[Bug c/121485] New: RISC-V Zvkned vector-scalar intrinsics use wrong EMUL (m1 for m2/m4/m8 variants of intrinsics)

[jussi.kivilinna at iki dot fi via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121485

            Bug ID: 121485
           Summary: RISC-V Zvkned vector-scalar intrinsics use wrong EMUL
                    (m1 for m2/m4/m8 variants of intrinsics)
           Product: gcc
           Version: 15.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jussi.kivilinna at iki dot fi
  Target Milestone: ---

I noticed issue with '__riscv_vaesz_vs_u32m1_u32m4' intrinsic with GCC
(riscv64) where intrinsic generated assembly with element group of 1 (m1)
instead of expected m4.

Following example:

  #include <riscv_vector.h>

  vuint32m4_t test_riscv_vaesz_vs_u32m1_u32m4(vuint32m4_t a, vuint32m1_t b, int
vl)
  {
    // Expecting vsetvli with "m4", gcc generates "m1"...
    return __riscv_vaesz_vs_u32m1_u32m4(a, b, vl);
  }

Results in:

  test_riscv_vaesz_vs_u32m1_u32m4:
        vsetvli zero,a0,e32,m1,ta,ma
        vaesz.vs        v8,v12
        ret

Expected assembly is (as generated by clang):

  test_riscv_vaesz_vs_u32m1_u32m4:
        vsetvli zero, a0, e32, m4, ta, ma
        vaesz.vs        v8, v12
        ret

Same issue is with m2/m4/m8 vector-scalar variants of
vaesz/vaesem/vaesef/vaesdm/vaesdf intrinsics. Tested with Godbolt riscv64 gcc
15.1.0 and gcc trunk: https://godbolt.org/z/4jGo74M3o