNTIL_AFTER_4_3(__popcountsi2)
+NOT_HERE_UNTIL_AFTER_4_3(__powidf2)
+NOT_HERE_UNTIL_AFTER_4_3(__powisf2)
+NOT_HERE_UNTIL_AFTER_4_3(__subdf3)
+NOT_HERE_UNTIL_AFTER_4_3(__subdf3vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__subsf3)
+NOT_HERE_UNTIL_AFTER_4_3(__subsf3vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__subvdi3)
+NOT_HERE_UNTIL_AFTER_4_3(__subvsi3)
+NOT_HERE_UNTIL_AFTER_4_3(__truncdfsf2)
+NOT_HERE_UNTIL_AFTER_4_3(__truncdfsf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__ucmpdi2)
+NOT_HERE_UNTIL_AFTER_4_3(__udivdi3)
+NOT_HERE_UNTIL_AFTER_4_3(__udivmoddi4)
+NOT_HERE_UNTIL_AFTER_4_3(__udivsi3)
+NOT_HERE_UNTIL_AFTER_4_3(__umoddi3)
+NOT_HERE_UNTIL_AFTER_4_3(__umodsi3)
+NOT_HERE_UNTIL_AFTER_4_3(__unorddf2)
+NOT_HERE_UNTIL_AFTER_4_3(__unorddf2vfp)
+NOT_HERE_UNTIL_AFTER_4_3(__unordsf2)
+NOT_HERE_UNTIL_AFTER_4_3(__unordsf2vfp)
+
+NOT_HERE_UNTIL_AFTER_4_3(__divmodsi4)
+NOT_HERE_UNTIL_AFTER_4_3(__udivmodsi4)
+#endif // __arm__ && __DYNAMIC__
+
+       
+       
+
+
 #else /* !__APPLE__ */
 
 extern int avoid_empty_file;
diff --git a/lib/arm/divmodsi4.S b/lib/arm/divmodsi4.S
new file mode 100644
index 0000000..cec39a7
--- /dev/null
+++ b/lib/arm/divmodsi4.S
@@ -0,0 +1,47 @@
+/*===-- divmodsi4.S - 32-bit signed integer divide and modulus ------------===//
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is dual licensed under the MIT and the University of Illinois Open
+ * Source Licenses. See LICENSE.TXT for details.
+ *
+ *===----------------------------------------------------------------------===//
+ *
+ * This file implements the __divmodsi4 (32-bit signed integer divide and
+ * modulus) function for the ARM architecture.  A naive digit-by-digit
+ * computation is employed for simplicity.
+ *
+ *===----------------------------------------------------------------------===*/
+
+#include "../assembly.h"
+
+#define ESTABLISH_FRAME    \
+    push   {r4-r7, lr}   ;\
+    add     r7,     sp, #12
+#define CLEAR_FRAME_AND_RETURN \
+    pop    {r4-r7, pc}
+
+.syntax unified
+.align 3
+DEFINE_COMPILERRT_FUNCTION(__divmodsi4)
+    ESTABLISH_FRAME
+//  Set aside the sign of the quotient and modulus, and the address for the
+//  modulus.
+    eor     r4,     r0, r1
+    mov     r5,     r0
+    mov     r6,     r2
+//  Take the absolute value of a and b via abs(x) = (x^(x >> 31)) - (x >> 31).
+    eor     ip,     r0, r0, asr #31
+    eor     lr,     r1, r1, asr #31
+    sub     r0,     ip, r0, asr #31
+    sub     r1,     lr, r1, asr #31
+//  Unsigned divmod:
+    bl      SYMBOL_NAME(__udivmodsi4)
+//  Apply the sign of quotient and modulus
+    ldr     r1,    [r6]
+    eor     r0,     r0, r4, asr #31
+    eor     r1,     r1, r5, asr #31
+    sub     r0,     r0, r4, asr #31
+    sub     r1,     r1, r5, asr #31
+    str     r1,    [r6]
+    CLEAR_FRAME_AND_RETURN
diff --git a/lib/arm/divsi3.S b/lib/arm/divsi3.S
new file mode 100644
index 0000000..00e6181
--- /dev/null
+++ b/lib/arm/divsi3.S
@@ -0,0 +1,41 @@
+/*===-- divsi3.S - 32-bit signed integer divide ---------------------------===//
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is dual licensed under the MIT and the University of Illinois Open
+ * Source Licenses. See LICENSE.TXT for details.
+ *
+ *===----------------------------------------------------------------------===//
+ *
+ * This file implements the __divsi3 (32-bit signed integer divide) function
+ * for the ARM architecture as a wrapper around the unsigned routine.
+ *
+ *===----------------------------------------------------------------------===*/
+
+#include "../assembly.h"
+
+#define ESTABLISH_FRAME \
+    push   {r4, r7, lr}    ;\
+    add     r7,     sp, #4
+#define CLEAR_FRAME_AND_RETURN \
+    pop    {r4, r7, pc}
+
+.syntax unified
+.align 3
+// Ok, APCS and AAPCS agree on 32 bit args, so it's safe to use the same routine.
+DEFINE_AEABI_FUNCTION_ALIAS(__aeabi_idiv, __divsi3)
+DEFINE_COMPILERRT_FUNCTION(__divsi3)
+    ESTABLISH_FRAME
+//  Set aside the sign of the quotient.
+    eor     r4,     r0, r1
+//  Take absolute value of a and b via abs(x) = (x^(x >> 31)) - (x >> 31).
+    eor     r2,     r0, r0, asr #31
+    eor     r3,     r1, r1, asr #31
+    sub     r0,     r2, r0, asr #31
+    sub     r1,     r3, r1, asr #31
+//  abs(a) / abs(b)
+    bl      SYMBOL_NAME(__udivsi3)
+//  Apply sign of quotient to result and return.
+    eor     r0,     r0, r4, asr #31
+    sub     r0,     r0, r4, asr #31
+    CLEAR_FRAME_AND_RETURN
diff --git a/lib/arm/modsi3.S b/lib/arm/modsi3.S
index 40ba856..a4cd2ee 100644
--- a/lib/arm/modsi3.S
+++ b/lib/arm/modsi3.S
@@ -1,36 +1,39 @@
-//===-------- modsi3.S - Implement modsi3 ---------------------------------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is dual licensed under the MIT and the University of Illinois Open
-// Source Licenses. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
+/*===-- modsi3.S - 32-bit signed integer modulus --------------------------===//
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is dual licensed under the MIT and the University of Illinois Open
+ * Source Licenses. See LICENSE.TXT for details.
+ *
+ *===----------------------------------------------------------------------===//
+ *
+ * This file implements the __modsi3 (32-bit signed integer modulus) function
+ * for the ARM architecture as a wrapper around the unsigned routine.
+ *
+ *===----------------------------------------------------------------------===*/
 
 #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}
-	
-
+#define ESTABLISH_FRAME \
+    push   {r4, r7, lr}    ;\
+    add     r7,     sp, #4
+#define CLEAR_FRAME_AND_RETURN \
+    pop    {r4, r7, pc}
 
