the generation of getelementptr instruction

Project Infrastructure LLVM Dev List Archives
1

Hi all,
I’m reading the language reference of LLVM here: http://llvm.org/docs/LangRef.html#i_getelementptr

In the first example, it shows that the generated IR is supposed to be something like:

%reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13


But when I copied the code and compile it myself (using v2.7), I got five getelementptr instructions, which are just like the equivalent code given later in the document. 

In the case above, why the resulting IRs are different? 

Thanks a lot.

 

Regards,

--Wenbin
2

Just a quirk of frontend code generation; you'll get the shown result with -O.

-Eli

3

Hi Wenbin,

I'm reading the language reference of LLVM here:
LLVM Language Reference Manual — LLVM 18.0.0git documentation

In the first example, it shows that the generated IR is supposed to be
something like:
%reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
But when I copied the code and compile it myself (using v2.7), I got
five getelementptr instructions, which are just like the equivalent
code given later in the document. In the case above, why the
resulting IRs are different?

if you enable optimizations in the frontend (-O1), the resulting LLVM
code will look just as shown in the example.

Regards,
Christoph