This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
RE: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- From: Matthew Fortune <Matthew dot Fortune at imgtec dot com>
- To: Steve Ellcey <Steve dot Ellcey at imgtec dot com>, Richard Sandiford <rdsandiford at googlemail dot com>
- Cc: "Maciej W. Rozycki" <macro at linux-mips dot org>, Richard Sandiford <richard dot sandiford at arm dot com>, Joseph Myers <joseph at codesourcery dot com>, "gcc-patches at gcc dot gnu dot org" <gcc-patches at gcc dot gnu dot org>, Catherine Moore <clm at codesourcery dot com>
- Date: Wed, 17 Jun 2015 21:09:07 +0000
- Subject: RE: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Authentication-results: sourceware.org; auth=none
- References: <4c25620c-546c-40ae-b330-3652fe25f791 at BAMAIL02 dot ba dot imgtec dot org> <alpine dot DEB dot 2 dot 10 dot 1506112002380 dot 15628 at digraph dot polyomino dot org dot uk> <alpine dot LFD dot 2 dot 11 dot 1506152042580 dot 5418 at eddie dot linux-mips dot org> <alpine dot DEB dot 2 dot 10 dot 1506152047390 dot 9772 at digraph dot polyomino dot org dot uk> <alpine dot LFD dot 2 dot 11 dot 1506152215570 dot 5418 at eddie dot linux-mips dot org> <alpine dot DEB dot 2 dot 10 dot 1506152157470 dot 9772 at digraph dot polyomino dot org dot uk> <alpine dot LFD dot 2 dot 11 dot 1506161240330 dot 5418 at eddie dot linux-mips dot org> <alpine dot DEB dot 2 dot 10 dot 1506161215400 dot 4909 at digraph dot polyomino dot org dot uk> <alpine dot LFD dot 2 dot 11 dot 1506161421560 dot 5418 at eddie dot linux-mips dot org> <87mvzy4nwu dot fsf at e105548-lin dot cambridge dot arm dot com> <1434562571 dot 18552 dot 6 dot camel at ubuntu-sellcey> <alpine dot LFD dot 2 dot 11 dot 1506171910160 dot 16441 at eddie dot linux-mips dot org> <87381qw4wz dot fsf at googlemail dot com> <1434574154 dot 18552 dot 15 dot camel at ubuntu-sellcey>
Steve Ellcey <Steve.Ellcey@imgtec.com> writes:
> On Wed, 2015-06-17 at 19:44 +0100, Richard Sandiford wrote:
>
> >
> > FWIW, to be specific, I think we're talking about every check except
> > the last two in mips.md:
> >
> > and the one mips-ps-3d.md:
> >
> > In particular, the two checks in mips.c should go.
>
> OK, here is a patch that does that.
>
> > But like I say, please do add a comment above the unfused patterns to say
> > why we can match NEG here without the HONOR_NANS test. It's going to be
> > far from obvious to anyone who hasn't read this thread.
>
> I put these comments in mips.md as part of the patch:
>
> ;; The various multiply accumulate instructions can be used even when
> ;; HONOR_NANS is true because while IEEE 754-2008 requires the negate
> ;; operation to negate the sign of a NAN and the MIPS neg instruction does
> ;; not do this, the multiply and add (or minus) parts of these instructions
> ;; have no requirement on how the sign of a NAN is handled and so the final
> ;; sign bit of the entire operation is undefined.
>
> > It's also worth pointing out (although it's probably obvious) that this is
> > effectively going to be a no-op change. We'd never have a NEG to combine
> > if we were honouring NANs and using the legacy pre-2008 mode. We should
> > still be accurate though. Especially when it's also less code :-)
>
> Actually, this wasn't obvious to me. I thought this must have some
> affect but I ran through the various options and architectures and, sure
> enough, I found no differences in the generated code.
>
> Here is my "prequel" patch. How does this look?
