On Sun, 22 Jun 2025 at 00:53, Richard Henderson
<richard.hender...@linaro.org> wrote:
>
> Add gen_helper_gvec_sudot_idx_4b as an expander which
> swaps arguments and uses helper_gvec_usdot_idx_4b.
>
> Signed-off-by: Richard Henderson <richard.hender...@linaro.org>
> ---
> target/arm/tcg/helper.h | 2 --
> target/arm/tcg/translate.h | 3 +++
> target/arm/tcg/gengvec.c | 6 ++++++
> target/arm/tcg/vec_helper.c | 1 -
> 4 files changed, 9 insertions(+), 3 deletions(-)
>
> diff --git a/target/arm/tcg/helper.h b/target/arm/tcg/helper.h
> index c459eba079..2702627eab 100644
> --- a/target/arm/tcg/helper.h
> +++ b/target/arm/tcg/helper.h
> @@ -622,8 +622,6 @@ DEF_HELPER_FLAGS_5(gvec_sdot_idx_4h, TCG_CALL_NO_RWG,
> void, ptr, ptr, ptr, ptr, i32)
> DEF_HELPER_FLAGS_5(gvec_udot_idx_4h, TCG_CALL_NO_RWG,
> void, ptr, ptr, ptr, ptr, i32)
> -DEF_HELPER_FLAGS_5(gvec_sudot_idx_4b, TCG_CALL_NO_RWG,
> - void, ptr, ptr, ptr, ptr, i32)
> DEF_HELPER_FLAGS_5(gvec_usdot_idx_4b, TCG_CALL_NO_RWG,
> void, ptr, ptr, ptr, ptr, i32)
>
> diff --git a/target/arm/tcg/translate.h b/target/arm/tcg/translate.h
> index 8cf6324e2f..90a03a4642 100644
> --- a/target/arm/tcg/translate.h
> +++ b/target/arm/tcg/translate.h
> @@ -627,6 +627,9 @@ void gen_gvec_urecpe(unsigned vece, uint32_t rd_ofs,
> uint32_t rn_ofs,
> void gen_gvec_ursqrte(unsigned vece, uint32_t rd_ofs, uint32_t rn_ofs,
> uint32_t opr_sz, uint32_t max_sz);
>
> +void gen_helper_gvec_sudot_idx_4b(TCGv_ptr d, TCGv_ptr n, TCGv_ptr m,
> + TCGv_ptr a, TCGv_i32 desc);
> +
> /*
> * Forward to the isar_feature_* tests given a DisasContext pointer.
> */
> diff --git a/target/arm/tcg/gengvec.c b/target/arm/tcg/gengvec.c
> index 01867f8ace..c182e5ab6f 100644
> --- a/target/arm/tcg/gengvec.c
> +++ b/target/arm/tcg/gengvec.c
> @@ -23,6 +23,12 @@
> #include "translate.h"
>
>
> +void gen_helper_gvec_sudot_idx_4b(TCGv_ptr d, TCGv_ptr n, TCGv_ptr m,
> + TCGv_ptr a, TCGv_i32 desc)
> +{
> + gen_helper_gvec_usdot_idx_4b(d, m, n, a, desc);
> +}
> +
How does this work? In the DO_DOT_IDX() helper functions
the 'm' argument is indexed, so I don't see how we can
just flip the n and m arguments around.
-- PMM