+.syntax unified
+.align 3
+DEFINE_COMPILERRT_FUNCTION(__modsi3)
+    ESTABLISH_FRAME
+    //  Set aside the sign of the dividend.
+    mov     r4,     r0
+    //  Take absolute value of a and b via abs(x) = (x^(x >> 31)) - (x >> 31).
+    eor     r2,     r0, r0, asr #31
+    eor     r3,     r1, r1, asr #31
+    sub     r0,     r2, r0, asr #31
+    sub     r1,     r3, r1, asr #31
+    //  abs(a) % abs(b)
+    bl     SYMBOL_NAME(__umodsi3)
+    //  Apply sign of dividend to result and return.
+    eor     r0,     r0, r4, asr #31
+    sub     r0,     r0, r4, asr #31
+    CLEAR_FRAME_AND_RETURN
diff --git a/lib/arm/udivmodsi4.S b/lib/arm/udivmodsi4.S
new file mode 100644
index 0000000..d164a75
--- /dev/null
+++ b/lib/arm/udivmodsi4.S
@@ -0,0 +1,80 @@
+/*===-- udivmodsi4.S - 32-bit unsigned integer divide and modulus ---------===//
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is dual licensed under the MIT and the University of Illinois Open
+ * Source Licenses. See LICENSE.TXT for details.
+ *
+ *===----------------------------------------------------------------------===//
+ *
+ * This file implements the __udivmodsi4 (32-bit unsigned integer divide and
+ * modulus) function for the ARM architecture.  A naive digit-by-digit
+ * computation is employed for simplicity.
+ *
+ *===----------------------------------------------------------------------===*/
+
+#include "../assembly.h"
+
+#define ESTABLISH_FRAME    \
+    push   {r4, r7, lr}   ;\
+    add     r7,     sp, #4
+#define CLEAR_FRAME_AND_RETURN \
+    pop    {r4, r7, pc}
+    
+#define a r0
+#define b r1
+#define i r3
+#define r r4
+#define q ip
+#define one lr
+
+.syntax unified
+.align 3
+DEFINE_COMPILERRT_FUNCTION(__udivmodsi4)
+//  We use a simple digit by digit algorithm; before we get into the actual 
+//  divide loop, we must calculate the left-shift amount necessary to align
+//  the MSB of the divisor with that of the dividend (If this shift is
+//  negative, then the result is zero, and we early out). We also conjure a
+//  bit mask of 1 to use in constructing the quotient, and initialize the
+//  quotient to zero.
+    ESTABLISH_FRAME
+    clz     r4,     a
+    tst     b,      b   // detect divide-by-zero
+    clz     r3,     b
+    mov     q,      #0
+    beq     LOCAL_LABEL(return)    // return 0 if b is zero.
+    mov     one,    #1
+    subs    i,      r3, r4
+    blt     LOCAL_LABEL(return)    // return 0 if MSB(a) < MSB(b)
+
+LOCAL_LABEL(mainLoop):
+//  This loop basically implements the following:
+//
+//  do {
+//      if (a >= b << i) {
+//          a -= b << i;
+//          q |= 1 << i;
+//          if (a == 0) break;
+//      }
+//  } while (--i)
+//
+//  Note that this does not perform the final iteration (i == 0); by doing it
+//  this way, we can merge the two branches which is a substantial win for
+//  such a tight loop on current ARM architectures.
+    subs    r,      a,  b, lsl i
+    orrhs   q,      q,one, lsl i
+    movhs   a,      r
+    subsne  i,      i, #1
+    bhi     LOCAL_LABEL(mainLoop)
+
+//  Do the final test subtraction and update of quotient (i == 0), as it is
+//  not performed in the main loop.
+    subs    r,      a,  b
+    orrhs   q,      #1
+    movhs   a,      r
+
+LOCAL_LABEL(return):
+//  Store the remainder, and move the quotient to r0, then return.
+    str     a,     [r2]
+    mov     r0,     q
+    CLEAR_FRAME_AND_RETURN
diff --git a/lib/arm/udivsi3.S b/lib/arm/udivsi3.S
new file mode 100644
index 0000000..6d89665
--- /dev/null
+++ b/lib/arm/udivsi3.S
@@ -0,0 +1,80 @@
+/*===-- udivsi3.S - 32-bit unsigned integer divide ------------------------===//
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is dual licensed under the MIT and the University of Illinois Open
+ * Source Licenses. See LICENSE.TXT for details.
+ *
+ *===----------------------------------------------------------------------===//
+ *
+ * This file implements the __udivsi3 (32-bit unsigned integer divide) 
+ * function for the ARM architecture.  A naive digit-by-digit computation is
+ * employed for simplicity.
+ *
+ *===----------------------------------------------------------------------===*/
+
+#include "../assembly.h"
+
+#define ESTABLISH_FRAME \
+    push   {r7, lr}    ;\
+    mov     r7,     sp
+#define CLEAR_FRAME_AND_RETURN \
+    pop    {r7, pc}
+
+#define a r0
+#define b r1
+#define r r2
+#define i r3
+#define q ip
+#define one lr
+
+.syntax unified
+.align 3
+// Ok, APCS and AAPCS agree on 32 bit args, so it's safe to use the same routine.
+DEFINE_AEABI_FUNCTION_ALIAS(__aeabi_uidiv, __udivsi3)
+DEFINE_COMPILERRT_FUNCTION(__udivsi3)
+//  We use a simple digit by digit algorithm; before we get into the actual 
+//  divide loop, we must calculate the left-shift amount necessary to align
+//  the MSB of the divisor with that of the dividend (If this shift is
+//  negative, then the result is zero, and we early out). We also conjure a
+//  bit mask of 1 to use in constructing the quotient, and initialize the
+//  quotient to zero.
+    ESTABLISH_FRAME
+    clz     r2,     a
+    tst     b,      b   // detect divide-by-zero
+    clz     r3,     b
+    mov     q,      #0
+    beq     LOCAL_LABEL(return)    // return 0 if b is zero.
+    mov     one,    #1
+    subs    i,      r3, r2
+    blt     LOCAL_LABEL(return)    // return 0 if MSB(a) < MSB(b)
+
+LOCAL_LABEL(mainLoop):
+//  This loop basically implements the following:
+//
+//  do {
+//      if (a >= b << i) {
+//          a -= b << i;
+//          q |= 1 << i;
+//          if (a == 0) break;
+//      }
+//  } while (--i)
+//
+//  Note that this does not perform the final iteration (i == 0); by doing it
+//  this way, we can merge the two branches which is a substantial win for
+//  such a tight loop on current ARM architectures.
+    subs    r,      a,  b, lsl i
+    orrhs   q,      q,one, lsl i
+    movhs   a,      r
+    subsne  i,      i, #1
+    bhi     LOCAL_LABEL(mainLoop)
+
+//  Do the final test subtraction and update of quotient (i == 0), as it is
+//  not performed in the main loop.
+    subs    r,      a,  b
+    orrhs   q,      #1
+
+LOCAL_LABEL(return):
+//  Move the quotient to r0 and return.
+    mov     r0,     q
+    CLEAR_FRAME_AND_RETURN
diff --git a/lib/arm/umodsi3.S b/lib/arm/umodsi3.S
new file mode 100644
index 0000000..3a2ab2b
--- /dev/null
+++ b/lib/arm/umodsi3.S
@@ -0,0 +1,58 @@
+/*===-- umodsi3.S - 32-bit unsigned integer modulus -----------------------===//
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is dual licensed under the MIT and the University of Illinois Open
+ * Source Licenses. See LICENSE.TXT for details.
+ *
+ *===----------------------------------------------------------------------===//
+ *
+ * This file implements the __umodsi3 (32-bit unsigned integer modulus) 
+ * function for the ARM architecture.  A naive digit-by-digit computation is
+ * employed for simplicity.
+ *
+ *===----------------------------------------------------------------------===*/
+
+#include "../assembly.h"
+
+#define a r0
+#define b r1
+#define r r2
+#define i r3
+
+.syntax unified
+.align 3
+DEFINE_COMPILERRT_FUNCTION(__umodsi3)
+//  We use a simple digit by digit algorithm; before we get into the actual 
+//  divide loop, we must calculate the left-shift amount necessary to align
+//  the MSB of the divisor with that of the dividend.
+    clz     r2,     a
+    tst     b,      b       // detect b == 0
+    clz     r3,     b
+    bxeq    lr              // return a if b == 0
+    subs    i,      r3, r2
+    bxlt    lr              // return a if MSB(a) < MSB(b)
+
+LOCAL_LABEL(mainLoop):
+//  This loop basically implements the following:
+//
+//  do {
+//      if (a >= b << i) {
+//          a -= b << i;
+//          if (a == 0) break;
+//      }
+//  } while (--i)
+//
+//  Note that this does not perform the final iteration (i == 0); by doing it
+//  this way, we can merge the two branches which is a substantial win for
+//  such a tight loop on current ARM architectures.
+    subs    r,      a,  b, lsl i
+    movhs   a,      r
+    subsne  i,      i, #1
+    bhi     LOCAL_LABEL(mainLoop)
+
+//  Do the final test subtraction and update of remainder (i == 0), as it is
+//  not performed in the main loop.
+    subs    r,      a,  b
+    movhs   a,      r
+    bx      lr
diff --git a/lib/ashldi3.c b/lib/ashldi3.c
index 1067e6f..adce4e2 100644
--- a/lib/ashldi3.c
+++ b/lib/ashldi3.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -18,7 +19,9 @@
 
