On Thu, Jun 6, 2024 at 4:49 PM Kong, Lingling <lingling.k...@intel.com> wrote:
>
> Enable ZU for IMUL (opcodes 0x69 and 0x6B) and SETcc.
>
> gcc/ChangeLog:
>
> * config/i386/i386-opts.h (enum apx_features):Add apx_zu.
> * config/i386/i386.h (TARGET_APX_ZU): Define.
> * config/i386/i386.md (*imulhi<mode>zu): New define_insn.
> (*setcc_<mode>_zu): Ditto.
> * config/i386/i386.opt: Add enum value for zu.
>
> gcc/testsuite/ChangeLog:
>
> * gcc.target/i386/apx-zu-1.c: New test.
> * gcc.target/i386/apx-zu-2.c: Ditto.
>
> Bootstrapped & regtested on x86-64-pc-linux-gnu with binutils 2.42 branch.
> OK for trunk?
Ok.
>
> ---
> gcc/config/i386/i386-opts.h | 3 +-
> gcc/config/i386/i386.h | 1 +
> gcc/config/i386/i386.md | 25 ++++++++++++++--
> gcc/config/i386/i386.opt | 3 ++
> gcc/testsuite/gcc.target/i386/apx-zu-1.c | 38 ++++++++++++++++++++++++
> gcc/testsuite/gcc.target/i386/apx-zu-2.c | 19 ++++++++++++
> 6 files changed, 86 insertions(+), 3 deletions(-) create mode 100644
> gcc/testsuite/gcc.target/i386/apx-zu-1.c
> create mode 100644 gcc/testsuite/gcc.target/i386/apx-zu-2.c
>
> diff --git a/gcc/config/i386/i386-opts.h b/gcc/config/i386/i386-opts.h index
> 5fcc4927978..c7ec0d9fd39 100644
> --- a/gcc/config/i386/i386-opts.h
> +++ b/gcc/config/i386/i386-opts.h
> @@ -142,8 +142,9 @@ enum apx_features {
> apx_ppx = 1 << 3,
> apx_nf = 1 << 4,
> apx_ccmp = 1 << 5,
> + apx_zu = 1 << 6,
> apx_all = apx_egpr | apx_push2pop2 | apx_ndd
> - | apx_ppx | apx_nf | apx_ccmp,
> + | apx_ppx | apx_nf | apx_ccmp | apx_zu,
> };
>
> #endif
> diff --git a/gcc/config/i386/i386.h b/gcc/config/i386/i386.h index
> 7051c6c13e4..dc1a1f44320 100644
> --- a/gcc/config/i386/i386.h
> +++ b/gcc/config/i386/i386.h
> @@ -57,6 +57,7 @@ see the files COPYING3 and COPYING.RUNTIME respectively.
> If not, see #define TARGET_APX_PPX (ix86_apx_features & apx_ppx) #define
> TARGET_APX_NF (ix86_apx_features & apx_nf)
> #define TARGET_APX_CCMP (ix86_apx_features & apx_ccmp)
> +#define TARGET_APX_ZU (ix86_apx_features & apx_zu)
>
> #include "config/vxworks-dummy.h"
>
> diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md index
> ffcf63e1cba..a2765f65754 100644
> --- a/gcc/config/i386/i386.md
> +++ b/gcc/config/i386/i386.md
> @@ -9967,6 +9967,19 @@
> (const_string "direct")))
> (set_attr "mode" "<MODE>")])
>
> +(define_insn "*imulhi<mode>zu"
> + [(set (match_operand:SWI48x 0 "register_operand" "=r,r")
> + (zero_extend:SWI48x
> + (mult:HI (match_operand:HI 1 "nonimmediate_operand" "%rm,rm")
> + (match_operand:HI 2 "immediate_operand" "K,n"))))
> + (clobber (reg:CC FLAGS_REG))]
> + "TARGET_APX_ZU"
> + "@
> + imulzu{w}\t{%2, %1, %w0|%w0, %1, %2}
> + imulzu{w}\t{%2, %1, %w0|%w0, %1, %2}"
> + [(set_attr "type" "imul")
> + (set_attr "mode" "HI")])
> +
> (define_insn "*mulsi3_1_zext"
> [(set (match_operand:DI 0 "register_operand" "=r,r,r")
> (zero_extend:DI
> @@ -18354,11 +18367,19 @@
> ;; For all sCOND expanders, also expand the compare or test insn that ;;
> generates cc0. Generate an equality comparison if `seq' or `sne'.
