diff mbox series

Fix unexpected Illegal instruction error on RISC-V.

Message ID 20240301145545.333810-2-liyongtai@iscas.ac.cn (mailing list archive)
State New, archived
Headers show
Series Fix unexpected Illegal instruction error on RISC-V. | expand

Commit Message

李永泰 March 1, 2024, 2:55 p.m. UTC
Avoid right-shifting by a negative number of bits when lmul is 8.

Signed-off-by: SiHuaN <liyongtai@iscas.ac.cn>
---
 target/riscv/vector_helper.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Comments

Philippe Mathieu-Daudé March 1, 2024, 3:51 p.m. UTC | #1
Hi SiHuaN,

On 1/3/24 15:55, SiHuaN wrote:
> Avoid right-shifting by a negative number of bits when lmul is 8.

FYI Demin posted a similar patch, see:
https://lore.kernel.org/qemu-devel/20240225174114.5298-1-demin.han@starfivetech.com/

> Signed-off-by: SiHuaN <liyongtai@iscas.ac.cn>
> ---
>   target/riscv/vector_helper.c | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
> index 84cec73eb2..f0158ea237 100644
> --- a/target/riscv/vector_helper.c
> +++ b/target/riscv/vector_helper.c
> @@ -53,10 +53,11 @@ target_ulong HELPER(vsetvl)(CPURISCVState *env, target_ulong s1,
>            * VLEN * LMUL >= SEW
>            * VLEN >> (8 - lmul) >= sew
>            * (vlenb << 3) >> (8 - lmul) >= sew
> +         * Considering that lmul may be 8, the following form cannot be used.
>            * vlenb >> (8 - 3 - lmul) >= sew
>            */
>           if (vlmul == 4 ||
> -            cpu->cfg.vlenb >> (8 - 3 - vlmul) < sew) {
> +            (cpu->cfg.vlenb << 3) >> (8 - vlmul) < sew) {
>               vill = true;
>           }
>       }
李永泰 March 1, 2024, 4:15 p.m. UTC | #2
Hi Philippe,

Thanks for the heads up. Sorry I didn't check for this before sending out my patch.
I'll track this in Demin's thread.


&gt; -----原始邮件-----
&gt; 发件人: "Philippe Mathieu-Daudé" <philmd@linaro.org>
&gt; 发送时间: 2024-03-01 23:51:03 (星期五)
&gt; 收件人: SiHuaN <liyongtai@iscas.ac.cn>, qemu-devel@nongnu.org
&gt; 抄送: "demin.han" <demin.han@starfivetech.com>, qemu-riscv <qemu-riscv@nongnu.org>, "Daniel Henrique Barboza" <dbarboza@ventanamicro.com>
&gt; 主题: Re: [PATCH] Fix unexpected Illegal instruction error on RISC-V.
&gt; 
&gt; Hi SiHuaN,
&gt; 
&gt; On 1/3/24 15:55, SiHuaN wrote:
&gt; &gt; Avoid right-shifting by a negative number of bits when lmul is 8.
&gt; 
&gt; FYI Demin posted a similar patch, see:
&gt; https://lore.kernel.org/qemu-devel/20240225174114.5298-1-demin.han@starfivetech.com/
&gt; 
&gt; &gt; Signed-off-by: SiHuaN <liyongtai@iscas.ac.cn>
&gt; &gt; ---
&gt; &gt;   target/riscv/vector_helper.c | 3 ++-
&gt; &gt;   1 file changed, 2 insertions(+), 1 deletion(-)
&gt; &gt; 
&gt; &gt; diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
&gt; &gt; index 84cec73eb2..f0158ea237 100644
&gt; &gt; --- a/target/riscv/vector_helper.c
&gt; &gt; +++ b/target/riscv/vector_helper.c
&gt; &gt; @@ -53,10 +53,11 @@ target_ulong HELPER(vsetvl)(CPURISCVState *env, target_ulong s1,
&gt; &gt;            * VLEN * LMUL &gt;= SEW
&gt; &gt;            * VLEN &gt;&gt; (8 - lmul) &gt;= sew
&gt; &gt;            * (vlenb &lt;&lt; 3) &gt;&gt; (8 - lmul) &gt;= sew
&gt; &gt; +         * Considering that lmul may be 8, the following form cannot be used.
&gt; &gt;            * vlenb &gt;&gt; (8 - 3 - lmul) &gt;= sew
&gt; &gt;            */
&gt; &gt;           if (vlmul == 4 ||
&gt; &gt; -            cpu-&gt;cfg.vlenb &gt;&gt; (8 - 3 - vlmul) &lt; sew) {
&gt; &gt; +            (cpu-&gt;cfg.vlenb &lt;&lt; 3) &gt;&gt; (8 - vlmul) &lt; sew) {
&gt; &gt;               vill = true;
&gt; &gt;           }
&gt; &gt;       }
</liyongtai@iscas.ac.cn></dbarboza@ventanamicro.com></qemu-riscv@nongnu.org></demin.han@starfivetech.com></liyongtai@iscas.ac.cn></philmd@linaro.org>
diff mbox series

Patch

diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
index 84cec73eb2..f0158ea237 100644
--- a/target/riscv/vector_helper.c
+++ b/target/riscv/vector_helper.c
@@ -53,10 +53,11 @@  target_ulong HELPER(vsetvl)(CPURISCVState *env, target_ulong s1,
          * VLEN * LMUL >= SEW
          * VLEN >> (8 - lmul) >= sew
          * (vlenb << 3) >> (8 - lmul) >= sew
+         * Considering that lmul may be 8, the following form cannot be used.
          * vlenb >> (8 - 3 - lmul) >= sew
          */
         if (vlmul == 4 ||
-            cpu->cfg.vlenb >> (8 - 3 - vlmul) < sew) {
+            (cpu->cfg.vlenb << 3) >> (8 - vlmul) < sew) {
             vill = true;
         }
     }