[fpc-devel] div creates wrong results on cortexm0
Jeppe Johansen
jeppe at j-software.dk
Sat Jan 26 15:20:30 CET 2019
Does the attached patch work?
On 1/26/19 12:20 PM, Michael Ring wrote:
> I thought I was going crazy when debugging code that worked fine on
> cortex m3/4 but created unexpected results on cortex m0.
>
> Turns out that the compiler creates incorrect assembler code for
> cortexm0 for div when used with constant expressions. Optimization
> level was -O1 and -O-
>
> Take this code:
>
> program divtest;
> var
> a,b,c,d : word;
> begin
> a := 64;
> b := 8;
> c := a div 8; //.LI4
> d := a div b; //.LI5
> end.
>
> looking at the assembler code it is very evident that no shift
> operations are done on cortex m0 (see .LI4). Is this a known issue or
> should I create a new one in bugzilla? (Search did not reveal a hit
> that looked like a match)
>
> Michael
>
> assembler cortex m0:
>
> .Ll2:
> mov r1,#64
> ldr r0,.Lj3
> strh r1,[r0]
> .Ll3:
> mov r1,#8
> ldr r0,.Lj4
> strh r1,[r0]
> .Ll4:
> ldr r0,.Lj3
> ldrh r0,[r0]
> uxth r0,r0
> ldr r1,.Lj6
> strh r0,[r1]
> .Ll5:
> ldr r0,.Lj3
> ldrh r1,[r0]
> ldr r0,.Lj4
> ldrh r0,[r0]
> bl fpc_div_longint
> uxth r0,r0
> ldr r1,.Lj9
> strh r0,[r1]
>
> assembler cortex m4:
>
> .Ll2:
> movs r1,#64
> ldr r0,.Lj3
> strh r1,[r0]
> .Ll3:
> movs r1,#8
> ldr r0,.Lj4
> strh r1,[r0]
> .Ll4:
> ldr r0,.Lj3
> ldrh r0,[r0]
> asrs r1,r0,#31
> add r0,r0,r1,lsr #29
> asrs r0,#3
> uxth r1,r0
> ldr r0,.Lj6
> strh r1,[r0]
> .Ll5:
> ldr r0,.Lj3
> ldrh r0,[r0]
> ldr r1,.Lj4
> ldrh r1,[r1]
> sdiv r0,r0,r1
> uxth r0,r0
> ldr r1,.Lj9
> strh r0,[r1]
-------------- next part --------------
A non-text attachment was scrubbed...
Name: div_patch.patch
Type: text/x-patch
Size: 1865 bytes
Desc: not available
URL: <http://lists.freepascal.org/pipermail/fpc-devel/attachments/20190126/8d71313b/attachment.bin>
More information about the fpc-devel
mailing list