[llvm-bugs] [Bug 51677] New: llvm.smul.fix.sat.i8(-128, -128, 0) equals -128

via llvm-bugs llvm-bugs at lists.llvm.org
Mon Aug 30 09:09:47 PDT 2021


            Bug ID: 51677
           Summary: llvm.smul.fix.sat.i8(-128, -128, 0) equals -128
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Common Code Generator Code
          Assignee: unassignedbugs at nondot.org
          Reporter: twostepted at gmail.com
                CC: llvm-bugs at lists.llvm.org

The result of this operation should be 127.  
This was confirmed in irc
> <LebedevRI> for scale=0, i would still say this is a bug

$ cat test.ll

; test.ll - smul.fix.sat i8min * i8min should == i8max
define i8 @main() {
    %res = call i8 @llvm.smul.fix.sat.i8(i8 -128, i8 -128, i32 0)
    %x = sub i8 %res, 127
    ret i8 %x

declare i8 @llvm.smul.fix.sat.i8(i8, i8, i32)

$ lli-12 test.ll
$ echo $?
$ cat test2.ll

; test2.ll - smul.fix.sat i8max * i8max should == i8max
define i8 @main() {
    %res = call i8 @llvm.smul.fix.sat.i8(i8 127, i8 127, i32 0)

    %x = sub i8 %res, 127
    ret i8 %x

declare i8 @llvm.smul.fix.sat.i8(i8, i8, i32)

$ lli-12 test2.ll
$ echo $?

So it seems like there is a signedness error when dealing with signed type
maximums and minimums.  I observed the same using i128.  This doesn't seem to
happend with the unsigned umul.fix.sat intrinsics.

The same results were observed using trunk.  

Not sure i've chosen the correct component 'Common Code Generator Code' in the
drop-down above.

You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20210830/99bf187f/attachment-0001.html>

More information about the llvm-bugs mailing list