public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [PATCH] x86: Enable FMA in unsigned SI to SF expanders
@ 2021-09-04 21:55 H.J. Lu
  2021-09-06  5:39 ` Hongtao Liu
  0 siblings, 1 reply; 4+ messages in thread
From: H.J. Lu @ 2021-09-04 21:55 UTC (permalink / raw)
  To: gcc-patches; +Cc: liuhongt, Uros Bizjak

Enable FMA in scalar/vector unsigned SI to SF expanders.

gcc/

	PR target/85819
	* config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
	Enable FMA.
	(ix86_expand_vector_convert_uns_vsivsf): Likewise.

gcc/testsuite/

	PR target/85819
	* gcc.target/i386/pr85819-1.c: New test.
	* gcc.target/i386/pr85819-2a.c: Likewise.
	* gcc.target/i386/pr85819-2b.c: Likewise.
	* gcc.target/i386/pr85819-2c.c: Likewise.
	* gcc.target/i386/pr85819-3.c: Likewise.
---
 gcc/config/i386/i386-expand.c              | 44 ++++++++++++++++------
 gcc/testsuite/gcc.target/i386/pr85819-1.c  | 11 ++++++
 gcc/testsuite/gcc.target/i386/pr85819-2a.c | 17 +++++++++
 gcc/testsuite/gcc.target/i386/pr85819-2b.c |  6 +++
 gcc/testsuite/gcc.target/i386/pr85819-2c.c |  7 ++++
 gcc/testsuite/gcc.target/i386/pr85819-3.c  | 18 +++++++++
 6 files changed, 91 insertions(+), 12 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-1.c
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2a.c
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2b.c
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2c.c
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-3.c

diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c
index 2500dbfa7fb..26263bbe1af 100644
--- a/gcc/config/i386/i386-expand.c
+++ b/gcc/config/i386/i386-expand.c
@@ -1851,12 +1851,21 @@ ix86_expand_convert_uns_sisf_sse (rtx target, rtx input)
   fp_lo = gen_reg_rtx (SFmode);
   emit_insn (gen_floatsisf2 (fp_hi, int_hi));
   emit_insn (gen_floatsisf2 (fp_lo, int_lo));
-  fp_hi = expand_simple_binop (SFmode, MULT, fp_hi, x, fp_hi,
-			       0, OPTAB_DIRECT);
-  fp_hi = expand_simple_binop (SFmode, PLUS, fp_hi, fp_lo, target,
-			       0, OPTAB_DIRECT);
-  if (!rtx_equal_p (target, fp_hi))
-    emit_move_insn (target, fp_hi);
+  if (TARGET_FMA || TARGET_AVX512F)
+    {
+      x = validize_mem (force_const_mem (SFmode, x));
+      fp_hi = gen_rtx_FMA (SFmode, fp_hi, x, fp_lo);
+      emit_move_insn (target, fp_hi);
+    }
+  else
+    {
+      fp_hi = expand_simple_binop (SFmode, MULT, fp_hi, x, fp_hi,
+				   0, OPTAB_DIRECT);
+      fp_hi = expand_simple_binop (SFmode, PLUS, fp_hi, fp_lo, target,
+				   0, OPTAB_DIRECT);
+      if (!rtx_equal_p (target, fp_hi))
+	emit_move_insn (target, fp_hi);
+    }
 }
 
 /* floatunsv{4,8}siv{4,8}sf2 expander.  Expand code to convert
@@ -1888,12 +1897,23 @@ ix86_expand_vector_convert_uns_vsivsf (rtx target, rtx val)
   real_ldexp (&TWO16r, &dconst1, 16);
   tmp[5] = const_double_from_real_value (TWO16r, SFmode);
   tmp[5] = force_reg (fltmode, ix86_build_const_vector (fltmode, 1, tmp[5]));
-  tmp[6] = expand_simple_binop (fltmode, MULT, tmp[4], tmp[5], NULL_RTX, 1,
-				OPTAB_DIRECT);
-  tmp[7] = expand_simple_binop (fltmode, PLUS, tmp[3], tmp[6], target, 1,
-				OPTAB_DIRECT);
-  if (tmp[7] != target)
-    emit_move_insn (target, tmp[7]);
+  unsigned vector_size = GET_MODE_SIZE (fltmode);
+  if (TARGET_FMA
+      || (TARGET_AVX512F && vector_size == 64)
+      || (TARGET_AVX512VL && (vector_size == 32 || vector_size == 16)))
+    {
+      tmp[6] = gen_rtx_FMA (fltmode, tmp[4], tmp[5], tmp[3]);
+      emit_move_insn (target, tmp[6]);
+    }
+  else
+    {
+      tmp[6] = expand_simple_binop (fltmode, MULT, tmp[4], tmp[5],
+				    NULL_RTX, 1, OPTAB_DIRECT);
+      tmp[7] = expand_simple_binop (fltmode, PLUS, tmp[3], tmp[6],
+				    target, 1, OPTAB_DIRECT);
+      if (tmp[7] != target)
+	emit_move_insn (target, tmp[7]);
+    }
 }
 
