llvm dev - May 2019 - "BL" or "B" when compiling ARM

Hi

I am recently doing research on binaries in ARM architecture.

I noticed that when I compile the source code using thumb instruction set,
llvm will use BL to jump to some basic blocks rather than B, which means
the jump target by BL could also be a basic block rather than function
start. Some disassemblers would directly set the target of BL as  a start
of a function, which is not right.

I am really curious when the compiler would choose to use BL and when the
compiler would choose to use B when the target is a basic block. I noticed
sometimes there might be no differences between these two branch
instructions. Many Thanks

Regards
Muhui
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20190523/24845a28/attachment.html>

Hi Muhui,

On Thu, 23 May 2019 at 12:23, Muhui Jiang via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> I am really curious when the compiler would choose to use BL and when the
compiler would choose to use B when the target is a basic block. I noticed
sometimes there might be no differences between these two branch instructions.
BL is one of the very few instructions that's 32-bits wide on CPUs
that only support the Thumb1 instruction set (usually Cortex-M0 these
days). Because of that it's got a much larger range than any other
branch so LLVM uses it instead of a string of intermediate branches
when a target block would be out of range of the usual unconditional
branches.

Cheers.

Tim.