llvm dev - Jun 2020 - Question about Constant expressions

Hi,

Why is the getelementptr part of the call instruction, instead of
being a separate instruction in the LLVM IR ?
The LLVM IR Builder adds it as a separate instruction, but it is
rolled into one call instruction in the .bc output.
I am just curious as to why it is done this way.

#include <stdint.h>
#include <stdio.h>
int test21(void);
int test21(void) {
        printf("Hello1\n");
        return 0;
}

define dso_local i32 @test21() #0 {
  %1 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([8 x
i8], [8 x i8]* @.str, i64 0, i64 0))
  ret i32 0
}

Hi James,


Like other constant expressions, e.g., i32 0, this GEP constant

expression are "shown" inlined. Constants are unique and exist

everywhere. I guess the most "important' use case I can think

of is to use such GEPs as part of other constant expressions,

e.g., initializers of constant globals.


Cheers,

   Johannes


On 5/31/20 3:49 AM, James Courtier-Dutton via llvm-dev
wrote:
> Hi,
>
> Why is the getelementptr part of the call instruction, instead of
> being a separate instruction in the LLVM IR ?
> The LLVM IR Builder adds it as a separate instruction, but it is
> rolled into one call instruction in the .bc output.
> I am just curious as to why it is done this way.
>
> #include <stdint.h>
> #include <stdio.h>
> int test21(void);
> int test21(void) {
>          printf("Hello1\n");
>          return 0;
> }
>
> define dso_local i32 @test21() #0 {
>    %1 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([8 x
> i8], [8 x i8]* @.str, i64 0, i64 0))
>    ret i32 0
> }
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20200531/71275c9f/attachment.html>

Hi,

What I meant was, why is it not represented by something like this instead:

define dso_local i32 @test21() #0 {
  %1 = i8* getelementptr inbounds ([8 x
i8], [8 x i8]* @.str, i64 0, i64 0)
  %2 = call i32 (i8*, ...) @printf(%1)
  ret i32 0
}

instead of:

define dso_local i32 @test21() #0 {
  %1 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([8 x
i8], [8 x i8]* @.str, i64 0, i64 0))
  ret i32 0
}

Some code I am writing would be a lot simpler if the first case was used.


On Sun, 31 May 2020 at 16:56, Johannes Doerfert
<johannesdoerfert at gmail.com> wrote:
>
> Hi James,
>
>
> Like other constant expressions, e.g., i32 0, this GEP constant
>
> expression are "shown" inlined. Constants are unique and exist
>
> everywhere. I guess the most "important' use case I can think
>
> of is to use such GEPs as part of other constant expressions,
>
> e.g., initializers of constant globals.
>
>
> Cheers,
>
>   Johannes
>
>
> On 5/31/20 3:49 AM, James Courtier-Dutton via llvm-dev wrote:
>
> Hi,
>
> Why is the getelementptr part of the call instruction, instead of
> being a separate instruction in the LLVM IR ?
> The LLVM IR Builder adds it as a separate instruction, but it is
> rolled into one call instruction in the .bc output.
> I am just curious as to why it is done this way.
>
> #include <stdint.h>
> #include <stdio.h>
> int test21(void);
> int test21(void) {
>         printf("Hello1\n");
>         return 0;
> }
>
> define dso_local i32 @test21() #0 {
>   %1 = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([8 x
> i8], [8 x i8]* @.str, i64 0, i64 0))
>   ret i32 0
> }
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev