[compiler-rt][builtins] Support builtins for armv8m.base

[fdischner]

Summary

This allows building the compiler builtins library for the Armv8-M
Baseline architecture. It can be built in the same way as other
baremetal targets using the appropriate '--target' flag
(e.g. --target=armv8m.base-eabi).

NOTE: As with the other Cortex-M targets, only the builtins library is supported. There is no support for sanitizers, etc.

The armv8m.base architecture is a superset of armv6m, so adding it to
the cmake files using thumb1_SOURCES is almost enough for it to compile.
Minor changes are needed to divsi3 and udivsi3, because armv8m.base does
have support for div instructions but not mov with an immediate operand.

[peter.smith]

Looks OK to me. I think we can simplify by always using the movs instruction.

compiler-rt/lib/builtins/arm/divsi3.S

40

In this case I think it is better to use movs r0,#0 universally as it is a 16-bit instruction and mov r0, #0 is a 32-bit instruction.

As we don't use the flags before returning and as flags are not preserved across prodcedure call boundaries there isn't an advantage to use the larger instruction.

Perhaps

// Use movs for compatibility with v8-m.base.
movs r0, #0

compiler-rt/lib/builtins/arm/udivsi3.S

35

Same here, we can always use movs

[fdischner]

As per review, simplify by always using movs.

[fdischner]

compiler-rt/lib/builtins/arm/divsi3.S

40

Thanks, that was my initial thought as well, but I wasn't sure if there was some advantage to using mov that I wasn't aware of. I've updated the patch to simplify this.

[peter.smith]

Thanks for the update, I'm happy with the changes. Best wait a bit to let anyone in the US timezone comment.

[MaskRay]

LGTM. IIUC this makes clang --target=armv8m.base-eabi -c compilable.

[fdischner]

In D143297#4107528, @MaskRay wrote:

LGTM. IIUC this makes clang --target=armv8m.base-eabi -c compilable.

I’m not sure I understand. Clang already supports compiling for the armv8m.base-eabi target. This just allows building the compiler-rt builtins library for this target. Or maybe that’s what you are saying?

[fdischner]

Is there something that still needs to be done for this to be committed? I'm new to LLVM and don't have commit access, so could use some help here.

[MaskRay]

In D143297#4130668, @fdischner wrote:

Is there something that still needs to be done for this to be committed? I'm new to LLVM and don't have commit access, so could use some help here.

You can provide your name and email address to be used in git commit --amend --author='... <...>' so that someone with commit access can push this patch for you.
Note: if you upload the patch via arc diff 'HEAD^' (https://llvm.org/docs/Phabricator.html), arc patch D143297 will obtain the name and email.

I think it'd be nice to adjust the commit message (Phabricator summary) to include information about how you configure compiler-rt and do the testing.
I was confused so I asked for confirmation that this fixed the build when someone uses clang --target=armv8m.base-eabi -c to assemble the file.

[fdischner]

Updating commit message

[fdischner]

I've updated the commit message to hopefully clear up the confusion and also re-uploaded the patch using arc diff --update D143297. In case the author still isn't correct with arc patch D143297, please use "Frank Dischner <[e-mail]>" as the author. Thanks!

compiler-rt/cmake/builtin-config-ix.cmake

 
 builtin_check_c_compiler_source(COMPILER_RT_HAS_ASM_LSE
 "
 asm(\".arch armv8-a+lse\");
 asm(\"cas w0, w1, [x2]\");
 ")
 
 set(ARM64 aarch64)
-set(ARM32 arm armhf armv4t armv5te armv6 armv6m armv7m armv7em armv7 armv7s armv7k armv8m.main armv8.1m.main)
+set(ARM32 arm armhf armv4t armv5te armv6 armv6m armv7m armv7em armv7 armv7s armv7k armv8m.base armv8m.main armv8.1m.main)
 set(AVR avr)
 set(HEXAGON hexagon)
 set(X86 i386)
 set(X86_64 x86_64)
 set(LOONGARCH64 loongarch64)
 set(MIPS32 mips mipsel)
 set(MIPS64 mips64 mips64el)
 set(PPC32 powerpc powerpcspe)

compiler-rt/lib/builtins/CMakeLists.txt

 set(arm64_SOURCES ${aarch64_SOURCES})
 set(arm64e_SOURCES ${aarch64_SOURCES})
 set(arm64_32_SOURCES ${aarch64_SOURCES})
 
 # macho_embedded archs
 set(armv6m_SOURCES ${thumb1_SOURCES})
 set(armv7m_SOURCES ${arm_SOURCES})
 set(armv7em_SOURCES ${arm_SOURCES})
+set(armv8m.base_SOURCES ${thumb1_SOURCES})
 set(armv8m.main_SOURCES ${arm_SOURCES})
 set(armv8.1m.main_SOURCES ${arm_SOURCES})
 
 # 8-bit AVR MCU
 set(avr_SOURCES
   avr/mulqi3.S
   avr/mulhi3.S
   avr/exit.S

compiler-rt/lib/builtins/arm/divsi3.S

 
 DEFINE_COMPILERRT_FUNCTION(__divsi3)
 #if __ARM_ARCH_EXT_IDIV__
    tst     r1,r1
    beq     LOCAL_LABEL(divzero)
    sdiv    r0, r0, r1
    bx      lr
 LOCAL_LABEL(divzero):
-   mov     r0,#0
+   // Use movs for compatibility with v8-m.base.
+   movs    r0,#0
    bx      lr
 #else
 ESTABLISH_FRAME
 //  Set aside the sign of the quotient.
 #  if defined(USE_THUMB_1)
     movs    r4,     r0
     eors    r4,     r1
 #  else

compiler-rt/lib/builtins/arm/udivsi3.S

 DEFINE_COMPILERRT_FUNCTION(__udivsi3)
 #if __ARM_ARCH_EXT_IDIV__
   tst     r1, r1
   beq     LOCAL_LABEL(divby0)
   udiv  r0, r0, r1
   bx    lr
 
 LOCAL_LABEL(divby0):
-  mov     r0, #0
+  // Use movs for compatibility with v8-m.base.
+  movs    r0, #0
 #  ifdef __ARM_EABI__
   b       __aeabi_idiv0
 #  else
   JMP(lr)
 #  endif
 
 #else // ! __ARM_ARCH_EXT_IDIV__
   cmp r1, #1