Re: [PATCH] target/avr: fix long address calculation

[Richard Henderson]

On 11/17/22 08:11, Pavel Dovgalyuk wrote:

AVR ELPMX instruction (and some others) use three registers to
form long 24-bit address from RAMPZ and two 8-bit registers.
RAMPZ stores shifted 8 bits like ff0000 to simplify address calculation.
This patch fixes full address calculation in function gen_get_addr
by changing the mess in offsets of deposit tcg instructions.

Signed-off-by: Pavel Dovgalyuk <pavel.dovgal...@ispras.ru>

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


r~

---
  target/avr/translate.c |    4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/target/avr/translate.c b/target/avr/translate.c
index e65b6008c0..c9a0a39c2d 100644
--- a/target/avr/translate.c
+++ b/target/avr/translate.c
@@ -1572,8 +1572,8 @@ static TCGv gen_get_addr(TCGv H, TCGv M, TCGv L)
  {
      TCGv addr = tcg_temp_new_i32();

- tcg_gen_deposit_tl(addr, M, H, 8, 8);

-    tcg_gen_deposit_tl(addr, L, addr, 8, 16);
+    tcg_gen_deposit_tl(addr, H, M, 8, 8);
+    tcg_gen_deposit_tl(addr, addr, L, 0, 8);

return addr;

  }