diff mbox series

i386: Support complex fma/conj_fma for _Float16.

Message ID DM4PR11MB5487B8EACAC36CFA1AB3460BEC8E9@DM4PR11MB5487.namprd11.prod.outlook.com
State New
Headers
Series i386: Support complex fma/conj_fma for _Float16. |

Commit Message

Li, Pan2 via Gcc-patches Nov. 5, 2021, 7:09 a.m. UTC 
  Hi,

This patch is to support cmla_optab, cmul_optab, cmla_conj_optab, cmul_conj_optab for vector _Float16.
Ok for master?

gcc/ChangeLog:

	* config/i386/sse.md (cmul<conj_op><mode>3): add new define_expand.
	(cmla<conj_op><mode>4): Likewise

gcc/testsuite/ChangeLog:

	* gcc.target/i386/avx512fp16-vector-complex-float.c: New test.
---
 gcc/config/i386/sse.md                        | 23 +++++++++++
 .../i386/avx512fp16-vector-complex-float.c    | 40 +++++++++++++++++++
 2 files changed, 63 insertions(+)
 create mode 100644 gcc/testsuite/gcc.target/i386/avx512fp16-vector-complex-float.c

--
2.18.1

Comments

Hongtao Liu Nov. 8, 2021, 1:47 a.m. UTC | #1 
On Fri, Nov 5, 2021 at 3:09 PM Kong, Lingling via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> Hi,
>
> This patch is to support cmla_optab, cmul_optab, cmla_conj_optab, cmul_conj_optab for vector _Float16.
> Ok for master?
LGTM.
> gcc/ChangeLog:
>
>         * config/i386/sse.md (cmul<conj_op><mode>3): add new define_expand.
>         (cmla<conj_op><mode>4): Likewise
>
> gcc/testsuite/ChangeLog:
>
>         * gcc.target/i386/avx512fp16-vector-complex-float.c: New test.
> ---
>  gcc/config/i386/sse.md                        | 23 +++++++++++
>  .../i386/avx512fp16-vector-complex-float.c    | 40 +++++++++++++++++++
>  2 files changed, 63 insertions(+)
>  create mode 100644 gcc/testsuite/gcc.target/i386/avx512fp16-vector-complex-float.c
>
> diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md index 0a7f5b178f9..8d3fef0a31a 100644
> --- a/gcc/config/i386/sse.md
> +++ b/gcc/config/i386/sse.md
> @@ -5922,6 +5922,12 @@
>          (UNSPEC_COMPLEX_FMUL "fmulc")
>          (UNSPEC_COMPLEX_FCMUL "fcmulc")])
>
> +(define_int_attr conj_op
> +       [(UNSPEC_COMPLEX_FMA "")
> +        (UNSPEC_COMPLEX_FCMA "_conj")
> +        (UNSPEC_COMPLEX_FMUL "")
> +        (UNSPEC_COMPLEX_FCMUL "_conj")])
> +
>  (define_mode_attr complexmove
>    [(V32HF "avx512f_loadv16sf")
>     (V16HF "avx512vl_loadv8sf")
> @@ -6003,6 +6009,15 @@
>    DONE;
>  })
>
> +(define_expand "cmla<conj_op><mode>4"
> +  [(set (match_operand:VF_AVX512FP16VL 0 "register_operand")
> +       (unspec:VF_AVX512FP16VL
> +           [(match_operand:VF_AVX512FP16VL 1 "vector_operand")
> +            (match_operand:VF_AVX512FP16VL 2 "vector_operand")
> +            (match_operand:VF_AVX512FP16VL 3 "vector_operand")]
> +            UNSPEC_COMPLEX_F_C_MA))]
> +  "TARGET_AVX512FP16")
> +
>  (define_insn "fma_<complexopname>_<mode><sdc_maskz_name><round_name>"
>    [(set (match_operand:VF_AVX512FP16VL 0 "register_operand" "=&v")
>         (unspec:VF_AVX512FP16VL
> @@ -6084,6 +6099,14 @@
>    [(set_attr "type" "ssemuladd")
>     (set_attr "mode" "<MODE>")])
>
> +(define_expand "cmul<conj_op><mode>3"
> +  [(set (match_operand:VF_AVX512FP16VL 0 "register_operand")
> +       (unspec:VF_AVX512FP16VL
> +         [(match_operand:VF_AVX512FP16VL 1 "vector_operand")
> +          (match_operand:VF_AVX512FP16VL 2 "vector_operand")]
> +          UNSPEC_COMPLEX_F_C_MUL))]
> +  "TARGET_AVX512FP16")
> +
>  (define_insn "<avx512>_<complexopname>_<mode><maskc_name><round_name>"
>    [(set (match_operand:VF_AVX512FP16VL 0 "register_operand" "=&v")
>           (unspec:VF_AVX512FP16VL
> diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-vector-complex-float.c b/gcc/testsuite/gcc.target/i386/avx512fp16-vector-complex-float.c
> new file mode 100644
> index 00000000000..bcb957f0de0
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/avx512fp16-vector-complex-float.c
> @@ -0,0 +1,40 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -mavx512fp16 -mavx512vl" } */
> +/* { dg-final { scan-assembler-times "vfmaddcph\[ \\t\]" 1 } } */
> +/* { dg-final { scan-assembler-not "vfmadd\[123]*ph\[ \\t\]"} } */
> +/* { dg-final { scan-assembler-not "vfmadd\[123]*sh\[ \\t\]"} } */
> +/* { dg-final { scan-assembler-times "vfcmaddcph\[ \\t\]" 1 } } */
> +/* { dg-final { scan-assembler-times "vfmulcph\[ \\t\]" 1 } } */
> +/* { dg-final { scan-assembler-times "vfcmulcph\[ \\t\]" 1 } } */
> +
> +#include<complex.h>
> +#define TYPE _Float16
> +#define N 16
> +
> +void fma0 (_Complex TYPE *a, _Complex TYPE *b,
> +           _Complex TYPE *c)
> +{
> +  for (int i = 0; i < N; i++)
> +    c[i] += a[i] * b[i];
> +}
> +
> +void fmaconj (_Complex TYPE a[restrict N], _Complex TYPE b[restrict N],
> +             _Complex TYPE c[restrict N])
> +{
> +  for (int i = 0; i < N; i++)
> +    c[i] += a[i] * ~b[i];
> +}
> +
> +void fmul (_Complex TYPE a[restrict N], _Complex TYPE b[restrict N],
> +          _Complex TYPE c[restrict N])
> +{
> +  for (int i = 0; i < N; i++)
> +    c[i] = a[i] * b[i];
> +}
> +
> +void fmulconj (_Complex TYPE a[restrict N], _Complex TYPE b[restrict N],
> +              _Complex TYPE c[restrict N])
> +{
> +  for (int i = 0; i < N; i++)
> +    c[i] = a[i] * ~b[i];
> +}
> --
> 2.18.1
>
diff mbox series