 /* Adjust a V*SFmode/V*DFmode value VAL so that *sfix_trunc* resp. fix_trunc*
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-1.c b/gcc/testsuite/gcc.target/i386/pr85819-1.c
new file mode 100644
index 00000000000..db02282d100
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-1.c
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mno-avx512f -mfma -mfpmath=sse" } */
+
+float
+foo (unsigned int x)
+{
+  return x;
+}
+
+/* { dg-final { scan-assembler "vfmadd132ss" { target ia32 } } } */
+/* { dg-final { scan-assembler "vcvtsi2ssq" { target { ! ia32 } } } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-2a.c b/gcc/testsuite/gcc.target/i386/pr85819-2a.c
new file mode 100644
index 00000000000..cea599fe416
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-2a.c
@@ -0,0 +1,17 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mno-avx512f -mavx2 -mfma -mfpmath=sse" } */
+
+typedef float To __attribute__ ((__vector_size__ (32)));
+typedef unsigned int From __attribute__ ((__vector_size__ (32)));
+
+#define A2(I) (float)a[I], (float)a[1+I]
+#define A4(I) A2(I), A2(2+I)
+#define A8(I) A4(I), A4(4+I)
+
+To
+f(From a)
+{
+  return __extension__ (To) {A8(0)};
+}
+
+/* { dg-final { scan-assembler "vfmadd132ps" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-2b.c b/gcc/testsuite/gcc.target/i386/pr85819-2b.c
new file mode 100644
index 00000000000..0750e56f29e
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-2b.c
@@ -0,0 +1,6 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx512vl -mfpmath=sse" } */
+
+#include "pr85819-2a.c"
+
+/* { dg-final { scan-assembler "vcvtudq2ps" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-2c.c b/gcc/testsuite/gcc.target/i386/pr85819-2c.c
new file mode 100644
index 00000000000..821166908da
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-2c.c
@@ -0,0 +1,7 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mno-fma -mno-avx2 -mno-avx512vl -mavx512f -mfpmath=sse" } */
+
+#include "pr85819-2a.c"
+
+/* { dg-final { scan-assembler-not "vcvtudq2ps" } } */
+/* { dg-final { scan-assembler-not "vfmadd132ps" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-3.c b/gcc/testsuite/gcc.target/i386/pr85819-3.c
new file mode 100644
index 00000000000..cd3bf9b8d35
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-3.c
@@ -0,0 +1,18 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx512f -mfpmath=sse" } */
+
+typedef float To __attribute__ ((__vector_size__ (64)));
+typedef unsigned int From __attribute__ ((__vector_size__ (64)));
+
+#define A2(I) (float)a[I], (float)a[1+I]
+#define A4(I) A2(I), A2(2+I)
+#define A8(I) A4(I), A4(4+I)
+#define A16(I) A8(I), A8(8+I)
+
+To
+f(From a)
+{
+  return __extension__ (To) {A16(0)};
+}
+
+/* { dg-final { scan-assembler "vcvtudq2ps" } } */
-- 
2.31.1


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] x86: Enable FMA in unsigned SI to SF expanders
  2021-09-04 21:55 [PATCH] x86: Enable FMA in unsigned SI to SF expanders H.J. Lu
@ 2021-09-06  5:39 ` Hongtao Liu
  2021-09-06 12:34   ` [PATCH v2] " H.J. Lu
  0 siblings, 1 reply; 4+ messages in thread
From: Hongtao Liu @ 2021-09-06  5:39 UTC (permalink / raw)
  To: H.J. Lu; +Cc: GCC Patches, liuhongt