 /* Precondition:  0 <= b < bits_in_dword */
 
-di_int
+ARM_EABI_FNALIAS(llsl, ashldi3);
+
+COMPILER_RT_ABI di_int
 __ashldi3(di_int a, si_int b)
 {
     const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT);
diff --git a/lib/ashrdi3.c b/lib/ashrdi3.c
index 94d46f1..03692a3 100644
--- a/lib/ashrdi3.c
+++ b/lib/ashrdi3.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -18,7 +19,9 @@
 
 /* Precondition:  0 <= b < bits_in_dword */
 
-di_int
+ARM_EABI_FNALIAS(lasr, ashrdi3);
+
+COMPILER_RT_ABI di_int
 __ashrdi3(di_int a, si_int b)
 {
     const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT);
diff --git a/lib/assembly.h b/lib/assembly.h
index 41c24d7..0ce83ac 100644
--- a/lib/assembly.h
+++ b/lib/assembly.h
@@ -22,33 +22,48 @@
 #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
+#define HIDDEN_DIRECTIVE .private_extern
+#define LOCAL_LABEL(name) L_##name
 #else
-#define SYMBOL_NAME(name) name
+#define HIDDEN_DIRECTIVE .hidden
+#define LOCAL_LABEL(name) .L_##name
 #endif
 
+#define GLUE2(a, b) a ## b
+#define GLUE(a, b) GLUE2(a, b)
+#define SYMBOL_NAME(name) GLUE(__USER_LABEL_PREFIX__, name)
+
 #ifdef VISIBILITY_HIDDEN
-#define DEFINE_COMPILERRT_FUNCTION(name) \
-  .globl SYMBOL_NAME(name) SEPARATOR     \
-  .private_extern SYMBOL_NAME(name) SEPARATOR    \
+#define DEFINE_COMPILERRT_FUNCTION(name)                   \
+  .globl SYMBOL_NAME(name) SEPARATOR                       \
+  HIDDEN_DIRECTIVE SYMBOL_NAME(name) SEPARATOR             \
   SYMBOL_NAME(name):
 #else
-#define DEFINE_COMPILERRT_FUNCTION(name) \
-  .globl SYMBOL_NAME(name) SEPARATOR     \
+#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    \
+#define DEFINE_COMPILERRT_PRIVATE_FUNCTION(name)           \
+  .globl SYMBOL_NAME(name) SEPARATOR                       \
+  HIDDEN_DIRECTIVE SYMBOL_NAME(name) SEPARATOR             \
   SYMBOL_NAME(name):
 
 #define DEFINE_COMPILERRT_PRIVATE_FUNCTION_UNMANGLED(name) \
-  .globl name SEPARATOR             \
-  .private_extern name SEPARATOR    \
+  .globl name SEPARATOR                                    \
+  HIDDEN_DIRECTIVE name SEPARATOR                          \
   name:
 
+#define DEFINE_COMPILERRT_FUNCTION_ALIAS(name, target)     \
+  .globl SYMBOL_NAME(name) SEPARATOR                       \
+  .set SYMBOL_NAME(name), SYMBOL_NAME(target) SEPARATOR
+
+#if defined (__ARM_EABI__)
+# define DEFINE_AEABI_FUNCTION_ALIAS(aeabi_name, name)      \
+  DEFINE_COMPILERRT_FUNCTION_ALIAS(aeabi_name, name)
+#else
+# define DEFINE_AEABI_FUNCTION_ALIAS(aeabi_name, name)
+#endif
+
 #endif /* COMPILERRT_ASSEMBLY_H */
diff --git a/lib/clzdi2.c b/lib/clzdi2.c
index 2132777..c281945 100644
--- a/lib/clzdi2.c
+++ b/lib/clzdi2.c
@@ -1,6 +1,6 @@
 /* ===-- clzdi2.c - Implement __clzdi2 -------------------------------------===
  *
- *      	       The LLVM Compiler Infrastructure
+ *               The LLVM Compiler Infrastructure
  *
  * This file is dual licensed under the MIT and the University of Illinois Open
  * Source Licenses. See LICENSE.TXT for details.
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -18,7 +19,7 @@
 
 /* Precondition: a != 0 */
 
-si_int
+COMPILER_RT_ABI si_int
 __clzdi2(di_int a)
 {
     dwords x;
diff --git a/lib/clzsi2.c b/lib/clzsi2.c
index d99047d..d0a6aea 100644
--- a/lib/clzsi2.c
+++ b/lib/clzsi2.c
@@ -1,6 +1,6 @@
 /* ===-- clzsi2.c - Implement __clzsi2 -------------------------------------===
  *
- *      	       The LLVM Compiler Infrastructure
+ *               The LLVM Compiler Infrastructure
  *
  * This file is dual licensed under the MIT and the University of Illinois Open
  * Source Licenses. See LICENSE.TXT for details.
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -18,7 +19,7 @@
 
