diff mbox series

[v3] RISC-V: Align IOR optimization MODE_CLASS condition to AND.

Message ID 20230419091820.3729443-1-pan2.li@intel.com
State New
Headers show
Series [v3] RISC-V: Align IOR optimization MODE_CLASS condition to AND. | expand

Commit Message

Li, Pan2 via Gcc-patches April 19, 2023, 9:18 a.m. UTC 
From: Pan Li <pan2.li@intel.com>

This patch aligned the MODE_CLASS condition of the IOR to the AND. Then
more MODE_CLASS besides SCALAR_INT can able to perform the optimization
A | (~A) -> -1 similar to AND operator. For example as below sample code.

vbool32_t test_shortcut_for_riscv_vmorn_case_5(vbool32_t v1, size_t vl)
{
  return __riscv_vmorn_mm_b32(v1, v1, vl);
}

Before this patch:
vsetvli  a5,zero,e8,mf4,ta,ma
vlm.v    v24,0(a1)
vsetvli  zero,a2,e8,mf4,ta,ma
vmorn.mm v24,v24,v24
vsetvli  a5,zero,e8,mf4,ta,ma
vsm.v    v24,0(a0)
ret

After this patch:
vsetvli zero,a2,e8,mf4,ta,ma
vmset.m v24
vsetvli a5,zero,e8,mf4,ta,ma
vsm.v   v24,0(a0)
ret

Or in RTL's perspective,
from:
(ior:VNx2BI (reg/v:VNx2BI 137 [ v1 ]) (not:VNx2BI (reg/v:VNx2BI 137 [ v1 ])))
to:
(const_vector:VNx2BI repeat [ (const_int 1 [0x1]) ])

The similar optimization like VMANDN has enabled already. There should
be no difference execpt the operator when compare the VMORN and VMANDN
for such kind of optimization. The patch aligns the IOR MODE_CLASS condition
of the simplification to the AND operator.

gcc/ChangeLog:

	* simplify-rtx.cc (simplify_context::simplify_binary_operation_1):
	  Align IOR (A | (~A) -> -1) optimization MODE_CLASS condition to AND.

gcc/testsuite/ChangeLog:

	* gcc.target/riscv/rvv/base/mask_insn_shortcut.c: Update check
	  condition.
	* gcc.target/riscv/simplify_ior_optimization.c: New test.

Signed-off-by: Pan Li <pan2.li@intel.com>
---
 gcc/simplify-rtx.cc                           |  4 +-
 .../riscv/rvv/base/mask_insn_shortcut.c       |  3 +-
 .../riscv/simplify_ior_optimization.c         | 50 +++++++++++++++++++
 3 files changed, 53 insertions(+), 4 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/riscv/simplify_ior_optimization.c

Comments

Kito Cheng April 19, 2023, 3:19 p.m. UTC | #1 
Committed, thanks :)

