target/riscv: Fix the check with vector register multiples of LMUL

[PATCH] target/riscv: Fix the check with vector register multiples of LMUL

Posted by Zhiwei Jiang 1 year, 5 months ago

In the original extract32(val, 0, lmul) logic, when lmul is 2 and val is v10 or v12,
there is an issue with this check condition. I think a simple mod operation is sufficient.

Signed-off-by: Zhiwei Jiang <jiangzw@tecorigin.com>
---
 target/riscv/insn_trans/trans_rvv.c.inc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/target/riscv/insn_trans/trans_rvv.c.inc b/target/riscv/insn_trans/trans_rvv.c.inc
index 3a3896ba06..e89b0f2b1e 100644
--- a/target/riscv/insn_trans/trans_rvv.c.inc
+++ b/target/riscv/insn_trans/trans_rvv.c.inc
@@ -118,7 +118,7 @@ static bool require_nf(int vd, int nf, int lmul)
  */
 static bool require_align(const int8_t val, const int8_t lmul)
 {
-    return lmul <= 0 || extract32(val, 0, lmul) == 0;
+    return lmul <= 0 || val % lmul == 0;
 }
 
 /*
-- 
2.17.1

Re: [PATCH] target/riscv: Fix the check with vector register multiples of LMUL

Posted by Alistair Francis 1 year, 5 months ago

On Fri, Jun 28, 2024 at 7:46 PM Zhiwei Jiang <jiangzw@tecorigin.com> wrote:
>
> In the original extract32(val, 0, lmul) logic, when lmul is 2 and val is v10 or v12,
> there is an issue with this check condition. I think a simple mod operation is sufficient.

Overall looks ok. Do you mind updating the commit message to describe
what exactly this fixes (as in what issue you had) and why this fixes
it

Alistair

>
> Signed-off-by: Zhiwei Jiang <jiangzw@tecorigin.com>
> ---
>  target/riscv/insn_trans/trans_rvv.c.inc | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/target/riscv/insn_trans/trans_rvv.c.inc b/target/riscv/insn_trans/trans_rvv.c.inc
> index 3a3896ba06..e89b0f2b1e 100644
> --- a/target/riscv/insn_trans/trans_rvv.c.inc
> +++ b/target/riscv/insn_trans/trans_rvv.c.inc
> @@ -118,7 +118,7 @@ static bool require_nf(int vd, int nf, int lmul)
>   */
>  static bool require_align(const int8_t val, const int8_t lmul)
>  {
> -    return lmul <= 0 || extract32(val, 0, lmul) == 0;
> +    return lmul <= 0 || val % lmul == 0;
>  }
>
>  /*
> --
> 2.17.1
>
>