 /* Precondition: a != 0 */
 
-si_int
+COMPILER_RT_ABI si_int
 __clzsi2(si_int a)
 {
     su_int x = (su_int)a;
diff --git a/lib/cmpdi2.c b/lib/cmpdi2.c
index 04bd7cf..999c3d2 100644
--- a/lib/cmpdi2.c
+++ b/lib/cmpdi2.c
@@ -6,20 +6,21 @@
  * Source Licenses. See LICENSE.TXT for details.
  *
  * ===----------------------------------------------------------------------===
- * 
+ *
  * This file implements __cmpdi2 for the compiler_rt library.
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
-/* Returns:  if (a <  b) returns 0
+/* Returns: if (a <  b) returns 0
 *           if (a == b) returns 1
 *           if (a >  b) returns 2
 */
 
-si_int
+COMPILER_RT_ABI si_int
 __cmpdi2(di_int a, di_int b)
 {
     dwords x;
diff --git a/lib/ctzdi2.c b/lib/ctzdi2.c
index f7ae50f..b3d37d0 100644
--- a/lib/ctzdi2.c
+++ b/lib/ctzdi2.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -18,7 +19,7 @@
 
 /* Precondition: a != 0 */
 
-si_int
+COMPILER_RT_ABI si_int
 __ctzdi2(di_int a)
 {
     dwords x;
diff --git a/lib/ctzsi2.c b/lib/ctzsi2.c
index 0c738fa..2ff0e5d 100644
--- a/lib/ctzsi2.c
+++ b/lib/ctzsi2.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -18,7 +19,7 @@
 
 /* Precondition: a != 0 */
 
-si_int
+COMPILER_RT_ABI si_int
 __ctzsi2(si_int a)
 {
     su_int x = (su_int)a;
diff --git a/lib/divdf3.c b/lib/divdf3.c
index 217d284..925abd5 100644
--- a/lib/divdf3.c
+++ b/lib/divdf3.c
@@ -15,10 +15,13 @@
 // underflow with correct rounding.
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define DOUBLE_PRECISION
 #include "fp_lib.h"
 
+ARM_EABI_FNALIAS(ddiv, divdf3);
+
 fp_t __divdf3(fp_t a, fp_t b) {
     
     const unsigned int aExponent = toRep(a) >> significandBits & maxExponent;
diff --git a/lib/divdi3.c b/lib/divdi3.c
index a1b83ea..d62df56 100644
--- a/lib/divdi3.c
+++ b/lib/divdi3.c
@@ -11,14 +11,15 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
-du_int __udivmoddi4(du_int a, du_int b, du_int* rem);
+du_int COMPILER_RT_ABI __udivmoddi4(du_int a, du_int b, du_int* rem);
 
 /* Returns: a / b */
 
-di_int
+COMPILER_RT_ABI di_int
 __divdi3(di_int a, di_int b)
 {
     const int bits_in_dword_m1 = (int)(sizeof(di_int) * CHAR_BIT) - 1;
diff --git a/lib/divmoddi4.c b/lib/divmoddi4.c
new file mode 100644
index 0000000..d3ca745
--- /dev/null
+++ b/lib/divmoddi4.c
@@ -0,0 +1,30 @@
+/*===-- divmoddi4.c - Implement __divmoddi4 --------------------------------===
+ *
+ *                    The LLVM Compiler Infrastructure
+ *
+ * This file is dual licensed under the MIT and the University of Illinois Open
+ * Source Licenses. See LICENSE.TXT for details.
+ *
+ * ===----------------------------------------------------------------------===
+ *
+ * This file implements __divmoddi4 for the compiler_rt library.
+ *
+ * ===----------------------------------------------------------------------===
+ */
+#include "abi.h"
+
+#include "int_lib.h"
+
+extern COMPILER_RT_ABI di_int __divdi3(di_int a, di_int b);
+
+ARM_EABI_FNALIAS(ldivmod, divmoddi4);
+
+/* Returns: a / b, *rem = a % b  */
+
+COMPILER_RT_ABI di_int
+__divmoddi4(di_int a, di_int b, di_int* rem)
+{
+  di_int d = __divdi3(a,b);
+  *rem = a - (d*b);
+  return d;
+}
diff --git a/lib/divmodsi4.c b/lib/divmodsi4.c
index 2ec3dd4..4dc1978 100644
--- a/lib/divmodsi4.c
+++ b/lib/divmodsi4.c
@@ -11,15 +11,16 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
-extern si_int __divsi3(si_int a, si_int b);
+extern COMPILER_RT_ABI si_int __divsi3(si_int a, si_int b);
 
 
 /* Returns: a / b, *rem = a % b  */
 
-si_int
+COMPILER_RT_ABI si_int
 __divmodsi4(si_int a, si_int b, si_int* rem)
 {
   si_int d = __divsi3(a,b);
diff --git a/lib/divsf3.c b/lib/divsf3.c
index b798cfb..b733307 100644
--- a/lib/divsf3.c
+++ b/lib/divsf3.c
@@ -15,10 +15,13 @@
 // underflow with correct rounding.
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define SINGLE_PRECISION
 #include "fp_lib.h"
 
+ARM_EABI_FNALIAS(fdiv, divsf3);
+
 fp_t __divsf3(fp_t a, fp_t b) {
     
     const unsigned int aExponent = toRep(a) >> significandBits & maxExponent;
diff --git a/lib/divsi3.c b/lib/divsi3.c
index c48ecf4..01ef274 100644
--- a/lib/divsi3.c
+++ b/lib/divsi3.c
@@ -11,14 +11,17 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
-su_int __udivsi3(su_int n, su_int d);
+su_int COMPILER_RT_ABI __udivsi3(su_int n, su_int d);
 
 /* Returns: a / b */
 
-si_int
+ARM_EABI_FNALIAS(idiv, divsi3);
+
+COMPILER_RT_ABI si_int
 __divsi3(si_int a, si_int b)
 {
     const int bits_in_word_m1 = (int)(sizeof(si_int) * CHAR_BIT) - 1;
diff --git a/lib/extendsfdf2.c b/lib/extendsfdf2.c
index db65acf..c0b628d 100644
--- a/lib/extendsfdf2.c
+++ b/lib/extendsfdf2.c
@@ -41,6 +41,8 @@
 #include <stdint.h>
 #include <limits.h>
 
+#include "abi.h"
+
 typedef float src_t;
 typedef uint32_t src_rep_t;
 #define SRC_REP_C UINT32_C
@@ -67,6 +69,8 @@ static inline dst_t dstFromRep(dst_rep_t x) {
 
 // End helper routines.  Conversion implementation follows.
 
