The optimizer prefers to have constants as the second operand,
so expand LT x,0 instead of GT 0,x.  This will not affect the
generated code at all.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 tcg/i386/tcg-target.c.inc | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc
index 4cd5d4276c..8260c35edd 100644
--- a/tcg/i386/tcg-target.c.inc
+++ b/tcg/i386/tcg-target.c.inc
@@ -4399,8 +4399,8 @@ static void expand_vec_sari(TCGType type, unsigned vece,
             /* Otherwise we will need to use a compare vs 0 to produce
              * the sign-extend, shift and merge.
              */
-            tcg_gen_cmp_vec(TCG_COND_GT, MO_64, t1,
-                            tcg_constant_vec(type, MO_64, 0), v1);
+            tcg_gen_cmp_vec(TCG_COND_LT, MO_64, t1, v1,
+                            tcg_constant_vec(type, MO_64, 0));
             tcg_gen_shri_vec(MO_64, v0, v1, imm);
             tcg_gen_shli_vec(MO_64, t1, t1, 64 - imm);
             tcg_gen_or_vec(MO_64, v0, v0, t1);
-- 
2.43.0

Hi Richard,

On 30/8/25 05:39, Richard Henderson wrote:
> The optimizer prefers to have constants as the second operand,

Should we document this somewhere?

> so expand LT x,0 instead of GT 0,x.  This will not affect the
> generated code at all.
> 
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>   tcg/i386/tcg-target.c.inc | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>

On 9/1/25 16:44, Philippe Mathieu-Daudé wrote:
> Hi Richard,
> 
> On 30/8/25 05:39, Richard Henderson wrote:
>> The optimizer prefers to have constants as the second operand,
> 
> Should we document this somewhere?

It's not a requirement.  It's just how tcg/optimize will canonicalize things.  Deviating 
from that unnecessarily just doesn't feel as clean.


r~