When clearing a bitfield we don't need to lead the
source register. Use deposit_z_i32() with the BFC
opcode to save a load_reg() call.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 target/arm/tcg/translate.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/target/arm/tcg/translate.c b/target/arm/tcg/translate.c
index b71ac2d0d5..1a6938d1b3 100644
--- a/target/arm/tcg/translate.c
+++ b/target/arm/tcg/translate.c
@@ -7255,7 +7255,7 @@ static bool trans_UBFX(DisasContext *s, arg_UBFX *a)
 static bool trans_BFCI(DisasContext *s, arg_BFCI *a)
 {
     int msb = a->msb, lsb = a->lsb;
-    TCGv_i32 t_in, t_rd;
+    TCGv_i32 t_rd;
     int width;
 
     if (!ENABLE_ARCH_6T2) {
@@ -7268,15 +7268,14 @@ static bool trans_BFCI(DisasContext *s, arg_BFCI *a)
     }
 
     width = msb + 1 - lsb;
+    t_rd = load_reg(s, a->rd);
     if (a->rn == 15) {
         /* BFC */
-        t_in = tcg_constant_i32(0);
+        tcg_gen_deposit_z_i32(t_rd, t_rd, lsb, width);
     } else {
         /* BFI */
-        t_in = load_reg(s, a->rn);
+        tcg_gen_deposit_i32(t_rd, t_rd, load_reg(s, a->rn), lsb, width);
     }
-    t_rd = load_reg(s, a->rd);
-    tcg_gen_deposit_i32(t_rd, t_rd, t_in, lsb, width);
     store_reg(s, a->rd, t_rd);
     return true;
 }
-- 
2.41.0

On Tue, 22 Aug 2023 at 10:51, Philippe Mathieu-Daudé <philmd@linaro.org> wrote:
>
> When clearing a bitfield we don't need to lead the

"load" ?

> source register. Use deposit_z_i32() with the BFC
> opcode to save a load_reg() call.
>
> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> ---
>  target/arm/tcg/translate.c | 9 ++++-----
>  1 file changed, 4 insertions(+), 5 deletions(-)
>
> diff --git a/target/arm/tcg/translate.c b/target/arm/tcg/translate.c
> index b71ac2d0d5..1a6938d1b3 100644
> --- a/target/arm/tcg/translate.c
> +++ b/target/arm/tcg/translate.c
> @@ -7255,7 +7255,7 @@ static bool trans_UBFX(DisasContext *s, arg_UBFX *a)
>  static bool trans_BFCI(DisasContext *s, arg_BFCI *a)
>  {
>      int msb = a->msb, lsb = a->lsb;
> -    TCGv_i32 t_in, t_rd;
> +    TCGv_i32 t_rd;
>      int width;
>
>      if (!ENABLE_ARCH_6T2) {
> @@ -7268,15 +7268,14 @@ static bool trans_BFCI(DisasContext *s, arg_BFCI *a)
>      }
>
>      width = msb + 1 - lsb;
> +    t_rd = load_reg(s, a->rd);
>      if (a->rn == 15) {
>          /* BFC */
> -        t_in = tcg_constant_i32(0);
> +        tcg_gen_deposit_z_i32(t_rd, t_rd, lsb, width);
>      } else {
>          /* BFI */
> -        t_in = load_reg(s, a->rn);
> +        tcg_gen_deposit_i32(t_rd, t_rd, load_reg(s, a->rn), lsb, width);
>      }
> -    t_rd = load_reg(s, a->rd);
> -    tcg_gen_deposit_i32(t_rd, t_rd, t_in, lsb, width);
>      store_reg(s, a->rd, t_rd);
>      return true;

The comment says we are saving a load_reg() call, but the
code change doesn't seem to do that. Before the change:
 * for BFC we call load_reg for rd
 * for BFI we call load_reg for rn and rd

After the change:
 * for BFC we call load_reg for rd
 * for BFI we call load_reg for rn and rd

So we're not saving any load_reg() calls as far as I can see ?

-- PMM

On 29/8/23 15:20, Peter Maydell wrote:
> On Tue, 22 Aug 2023 at 10:51, Philippe Mathieu-Daudé <philmd@linaro.org> wrote:
>>
>> When clearing a bitfield we don't need to lead the
> 
> "load" ?
> 
>> source register. Use deposit_z_i32() with the BFC
>> opcode to save a load_reg() call.
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>> ---
>>   target/arm/tcg/translate.c | 9 ++++-----
>>   1 file changed, 4 insertions(+), 5 deletions(-)
>>
>> diff --git a/target/arm/tcg/translate.c b/target/arm/tcg/translate.c
>> index b71ac2d0d5..1a6938d1b3 100644
>> --- a/target/arm/tcg/translate.c
>> +++ b/target/arm/tcg/translate.c
>> @@ -7255,7 +7255,7 @@ static bool trans_UBFX(DisasContext *s, arg_UBFX *a)
>>   static bool trans_BFCI(DisasContext *s, arg_BFCI *a)
>>   {
>>       int msb = a->msb, lsb = a->lsb;
>> -    TCGv_i32 t_in, t_rd;
>> +    TCGv_i32 t_rd;
>>       int width;
>>
>>       if (!ENABLE_ARCH_6T2) {
>> @@ -7268,15 +7268,14 @@ static bool trans_BFCI(DisasContext *s, arg_BFCI *a)
>>       }
>>
>>       width = msb + 1 - lsb;
>> +    t_rd = load_reg(s, a->rd);
>>       if (a->rn == 15) {
>>           /* BFC */
>> -        t_in = tcg_constant_i32(0);
>> +        tcg_gen_deposit_z_i32(t_rd, t_rd, lsb, width);
>>       } else {
>>           /* BFI */
>> -        t_in = load_reg(s, a->rn);
>> +        tcg_gen_deposit_i32(t_rd, t_rd, load_reg(s, a->rn), lsb, width);
>>       }
>> -    t_rd = load_reg(s, a->rd);
>> -    tcg_gen_deposit_i32(t_rd, t_rd, t_in, lsb, width);
>>       store_reg(s, a->rd, t_rd);
>>       return true;
> 
> The comment says we are saving a load_reg() call, but the
> code change doesn't seem to do that. Before the change:
>   * for BFC we call load_reg for rd
>   * for BFI we call load_reg for rn and rd
> 
> After the change:
>   * for BFC we call load_reg for rd
>   * for BFI we call load_reg for rn and rd
> 
> So we're not saving any load_reg() calls as far as I can see ?

Indeed you are right.

The optimizations are in tcg_gen_deposit_z_i32() vs using
tcg_gen_deposit_i32(). I'll reword the description.

Thanks,

Phil.

On 8/22/23 02:51, Philippe Mathieu-Daudé wrote:
> When clearing a bitfield we don't need to lead the
> source register. Use deposit_z_i32() with the BFC
> opcode to save a load_reg() call.
> 
> Signed-off-by: Philippe Mathieu-Daudé<philmd@linaro.org>
> ---
>   target/arm/tcg/translate.c | 9 ++++-----
>   1 file changed, 4 insertions(+), 5 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~