GCC Miscompilation of vector shift

[bugproxy]

Vector shift miscompilation

---Steps to Reproduce---
   t.c:
__vector int
foo (__vector int a, int s)
{
  return a << (s + 3);
}

gcc -O3 -march=z13 t.c -S -mzvector

foo:
        veslf %v24,%v24,0(%r2) <---- +3 is omitted
        br %r14

Contact Information = <email address hidden>

Userspace tool common name: GCC

The userspace tool has the following bit modes: 64

Backported fix has been applied to GCC 8 branch as r273493:

https://gcc.gnu.org/viewcvs?rev=273493&root=gcc&view=rev

S/390: Fix vector shift count operand

We currently use subst definitions to handle the different variants of shift
count operands. Unfortunately, in the vector shift pattern the shift count
operand is used directly. Without it being adjusted for the 'subst' variants the
displacement value is omitted resulting in a wrong shift count being applied.

This patch needs to be applied to older branches as well.

gcc/ChangeLog:

2019-07-15 Andreas Krebbel <email address hidden>

Backport from mainline
2019-07-01 Andreas Krebbel <email address hidden>

* config/s390/vector.md: Fix shift count operand printing.

gcc/testsuite/ChangeLog:

2019-07-15 Andreas Krebbel <email address hidden>

Backport from mainline
2019-07-01 Andreas Krebbel <email address hidden>

* gcc.target/s390/vector/vec-shift-2.c: New test.

Added:
branches/gcc-8-branch/gcc/testsuite/gcc.target/s390/vector/vec-shift-2.c
Modified:
branches/gcc-8-branch/gcc/ChangeLog
branches/gcc-8-branch/gcc/config/s390/vector.md
branches/gcc-8-branch/gcc/testsuite/ChangeLog

[Matthias Klose (doko)]

that's on the gcc-7 branch as well. Is it needed on older branches as well?

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2019-07-24 03:53 EDT-------
(In reply to comment #5)
> that's on the gcc-7 branch as well. Is it needed on older branches as well?

Yes, the problem is in all GCC branches starting with GCC 6.

[Matthias Klose (doko)]

the fix for GCC 6 may be not needed, as it's not the default compiler in any supported release. However GCC 5 needs the fix for 18.06 LTS.

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2019-08-01 02:49 EDT-------
(In reply to comment #7)
> the fix for GCC 6 may be not needed, as it's not the default compiler in any
> supported release. However GCC 5 needs the fix for 18.06 LTS.

GCC 5 is ok as is. The patch introducing the problem went into GCC 6.

[Dimitri John Ledkov (xnox)]

Adjusted statuses as per above comments.

[Dimitri John Ledkov (xnox)]

I've updated lots of series tasks and statuses.
Probably best viewed on launchpad https://bugs.launchpad.net/ubuntu/+source/gcc-8/+bug/1837522

Summary:

gcc-7 & gcc-8 are fixed in Eoan and Focal

Disco will go end of life soon, and is marked as Won't Fix, as there is not enough time to complete toolchain SRU prior to its End of Life.

Bionic update is in progress, with test rebuild results being analyzed. However, gcc 7.5 was recently announced for November. We are evaluating doing two gcc-7 SRUs with current results, and then later with 7.5. Or we might postpone currently inflight gcc-7 SRU and include this fix with the gcc 7.5 targetting security pocket.

[Dimitri John Ledkov (xnox)]

gcc-7.5 is now in bionic-updates.

[Dimitri John Ledkov (xnox)]

gcc-8.4 update cycle has now started at https://bugs.launchpad.net/ubuntu/+source/gcc-8/+bug/1866835

it will take a while, but hopefully a lot less than the last 7.5 update which hit a lot of QA automation roadblocks, which we have now resolved.

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2020-04-09 02:57 EDT-------
IBM Bugzilla status->closed. Fix Released for all requested packages