On Wed, Apr 19, 2023 at 5:19 PM Pan Li via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> From: Pan Li <pan2.li@intel.com>
>
> This patch aligned the MODE_CLASS condition of the IOR to the AND. Then
> more MODE_CLASS besides SCALAR_INT can able to perform the optimization
> A | (~A) -> -1 similar to AND operator. For example as below sample code.
>
> vbool32_t test_shortcut_for_riscv_vmorn_case_5(vbool32_t v1, size_t vl)
> {
>   return __riscv_vmorn_mm_b32(v1, v1, vl);
> }
>
> Before this patch:
> vsetvli  a5,zero,e8,mf4,ta,ma
> vlm.v    v24,0(a1)
> vsetvli  zero,a2,e8,mf4,ta,ma
> vmorn.mm v24,v24,v24
> vsetvli  a5,zero,e8,mf4,ta,ma
> vsm.v    v24,0(a0)
> ret
>
> After this patch:
> vsetvli zero,a2,e8,mf4,ta,ma
> vmset.m v24
> vsetvli a5,zero,e8,mf4,ta,ma
> vsm.v   v24,0(a0)
> ret
>
> Or in RTL's perspective,
> from:
> (ior:VNx2BI (reg/v:VNx2BI 137 [ v1 ]) (not:VNx2BI (reg/v:VNx2BI 137 [ v1 ])))
> to:
> (const_vector:VNx2BI repeat [ (const_int 1 [0x1]) ])
>
> The similar optimization like VMANDN has enabled already. There should
> be no difference execpt the operator when compare the VMORN and VMANDN
> for such kind of optimization. The patch aligns the IOR MODE_CLASS condition
> of the simplification to the AND operator.
>
> gcc/ChangeLog:
>
>         * simplify-rtx.cc (simplify_context::simplify_binary_operation_1):
>           Align IOR (A | (~A) -> -1) optimization MODE_CLASS condition to AND.
>
> gcc/testsuite/ChangeLog:
>
>         * gcc.target/riscv/rvv/base/mask_insn_shortcut.c: Update check
>           condition.
>         * gcc.target/riscv/simplify_ior_optimization.c: New test.
>
> Signed-off-by: Pan Li <pan2.li@intel.com>
> ---
>  gcc/simplify-rtx.cc                           |  4 +-
>  .../riscv/rvv/base/mask_insn_shortcut.c       |  3 +-
>  .../riscv/simplify_ior_optimization.c         | 50 +++++++++++++++++++
>  3 files changed, 53 insertions(+), 4 deletions(-)
>  create mode 100644 gcc/testsuite/gcc.target/riscv/simplify_ior_optimization.c
>
> diff --git a/gcc/simplify-rtx.cc b/gcc/simplify-rtx.cc
> index c57ff3320ee..d4aeebc7a5f 100644
> --- a/gcc/simplify-rtx.cc
> +++ b/gcc/simplify-rtx.cc
> @@ -3370,8 +3370,8 @@ simplify_context::simplify_binary_operation_1 (rtx_code code,
>        if (((GET_CODE (op0) == NOT && rtx_equal_p (XEXP (op0, 0), op1))
>            || (GET_CODE (op1) == NOT && rtx_equal_p (XEXP (op1, 0), op0)))
>           && ! side_effects_p (op0)
> -         && SCALAR_INT_MODE_P (mode))
> -       return constm1_rtx;
> +         && GET_MODE_CLASS (mode) != MODE_CC)
> +       return CONSTM1_RTX (mode);
>
>        /* (ior A C) is C if all bits of A that might be nonzero are on in C.  */
>        if (CONST_INT_P (op1)
> diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/mask_insn_shortcut.c b/gcc/testsuite/gcc.target/riscv/rvv/base/mask_insn_shortcut.c
> index 83cc4a1b5a5..57d0241675a 100644
> --- a/gcc/testsuite/gcc.target/riscv/rvv/base/mask_insn_shortcut.c
> +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/mask_insn_shortcut.c
> @@ -233,9 +233,8 @@ vbool64_t test_shortcut_for_riscv_vmxnor_case_6(vbool64_t v1, size_t vl) {
>  /* { dg-final { scan-assembler-not {vmxor\.mm\s+v[0-9]+,\s*v[0-9]+} } } */
>  /* { dg-final { scan-assembler-not {vmor\.mm\s+v[0-9]+,\s*v[0-9]+} } } */
>  /* { dg-final { scan-assembler-not {vmnor\.mm\s+v[0-9]+,\s*v[0-9]+} } } */
> -/* { dg-final { scan-assembler-times {vmorn\.mm\s+v[0-9]+,\s*v[0-9]+,\s*v[0-9]+} 7 } } */
>  /* { dg-final { scan-assembler-not {vmxnor\.mm\s+v[0-9]+,\s*v[0-9]+} } } */
>  /* { dg-final { scan-assembler-times {vmclr\.m\s+v[0-9]+} 14 } } */
> -/* { dg-final { scan-assembler-times {vmset\.m\s+v[0-9]+} 7 } } */
> +/* { dg-final { scan-assembler-times {vmset\.m\s+v[0-9]+} 14 } } */
>  /* { dg-final { scan-assembler-times {vmmv\.m\s+v[0-9]+,\s*v[0-9]+} 14 } } */
>  /* { dg-final { scan-assembler-times {vmnot\.m\s+v[0-9]+,\s*v[0-9]+} 14 } } */
> diff --git a/gcc/testsuite/gcc.target/riscv/simplify_ior_optimization.c b/gcc/testsuite/gcc.target/riscv/simplify_ior_optimization.c
> new file mode 100644
> index 00000000000..ec3bd0baf03
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/simplify_ior_optimization.c
> @@ -0,0 +1,50 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv64gc -mabi=lp64 -O2" } */
> +
> +#include <stdint.h>
> +
> +uint8_t test_simplify_ior_scalar_case_0 (uint8_t a)
> +{
> +  return a | ~a;
> +}
> +
> +uint16_t test_simplify_ior_scalar_case_1 (uint16_t a)
> +{
> +  return a | ~a;
> +}
> +
> +uint32_t test_simplify_ior_scalar_case_2 (uint32_t a)
> +{
> +  return a | ~a;
> +}
> +
> +uint64_t test_simplify_ior_scalar_case_3 (uint64_t a)
> +{
> +  return a | ~a;
> +}
> +
> +int8_t test_simplify_ior_scalar_case_4 (int8_t a)
> +{
> +  return a | ~a;
> +}
> +
> +int16_t test_simplify_ior_scalar_case_5 (int16_t a)
> +{
> +  return a | ~a;
> +}
> +
> +int32_t test_simplify_ior_scalar_case_6 (int32_t a)
> +{
> +  return a | ~a;
> +}
> +
> +int64_t test_simplify_ior_scalar_case_7 (int64_t a)
> +{
> +  return a | ~a;
> +}
> +
> +/* { dg-final { scan-assembler-times {li\s+a[0-9]+,\s*-1} 6 } } */
> +/* { dg-final { scan-assembler-times {li\s+a[0-9]+,\s*255} 1 } } */
> +/* { dg-final { scan-assembler-times {li\s+a[0-9]+,\s*65536} 1 } } */
> +/* { dg-final { scan-assembler-not {or\s+a[0-9]+} } } */
> +/* { dg-final { scan-assembler-not {not\s+a[0-9]+} } } */
> --
> 2.34.1
>
diff mbox series