On Sun, Sep 5, 2021 at 5:56 AM H.J. Lu via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> Enable FMA in scalar/vector unsigned SI to SF expanders.
>
> gcc/
>
>         PR target/85819
>         * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
>         Enable FMA.
>         (ix86_expand_vector_convert_uns_vsivsf): Likewise.
>
> gcc/testsuite/
>
>         PR target/85819
>         * gcc.target/i386/pr85819-1.c: New test.
>         * gcc.target/i386/pr85819-2a.c: Likewise.
>         * gcc.target/i386/pr85819-2b.c: Likewise.
>         * gcc.target/i386/pr85819-2c.c: Likewise.
>         * gcc.target/i386/pr85819-3.c: Likewise.
> ---
>  gcc/config/i386/i386-expand.c              | 44 ++++++++++++++++------
>  gcc/testsuite/gcc.target/i386/pr85819-1.c  | 11 ++++++
>  gcc/testsuite/gcc.target/i386/pr85819-2a.c | 17 +++++++++
>  gcc/testsuite/gcc.target/i386/pr85819-2b.c |  6 +++
>  gcc/testsuite/gcc.target/i386/pr85819-2c.c |  7 ++++
>  gcc/testsuite/gcc.target/i386/pr85819-3.c  | 18 +++++++++
>  6 files changed, 91 insertions(+), 12 deletions(-)
>  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-1.c
>  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2a.c
>  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2b.c
>  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2c.c
>  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-3.c
>
> diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c
> index 2500dbfa7fb..26263bbe1af 100644
> --- a/gcc/config/i386/i386-expand.c
> +++ b/gcc/config/i386/i386-expand.c
> @@ -1851,12 +1851,21 @@ ix86_expand_convert_uns_sisf_sse (rtx target, rtx input)
>    fp_lo = gen_reg_rtx (SFmode);
>    emit_insn (gen_floatsisf2 (fp_hi, int_hi));
>    emit_insn (gen_floatsisf2 (fp_lo, int_lo));
> -  fp_hi = expand_simple_binop (SFmode, MULT, fp_hi, x, fp_hi,
> -                              0, OPTAB_DIRECT);
> -  fp_hi = expand_simple_binop (SFmode, PLUS, fp_hi, fp_lo, target,
> -                              0, OPTAB_DIRECT);
> -  if (!rtx_equal_p (target, fp_hi))
> -    emit_move_insn (target, fp_hi);
> +  if (TARGET_FMA || TARGET_AVX512F)
Looking at the expander floatunssi<mode>2, the || in the condition
should never be hit since we have direct vcvtsi2s[sd] instruction
under TARGET_AVX512F.
> +    {
> +      x = validize_mem (force_const_mem (SFmode, x));
> +      fp_hi = gen_rtx_FMA (SFmode, fp_hi, x, fp_lo);
> +      emit_move_insn (target, fp_hi);
> +    }
> +  else
> +    {
> +      fp_hi = expand_simple_binop (SFmode, MULT, fp_hi, x, fp_hi,
> +                                  0, OPTAB_DIRECT);
> +      fp_hi = expand_simple_binop (SFmode, PLUS, fp_hi, fp_lo, target,
> +                                  0, OPTAB_DIRECT);
> +      if (!rtx_equal_p (target, fp_hi))
> +       emit_move_insn (target, fp_hi);
> +    }
>  }
>
>  /* floatunsv{4,8}siv{4,8}sf2 expander.  Expand code to convert
> @@ -1888,12 +1897,23 @@ ix86_expand_vector_convert_uns_vsivsf (rtx target, rtx val)
>    real_ldexp (&TWO16r, &dconst1, 16);
>    tmp[5] = const_double_from_real_value (TWO16r, SFmode);
>    tmp[5] = force_reg (fltmode, ix86_build_const_vector (fltmode, 1, tmp[5]));
> -  tmp[6] = expand_simple_binop (fltmode, MULT, tmp[4], tmp[5], NULL_RTX, 1,
> -                               OPTAB_DIRECT);
> -  tmp[7] = expand_simple_binop (fltmode, PLUS, tmp[3], tmp[6], target, 1,
> -                               OPTAB_DIRECT);
> -  if (tmp[7] != target)
> -    emit_move_insn (target, tmp[7]);
> +  unsigned vector_size = GET_MODE_SIZE (fltmode);
> +  if (TARGET_FMA
> +      || (TARGET_AVX512F && vector_size == 64)
> +      || (TARGET_AVX512VL && (vector_size == 32 || vector_size == 16)))
Similar here for the last two || in the condition.
> +    {
> +      tmp[6] = gen_rtx_FMA (fltmode, tmp[4], tmp[5], tmp[3]);
> +      emit_move_insn (target, tmp[6]);
> +    }
> +  else
> +    {
> +      tmp[6] = expand_simple_binop (fltmode, MULT, tmp[4], tmp[5],
> +                                   NULL_RTX, 1, OPTAB_DIRECT);
> +      tmp[7] = expand_simple_binop (fltmode, PLUS, tmp[3], tmp[6],
> +                                   target, 1, OPTAB_DIRECT);
> +      if (tmp[7] != target)
> +       emit_move_insn (target, tmp[7]);
> +    }
>  }
>
>  /* Adjust a V*SFmode/V*DFmode value VAL so that *sfix_trunc* resp. fix_trunc*
> diff --git a/gcc/testsuite/gcc.target/i386/pr85819-1.c b/gcc/testsuite/gcc.target/i386/pr85819-1.c
> new file mode 100644
> index 00000000000..db02282d100
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/pr85819-1.c
> @@ -0,0 +1,11 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -mno-avx512f -mfma -mfpmath=sse" } */
> +
> +float
> +foo (unsigned int x)
> +{
> +  return x;
> +}
> +
> +/* { dg-final { scan-assembler "vfmadd132ss" { target ia32 } } } */
> +/* { dg-final { scan-assembler "vcvtsi2ssq" { target { ! ia32 } } } } */
> diff --git a/gcc/testsuite/gcc.target/i386/pr85819-2a.c b/gcc/testsuite/gcc.target/i386/pr85819-2a.c
> new file mode 100644
> index 00000000000..cea599fe416
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/pr85819-2a.c
> @@ -0,0 +1,17 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -mno-avx512f -mavx2 -mfma -mfpmath=sse" } */
> +
> +typedef float To __attribute__ ((__vector_size__ (32)));
> +typedef unsigned int From __attribute__ ((__vector_size__ (32)));
> +
> +#define A2(I) (float)a[I], (float)a[1+I]
> +#define A4(I) A2(I), A2(2+I)
> +#define A8(I) A4(I), A4(4+I)
> +
> +To
> +f(From a)
> +{
> +  return __extension__ (To) {A8(0)};
> +}
> +
> +/* { dg-final { scan-assembler "vfmadd132ps" } } */
> diff --git a/gcc/testsuite/gcc.target/i386/pr85819-2b.c b/gcc/testsuite/gcc.target/i386/pr85819-2b.c
> new file mode 100644
> index 00000000000..0750e56f29e
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/pr85819-2b.c
> @@ -0,0 +1,6 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -mavx512vl -mfpmath=sse" } */
> +
> +#include "pr85819-2a.c"
> +
> +/* { dg-final { scan-assembler "vcvtudq2ps" } } */
> diff --git a/gcc/testsuite/gcc.target/i386/pr85819-2c.c b/gcc/testsuite/gcc.target/i386/pr85819-2c.c
> new file mode 100644
> index 00000000000..821166908da
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/pr85819-2c.c
> @@ -0,0 +1,7 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -mno-fma -mno-avx2 -mno-avx512vl -mavx512f -mfpmath=sse" } */
> +
> +#include "pr85819-2a.c"
> +
> +/* { dg-final { scan-assembler-not "vcvtudq2ps" } } */
> +/* { dg-final { scan-assembler-not "vfmadd132ps" } } */
> diff --git a/gcc/testsuite/gcc.target/i386/pr85819-3.c b/gcc/testsuite/gcc.target/i386/pr85819-3.c
> new file mode 100644
> index 00000000000..cd3bf9b8d35
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/pr85819-3.c
> @@ -0,0 +1,18 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -mavx512f -mfpmath=sse" } */
> +
> +typedef float To __attribute__ ((__vector_size__ (64)));
> +typedef unsigned int From __attribute__ ((__vector_size__ (64)));
> +
> +#define A2(I) (float)a[I], (float)a[1+I]
> +#define A4(I) A2(I), A2(2+I)
> +#define A8(I) A4(I), A4(4+I)
> +#define A16(I) A8(I), A8(8+I)
> +
> +To
> +f(From a)
> +{
> +  return __extension__ (To) {A16(0)};
> +}
> +
> +/* { dg-final { scan-assembler "vcvtudq2ps" } } */
> --
> 2.31.1
>


