Hi,

This relaxes an overzealous assert that required the fpm_t argument to
be in DImode when expanding FP8 intrinsics.  Of course this fails to
account for modeless const_ints.

Bootstrapped/regtested on aarch64-linux-gnu, OK for trunk and backport
to GCC 15?

Thanks,
Alex

gcc/ChangeLog:

        PR target/120986
        * config/aarch64/aarch64-sve-builtins.cc
        (function_expander::expand): Relax fpm_t assert to allow
        modeless const_ints.

gcc/testsuite/ChangeLog:

        PR target/120986
        * gcc.target/aarch64/torture/pr120986-2.c: New test.
---
 gcc/config/aarch64/aarch64-sve-builtins.cc            | 5 +++--
 gcc/testsuite/gcc.target/aarch64/torture/pr120986-2.c | 7 +++++++
 2 files changed, 10 insertions(+), 2 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/aarch64/torture/pr120986-2.c

diff --git a/gcc/config/aarch64/aarch64-sve-builtins.cc b/gcc/config/aarch64/aarch64-sve-builtins.cc
index 2b627a95060..19d6e36d948 100644
--- a/gcc/config/aarch64/aarch64-sve-builtins.cc
+++ b/gcc/config/aarch64/aarch64-sve-builtins.cc
@@ -4589,8 +4589,9 @@ function_expander::expand ()
     {
       /* The last element of these functions is always an fpm_t that must be
          written to FPMR before the call to the instruction itself. */
-      gcc_assert (args.last ()->mode == DImode);
-      emit_move_insn (gen_rtx_REG (DImode, FPM_REGNUM), args.last ());
+      rtx fpm = args.last ();
+      gcc_assert (CONST_INT_P (fpm) || GET_MODE (fpm) == DImode);
+      emit_move_insn (gen_rtx_REG (DImode, FPM_REGNUM), fpm);
     }
   return base->expand (*this);
 }
diff --git a/gcc/testsuite/gcc.target/aarch64/torture/pr120986-2.c b/gcc/testsuite/gcc.target/aarch64/torture/pr120986-2.c
new file mode 100644
index 00000000000..1218dead9dd
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/torture/pr120986-2.c
@@ -0,0 +1,7 @@
+/* { dg-do compile } */
+/* { dg-options "-march=armv8.2-a+sve2+fp8dot2" } */
+#include <arm_sve.h>
+svfloat16_t foo(svfloat16_t a, svmfloat8_t b, svmfloat8_t c)
+{
+  return svdot_lane_fpm (a, b, c, 0, 0);
+}

Reply via email to