>
> Steve Ellcey
> sellcey@imgtec.com
>
> 2015-06-17 Steve Ellcey <sellcey@imgtec.com>
>
> * config/mips/mips.c (mips_rtx_costs): Remove HONOR_NAN check.
> * config/mips/mips.md (*madd4<mode>): Ditto.
> (*nmadd3<mode>) Ditto.
> (*nmadd4<mode>_fastmath): Ditto.
> (*nmadd3<mode>_fastmath): Ditto.
> (*nmsub4<mode>): Ditto.
> (*nmsub3<mode>): Ditto.
> (*nmsub4<mode>_fastmath): Ditto.
> (*nmsub3<mode>_fastmath): Ditto.
I'd like to delegate to Maciej to approve this and the actual
fp-contract patch. I have nothing of value to add to this
discussion but to say it is clearly comprehensive.
Thanks,
Matthew
> diff --git a/gcc/config/mips/mips.c b/gcc/config/mips/mips.c
> index d427c0c..1c837cf 100644
> --- a/gcc/config/mips/mips.c
> +++ b/gcc/config/mips/mips.c
> @@ -4069,7 +4069,6 @@ mips_rtx_costs (rtx x, int code, int outer_code, int opno
> ATTRIBUTE_UNUSED,
> if (float_mode_p
> && (ISA_HAS_NMADD4_NMSUB4 || ISA_HAS_NMADD3_NMSUB3)
> && TARGET_FUSED_MADD
> - && !HONOR_NANS (mode)
> && !HONOR_SIGNED_ZEROS (mode))
> {
> /* See if we can use NMADD or NMSUB. See mips.md for the
> @@ -4137,7 +4136,6 @@ mips_rtx_costs (rtx x, int code, int outer_code, int opno
> ATTRIBUTE_UNUSED,
> if (float_mode_p
> && (ISA_HAS_NMADD4_NMSUB4 || ISA_HAS_NMADD3_NMSUB3)
> && TARGET_FUSED_MADD
> - && !HONOR_NANS (mode)
> && HONOR_SIGNED_ZEROS (mode))
> {
> /* See if we can use NMADD or NMSUB. See mips.md for the
> diff --git a/gcc/config/mips/mips.md b/gcc/config/mips/mips.md
> index 0a23fa2..f6912e1 100644
> --- a/gcc/config/mips/mips.md
> +++ b/gcc/config/mips/mips.md
> @@ -2475,6 +2475,13 @@
>
> ;; Floating point multiply accumulate instructions.
>
> +;; The various multiply accumulate instructions can be used even when
> +;; HONOR_NANS is true because while IEEE 754-2008 requires the negate
> +;; operation to negate the sign of a NAN and the MIPS neg instruction does
> +;; not do this, the multiply and add (or minus) parts of these instructions
> +;; have no requirement on how the sign of a NAN is handled and so the final
> +;; sign bit of the entire operation is undefined.
