flat assembler
Message board for the users of flat assembler.
Index
> Compiler Internals > pshufb opcode |
Author |
|
IntermittentP 24 Sep 2015, 14:15
When using the pshufb opcode I cannot get it to work with mem as the source. Works ok when xmm* used. Any thoughts.
|
|||
24 Sep 2015, 14:15 |
|
IntermittentP 24 Sep 2015, 15:40
Thanks I_inc
I've tried the your suggestions but to no avail, trying dqword on the source, dest (and even both). Another opcode, psignb, which like pshufb, can also operate on 64 bits also fails. movdqu which is only 128 bits works ok. I'm using 1.71.39. |
|||
24 Sep 2015, 15:40 |
|
l_inc 24 Sep 2015, 15:48
IntermittentP
Please provide your code. Preferably a minimal example with the problem. _________________ Faith is a superposition of knowledge and fallacy |
|||
24 Sep 2015, 15:48 |
|
IntermittentP 24 Sep 2015, 16:21
Using psignb produces ... 66 0F 38 08 05 08 20 40 00
66 0F 38 08 seems correct 05 must be the xxm register. I used xxm0 but have been unable to decode the 05 08 20 40 00 is the correct address for the mask. This data is produced by listing.exe from the .fas file. |
|||
24 Sep 2015, 16:21 |
|
IntermittentP 24 Sep 2015, 16:27
Apologies. Of course in this instance the 05 is the dest.
|
|||
24 Sep 2015, 16:27 |
|
IntermittentP 24 Sep 2015, 16:29
Will provide code soonest. Am summoned atm!
|
|||
24 Sep 2015, 16:29 |
|
l_inc 24 Sep 2015, 16:58
IntermittentP
Quote: Apologies. Of course in this instance the 05 is the dest. 05 is the ModR/M byte, which specifies both the source and the destination. 05 stands for an operation with a register AL/AX/EAX/MM0/XMM0 and memory addressed by a displacement that follows. _________________ Faith is a superposition of knowledge and fallacy |
|||
24 Sep 2015, 16:58 |
|
IntermittentP 25 Sep 2015, 10:16
Thanks for that info I_inc, I'll try to find the info in the docs.
I minimised the code to present it here by bulk deletion of lines. As I was only experimenting I didn't store the original (Flatasm resaved on run). When the code was minimised the line ran with no problems which is a "good thing" but of course it would have been better to have found the problem. I'll play around with it and come back if I find anything interesting! Thanks again I_inc. |
|||
25 Sep 2015, 10:16 |
|
revolution 25 Sep 2015, 10:43
IntermittentP: If you show the source code we can probably help very easily. Please reduce it to just the lines of interest.
|
|||
25 Sep 2015, 10:43 |
|
< Last Thread | Next Thread > |
Forum Rules:
|
Copyright © 1999-2024, Tomasz Grysztar. Also on GitHub, YouTube.
Website powered by rwasa.