[pushed] aarch64: Use force_lowpart_subreg in a BFI splitter [PR119133]

[Richard Sandiford]

lowpart_subreg ICEs are the gift that keeps giving.  This is another
case where we need to use force_lowpart_subreg instead, to handle
cases where the input is already a subreg and where the combined
subreg is not allowed as a single operation.

We don't need to check can_create_pseudo_p since the input should
be a hard register rather than a subreg if !can_create_pseudo_p.

Tested on aarch64-linux-gnu & pushed as hopefuly uncontroversial.

Richard


gcc/
        PR target/119133
        * config/aarch64/aarch64.md
        (*aarch64_bfi<GPI:mode><ALLX:mode>_<SUBDI_BITS>): Use
        force_lowpart_subreg.

gcc/testsuite/
        PR target/119133
        * gcc.dg/torture/pr119133.c: New test.

diff --git a/gcc/config/aarch64/aarch64.md b/gcc/config/aarch64/aarch64.md
index b10059e4f58..03188a64ab1 100644
--- a/gcc/config/aarch64/aarch64.md
+++ b/gcc/config/aarch64/aarch64.md
@@ -6435,8 +6435,8 @@ (define_insn_and_split 
"*aarch64_bfi<GPI:mode><ALLX:mode>_<SUBDI_BITS>"
                          (match_dup 1))
        (match_dup 2))]
   {
-    operands[2] = lowpart_subreg (<GPI:MODE>mode, operands[2],
-                                 <ALLX:MODE>mode);
+    operands[2] = force_lowpart_subreg (<GPI:MODE>mode, operands[2],
+                                       <ALLX:MODE>mode);
   }
   [(set_attr "type" "bfm,neon_ins_q,neon_ins_q")
    (set_attr "arch" "*,simd,simd")]
diff --git a/gcc/testsuite/gcc.dg/torture/pr119133.c 
b/gcc/testsuite/gcc.dg/torture/pr119133.c
new file mode 100644
index 00000000000..78cdda69afe
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/pr119133.c
@@ -0,0 +1,8 @@
+/* { dg-additional-options "-fno-tree-ter" } */
+
+int
+foo(_Float16 f, int i)
+{
+  __builtin_memcpy(&i, &f, 2);
+  return i;
+}
-- 
2.25.1