-- 
BR,
Hongtao

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [PATCH v2] x86: Enable FMA in unsigned SI to SF expanders
  2021-09-06  5:39 ` Hongtao Liu
@ 2021-09-06 12:34   ` H.J. Lu
  2021-09-07  1:15     ` Hongtao Liu
  0 siblings, 1 reply; 4+ messages in thread
From: H.J. Lu @ 2021-09-06 12:34 UTC (permalink / raw)
  To: Hongtao Liu; +Cc: GCC Patches, liuhongt

[-- Attachment #1: Type: text/plain, Size: 4676 bytes --]

On Sun, Sep 5, 2021 at 10:34 PM Hongtao Liu <crazylht@gmail.com> wrote:
>
> On Sun, Sep 5, 2021 at 5:56 AM H.J. Lu via Gcc-patches
> <gcc-patches@gcc.gnu.org> wrote:
> >
> > Enable FMA in scalar/vector unsigned SI to SF expanders.
> >
> > gcc/
> >
> >         PR target/85819
> >         * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
> >         Enable FMA.
> >         (ix86_expand_vector_convert_uns_vsivsf): Likewise.
> >
> > gcc/testsuite/
> >
> >         PR target/85819
> >         * gcc.target/i386/pr85819-1.c: New test.
> >         * gcc.target/i386/pr85819-2a.c: Likewise.
> >         * gcc.target/i386/pr85819-2b.c: Likewise.
> >         * gcc.target/i386/pr85819-2c.c: Likewise.
> >         * gcc.target/i386/pr85819-3.c: Likewise.
> > ---
> >  gcc/config/i386/i386-expand.c              | 44 ++++++++++++++++------
> >  gcc/testsuite/gcc.target/i386/pr85819-1.c  | 11 ++++++
> >  gcc/testsuite/gcc.target/i386/pr85819-2a.c | 17 +++++++++
> >  gcc/testsuite/gcc.target/i386/pr85819-2b.c |  6 +++
> >  gcc/testsuite/gcc.target/i386/pr85819-2c.c |  7 ++++
> >  gcc/testsuite/gcc.target/i386/pr85819-3.c  | 18 +++++++++
> >  6 files changed, 91 insertions(+), 12 deletions(-)
> >  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-1.c
> >  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2a.c
> >  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2b.c
> >  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2c.c
> >  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-3.c
> >
> > diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c
> > index 2500dbfa7fb..26263bbe1af 100644
> > --- a/gcc/config/i386/i386-expand.c
> > +++ b/gcc/config/i386/i386-expand.c
> > @@ -1851,12 +1851,21 @@ ix86_expand_convert_uns_sisf_sse (rtx target, rtx input)
> >    fp_lo = gen_reg_rtx (SFmode);
> >    emit_insn (gen_floatsisf2 (fp_hi, int_hi));
> >    emit_insn (gen_floatsisf2 (fp_lo, int_lo));
> > -  fp_hi = expand_simple_binop (SFmode, MULT, fp_hi, x, fp_hi,
> > -                              0, OPTAB_DIRECT);
> > -  fp_hi = expand_simple_binop (SFmode, PLUS, fp_hi, fp_lo, target,
> > -                              0, OPTAB_DIRECT);
> > -  if (!rtx_equal_p (target, fp_hi))
> > -    emit_move_insn (target, fp_hi);
> > +  if (TARGET_FMA || TARGET_AVX512F)
> Looking at the expander floatunssi<mode>2, the || in the condition
> should never be hit since we have direct vcvtsi2s[sd] instruction
> under TARGET_AVX512F.