+ARM_EABI_FNALIAS(f2d, extendsfdf2);
+
 dst_t __extendsfdf2(src_t a) {
     
     // Various constants whose values follow from the type parameters.
diff --git a/lib/ffsdi2.c b/lib/ffsdi2.c
index 8a1c68c..89f1b7b 100644
--- a/lib/ffsdi2.c
+++ b/lib/ffsdi2.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -18,7 +19,7 @@
  * the value zero if a is zero. The least significant bit is index one.
  */
 
-si_int
+COMPILER_RT_ABI si_int
 __ffsdi2(di_int a)
 {
     dwords x;
diff --git a/lib/fixdfdi.c b/lib/fixdfdi.c
index 3a760ff..85a456d 100644
--- a/lib/fixdfdi.c
+++ b/lib/fixdfdi.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -23,6 +24,8 @@
 
 /* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */
 
+ARM_EABI_FNALIAS(d2lz, fixdfdi);
+
 di_int
 __fixdfdi(double a)
 {
diff --git a/lib/fixdfsi.c b/lib/fixdfsi.c
index 98062ab..fbcf147 100644
--- a/lib/fixdfsi.c
+++ b/lib/fixdfsi.c
@@ -12,10 +12,15 @@
 // conversion is undefined for out of range values in the C standard.
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define DOUBLE_PRECISION
 #include "fp_lib.h"
 
+#include "int_lib.h"
+
+ARM_EABI_FNALIAS(d2iz, fixdfsi);
+
 int __fixdfsi(fp_t a) {
     
     // Break a into sign, exponent, significand
diff --git a/lib/fixsfdi.c b/lib/fixsfdi.c
index 1a85306..d80e33e 100644
--- a/lib/fixsfdi.c
+++ b/lib/fixsfdi.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -23,7 +24,9 @@
 
 /* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */
 
-di_int
+ARM_EABI_FNALIAS(d2lz, fixsfdi);
+
+COMPILER_RT_ABI di_int
 __fixsfdi(float a)
 {
     float_bits fb;
diff --git a/lib/fixsfsi.c b/lib/fixsfsi.c
index b68471d..67749a5 100644
--- a/lib/fixsfsi.c
+++ b/lib/fixsfsi.c
@@ -12,12 +12,15 @@
 // conversion is undefined for out of range values in the C standard.
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define SINGLE_PRECISION
 #include "fp_lib.h"
 
-int __fixsfsi(fp_t a) {
-    
+ARM_EABI_FNALIAS(f2iz, fixsfsi);
+
+COMPILER_RT_ABI int
+__fixsfsi(fp_t a) {
     // Break a into sign, exponent, significand
     const rep_t aRep = toRep(a);
     const rep_t aAbs = aRep & absMask;
diff --git a/lib/fixunsdfdi.c b/lib/fixunsdfdi.c
index 1c78e2b..d80b84a 100644
--- a/lib/fixunsdfdi.c
+++ b/lib/fixunsdfdi.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -26,7 +27,9 @@
 
 /* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */
 
-du_int
+ARM_EABI_FNALIAS(d2ulz, fixunsdfdi);
+
+COMPILER_RT_ABI du_int
 __fixunsdfdi(double a)
 {
     double_bits fb;
diff --git a/lib/fixunsdfsi.c b/lib/fixunsdfsi.c
index e029891..ecdfb5d 100644
--- a/lib/fixunsdfsi.c
+++ b/lib/fixunsdfsi.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -26,7 +27,9 @@
 
 /* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */
 
-su_int
+ARM_EABI_FNALIAS(d2uiz, fixunsdfsi);
+
+COMPILER_RT_ABI su_int
 __fixunsdfsi(double a)
 {
     double_bits fb;
diff --git a/lib/fixunssfdi.c b/lib/fixunssfdi.c
index 3b1bc4a..1506420 100644
--- a/lib/fixunssfdi.c
+++ b/lib/fixunssfdi.c
@@ -11,9 +11,9 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
-
 /* Returns: convert a to a unsigned long long, rounding toward zero.
  *          Negative values all become zero.
  */
@@ -26,7 +26,9 @@
 
 /* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */
 
-du_int
+ARM_EABI_FNALIAS(f2ulz, fixunssfdi);
+
+COMPILER_RT_ABI du_int
 __fixunssfdi(float a)
 {
     float_bits fb;
diff --git a/lib/fixunssfsi.c b/lib/fixunssfsi.c
index 023d7b2..dbaa511 100644
--- a/lib/fixunssfsi.c
+++ b/lib/fixunssfsi.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -26,7 +27,9 @@
 
 /* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */
 
-su_int
+ARM_EABI_FNALIAS(f2uiz, fixunssfsi);
+
+COMPILER_RT_ABI su_int
 __fixunssfsi(float a)
 {
     float_bits fb;
diff --git a/lib/floatdidf.c b/lib/floatdidf.c
index cad354a..5ba9526 100644
--- a/lib/floatdidf.c
+++ b/lib/floatdidf.c
@@ -11,6 +11,7 @@
  *
  *===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 #include <float.h>
@@ -23,13 +24,15 @@
 
 /* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */
 
+ARM_EABI_FNALIAS(l2d, floatdidf);
+
 #ifndef __SOFT_FP__
 /* Support for systems that have hardware floating-point; we'll set the inexact flag
  * as a side-effect of this computation.
  */
 #include <stdint.h>
 
-double
+COMPILER_RT_ABI double
 __floatdidf(di_int a)
 {
 	static const double twop52 = 0x1.0p52;
@@ -49,7 +52,7 @@ __floatdidf(di_int a)
  * set, and we don't want to code-gen to an unknown soft-float implementation.
  */
 
-double
+COMPILER_RT_ABI double
 __floatdidf(di_int a)
 {
     if (a == 0)
diff --git a/lib/floatdisf.c b/lib/floatdisf.c
index 71d603b..4dc13ca 100644
--- a/lib/floatdisf.c
+++ b/lib/floatdisf.c
@@ -12,7 +12,7 @@
  *===----------------------------------------------------------------------===
  */
 
-#include "int_lib.h"
+#include "abi.h"
 #include <float.h>
 
 /* Returns: convert a to a float, rounding toward even.*/
@@ -23,7 +23,11 @@
 
 /* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */
 
-float
+#include "int_lib.h"
+
+ARM_EABI_FNALIAS(l2f, floatdisf);
+
+COMPILER_RT_ABI float
 __floatdisf(di_int a)
 {
     if (a == 0)
diff --git a/lib/floatsidf.c b/lib/floatsidf.c
index 85facea..7227335 100644
--- a/lib/floatsidf.c
+++ b/lib/floatsidf.c
@@ -12,10 +12,15 @@
 // mode.
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define DOUBLE_PRECISION
 #include "fp_lib.h"
 
+#include "int_lib.h"
+
+ARM_EABI_FNALIAS(i2d, floatsidf);
+
 fp_t __floatsidf(int a) {
     
     const int aWidth = sizeof a * CHAR_BIT;
diff --git a/lib/floatsisf.c b/lib/floatsisf.c
index d1bb460..e5250ff 100644
--- a/lib/floatsisf.c
+++ b/lib/floatsisf.c
@@ -12,10 +12,15 @@
 // mode.
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define SINGLE_PRECISION
 #include "fp_lib.h"
 
