[Bug target/105213] New: Extend __builtin_{un,}pack_{longdouble,ibm128} valid for soft-float

[Bug target/105213] New: Extend __builtin_{un,}pack_{longdouble,ibm128} valid for soft-float

[linkw at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105213

            Bug ID: 105213
           Summary: Extend __builtin_{un,}pack_{longdouble,ibm128} valid
                    for soft-float
           Product: gcc
           Version: 12.0
               URL: https://gcc.gnu.org/pipermail/gcc-patches/2021-Decembe
                    r/586712.html,https://gcc.gnu.org/pipermail/gcc-patche
                    s/2022-March/591147.html
            Status: UNCONFIRMED
          Keywords: accepts-invalid, ice-on-invalid-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: linkw at gcc dot gnu.org
                CC: asolokha at gmx dot com, bergner at gcc dot gnu.org,
                    linkw at gcc dot gnu.org, segher at gcc dot gnu.org
        Depends on: 103623
  Target Milestone: ---
            Target: powerpc-*-linux-gnu

+++ This bug was initially created as a clone of Bug #103623 +++

As PR103623 shows, we don't support __builtin_{un,}pack_{longdouble,ibm128} for
soft-float before, as Segher and Peter pointed out, we should make it
available.

Segher posted one patch to fix the reload ICE like:

===
diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md
index fdfbc6566a5c..f05b8358ba0a 100644
--- a/gcc/config/rs6000/rs6000.md
+++ b/gcc/config/rs6000/rs6000.md
@@ -14580,10 +14580,10 @@ (define_insn_and_split "unpack<mode>_dm"
   [(set_attr "type" "fp,fpstore,mtvsr,mfvsr,store")])

 (define_insn_and_split "unpack<mode>_nodm"
-  [(set (match_operand:<FP128_64> 0 "nonimmediate_operand" "=d,m")
+  [(set (match_operand:<FP128_64> 0 "nonimmediate_operand" "=d,m,m")
        (unspec:<FP128_64>
-        [(match_operand:FMOVE128 1 "register_operand" "d,d")
-         (match_operand:QI 2 "const_0_to_1_operand" "i,i")]
+        [(match_operand:FMOVE128 1 "register_operand" "d,d,r")
+         (match_operand:QI 2 "const_0_to_1_operand" "i,i,i")]
         UNSPEC_UNPACK_128BIT))]
   "(!TARGET_POWERPC64 || !TARGET_DIRECT_MOVE) && FLOAT128_2REG_P (<MODE>mode)"
   "#"
@@ -14600,7 +14600,7 @@ (define_insn_and_split "unpack<mode>_nodm"

   operands[3] = gen_rtx_REG (<FP128_64>mode, fp_regno);
 }
-  [(set_attr "type" "fp,fpstore")])
+  [(set_attr "type" "fp,fpstore,store")])

 (define_insn_and_split "pack<mode>"
   [(set (match_operand:FMOVE128 0 "register_operand" "=&d")
===

It needs more test coverage and documentation change in GCC13.


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103623
[Bug 103623] [12 Regression] error: unable to generate reloads (ICE in
curr_insn_transform, at lra-constraints.c:4132), or error: insn does not
satisfy its constraints (ICE in extract_constrain_insn_cached, at recog.c:2682)

[Bug target/105213] Extend __builtin_{un,}pack_{longdouble,ibm128} valid for soft-float

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105213

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement

[Bug target/105213] Extend __builtin_{un,}pack_{longdouble,ibm128} valid for soft-float

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105213

--- Comment #1 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Segher Boessenkool <segher@gcc.gnu.org>:

https://gcc.gnu.org/g:0508f0b810985f4a8543ce44701ec7330ef29796

commit r12-8091-g0508f0b810985f4a8543ce44701ec7330ef29796
Author: Segher Boessenkool <segher@kernel.crashing.org>
Date:   Mon Apr 11 15:43:38 2022 +0000

    rs6000: Fix unpack for no-direct-move (PR103623)

    The _dm alternative works fine for soft-float, but the _nodm variant
    pattern is missing that alternative.  So, let's add that.

    There probably should be an r,r,i alternative as well (or we can make it
    rm,r,i), but that is for later.

    2022-04-11  Segher Boessenkool  <segher@kernel.crashing.org>

            PR target/105213
            PR target/103623
            * config/rs6000/rs6000.md (unpack<mode>_nodm): Add m,r,i
alternative.

[Bug target/105213] Extend __builtin_{un,}pack_{longdouble,ibm128} valid for soft-float

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105213

Segher Boessenkool <segher at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |FIXED
           Assignee|unassigned at gcc dot gnu.org      |segher at gcc dot gnu.org

--- Comment #2 from Segher Boessenkool <segher at gcc dot gnu.org> ---
Fixed.