ia32: Disallow mode(V1TI) [PR103020]

  Hi!

As discussed in the PR, TImode isn't supported for -m32 on x86 (for the same
reason as on most 32-bit targets, no support for > 2 * BITS_PER_WORD
precision integers), but since PR32280 V1TImode is allowed with -msse in SSE
regs, V2TImode with -mavx or V4TImode with -mavx512f.
typedef __int128 V __attribute__((vector_size ({16,32,64}));
will not work, neither typedef int I __attribute__((mode(TI)));
but mode(V1TI), mode(V2TI) etc. are accepted with a warning when those
ISAs are enabled.  But they are certainly not fully supported, for some
optabs maybe, but most of them will not.  And, veclower lowering those ops
to TImode scalar operations will not work either because TImode isn't
supported.

So, this patch keeps V1TImode etc. in VALID*_MODE macros so that we can use
it in certain instructions, but disallows it in
targetm.vector_mode_supported_p, so that we don't offer those modes to the
user as supported.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2021-11-02  Jakub Jelinek  <jakub@redhat.com>

	PR target/103020
	* config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
	modes with TImode inner mode if 32-bit.

	* gcc.target/i386/pr103020.c: New test.


	Jakub
  

On Tue, Nov 2, 2021 at 9:41 AM Jakub Jelinek <jakub@redhat.com> wrote:
>
> Hi!
>
> As discussed in the PR, TImode isn't supported for -m32 on x86 (for the same
> reason as on most 32-bit targets, no support for > 2 * BITS_PER_WORD
> precision integers), but since PR32280 V1TImode is allowed with -msse in SSE
> regs, V2TImode with -mavx or V4TImode with -mavx512f.
> typedef __int128 V __attribute__((vector_size ({16,32,64}));
> will not work, neither typedef int I __attribute__((mode(TI)));
> but mode(V1TI), mode(V2TI) etc. are accepted with a warning when those
> ISAs are enabled.  But they are certainly not fully supported, for some
> optabs maybe, but most of them will not.  And, veclower lowering those ops
> to TImode scalar operations will not work either because TImode isn't
> supported.
>
> So, this patch keeps V1TImode etc. in VALID*_MODE macros so that we can use
> it in certain instructions, but disallows it in
> targetm.vector_mode_supported_p, so that we don't offer those modes to the
> user as supported.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
>
> 2021-11-02  Jakub Jelinek  <jakub@redhat.com>
>
>         PR target/103020
>         * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
>         modes with TImode inner mode if 32-bit.
>
>         * gcc.target/i386/pr103020.c: New test.

OK.

Thanks,
Uros.

>
> --- gcc/config/i386/i386.c.jj   2021-10-28 11:29:01.827722053 +0200
> +++ gcc/config/i386/i386.c      2021-11-01 11:01:44.123587169 +0100
> @@ -21989,6 +21989,10 @@ ix86_libgcc_floating_mode_supported_p (s
>  static bool
>  ix86_vector_mode_supported_p (machine_mode mode)
>  {
> +  /* For ia32, scalar TImode isn't supported and so V1TImode shouldn't be
> +     either.  */
> +  if (!TARGET_64BIT && GET_MODE_INNER (mode) == TImode)
> +    return false;
>    if (TARGET_SSE && VALID_SSE_REG_MODE (mode))
>      return true;
>    if (TARGET_SSE2 && VALID_SSE2_REG_MODE (mode))
> --- gcc/testsuite/gcc.target/i386/pr103020.c.jj 2021-11-01 11:03:34.498017247 +0100
> +++ gcc/testsuite/gcc.target/i386/pr103020.c    2021-11-01 11:11:10.794527161 +0100
> @@ -0,0 +1,11 @@
> +/* PR target/103020 */
> +/* { dg-do compile { target { ! int128 } } } */
> +/* { dg-additional-options "-mavx512f" } */
> +
> +typedef int TI __attribute__((mode (TI)));     /* { dg-error "unable to emulate" } */
> +typedef int V1TI __attribute__((mode (V1TI))); /* { dg-error "unable to emulate" } */
> +typedef int V2TI __attribute__((mode (V2TI))); /* { dg-error "unable to emulate" } */
> +typedef int V4TI __attribute__((mode (V4TI))); /* { dg-error "unable to emulate" } */
> +/* { dg-warning "is deprecated" "V1TI" { target *-*-* } .-3 } */
> +/* { dg-warning "is deprecated" "V2TI" { target *-*-* } .-3 } */
> +/* { dg-warning "is deprecated" "V4TI" { target *-*-* } .-3 } */
>
>         Jakub
>