Re: [PATCH v6 0/9] riscv: set vstart_eq_zero on mark_vs_dirty

[Alistair Francis]

On Thu, Feb 22, 2024 at 7:33 AM Daniel Henrique Barboza
<dbarboza@ventanamicro.com> wrote:
>
> Hi,
>
> In this version 2 new patches were added:
>
> - patch 5 eliminates the 'cpu_vl' global, and do_vsetvl() now loads 'vl'
>   directly from env. This was suggested by Richard in the v5 review;
>
> - patch 9 does a change in how we're doing the loops in ldst helpers.
>   This was also proposed by Richard but back in v2.
>
> Patch 9 is not related to what we're fixing here but let's fold it in
> and avoid leaving any code suggestions behind.
>
> Series based on alistair/riscv-to-apply.next.
>
> Patches missing acks/reviews: 5 and 9
>
> Changes from v5:
> - patch 5 (new): remove 'cpu_vl' global
> - patch 9 (new): change the loop in ldst helpers
> - v5 link: 
> https://lore.kernel.org/qemu-riscv/20240221022252.252872-1-dbarboza@ventanamicro.com/
>
> Daniel Henrique Barboza (8):
>   trans_rvv.c.inc: mark_vs_dirty() before loads and stores
>   trans_rvv.c.inc: remove 'is_store' bool from load/store fns
>   target/riscv: remove 'over' brconds from vector trans
>   target/riscv/translate.c: remove 'cpu_vstart' global
>   target/riscv: remove 'cpu_vl' global
>   target/riscv/vector_helper.c: set vstart = 0 in GEN_VEXT_VSLIDEUP_VX()
>   trans_rvv.c.inc: remove redundant mark_vs_dirty() calls
>   target/riscv/vector_helper.c: optimize loops in ldst helpers
>
> Ivan Klokov (1):
>   target/riscv: Clear vstart_qe_zero flag

Looks good! Just conflicts with the Ztso extension support that was
recently added in the RISC-V tree

Alistair

>
>  target/riscv/insn_trans/trans_rvbf16.c.inc |  18 +-
>  target/riscv/insn_trans/trans_rvv.c.inc    | 294 ++++++---------------
>  target/riscv/insn_trans/trans_rvvk.c.inc   |  30 +--
>  target/riscv/translate.c                   |  11 +-
>  target/riscv/vector_helper.c               |   7 +-
>  5 files changed, 104 insertions(+), 256 deletions(-)
>
> --
> 2.43.2
>
>