+#include "int_lib.h"
+
+ARM_EABI_FNALIAS(i2f, floatsisf);
+
 fp_t __floatsisf(int a) {
     
     const int aWidth = sizeof a * CHAR_BIT;
diff --git a/lib/floatundidf.c b/lib/floatundidf.c
index 506fc3c..e74e9d8 100644
--- a/lib/floatundidf.c
+++ b/lib/floatundidf.c
@@ -12,7 +12,7 @@
  * ===----------------------------------------------------------------------===
  */
 
-#include "int_lib.h"
+#include "abi.h"
 #include <float.h>
 
 /* Returns: convert a to a double, rounding toward even. */
@@ -23,6 +23,10 @@
 
 /* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */
 
+#include "int_lib.h"
+
+ARM_EABI_FNALIAS(ul2d, floatundidf);
+
 #ifndef __SOFT_FP__
 /* Support for systems that have hardware floating-point; we'll set the inexact flag
  * as a side-effect of this computation.
@@ -30,7 +34,7 @@
 
 #include <stdint.h>
 
-double
+COMPILER_RT_ABI double
 __floatundidf(du_int a)
 {
 	static const double twop52 = 0x1.0p52;
@@ -52,7 +56,7 @@ __floatundidf(du_int a)
  * set, and we don't want to code-gen to an unknown soft-float implementation.
  */ 
 
-double
+COMPILER_RT_ABI double
 __floatundidf(du_int a)
 {
     if (a == 0)
diff --git a/lib/floatundisf.c b/lib/floatundisf.c
index 55e4023..eea45a7 100644
--- a/lib/floatundisf.c
+++ b/lib/floatundisf.c
@@ -12,7 +12,7 @@
  *===----------------------------------------------------------------------===
  */
 
-#include "int_lib.h"
+#include "abi.h"
 #include <float.h>
 
 /* Returns: convert a to a float, rounding toward even. */
@@ -23,7 +23,11 @@
 
 /* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */
 
-float
+#include "int_lib.h"
+
+ARM_EABI_FNALIAS(ul2f, floatundisf);
+
+COMPILER_RT_ABI float
 __floatundisf(du_int a)
 {
     if (a == 0)
diff --git a/lib/floatunsidf.c b/lib/floatunsidf.c
index 0f473aa..3756299 100644
--- a/lib/floatunsidf.c
+++ b/lib/floatunsidf.c
@@ -12,10 +12,15 @@
 // mode.
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define DOUBLE_PRECISION
 #include "fp_lib.h"
 
+#include "int_lib.h"
+
+ARM_EABI_FNALIAS(ui2d, floatunsidf);
+
 fp_t __floatunsidf(unsigned int a) {
     
     const int aWidth = sizeof a * CHAR_BIT;
diff --git a/lib/floatunsisf.c b/lib/floatunsisf.c
index 48eff93..14ef103 100644
--- a/lib/floatunsisf.c
+++ b/lib/floatunsisf.c
@@ -12,10 +12,15 @@
 // mode.
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define SINGLE_PRECISION
 #include "fp_lib.h"
 
+#include "int_lib.h"
+
+ARM_EABI_FNALIAS(ui2f, floatunsisf);
+
 fp_t __floatunsisf(unsigned int a) {
     
     const int aWidth = sizeof a * CHAR_BIT;
diff --git a/lib/lshrdi3.c b/lib/lshrdi3.c
index 84525b7..911edb1 100644
--- a/lib/lshrdi3.c
+++ b/lib/lshrdi3.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -18,7 +19,9 @@
 
 /* Precondition:  0 <= b < bits_in_dword */
 
-di_int
+ARM_EABI_FNALIAS(llsr, lshrdi3);
+
+COMPILER_RT_ABI di_int
 __lshrdi3(di_int a, si_int b)
 {
     const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT);
diff --git a/lib/moddi3.c b/lib/moddi3.c
index 3b350dd..af0a8083 100644
--- a/lib/moddi3.c
+++ b/lib/moddi3.c
@@ -11,14 +11,15 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
-du_int __udivmoddi4(du_int a, du_int b, du_int* rem);
+COMPILER_RT_ABI du_int __udivmoddi4(du_int a, du_int b, du_int* rem);
 
 /* Returns: a % b */
 
-di_int
+COMPILER_RT_ABI di_int
 __moddi3(di_int a, di_int b)
 {
     const int bits_in_dword_m1 = (int)(sizeof(di_int) * CHAR_BIT) - 1;
diff --git a/lib/modsi3.c b/lib/modsi3.c
index 70d38a6..05ce806 100644
--- a/lib/modsi3.c
+++ b/lib/modsi3.c
@@ -11,14 +11,15 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
-su_int __divsi3(si_int a, si_int b);
+su_int COMPILER_RT_ABI __divsi3(si_int a, si_int b);
 
 /* Returns: a % b */
 
-si_int
+COMPILER_RT_ABI si_int
 __modsi3(si_int a, si_int b)
 {
     return a - __divsi3(a, b) * b;
diff --git a/lib/muldf3.c b/lib/muldf3.c
index 85672e5..f402cfb 100644
--- a/lib/muldf3.c
+++ b/lib/muldf3.c
@@ -11,11 +11,15 @@
 // with the IEEE-754 default rounding (to nearest, ties to even).
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define DOUBLE_PRECISION
 #include "fp_lib.h"
 
-fp_t __muldf3(fp_t a, fp_t b) {
+ARM_EABI_FNALIAS(dmul, muldf3);
+
+COMPILER_RT_ABI fp_t
+__muldf3(fp_t a, fp_t b) {
     
     const unsigned int aExponent = toRep(a) >> significandBits & maxExponent;
     const unsigned int bExponent = toRep(b) >> significandBits & maxExponent;
diff --git a/lib/muldi3.c b/lib/muldi3.c
index 4363732..e6322bf 100644
--- a/lib/muldi3.c
+++ b/lib/muldi3.c
@@ -10,7 +10,8 @@
  * This file implements __muldi3 for the compiler_rt library.
  *
  * ===----------------------------------------------------------------------===
- */ 
+ */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -40,7 +41,9 @@ __muldsi3(su_int a, su_int b)
 
 /* Returns: a * b */
 
-di_int
+ARM_EABI_FNALIAS(lmul, muldi3);
+
+COMPILER_RT_ABI di_int
 __muldi3(di_int a, di_int b)
 {
     dwords x;
diff --git a/lib/mulsf3.c b/lib/mulsf3.c
index dd4ce11..bf46e14 100644
--- a/lib/mulsf3.c
+++ b/lib/mulsf3.c
@@ -11,11 +11,15 @@
 // with the IEEE-754 default rounding (to nearest, ties to even).
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define SINGLE_PRECISION
 #include "fp_lib.h"
 