> +
> (define_insn "*madd4<mode>"
> [(set (match_operand:ANYF 0 "register_operand" "=f")
> (plus:ANYF (mult:ANYF (match_operand:ANYF 1 "register_operand" "f")
> @@ -2533,8 +2540,7 @@
> (match_operand:ANYF 3 "register_operand" "f"))))]
> "ISA_HAS_NMADD4_NMSUB4
> && TARGET_FUSED_MADD
> - && HONOR_SIGNED_ZEROS (<MODE>mode)
> - && !HONOR_NANS (<MODE>mode)"
> + && HONOR_SIGNED_ZEROS (<MODE>mode)"
> "nmadd.<fmt>\t%0,%3,%1,%2"
> [(set_attr "type" "fmadd")
> (set_attr "mode" "<UNITMODE>")])
> @@ -2547,8 +2553,7 @@
> (match_operand:ANYF 3 "register_operand" "0"))))]
> "ISA_HAS_NMADD3_NMSUB3
> && TARGET_FUSED_MADD
> - && HONOR_SIGNED_ZEROS (<MODE>mode)
> - && !HONOR_NANS (<MODE>mode)"
> + && HONOR_SIGNED_ZEROS (<MODE>mode)"
> "nmadd.<fmt>\t%0,%1,%2"
> [(set_attr "type" "fmadd")
> (set_attr "mode" "<UNITMODE>")])
> @@ -2561,8 +2566,7 @@
> (match_operand:ANYF 3 "register_operand" "f")))]
> "ISA_HAS_NMADD4_NMSUB4
> && TARGET_FUSED_MADD
> - && !HONOR_SIGNED_ZEROS (<MODE>mode)
> - && !HONOR_NANS (<MODE>mode)"
> + && !HONOR_SIGNED_ZEROS (<MODE>mode)"
> "nmadd.<fmt>\t%0,%3,%1,%2"
> [(set_attr "type" "fmadd")
> (set_attr "mode" "<UNITMODE>")])
> @@ -2575,8 +2579,7 @@
> (match_operand:ANYF 3 "register_operand" "0")))]
> "ISA_HAS_NMADD3_NMSUB3
> && TARGET_FUSED_MADD
> - && !HONOR_SIGNED_ZEROS (<MODE>mode)
> - && !HONOR_NANS (<MODE>mode)"
> + && !HONOR_SIGNED_ZEROS (<MODE>mode)"
> "nmadd.<fmt>\t%0,%1,%2"
> [(set_attr "type" "fmadd")
> (set_attr "mode" "<UNITMODE>")])
> @@ -2589,8 +2592,7 @@
> (match_operand:ANYF 1 "register_operand" "f"))))]
> "ISA_HAS_NMADD4_NMSUB4
> && TARGET_FUSED_MADD
> - && HONOR_SIGNED_ZEROS (<MODE>mode)
> - && !HONOR_NANS (<MODE>mode)"
> + && HONOR_SIGNED_ZEROS (<MODE>mode)"
> "nmsub.<fmt>\t%0,%1,%2,%3"
> [(set_attr "type" "fmadd")
> (set_attr "mode" "<UNITMODE>")])
> @@ -2603,8 +2605,7 @@
> (match_operand:ANYF 1 "register_operand" "0"))))]
> "ISA_HAS_NMADD3_NMSUB3
> && TARGET_FUSED_MADD
> - && HONOR_SIGNED_ZEROS (<MODE>mode)
> - && !HONOR_NANS (<MODE>mode)"
> + && HONOR_SIGNED_ZEROS (<MODE>mode)"
> "nmsub.<fmt>\t%0,%1,%2"
> [(set_attr "type" "fmadd")
> (set_attr "mode" "<UNITMODE>")])
> @@ -2617,8 +2618,7 @@
> (match_operand:ANYF 3 "register_operand" "f"))))]
> "ISA_HAS_NMADD4_NMSUB4
> && TARGET_FUSED_MADD
> - && !HONOR_SIGNED_ZEROS (<MODE>mode)
> - && !HONOR_NANS (<MODE>mode)"
> + && !HONOR_SIGNED_ZEROS (<MODE>mode)"
> "nmsub.<fmt>\t%0,%1,%2,%3"
> [(set_attr "type" "fmadd")
> (set_attr "mode" "<UNITMODE>")])
> @@ -2631,8 +2631,7 @@
> (match_operand:ANYF 3 "register_operand" "0"))))]
> "ISA_HAS_NMADD3_NMSUB3
> && TARGET_FUSED_MADD
> - && !HONOR_SIGNED_ZEROS (<MODE>mode)
> - && !HONOR_NANS (<MODE>mode)"
> + && !HONOR_SIGNED_ZEROS (<MODE>mode)"
> "nmsub.<fmt>\t%0,%1,%2"
> [(set_attr "type" "fmadd")
> (set_attr "mode" "<UNITMODE>")])
>
>
- References:
- [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd
- Re: [Patch, MIPS] Enable fp-contract on MIPS and update -mfused-madd