Fixed.

> > +    {
> > +      x = validize_mem (force_const_mem (SFmode, x));
> > +      fp_hi = gen_rtx_FMA (SFmode, fp_hi, x, fp_lo);
> > +      emit_move_insn (target, fp_hi);
> > +    }
> > +  else
> > +    {
> > +      fp_hi = expand_simple_binop (SFmode, MULT, fp_hi, x, fp_hi,
> > +                                  0, OPTAB_DIRECT);
> > +      fp_hi = expand_simple_binop (SFmode, PLUS, fp_hi, fp_lo, target,
> > +                                  0, OPTAB_DIRECT);
> > +      if (!rtx_equal_p (target, fp_hi))
> > +       emit_move_insn (target, fp_hi);
> > +    }
> >  }
> >
> >  /* floatunsv{4,8}siv{4,8}sf2 expander.  Expand code to convert
> > @@ -1888,12 +1897,23 @@ ix86_expand_vector_convert_uns_vsivsf (rtx target, rtx val)
> >    real_ldexp (&TWO16r, &dconst1, 16);
> >    tmp[5] = const_double_from_real_value (TWO16r, SFmode);
> >    tmp[5] = force_reg (fltmode, ix86_build_const_vector (fltmode, 1, tmp[5]));
> > -  tmp[6] = expand_simple_binop (fltmode, MULT, tmp[4], tmp[5], NULL_RTX, 1,
> > -                               OPTAB_DIRECT);
> > -  tmp[7] = expand_simple_binop (fltmode, PLUS, tmp[3], tmp[6], target, 1,
> > -                               OPTAB_DIRECT);
> > -  if (tmp[7] != target)
> > -    emit_move_insn (target, tmp[7]);
> > +  unsigned vector_size = GET_MODE_SIZE (fltmode);
> > +  if (TARGET_FMA
> > +      || (TARGET_AVX512F && vector_size == 64)
> > +      || (TARGET_AVX512VL && (vector_size == 32 || vector_size == 16)))
> Similar here for the last two || in the condition.

Fixed.

> > +    {
> > +      tmp[6] = gen_rtx_FMA (fltmode, tmp[4], tmp[5], tmp[3]);
> > +      emit_move_insn (target, tmp[6]);
> > +    }
> > +  else
> > +    {
> > +      tmp[6] = expand_simple_binop (fltmode, MULT, tmp[4], tmp[5],
> > +                                   NULL_RTX, 1, OPTAB_DIRECT);
> > +      tmp[7] = expand_simple_binop (fltmode, PLUS, tmp[3], tmp[6],
> > +                                   target, 1, OPTAB_DIRECT);
> > +      if (tmp[7] != target)
> > +       emit_move_insn (target, tmp[7]);
> > +    }
> >  }
> >
> >  /* Adjust a V*SFmode/V*DFmode value VAL so that *sfix_trunc* resp. fix_trunc*

Here is the v2 patch.

-- 
H.J.

[-- Attachment #2: v2-0001-x86-Enable-FMA-in-unsigned-SI-to-SF-expanders.patch --]
[-- Type: text/x-patch, Size: 7004 bytes --]

From 35dac257d711aa9b37604d30c9d41abaf94b8650 Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Sat, 4 Sep 2021 07:48:43 -0700
Subject: [PATCH v2] x86: Enable FMA in unsigned SI to SF expanders

Enable FMA in scalar/vector unsigned SI to SF expanders.  Don't check
TARGET_AVX512F which has vcvtusi2ss and vcvtudq2ps instructions.

gcc/

	PR target/85819
	* config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
	Enable FMA.
	(ix86_expand_vector_convert_uns_vsivsf): Likewise.

