Hello Jiri,
W dniu 28.08.2024 o 06:35, Jiri Slaby pisze:
> On 27. 08. 24, 17:38, Lech Perczak wrote:
>> Only the single expansion of GENMASK proves very hard to analyze in the
>> diff - so I double-checked with a calculator,
>> though all BIT() expansions do match, as does the updated definition.
>> of SC16IS7XX_LSR_BRK_ERROR_MASK.
>
> You can diff also the generated .o fed through objdump -d (generated .s
> usually contains line numbers).
I see that Greg already got the series into tty-next, but for the sake of completeness,
I did the check anyway.
This also came to my mind after I posted diff of preprocessed source ;-)
The diff is way, way smaller, but functionality is the same.
On amd64, the sign extension used in BIT() is visible:
--- sc16is7xx.c.asm.1 2024-09-04 16:06:02.523038659 +0200
+++ sc16is7xx.c.asm.2 2024-09-04 16:05:42.103938886 +0200
@@ -289,17 +289,16 @@
1f9: c7 44 24 04 00 00 00 movl $0x0,0x4(%rsp)
200: 00
201: e8 00 00 00 00 call 206 <sc16is7xx_tx_empty+0x36>
- 206: 8b 44 24 04 mov 0x4(%rsp),%eax
- 20a: c1 e8 06 shr $0x6,%eax
- 20d: 83 e0 01 and $0x1,%eax
- 210: 48 8b 54 24 08 mov 0x8(%rsp),%rdx
- 215: 65 48 2b 14 25 28 00 sub %gs:0x28,%rdx
- 21c: 00 00
- 21e: 75 09 jne 229 <sc16is7xx_tx_empty+0x59>
- 220: 48 83 c4 10 add $0x10,%rsp
- 224: e9 00 00 00 00 jmp 229 <sc16is7xx_tx_empty+0x59>
- 229: e8 00 00 00 00 call 22e <sc16is7xx_tx_empty+0x5e>
- 22e: 66 90 xchg %ax,%ax
+ 206: 0f b6 44 24 04 movzbl 0x4(%rsp),%eax
+ 20b: 48 c1 e8 06 shr $0x6,%rax
+ 20f: 83 e0 01 and $0x1,%eax
+ 212: 48 8b 54 24 08 mov 0x8(%rsp),%rdx
+ 217: 65 48 2b 14 25 28 00 sub %gs:0x28,%rdx
+ 21e: 00 00
+ 220: 75 09 jne 22b <sc16is7xx_tx_empty+0x5b>
+ 222: 48 83 c4 10 add $0x10,%rsp
+ 226: e9 00 00 00 00 jmp 22b <sc16is7xx_tx_empty+0x5b>
+ 22b: e8 00 00 00 00 call 230 <sc16is7xx_tx_empty+0x60>
230: 90 nop
231: 90 nop
232: 90 nop
>
> regards,
--
Pozdrawiam/With kind regards,
Lech Perczak
Sr. Software Engineer
Camlin Technologies Poland Limited Sp. z o.o.