-fp_t __mulsf3(fp_t a, fp_t b) {
+ARM_EABI_FNALIAS(fmul, mulsf3);
+
+COMPILER_RT_ABI fp_t
+__mulsf3(fp_t a, fp_t b) {
     
     const unsigned int aExponent = toRep(a) >> significandBits & maxExponent;
     const unsigned int bExponent = toRep(b) >> significandBits & maxExponent;
diff --git a/lib/negdf2.c b/lib/negdf2.c
index aeae2e8..b47f397 100644
--- a/lib/negdf2.c
+++ b/lib/negdf2.c
@@ -10,10 +10,13 @@
 // This file implements double-precision soft-float negation.
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define DOUBLE_PRECISION
 #include "fp_lib.h"
 
+ARM_EABI_FNALIAS(dneg, negdf2);
+
 fp_t __negdf2(fp_t a) {
     return fromRep(toRep(a) ^ signBit);
 }
diff --git a/lib/negsf2.c b/lib/negsf2.c
index d211f7c..98f9fc0 100644
--- a/lib/negsf2.c
+++ b/lib/negsf2.c
@@ -10,10 +10,14 @@
 // This file implements single-precision soft-float negation.
 //
 //===----------------------------------------------------------------------===//
+#include "abi.h"
 
 #define SINGLE_PRECISION
 #include "fp_lib.h"
 
-fp_t __negsf2(fp_t a) {
+ARM_EABI_FNALIAS(fneg, negsf2);
+
+COMPILER_RT_ABI fp_t
+__negsf2(fp_t a) {
     return fromRep(toRep(a) ^ signBit);
 }
diff --git a/lib/negvdi2.c b/lib/negvdi2.c
index 2851d26..aafaa9d 100644
--- a/lib/negvdi2.c
+++ b/lib/negvdi2.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 #include <stdlib.h>
@@ -19,7 +20,7 @@
 
 /* Effects: aborts if -a overflows */
 
-di_int
+COMPILER_RT_ABI di_int
 __negvdi2(di_int a)
 {
     const di_int MIN = (di_int)1 << ((int)(sizeof(di_int) * CHAR_BIT)-1);
diff --git a/lib/negvsi2.c b/lib/negvsi2.c
index 1a83347..559ea18 100644
--- a/lib/negvsi2.c
+++ b/lib/negvsi2.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 #include <stdlib.h>
@@ -19,7 +20,7 @@
 
 /* Effects: aborts if -a overflows */
 
-si_int
+COMPILER_RT_ABI si_int
 __negvsi2(si_int a)
 {
     const si_int MIN = (si_int)1 << ((int)(sizeof(si_int) * CHAR_BIT)-1);
diff --git a/lib/paritydi2.c b/lib/paritydi2.c
index 9d349b8..e7bebf6 100644
--- a/lib/paritydi2.c
+++ b/lib/paritydi2.c
@@ -11,14 +11,15 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
 /* Returns: 1 if number of bits is odd else returns 0 */
 
-si_int __paritysi2(si_int a);
+si_int COMPILER_RT_ABI __paritysi2(si_int a);
 
-si_int
+COMPILER_RT_ABI si_int
 __paritydi2(di_int a)
 {
     dwords x;
diff --git a/lib/paritysi2.c b/lib/paritysi2.c
index 76de328..64d509f 100644
--- a/lib/paritysi2.c
+++ b/lib/paritysi2.c
@@ -11,12 +11,13 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
 /* Returns: 1 if number of bits is odd else returns 0 */
 
-si_int
+COMPILER_RT_ABI si_int
 __paritysi2(si_int a)
 {
     su_int x = (su_int)a;
diff --git a/lib/popcountdi2.c b/lib/popcountdi2.c
index 0778a66..136fc04 100644
--- a/lib/popcountdi2.c
+++ b/lib/popcountdi2.c
@@ -11,12 +11,13 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
 /* Returns: count of 1 bits */
 
-si_int
+COMPILER_RT_ABI si_int
 __popcountdi2(di_int a)
 {
     du_int x2 = (du_int)a;
diff --git a/lib/popcountsi2.c b/lib/popcountsi2.c
index e425b0b..bfaa3ff 100644
--- a/lib/popcountsi2.c
+++ b/lib/popcountsi2.c
@@ -11,12 +11,13 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
 /* Returns: count of 1 bits */
 
-si_int
+COMPILER_RT_ABI si_int
 __popcountsi2(si_int a)
 {
     su_int x = (su_int)a;
diff --git a/lib/powidf2.c b/lib/powidf2.c
index 0200e1e..2e211eb 100644
--- a/lib/powidf2.c
+++ b/lib/powidf2.c
@@ -11,12 +11,13 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
 /* Returns: a ^ b */
 
-double
+COMPILER_RT_ABI double
 __powidf2(double a, si_int b)
 {
     const int recip = b < 0;
diff --git a/lib/powisf2.c b/lib/powisf2.c
index c834b96..e6b43b3 100644
--- a/lib/powisf2.c
+++ b/lib/powisf2.c
@@ -11,12 +11,13 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
 /* Returns: a ^ b */
 
-float
+COMPILER_RT_ABI float
 __powisf2(float a, si_int b)
 {
     const int recip = b < 0;
diff --git a/lib/subdf3.c b/lib/subdf3.c
new file mode 100644
index 0000000..825e3c6
--- /dev/null
+++ b/lib/subdf3.c
@@ -0,0 +1,30 @@
+//===-- lib/adddf3.c - Double-precision subtraction ---------------*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is dual licensed under the MIT and the University of Illinois Open
+// Source Licenses. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements double-precision soft-float subtraction with the
+// IEEE-754 default rounding (to nearest, ties to even).
+//
+//===----------------------------------------------------------------------===//
+#include "abi.h"
+
+#define DOUBLE_PRECISION
+#include "fp_lib.h"
+
+fp_t COMPILER_RT_ABI __adddf3(fp_t a, fp_t b);
+
+
+ARM_EABI_FNALIAS(dsub, subdf3);
+
+// Subtraction; flip the sign bit of b and add.
+COMPILER_RT_ABI fp_t
+__subdf3(fp_t a, fp_t b) {
+    return __adddf3(a, fromRep(toRep(b) ^ signBit));
+}
+
+/* FIXME: rsub for ARM EABI */
diff --git a/lib/subsf3.c b/lib/subsf3.c
new file mode 100644
index 0000000..625376a
--- /dev/null
+++ b/lib/subsf3.c
@@ -0,0 +1,29 @@
+//===-- lib/subsf3.c - Single-precision subtraction ---------------*- C -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is dual licensed under the MIT and the University of Illinois Open
+// Source Licenses. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements single-precision soft-float subtraction with the
+// IEEE-754 default rounding (to nearest, ties to even).
+//
+//===----------------------------------------------------------------------===//
+#include "abi.h"
+
+#define SINGLE_PRECISION
+#include "fp_lib.h"
+
+fp_t COMPILER_RT_ABI __addsf3(fp_t a, fp_t b);
+
+ARM_EABI_FNALIAS(fsub, subsf3);
+
+// Subtraction; flip the sign bit of b and add.
+COMPILER_RT_ABI fp_t
+__subsf3(fp_t a, fp_t b) {
+    return __addsf3(a, fromRep(toRep(b) ^ signBit));
+}
+
+/* FIXME: rsub for ARM EABI */
diff --git a/lib/subvdi3.c b/lib/subvdi3.c
index 17f55d0..36b51ad 100644
--- a/lib/subvdi3.c
+++ b/lib/subvdi3.c
@@ -1,6 +1,6 @@
 /* ===-- subvdi3.c - Implement __subvdi3 -----------------------------------===
  *
- *      	       The LLVM Compiler Infrastructure
+ *                The LLVM Compiler Infrastructure
  *
  * This file is dual licensed under the MIT and the University of Illinois Open
  * Source Licenses. See LICENSE.TXT for details.
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 #include <stdlib.h>
@@ -19,7 +20,7 @@
 
