GNU Arm Embedded Toolchain

MOV instruction always assembled as MOVW

Asked by Dan Lewis

There are two versions of the MOV instruction. An instruction like

      MOV R0,1

using an 8-bit constant should assemble into a 16-bit instruction, while an instruction like

     MOVW R0,10000 // May also be written as MOV.W R0,1000

assembles into a 32-bit instruction in order to be able to use a 16-bit constant.

However, I find that the first example above generates:

     MOV.W R0,1

which is a 32-bit instruction.

Shouldn't the assembler use the smallest possible representation?

Thanks!
Dan

Question information

Language:
English Edit question
Status:
Answered
For:
GNU Arm Embedded Toolchain Edit question
Assignee:
No assignee Edit question
Last query:
Last reply:
Revision history for this message
Thomas Preud'homme (thomas-preudhomme) said : 		#1

Hi Dan,

I can confirm your issue. I agree that it would be nice to use the smallest representation possible for a mov but the code is not currently written to have that kind of behavior and rewriting it would require a substential amount of work. I'm sure this is not the answer you are looking for but right now your only solution is to use movs instead.

Best regards.

Revision history for this message
Dan Lewis (danielwlewis) said : 		#2

Hi Thomas,

After I posted my inquiry, I did a bit more research and discovered that the only 16-bit Thumb versions of the MOV immediate instruction are such that (1) outside an IT block, only MOVS exists, and (2) inside an IT block, only MOV<c> exists, where <c> is a condition code.

In other words, you're correct: There is no 16-bit version of a MOV immediate that can be used outside an IT block that does not affect the flags.

Reference: Section A6.7.75 of ARM®v7-M Architecture Reference Manual (Feb 2010).

Best,

Dan

Can you help with this problem?

Provide an answer of your own, or ask Dan Lewis for more information if necessary.

To post a message you must log in.