[Bug tree-optimization/68333] New: [6 Regression] FAIL: gcc.dg/vect/slp-multitypes-4.c -flto -ffat-lto-objects scan-tree-dump-times vect "vectorized 1 loops" 1

[jgreenhalgh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68333

            Bug ID: 68333
           Summary: [6 Regression] FAIL: gcc.dg/vect/slp-multitypes-4.c
                    -flto -ffat-lto-objects  scan-tree-dump-times vect
                    "vectorized 1 loops" 1
           Product: gcc
           Version: 6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jgreenhalgh at gcc dot gnu.org
                CC: dje at gcc dot gnu.org, michael.collison at linaro dot org,
                    rguenth at gcc dot gnu.org
  Target Milestone: ---
            Target: powerpc64-unknown-linux-gnu

This failure, along with:

FAIL: gcc.dg/vect/slp-multitypes-4.c -flto -ffat-lto-objects 
scan-tree-dump-times vect "vectorizing stmts using SLP" 1
FAIL: gcc.dg/vect/slp-multitypes-4.c scan-tree-dump-times vect "vectorized 1
loops" 1
FAIL: gcc.dg/vect/slp-multitypes-4.c scan-tree-dump-times vect "vectorizing
stmts using SLP" 1
FAIL: gcc.dg/vect/slp-multitypes-5.c -flto -ffat-lto-objects 
scan-tree-dump-times vect "vectorized 1 loops" 1
FAIL: gcc.dg/vect/slp-multitypes-5.c -flto -ffat-lto-objects 
scan-tree-dump-times vect "vectorizing stmts using SLP" 1
FAIL: gcc.dg/vect/slp-multitypes-5.c scan-tree-dump-times vect "vectorized 1
loops" 1
FAIL: gcc.dg/vect/slp-multitypes-5.c scan-tree-dump-times vect "vectorizing
stmts using SLP" 1
FAIL: gcc.dg/vect/slp-perm-8.c -flto -ffat-lto-objects  scan-tree-dump-times
vect "vectorized 1 loops" 1
FAIL: gcc.dg/vect/slp-perm-8.c scan-tree-dump-times vect "vectorized 1 loops" 1
FAIL: gcc.dg/vect/vect-125.c -flto -ffat-lto-objects  scan-tree-dump vect
"vectorized 1 loops"
FAIL: gcc.dg/vect/vect-125.c scan-tree-dump vect "vectorized 1 loops"

Seems to have started some time in May, near vectorization patches added on
2015-05-22. On IRC:

<richi> Rhy0lite: but yes, widen-sum pattern might be detected but not
supported afterwards as well
<jgreenhalgh> richi: On AArch64 it was slp-multitypes-4.c:16:3: note: not
vectorized: relevant stmt not supported: patt_127 = _7 w+ 1;
<richi> ah, caused by 2015-05-22, (vect_recog_widen_sum_pattern): Likewise.
<richi> formerly only detected for reductions now also in other places
<richi> there is a PR for the dot_prod case (and the sad case as well)
<Rhy0lite> richi: same note in PPC dump as jgreenhalgh showed
<richi> hum, vect_recog_widen_sum_pattern misses an optab check.
<Rhy0lite> note: not vectorized: relevant stmt not supported: patt_127 = _7 w+
1;
<richi> ok, can you open a new PR please?  looks like a different issue than
the dot_prod/sad issue

The failure currently shows up for the rs6000 and ia64 backends, but Michael
Collison's proposed patch adding support for vector widening add patterns would
introduce it to the aarch64 port (
https://gcc.gnu.org/ml/gcc-patches/2015-11/msg00898.html ).

With Michael's patch applied, I see the following in the dumps for vect-125.c
on AArch64:

        vect-125.c:10:3: note: vect_recog_widen_sum_pattern: detected: patt_58
= _13 w+ _24;
        vect-125.c:10:3: note: pattern recognized: patt_58 = _13 w+ _24;
        [...snip...]
        vect-125.c:10:3: note: ==> examining pattern statement: patt_58 = _13
w+ _24;
        vect-125.c:10:3: note: vect_is_simple_use: operand _13
        vect-125.c:10:3: note: def_stmt: _13 = *_12;
        vect-125.c:10:3: note: type of def: internal
        vect-125.c:10:3: note: not vectorized: relevant stmt not supported:
patt_58 = _13 w+ _24;