[Bultin][ARM] Make aeabi_memset be Thumb1 compatible

[weimingz]

Summary

Also fix some instructions to conform with the syntax of ARM reference manual.
For example, muls Rd, Rn, Rd and orrs Rd, Rm

[kubamracek]

While I don't really understand the assembly details, I'm confirming that this fixes the build failure on Darwin.

[rengolin]

All encodings seem to be Thumb1, not just ARMv6M. LGTM. Thanks!

The commit title should be a bit more generic, as this is not just about memset. You can describe the changes on the body of the message.

[compnerd]

That muls requirement that Rd and Rm be the same is dumb.

[rengolin]

In D28971#652414, @compnerd wrote:

That muls requirement that Rd and Rm be the same is dumb.

Designing ISAs is hard. :)

compiler-rt/trunk/lib/builtins/CMakeLists.txt

set(thumb1_EABI_SOURCES
   arm/aeabi_cfcmpeq_check_nan.c
   arm/aeabi_div0.c
   arm/aeabi_drsub.c
   arm/aeabi_fcmp.S
   arm/aeabi_frsub.c
   arm/aeabi_idivmod.S
   arm/aeabi_memcmp.S
   arm/aeabi_memcpy.S
+  arm/aeabi_memset.S
   arm/aeabi_memmove.S
   arm/aeabi_uidivmod.S)
 
 set(arm_Thumb1_JT_SOURCES
   arm/switch16.S
   arm/switch32.S
   arm/switch8.S
   arm/switchu8.S)

compiler-rt/trunk/lib/builtins/arm/aeabi_idivmod.S

#endif
 
         .syntax unified
         .p2align 2
 DEFINE_COMPILERRT_FUNCTION(__aeabi_idivmod)
 #if __ARM_ARCH_ISA_THUMB == 1
         push    {r0, r1, lr}
         bl      SYMBOL_NAME(__divsi3)
         pop     {r1, r2, r3} // now r0 = quot, r1 = num, r2 = denom
-        muls    r2, r2, r0   // r2 = quot * denom
+        muls    r2, r0, r2   // r2 = quot * denom
         subs    r1, r1, r2
         JMP     (r3)
 #else
         push    { lr }
         sub     sp, sp, #4
         mov     r2, sp
 #if defined(__MINGW32__)
         mov     r3, r0

compiler-rt/trunk/lib/builtins/arm/aeabi_memset.S

DEFINE_COMPILERRT_FUNCTION(__aeabi_memset)
         b       memset
 END_COMPILERRT_FUNCTION(__aeabi_memset)
 
 DEFINE_AEABI_FUNCTION_ALIAS(__aeabi_memset4, __aeabi_memset)
 DEFINE_AEABI_FUNCTION_ALIAS(__aeabi_memset8, __aeabi_memset)
 
 DEFINE_COMPILERRT_FUNCTION(__aeabi_memclr)
         mov     r2, r1
-        mov     r1, #0
+        movs    r1, #0
         b       memset
 END_COMPILERRT_FUNCTION(__aeabi_memclr)
 
 DEFINE_AEABI_FUNCTION_ALIAS(__aeabi_memclr4, __aeabi_memclr)
 DEFINE_AEABI_FUNCTION_ALIAS(__aeabi_memclr8, __aeabi_memclr)
 
 NO_EXEC_STACK_DIRECTIVE
 

compiler-rt/trunk/lib/builtins/arm/aeabi_uidivmod.S

#endif
         .p2align 2
 DEFINE_COMPILERRT_FUNCTION(__aeabi_uidivmod)
 #if __ARM_ARCH_ISA_THUMB == 1
         cmp     r0, r1
         bcc     LOCAL_LABEL(case_denom_larger)
         push    {r0, r1, lr}
         bl      SYMBOL_NAME(__aeabi_uidiv)
         pop     {r1, r2, r3}
-        muls    r2, r2, r0 // r2 = quot * denom
+        muls    r2, r0, r2 // r2 = quot * denom
         subs    r1, r1, r2
         JMP     (r3)
 LOCAL_LABEL(case_denom_larger):
         movs    r1, r0
         movs    r0, #0
         JMP     (lr)
 #else
         push    { lr }

compiler-rt/trunk/lib/builtins/arm/comparesf2.S

#endif
 
     // 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.
 #if __ARM_ARCH_ISA_THUMB == 1
     lsrs    r6,     r3, #1
-    orrs    r6,     r2, r6
+    orrs    r6,     r2
 #else
     orrs    r12,    r2, r3, lsr #1
 #endif
     // 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.
 #if __ARM_ARCH_ISA_THUMB == 1
     beq     1f
     movs    r6,     r0
#if defined(COMPILER_RT_ARMHF_TARGET)
     vmov r0, s0
     vmov r1, s1
 #endif
 #if __ARM_ARCH_ISA_THUMB == 1
     push    {r6, lr}
     lsls    r2,        r0, #1
     lsls    r3,        r1, #1
     lsrs    r6,        r3, #1
-    orrs    r6,        r2, r6
+    orrs    r6,        r2
     beq     1f
     movs    r6,        r0
     eors    r6,        r1
 1:
     bmi     2f
     subs    r0,        r2, r3
 2:
     bhs     3f