gcc/testsuite/

	PR target/85819
	* gcc.target/i386/pr85819-1a.c: New test.
	* gcc.target/i386/pr85819-1b.c: Likewise.
	* gcc.target/i386/pr85819-2a.c: Likewise.
	* gcc.target/i386/pr85819-2b.c: Likewise.
	* gcc.target/i386/pr85819-2c.c: Likewise.
	* gcc.target/i386/pr85819-3.c: Likewise.
---
 gcc/config/i386/i386-expand.c              | 41 +++++++++++++++-------
 gcc/testsuite/gcc.target/i386/pr85819-1a.c | 11 ++++++
 gcc/testsuite/gcc.target/i386/pr85819-1b.c |  6 ++++
 gcc/testsuite/gcc.target/i386/pr85819-2a.c | 17 +++++++++
 gcc/testsuite/gcc.target/i386/pr85819-2b.c |  6 ++++
 gcc/testsuite/gcc.target/i386/pr85819-2c.c |  7 ++++
 gcc/testsuite/gcc.target/i386/pr85819-3.c  | 18 ++++++++++
 7 files changed, 94 insertions(+), 12 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-1a.c
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-1b.c
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2a.c
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2b.c
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2c.c
 create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-3.c

diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c
index 273a0ba8e3d..3f90f67a994 100644
--- a/gcc/config/i386/i386-expand.c
+++ b/gcc/config/i386/i386-expand.c
@@ -1851,12 +1851,21 @@ ix86_expand_convert_uns_sisf_sse (rtx target, rtx input)
   fp_lo = gen_reg_rtx (SFmode);
   emit_insn (gen_floatsisf2 (fp_hi, int_hi));
   emit_insn (gen_floatsisf2 (fp_lo, int_lo));
-  fp_hi = expand_simple_binop (SFmode, MULT, fp_hi, x, fp_hi,
-			       0, OPTAB_DIRECT);
-  fp_hi = expand_simple_binop (SFmode, PLUS, fp_hi, fp_lo, target,
-			       0, OPTAB_DIRECT);
-  if (!rtx_equal_p (target, fp_hi))
-    emit_move_insn (target, fp_hi);
+  if (TARGET_FMA)
+    {
+      x = validize_mem (force_const_mem (SFmode, x));
+      fp_hi = gen_rtx_FMA (SFmode, fp_hi, x, fp_lo);
+      emit_move_insn (target, fp_hi);
+    }
+  else
+    {
+      fp_hi = expand_simple_binop (SFmode, MULT, fp_hi, x, fp_hi,
+				   0, OPTAB_DIRECT);
+      fp_hi = expand_simple_binop (SFmode, PLUS, fp_hi, fp_lo, target,
+				   0, OPTAB_DIRECT);
+      if (!rtx_equal_p (target, fp_hi))
+	emit_move_insn (target, fp_hi);
+    }
 }
 
 /* floatunsv{4,8}siv{4,8}sf2 expander.  Expand code to convert
@@ -1888,12 +1897,20 @@ ix86_expand_vector_convert_uns_vsivsf (rtx target, rtx val)
   real_ldexp (&TWO16r, &dconst1, 16);
   tmp[5] = const_double_from_real_value (TWO16r, SFmode);
   tmp[5] = force_reg (fltmode, ix86_build_const_vector (fltmode, 1, tmp[5]));
-  tmp[6] = expand_simple_binop (fltmode, MULT, tmp[4], tmp[5], NULL_RTX, 1,
-				OPTAB_DIRECT);
-  tmp[7] = expand_simple_binop (fltmode, PLUS, tmp[3], tmp[6], target, 1,
-				OPTAB_DIRECT);
-  if (tmp[7] != target)
-    emit_move_insn (target, tmp[7]);
+  if (TARGET_FMA)
+    {
+      tmp[6] = gen_rtx_FMA (fltmode, tmp[4], tmp[5], tmp[3]);
+      emit_move_insn (target, tmp[6]);
+    }
+  else
+    {
+      tmp[6] = expand_simple_binop (fltmode, MULT, tmp[4], tmp[5],
+				    NULL_RTX, 1, OPTAB_DIRECT);
+      tmp[7] = expand_simple_binop (fltmode, PLUS, tmp[3], tmp[6],
+				    target, 1, OPTAB_DIRECT);
+      if (tmp[7] != target)
+	emit_move_insn (target, tmp[7]);
+    }
 }
 