 /* Effects: aborts if a - b overflows */
 
-di_int
+COMPILER_RT_ABI di_int
 __subvdi3(di_int a, di_int b)
 {
     di_int s = a - b;
diff --git a/lib/subvsi3.c b/lib/subvsi3.c
index 51bb454..03983f7 100644
--- a/lib/subvsi3.c
+++ b/lib/subvsi3.c
@@ -1,6 +1,6 @@
 /* ===-- subvsi3.c - Implement __subvsi3 -----------------------------------===
  *
- *      	       The LLVM Compiler Infrastructure
+ *                The LLVM Compiler Infrastructure
  *
  * This file is dual licensed under the MIT and the University of Illinois Open
  * Source Licenses. See LICENSE.TXT for details.
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 #include <stdlib.h>
@@ -19,7 +20,7 @@
 
 /* Effects: aborts if a - b overflows */
 
-si_int
+COMPILER_RT_ABI si_int
 __subvsi3(si_int a, si_int b)
 {
     si_int s = a - b;
diff --git a/lib/truncdfsf2.c b/lib/truncdfsf2.c
index 92609fb..1dbf02f 100644
--- a/lib/truncdfsf2.c
+++ b/lib/truncdfsf2.c
@@ -41,6 +41,8 @@
 #include <limits.h>
 #include <stdbool.h>
 
+#include "abi.h"
+
 typedef double src_t;
 typedef uint64_t src_rep_t;
 #define SRC_REP_C UINT64_C
@@ -66,7 +68,10 @@ static inline dst_t dstFromRep(dst_rep_t x) {
 
 // End helper routines.  Conversion implementation follows.
 
-dst_t __truncdfsf2(src_t a) {
+ARM_EABI_FNALIAS(d2f, truncdfsf2);
+
+COMPILER_RT_ABI dst_t
+__truncdfsf2(src_t a) {
     
     // Various constants whose values follow from the type parameters.
     // Any reasonable optimizer will fold and propagate all of these.
diff --git a/lib/ucmpdi2.c b/lib/ucmpdi2.c
index ead7267..f2d3f99 100644
--- a/lib/ucmpdi2.c
+++ b/lib/ucmpdi2.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -19,7 +20,7 @@
  *           if (a >  b) returns 2
  */
 
-si_int
+COMPILER_RT_ABI si_int
 __ucmpdi2(du_int a, du_int b)
 {
     udwords x;
diff --git a/lib/udivdi3.c b/lib/udivdi3.c
index 642f2fb..bbd551a 100644
--- a/lib/udivdi3.c
+++ b/lib/udivdi3.c
@@ -11,14 +11,15 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
-du_int __udivmoddi4(du_int a, du_int b, du_int* rem);
+du_int COMPILER_RT_ABI __udivmoddi4(du_int a, du_int b, du_int* rem);
 
 /* Returns: a / b */
 
-du_int
+COMPILER_RT_ABI du_int
 __udivdi3(du_int a, du_int b)
 {
     return __udivmoddi4(a, b, 0);
diff --git a/lib/udivmoddi4.c b/lib/udivmoddi4.c
index 693736f..c5db21c 100644
--- a/lib/udivmoddi4.c
+++ b/lib/udivmoddi4.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -20,7 +21,9 @@
 
 /* Translated from Figure 3-40 of The PowerPC Compiler Writer's Guide */
 
-du_int
+ARM_EABI_FNALIAS(uldivmod, udivmoddi4);
+
+COMPILER_RT_ABI du_int
 __udivmoddi4(du_int a, du_int b, du_int* rem)
 {
     const unsigned n_uword_bits = sizeof(su_int) * CHAR_BIT;
diff --git a/lib/udivmodsi4.c b/lib/udivmodsi4.c
index 38b5bd4..2a3ee27 100644
--- a/lib/udivmodsi4.c
+++ b/lib/udivmodsi4.c
@@ -11,20 +11,21 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
-extern su_int __udivsi3(su_int n, su_int d);
+extern su_int COMPILER_RT_ABI __udivsi3(su_int n, su_int d);
 
 
 /* Returns: a / b, *rem = a % b  */
 
-su_int 
+COMPILER_RT_ABI su_int
 __udivmodsi4(su_int a, su_int b, su_int* rem)
 {
   si_int d = __udivsi3(a,b);
   *rem = a - (d*b);
-  return d; 
+  return d;
 }
 
 
diff --git a/lib/udivsi3.c b/lib/udivsi3.c
index 476f2bc..721ae89 100644
--- a/lib/udivsi3.c
+++ b/lib/udivsi3.c
@@ -11,6 +11,7 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
@@ -18,7 +19,9 @@
 
 /* Translated from Figure 3-40 of The PowerPC Compiler Writer's Guide */
 
-su_int
+ARM_EABI_FNALIAS(uidiv, udivsi3);
+
+COMPILER_RT_ABI su_int
 __udivsi3(su_int n, su_int d)
 {
     const unsigned n_uword_bits = sizeof(su_int) * CHAR_BIT;
diff --git a/lib/umoddi3.c b/lib/umoddi3.c
index adb4b0e..9de1a64 100644
--- a/lib/umoddi3.c
+++ b/lib/umoddi3.c
@@ -11,14 +11,15 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
-du_int __udivmoddi4(du_int a, du_int b, du_int* rem);
+du_int COMPILER_RT_ABI __udivmoddi4(du_int a, du_int b, du_int* rem);
 
 /* Returns: a % b */
 
-du_int
+COMPILER_RT_ABI du_int
 __umoddi3(du_int a, du_int b)
 {
     du_int r;
diff --git a/lib/umodsi3.c b/lib/umodsi3.c
index d3aaaeae..569b7fc 100644
--- a/lib/umodsi3.c
+++ b/lib/umodsi3.c
@@ -11,14 +11,15 @@
  *
  * ===----------------------------------------------------------------------===
  */
+#include "abi.h"
 
 #include "int_lib.h"
 
 /* Returns: a % b */
 
-su_int __udivsi3(su_int a, su_int b);
+su_int COMPILER_RT_ABI __udivsi3(su_int a, su_int b);
 
-su_int
+COMPILER_RT_ABI su_int
 __umodsi3(su_int a, su_int b)
 {
     return a - __udivsi3(a, b) * b;