On 02/06/17 00:54, Michael Collison wrote:
> This patch improves code generation for relational compares against zero that
> are not merged into a subtract instruction. This patch improves the >= and <
> cases.
>
> An example of the '<' case:
>
> int lt (int x, int y)
> {
> if ((x - y) < 0)
> return 10;
>
> return 0;
> }
>
> Trunk generates:
>
> lt:
> sub w1, w0, w1
> mov w0, 10
> cmp w1, 0
> csel w0, w0, wzr, lt
> ret
>
> With the patch we can eliminate the redundant subtract and now generate:
>
> lt:
> cmp w0, w1
> mov w0, 10
> csel w0, w0, wzr, mi
> ret
>
> Bootstrapped and tested on aarch64-linux-gnu. Okay for trunk?
>
> 2017-06-01 Michael Collison <[email protected]>
>
> * config/aarch64/aarch64-simd.md(aarch64_sub<mode>_compare0):
> New pattern.
> * testsuite/gcc.target/aarch64/cmp-2.c: New testcase.
>
>
OK.
R.
> pr7261.patch.patch
>
>
> diff --git a/gcc/config/aarch64/aarch64.md b/gcc/config/aarch64/aarch64.md
> index 51368e2..b90c728 100644
> --- a/gcc/config/aarch64/aarch64.md
> +++ b/gcc/config/aarch64/aarch64.md
> @@ -2011,6 +2011,17 @@
> [(set_attr "type" "alus_sreg,alus_imm,alus_imm")]
> )
>
> +(define_insn "aarch64_sub<mode>_compare0"
> + [(set (reg:CC_NZ CC_REGNUM)
> + (compare:CC_NZ
> + (minus:GPI (match_operand:GPI 0 "register_operand" "r")
> + (match_operand:GPI 1 "aarch64_plus_operand" "r"))
> + (const_int 0)))]
> + ""
> + "cmp\\t%<w>0, %<w>1"
> + [(set_attr "type" "alus_sreg")]
> +)
> +
> (define_insn "*compare_neg<mode>"
> [(set (reg:CC_Z CC_REGNUM)
> (compare:CC_Z
> diff --git a/gcc/testsuite/gcc.target/aarch64/cmp-2.c
> b/gcc/testsuite/gcc.target/aarch64/cmp-2.c
> new file mode 100644
> index 0000000..1201664
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/aarch64/cmp-2.c
> @@ -0,0 +1,21 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2" } */
> +
> +int lt (int x, int y)
> +{
> + if ((x - y) < 0)
> + return 10;
> +
> + return 0;
> +}
> +
> +int ge (int x, int y)
> +{
> + if ((x - y) >= 0)
> + return 10;
> +
> + return 0;
> +}
> +
> +/* { dg-final { scan-assembler-times "csel\t" 2 } } */
> +/* { dg-final { scan-assembler-not "sub\t" } } */
>