 /* Adjust a V*SFmode/V*DFmode value VAL so that *sfix_trunc* resp. fix_trunc*
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-1a.c b/gcc/testsuite/gcc.target/i386/pr85819-1a.c
new file mode 100644
index 00000000000..db02282d100
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-1a.c
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mno-avx512f -mfma -mfpmath=sse" } */
+
+float
+foo (unsigned int x)
+{
+  return x;
+}
+
+/* { dg-final { scan-assembler "vfmadd132ss" { target ia32 } } } */
+/* { dg-final { scan-assembler "vcvtsi2ssq" { target { ! ia32 } } } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-1b.c b/gcc/testsuite/gcc.target/i386/pr85819-1b.c
new file mode 100644
index 00000000000..f1408c4f463
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-1b.c
@@ -0,0 +1,6 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx512f -mfpmath=sse" } */
+
+#include "pr85819-1a.c"
+
+/* { dg-final { scan-assembler "vcvtusi2ss" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-2a.c b/gcc/testsuite/gcc.target/i386/pr85819-2a.c
new file mode 100644
index 00000000000..cea599fe416
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-2a.c
@@ -0,0 +1,17 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mno-avx512f -mavx2 -mfma -mfpmath=sse" } */
+
+typedef float To __attribute__ ((__vector_size__ (32)));
+typedef unsigned int From __attribute__ ((__vector_size__ (32)));
+
+#define A2(I) (float)a[I], (float)a[1+I]
+#define A4(I) A2(I), A2(2+I)
+#define A8(I) A4(I), A4(4+I)
+
+To
+f(From a)
+{
+  return __extension__ (To) {A8(0)};
+}
+
+/* { dg-final { scan-assembler "vfmadd132ps" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-2b.c b/gcc/testsuite/gcc.target/i386/pr85819-2b.c
new file mode 100644
index 00000000000..0750e56f29e
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-2b.c
@@ -0,0 +1,6 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx512vl -mfpmath=sse" } */
+
+#include "pr85819-2a.c"
+
+/* { dg-final { scan-assembler "vcvtudq2ps" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-2c.c b/gcc/testsuite/gcc.target/i386/pr85819-2c.c
new file mode 100644
index 00000000000..821166908da
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-2c.c
@@ -0,0 +1,7 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mno-fma -mno-avx2 -mno-avx512vl -mavx512f -mfpmath=sse" } */
+
+#include "pr85819-2a.c"
+
+/* { dg-final { scan-assembler-not "vcvtudq2ps" } } */
+/* { dg-final { scan-assembler-not "vfmadd132ps" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr85819-3.c b/gcc/testsuite/gcc.target/i386/pr85819-3.c
new file mode 100644
index 00000000000..cd3bf9b8d35
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr85819-3.c
@@ -0,0 +1,18 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx512f -mfpmath=sse" } */
+
+typedef float To __attribute__ ((__vector_size__ (64)));
+typedef unsigned int From __attribute__ ((__vector_size__ (64)));
+
+#define A2(I) (float)a[I], (float)a[1+I]
+#define A4(I) A2(I), A2(2+I)
+#define A8(I) A4(I), A4(4+I)
+#define A16(I) A8(I), A8(8+I)
+
+To
+f(From a)
+{
+  return __extension__ (To) {A16(0)};
+}
+
+/* { dg-final { scan-assembler "vcvtudq2ps" } } */
-- 
2.31.1


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH v2] x86: Enable FMA in unsigned SI to SF expanders
  2021-09-06 12:34   ` [PATCH v2] " H.J. Lu
@ 2021-09-07  1:15     ` Hongtao Liu
  0 siblings, 0 replies; 4+ messages in thread
From: Hongtao Liu @ 2021-09-07  1:15 UTC (permalink / raw)
  To: H.J. Lu; +Cc: GCC Patches, liuhongt