Patch

diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md index 0a7f5b178f9..8d3fef0a31a 100644
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -5922,6 +5922,12 @@ 
 	 (UNSPEC_COMPLEX_FMUL "fmulc")
 	 (UNSPEC_COMPLEX_FCMUL "fcmulc")])
 
+(define_int_attr conj_op
+	[(UNSPEC_COMPLEX_FMA "")
+	 (UNSPEC_COMPLEX_FCMA "_conj")
+	 (UNSPEC_COMPLEX_FMUL "")
+	 (UNSPEC_COMPLEX_FCMUL "_conj")])
+
 (define_mode_attr complexmove
   [(V32HF "avx512f_loadv16sf")
    (V16HF "avx512vl_loadv8sf")
@@ -6003,6 +6009,15 @@ 
   DONE;
 })
 
+(define_expand "cmla<conj_op><mode>4"
+  [(set (match_operand:VF_AVX512FP16VL 0 "register_operand")
+	(unspec:VF_AVX512FP16VL
+	    [(match_operand:VF_AVX512FP16VL 1 "vector_operand")
+	     (match_operand:VF_AVX512FP16VL 2 "vector_operand")
+	     (match_operand:VF_AVX512FP16VL 3 "vector_operand")]
+	     UNSPEC_COMPLEX_F_C_MA))]
+  "TARGET_AVX512FP16")
+
 (define_insn "fma_<complexopname>_<mode><sdc_maskz_name><round_name>"
   [(set (match_operand:VF_AVX512FP16VL 0 "register_operand" "=&v")
 	(unspec:VF_AVX512FP16VL
@@ -6084,6 +6099,14 @@ 
   [(set_attr "type" "ssemuladd")
    (set_attr "mode" "<MODE>")])
 
+(define_expand "cmul<conj_op><mode>3"
+  [(set (match_operand:VF_AVX512FP16VL 0 "register_operand")
+	(unspec:VF_AVX512FP16VL
+	  [(match_operand:VF_AVX512FP16VL 1 "vector_operand")
+	   (match_operand:VF_AVX512FP16VL 2 "vector_operand")]
+	   UNSPEC_COMPLEX_F_C_MUL))]
+  "TARGET_AVX512FP16")
+
 (define_insn "<avx512>_<complexopname>_<mode><maskc_name><round_name>"
   [(set (match_operand:VF_AVX512FP16VL 0 "register_operand" "=&v")
 	  (unspec:VF_AVX512FP16VL
diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-vector-complex-float.c b/gcc/testsuite/gcc.target/i386/avx512fp16-vector-complex-float.c
new file mode 100644
index 00000000000..bcb957f0de0
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/avx512fp16-vector-complex-float.c
@@ -0,0 +1,40 @@ 
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx512fp16 -mavx512vl" } */
+/* { dg-final { scan-assembler-times "vfmaddcph\[ \\t\]" 1 } } */
+/* { dg-final { scan-assembler-not "vfmadd\[123]*ph\[ \\t\]"} } */
+/* { dg-final { scan-assembler-not "vfmadd\[123]*sh\[ \\t\]"} } */
+/* { dg-final { scan-assembler-times "vfcmaddcph\[ \\t\]" 1 } } */
+/* { dg-final { scan-assembler-times "vfmulcph\[ \\t\]" 1 } } */
+/* { dg-final { scan-assembler-times "vfcmulcph\[ \\t\]" 1 } } */
+
+#include<complex.h>
+#define TYPE _Float16
+#define N 16
+
+void fma0 (_Complex TYPE *a, _Complex TYPE *b,
+           _Complex TYPE *c)
+{
+  for (int i = 0; i < N; i++)
+    c[i] += a[i] * b[i];
+}
+
+void fmaconj (_Complex TYPE a[restrict N], _Complex TYPE b[restrict N],
+	      _Complex TYPE c[restrict N])
+{
+  for (int i = 0; i < N; i++)
+    c[i] += a[i] * ~b[i];
+}
+
+void fmul (_Complex TYPE a[restrict N], _Complex TYPE b[restrict N],
+	   _Complex TYPE c[restrict N])
+{
+  for (int i = 0; i < N; i++)
+    c[i] = a[i] * b[i];
+}
+
+void fmulconj (_Complex TYPE a[restrict N], _Complex TYPE b[restrict N],
+	       _Complex TYPE c[restrict N])
+{
+  for (int i = 0; i < N; i++)
+    c[i] = a[i] * ~b[i];
+}