Patch

diff --git a/gcc/simplify-rtx.cc b/gcc/simplify-rtx.cc
index c57ff3320ee..d4aeebc7a5f 100644
--- a/gcc/simplify-rtx.cc
+++ b/gcc/simplify-rtx.cc
@@ -3370,8 +3370,8 @@  simplify_context::simplify_binary_operation_1 (rtx_code code,
       if (((GET_CODE (op0) == NOT && rtx_equal_p (XEXP (op0, 0), op1))
 	   || (GET_CODE (op1) == NOT && rtx_equal_p (XEXP (op1, 0), op0)))
 	  && ! side_effects_p (op0)
-	  && SCALAR_INT_MODE_P (mode))
-	return constm1_rtx;
+	  && GET_MODE_CLASS (mode) != MODE_CC)
+	return CONSTM1_RTX (mode);
 
       /* (ior A C) is C if all bits of A that might be nonzero are on in C.  */
       if (CONST_INT_P (op1)
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/mask_insn_shortcut.c b/gcc/testsuite/gcc.target/riscv/rvv/base/mask_insn_shortcut.c
index 83cc4a1b5a5..57d0241675a 100644
--- a/gcc/testsuite/gcc.target/riscv/rvv/base/mask_insn_shortcut.c
+++ b/gcc/testsuite/gcc.target/riscv/rvv/base/mask_insn_shortcut.c
@@ -233,9 +233,8 @@  vbool64_t test_shortcut_for_riscv_vmxnor_case_6(vbool64_t v1, size_t vl) {
 /* { dg-final { scan-assembler-not {vmxor\.mm\s+v[0-9]+,\s*v[0-9]+} } } */
 /* { dg-final { scan-assembler-not {vmor\.mm\s+v[0-9]+,\s*v[0-9]+} } } */
 /* { dg-final { scan-assembler-not {vmnor\.mm\s+v[0-9]+,\s*v[0-9]+} } } */
-/* { dg-final { scan-assembler-times {vmorn\.mm\s+v[0-9]+,\s*v[0-9]+,\s*v[0-9]+} 7 } } */
 /* { dg-final { scan-assembler-not {vmxnor\.mm\s+v[0-9]+,\s*v[0-9]+} } } */
 /* { dg-final { scan-assembler-times {vmclr\.m\s+v[0-9]+} 14 } } */
-/* { dg-final { scan-assembler-times {vmset\.m\s+v[0-9]+} 7 } } */
+/* { dg-final { scan-assembler-times {vmset\.m\s+v[0-9]+} 14 } } */
 /* { dg-final { scan-assembler-times {vmmv\.m\s+v[0-9]+,\s*v[0-9]+} 14 } } */
 /* { dg-final { scan-assembler-times {vmnot\.m\s+v[0-9]+,\s*v[0-9]+} 14 } } */
diff --git a/gcc/testsuite/gcc.target/riscv/simplify_ior_optimization.c b/gcc/testsuite/gcc.target/riscv/simplify_ior_optimization.c
new file mode 100644
index 00000000000..ec3bd0baf03
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/simplify_ior_optimization.c
@@ -0,0 +1,50 @@ 
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gc -mabi=lp64 -O2" } */
+
+#include <stdint.h>
+
+uint8_t test_simplify_ior_scalar_case_0 (uint8_t a)
+{
+  return a | ~a;
+}
+
+uint16_t test_simplify_ior_scalar_case_1 (uint16_t a)
+{
+  return a | ~a;
+}
+
+uint32_t test_simplify_ior_scalar_case_2 (uint32_t a)
+{
+  return a | ~a;
+}
+
+uint64_t test_simplify_ior_scalar_case_3 (uint64_t a)
+{
+  return a | ~a;
+}
+
+int8_t test_simplify_ior_scalar_case_4 (int8_t a)
+{
+  return a | ~a;
+}
+
+int16_t test_simplify_ior_scalar_case_5 (int16_t a)
+{
+  return a | ~a;
+}
+
+int32_t test_simplify_ior_scalar_case_6 (int32_t a)
+{
+  return a | ~a;
+}
+
+int64_t test_simplify_ior_scalar_case_7 (int64_t a)
+{
+  return a | ~a;
+}
+
+/* { dg-final { scan-assembler-times {li\s+a[0-9]+,\s*-1} 6 } } */
+/* { dg-final { scan-assembler-times {li\s+a[0-9]+,\s*255} 1 } } */
+/* { dg-final { scan-assembler-times {li\s+a[0-9]+,\s*65536} 1 } } */
+/* { dg-final { scan-assembler-not {or\s+a[0-9]+} } } */
+/* { dg-final { scan-assembler-not {not\s+a[0-9]+} } } */