On Mon, Sep 6, 2021 at 8:35 PM H.J. Lu <hjl.tools@gmail.com> wrote:
>
> On Sun, Sep 5, 2021 at 10:34 PM Hongtao Liu <crazylht@gmail.com> wrote:
> >
> > On Sun, Sep 5, 2021 at 5:56 AM H.J. Lu via Gcc-patches
> > <gcc-patches@gcc.gnu.org> wrote:
> > >
> > > Enable FMA in scalar/vector unsigned SI to SF expanders.
> > >
> > > gcc/
> > >
> > >         PR target/85819
> > >         * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
> > >         Enable FMA.
> > >         (ix86_expand_vector_convert_uns_vsivsf): Likewise.
> > >
> > > gcc/testsuite/
> > >
> > >         PR target/85819
> > >         * gcc.target/i386/pr85819-1.c: New test.
> > >         * gcc.target/i386/pr85819-2a.c: Likewise.
> > >         * gcc.target/i386/pr85819-2b.c: Likewise.
> > >         * gcc.target/i386/pr85819-2c.c: Likewise.
> > >         * gcc.target/i386/pr85819-3.c: Likewise.
> > > ---
> > >  gcc/config/i386/i386-expand.c              | 44 ++++++++++++++++------
> > >  gcc/testsuite/gcc.target/i386/pr85819-1.c  | 11 ++++++
> > >  gcc/testsuite/gcc.target/i386/pr85819-2a.c | 17 +++++++++
> > >  gcc/testsuite/gcc.target/i386/pr85819-2b.c |  6 +++
> > >  gcc/testsuite/gcc.target/i386/pr85819-2c.c |  7 ++++
> > >  gcc/testsuite/gcc.target/i386/pr85819-3.c  | 18 +++++++++
> > >  6 files changed, 91 insertions(+), 12 deletions(-)
> > >  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-1.c
> > >  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2a.c
> > >  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2b.c
> > >  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-2c.c
> > >  create mode 100644 gcc/testsuite/gcc.target/i386/pr85819-3.c
> > >
> > > diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c
> > > index 2500dbfa7fb..26263bbe1af 100644
> > > --- a/gcc/config/i386/i386-expand.c
> > > +++ b/gcc/config/i386/i386-expand.c
> > > @@ -1851,12 +1851,21 @@ ix86_expand_convert_uns_sisf_sse (rtx target, rtx input)
> > >    fp_lo = gen_reg_rtx (SFmode);
> > >    emit_insn (gen_floatsisf2 (fp_hi, int_hi));
> > >    emit_insn (gen_floatsisf2 (fp_lo, int_lo));
> > > -  fp_hi = expand_simple_binop (SFmode, MULT, fp_hi, x, fp_hi,
> > > -                              0, OPTAB_DIRECT);
> > > -  fp_hi = expand_simple_binop (SFmode, PLUS, fp_hi, fp_lo, target,
> > > -                              0, OPTAB_DIRECT);
> > > -  if (!rtx_equal_p (target, fp_hi))
> > > -    emit_move_insn (target, fp_hi);
> > > +  if (TARGET_FMA || TARGET_AVX512F)
> > Looking at the expander floatunssi<mode>2, the || in the condition
> > should never be hit since we have direct vcvtsi2s[sd] instruction
> > under TARGET_AVX512F.
>
> Fixed.
>
> > > +    {
> > > +      x = validize_mem (force_const_mem (SFmode, x));
> > > +      fp_hi = gen_rtx_FMA (SFmode, fp_hi, x, fp_lo);
> > > +      emit_move_insn (target, fp_hi);
> > > +    }
> > > +  else
> > > +    {
> > > +      fp_hi = expand_simple_binop (SFmode, MULT, fp_hi, x, fp_hi,
> > > +                                  0, OPTAB_DIRECT);
> > > +      fp_hi = expand_simple_binop (SFmode, PLUS, fp_hi, fp_lo, target,
> > > +                                  0, OPTAB_DIRECT);
> > > +      if (!rtx_equal_p (target, fp_hi))
> > > +       emit_move_insn (target, fp_hi);
> > > +    }
> > >  }
> > >
> > >  /* floatunsv{4,8}siv{4,8}sf2 expander.  Expand code to convert
> > > @@ -1888,12 +1897,23 @@ ix86_expand_vector_convert_uns_vsivsf (rtx target, rtx val)
> > >    real_ldexp (&TWO16r, &dconst1, 16);
> > >    tmp[5] = const_double_from_real_value (TWO16r, SFmode);
> > >    tmp[5] = force_reg (fltmode, ix86_build_const_vector (fltmode, 1, tmp[5]));
> > > -  tmp[6] = expand_simple_binop (fltmode, MULT, tmp[4], tmp[5], NULL_RTX, 1,
> > > -                               OPTAB_DIRECT);
> > > -  tmp[7] = expand_simple_binop (fltmode, PLUS, tmp[3], tmp[6], target, 1,
> > > -                               OPTAB_DIRECT);
> > > -  if (tmp[7] != target)
> > > -    emit_move_insn (target, tmp[7]);
> > > +  unsigned vector_size = GET_MODE_SIZE (fltmode);
> > > +  if (TARGET_FMA
> > > +      || (TARGET_AVX512F && vector_size == 64)
> > > +      || (TARGET_AVX512VL && (vector_size == 32 || vector_size == 16)))
> > Similar here for the last two || in the condition.
>
> Fixed.
>
> > > +    {
> > > +      tmp[6] = gen_rtx_FMA (fltmode, tmp[4], tmp[5], tmp[3]);
> > > +      emit_move_insn (target, tmp[6]);
> > > +    }
> > > +  else
> > > +    {
> > > +      tmp[6] = expand_simple_binop (fltmode, MULT, tmp[4], tmp[5],
> > > +                                   NULL_RTX, 1, OPTAB_DIRECT);
> > > +      tmp[7] = expand_simple_binop (fltmode, PLUS, tmp[3], tmp[6],
> > > +                                   target, 1, OPTAB_DIRECT);
> > > +      if (tmp[7] != target)
> > > +       emit_move_insn (target, tmp[7]);
> > > +    }
> > >  }
> > >
> > >  /* Adjust a V*SFmode/V*DFmode value VAL so that *sfix_trunc* resp. fix_trunc*
>
> Here is the v2 patch.
LGTM.
>
> --
> H.J.



-- 
BR,
Hongtao

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2021-09-07  1:10 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-09-04 21:55 [PATCH] x86: Enable FMA in unsigned SI to SF expanders H.J. Lu
2021-09-06  5:39 ` Hongtao Liu
2021-09-06 12:34   ` [PATCH v2] " H.J. Lu
2021-09-07  1:15     ` Hongtao Liu

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).