>
> +(define_insn "*setcc_<mode>_zu"
> + [(set (match_operand:SWI248 0 "register_operand" "=r")
> + (match_operator:SWI248 1 "ix86_comparison_operator"
> + [(reg FLAGS_REG) (const_int 0)]))]
> + "TARGET_APX_ZU"
> + "setzu%C1\t%b0"
> + [(set_attr "type" "setcc")])
> +
> (define_insn_and_split "*setcc_di_1"
> [(set (match_operand:DI 0 "register_operand" "=q")
> (match_operator:DI 1 "ix86_comparison_operator"
> [(reg FLAGS_REG) (const_int 0)]))]
> - "TARGET_64BIT && !TARGET_PARTIAL_REG_STALL"
> + "!TARGET_APX_ZU && TARGET_64BIT && !TARGET_PARTIAL_REG_STALL"
> "#"
> "&& reload_completed"
> [(set (match_dup 2) (match_dup 1))
> @@ -18391,7 +18412,7 @@
> [(set (match_operand:SWI24 0 "register_operand" "=q")
> (match_operator:SWI24 1 "ix86_comparison_operator"
> [(reg FLAGS_REG) (const_int 0)]))]
> - "!TARGET_PARTIAL_REG_STALL
> + "!TARGET_APX_ZU && !TARGET_PARTIAL_REG_STALL
> && (!TARGET_ZERO_EXTEND_WITH_AND || optimize_function_for_size_p (cfun))"
> "#"
> "&& reload_completed"
> diff --git a/gcc/config/i386/i386.opt b/gcc/config/i386/i386.opt index
> 7017cc87cec..353fffb2343 100644
> --- a/gcc/config/i386/i386.opt
> +++ b/gcc/config/i386/i386.opt
> @@ -1342,6 +1342,9 @@ Enum(apx_features) String(nf) Value(apx_nf) Set(6)
> EnumValue
> Enum(apx_features) String(ccmp) Value(apx_ccmp) Set(7)
>
> +EnumValue
> +Enum(apx_features) String(zu) Value(apx_zu) Set(8)
> +
> EnumValue
> Enum(apx_features) String(all) Value(apx_all) Set(1)
>
> diff --git a/gcc/testsuite/gcc.target/i386/apx-zu-1.c
> b/gcc/testsuite/gcc.target/i386/apx-zu-1.c
> new file mode 100644
> index 00000000000..927a87673a7
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/apx-zu-1.c
> @@ -0,0 +1,38 @@
> +/* { dg-do compile { target { ! ia32 } } } */
> +/* { dg-options "-mapxf -march=x86-64 -O2" } */
> +/* { dg-final { scan-assembler-not "setle"} } */
> +/* { dg-final { scan-assembler-not "setge"} } */
> +/* { dg-final { scan-assembler-not "sete"} } */
> +/* { dg-final { scan-assembler-not "xor"} } */
> +/* { dg-final { scan-assembler-times "setzune" 1} } */
> +/* { dg-final { scan-assembler-times "setzule" 1} } */
> +/* { dg-final { scan-assembler-times "setzue" 1} } */
> +/* { dg-final { scan-assembler-times "setzuge" 1} } */
> +/* { dg-final { scan-assembler "imulzu"} } */
> +long long foo0 (int a) {
> + return a == 0 ? 0 : 1;
> +}
> +
> +long foo1 (int a, int b)
> +{
> + return a > b ? 0 : 1;
> +}
> +
> +int foo2 (int a, int b)
> +{
> + return a != b ? 0 : 1;
> +}
> +
> +short foo3 (int a, int b)
> +{
> + return a < b ? 0 : 1;
> +}
> +
> +unsigned long
> +f1(unsigned short x)
> +{
> + unsigned short a;
> + a = x * 1000;
> + return a;
> +}
> diff --git a/gcc/testsuite/gcc.target/i386/apx-zu-2.c
> b/gcc/testsuite/gcc.target/i386/apx-zu-2.c
> new file mode 100644
> index 00000000000..3ee04495d98
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/apx-zu-2.c
> @@ -0,0 +1,19 @@
> +/* { dg-do run { target { ! ia32 } } } */
> +/* { dg-require-effective-target apxf } */
> +/* { dg-options "-mapxf -march=x86-64 -O2" } */
> +#include "apx-zu-1.c"
> +
> +int main(void)
> +{
> + if (foo0 (0))
> + __builtin_abort ();
> + if (foo1 (3, 2))
> + __builtin_abort ();
> + if (foo2 (3, 2))
> + __builtin_abort ();
> + if (foo3 (2, 3))
> + __builtin_abort ();
> + if (f1 (2) != 2000)
> + __builtin_abort ();
> + return 0;
> +}
> --
> 2.31.1
>
--
BR,
Hongtao
