Alex Coplan <alex.cop...@arm.com> writes:
> Hi,
>
> As the testcase shows, this pattern had an incorrect constraint leading
> to GCC's output getting rejected by the assembler.
>
> This patch fixes the constraint accordingly.
>
> The test is split into two: one that can run without bf16 support from
> the assembler and another that checks that the output actually assembles
> when such support is available.
>
> Bootstrapped/regtested on aarch64-linux-gnu.
>
> OK for GCC 13? Or better to wait for next stage 1? What about backports?
OK for GCC 13 & backports, thanks.
Richard
>
> Thanks,
> Alex
>
> gcc/ChangeLog:
>
> PR target/104921
> * config/aarch64/aarch64-simd.md (aarch64_bfmlal<bt>_lane<q>v4sf):
> Use correct constraint for operand 3.
>
> gcc/testsuite/ChangeLog:
>
> PR target/104921
> * gcc.target/aarch64/pr104921-1.c: New test.
> * gcc.target/aarch64/pr104921-2.c: New test.
> * gcc.target/aarch64/pr104921.x: Include file for new tests.
>
> diff --git a/gcc/config/aarch64/aarch64-simd.md
> b/gcc/config/aarch64/aarch64-simd.md
> index 7f212bf37cd..dd5eed387f2 100644
> --- a/gcc/config/aarch64/aarch64-simd.md
> +++ b/gcc/config/aarch64/aarch64-simd.md
> @@ -9153,7 +9153,7 @@ (define_insn "aarch64_bfmlal<bt>_lane<q>v4sf"
> [(set (match_operand:V4SF 0 "register_operand" "=w")
> (plus: V4SF (match_operand:V4SF 1 "register_operand" "0")
> (unspec:V4SF [(match_operand:V8BF 2 "register_operand"
> "w")
> - (match_operand:VBF 3 "register_operand"
> "w")
> + (match_operand:VBF 3 "register_operand"
> "x")
> (match_operand:SI 4 "const_int_operand"
> "n")]
> BF_MLA)))]
> "TARGET_BF16_SIMD"
> diff --git a/gcc/testsuite/gcc.target/aarch64/pr104921-1.c
> b/gcc/testsuite/gcc.target/aarch64/pr104921-1.c
> new file mode 100644
> index 00000000000..dcf6fe7d90d
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/aarch64/pr104921-1.c
> @@ -0,0 +1,12 @@
> +/* { dg-do compile } */
> +/* { dg-additional-options "-O2 -march=armv8.2-a+bf16 -std=gnu99
> -save-temps" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> +
> +#include "pr104921.x"
> +
> +/*
> +**foo:
> +** mov v([0-9]|1[0-5])\.8b, v16\.8b
> +** bfmlalb v0\.4s, v1\.8h, v([0-9]|1[0-5])\.h\[0\]
> +** ret
> +*/
> diff --git a/gcc/testsuite/gcc.target/aarch64/pr104921-2.c
> b/gcc/testsuite/gcc.target/aarch64/pr104921-2.c
> new file mode 100644
> index 00000000000..211fcd0aca9
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/aarch64/pr104921-2.c
> @@ -0,0 +1,6 @@
> +/* { dg-do assemble } */
> +/* { dg-add-options arm_v8_2a_bf16_neon } */
> +/* { dg-additional-options "-O2 -std=gnu99" } */
> +/* { dg-require-effective-target arm_v8_2a_bf16_neon_ok } */
> +
> +#include "pr104921.x"
> diff --git a/gcc/testsuite/gcc.target/aarch64/pr104921.x
> b/gcc/testsuite/gcc.target/aarch64/pr104921.x
> new file mode 100644
> index 00000000000..1e1a6f24e22
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/aarch64/pr104921.x
> @@ -0,0 +1,9 @@
> +#include <arm_neon.h>
> +
> +float32x4_t
> +foo(float32x4_t x, bfloat16x8_t a)
> +{
> + register bfloat16x4_t b asm ("v16");
> + asm volatile ("" : "=w"(b));
> + return vbfmlalbq_lane_f32 (x, a, b, 0);
> +}
