Forum : ST7/STM8
| Post Information | Post |
|---|---|
|
January 25, 2010 - 4:29pm
|
hi I have found a problem that I don´t understand: I have a switch case sentence with an enum:
enum codigos_funcion{codigo_direccionar=0, codigo_comienzo_direccionamiento=1,
codigo_fin_direccionamiento=2, codigo_guarda_peso_averia=3,
codigo_carga_peso_averia=4, codigo_carga_peso=5, codigo_conecta_unidad=6, codigo_desconecta_unidad=7, codigo_valor_unidad=8,
codigo_error=9, codigo_pregunta_estado_sistema=10,
max_codigos};
u8 buffer_recepcion_trama[BYTES_DATOS_EN_LA_TRAMA+2]={0,0,0,0,0,0,0,0};
enum eventos_maquina_modbus_esclavo processing_required_action(void)
{
enum codigos_funcion codigo_funcion_recibido;
enum eventos_maquina_modbus_esclavo evento_maquina_modbus_esclavo_procesado;
u8 peso_averia_recibido[4];
codigo_funcion_recibido = buffer_recepcion_trama[1];
//guarda_en_la_eeprom(3,codigo_funcion_recibido);
switch (codigo_funcion_recibido)
{
...
...
...
I receive the "codigo_funcion_recibido" in a buffer and I then I use it in the switch but the thing is that it doesn´t work. However if I add the function that is commented, which saves in eeprom this "codigo_funcion_redibido" then it works. I have been checking the .lst and it is: with that function included (it works):
; FUNCTION ?processing_required_action (BEGIN)
; SOURCE LINE # 27
0000 5206 SUB SP,#006H
; SOURCE LINE # 34
0002 B600 F LD A,buffer_recepcion_trama + 01H
0004 5F CLRW X
0005 97 LD XL,A
0006 1F01 F LDW (001H,SP),X ; [ codigo_funcion_recibido ]
; SOURCE LINE # 36
0008 AE0003 LDW X,#00003H
000B 8D000000 F CALLF ?guarda_en_la_eeprom
; SOURCE LINE # 38
000F 1E01 F LDW X,(001H,SP) ; [ codigo_funcion_recibido ]
0011 2750 JREQ ?CASE_0108
0013 A30001 CPW X,#00001H
0016 2730 JREQ ?CASE_0109
0018 A30002 CPW X,#00002H
001B 2761 JREQ ?CASE_0110
001D A30003 CPW X,#00003H
0020 276A JREQ ?CASE_0111
0022 A30004 CPW X,#00004H
0025 2603 JRNE ?LAB_0184
0027 CC0000 F JP ?NXT_0604
002A ?LAB_0184:
002A A30005 CPW X,#00005H
002D 2603 JRNE ?LAB_0185
002F CC0000 F JP ?NXT_0604
0032 ?LAB_0185:
0032 A30006 CPW X,#00006H
0035 276F JREQ ?CASE_0114
0037 A30007 CPW X,#00007H
003A 2775 JREQ ?CASE_0115
003C A30008 CPW X,#00008H
003F 2778 JREQ ?NXT_0604
0041 A3000A CPW X,#0000AH
0044 2673 JRNE ?NXT_0604
0046 2071 JRA ?NXT_0604
0048 ?CASE_0109:
...
...
...
with that function commented (it doesn´t work):
; FUNCTION ?processing_required_action (BEGIN)
; SOURCE LINE # 27
0000 5206 SUB SP,#006H
; SOURCE LINE # 34
0002 B600 F LD A,buffer_recepcion_trama + 01H
0004 5F CLRW X
0005 97 LD XL,A
0006 1F01 F LDW (001H,SP),X ; [ codigo_funcion_recibido ]
; SOURCE LINE # 38
0008 4D TNZ A
0009 2746 JREQ ?CASE_0108
000B 4A DEC A
000C 2728 JREQ ?CASE_0109
000E A102 CP A,#002H
0010 275A JREQ ?CASE_0110
0012 A103 CP A,#003H
0014 2764 JREQ ?CASE_0111
0016 A104 CP A,#004H
0018 2603 JRNE ?LAB_0184
001A CC0000 F JP ?NXT_0604
001D ?LAB_0184:
001D A105 CP A,#005H
001F 2603 JRNE ?LAB_0185
0021 CC0000 F JP ?NXT_0604
0024 ?LAB_0185:
0024 A106 CP A,#006H
0026 276C JREQ ?CASE_0114
0028 A107 CP A,#007H
002A 2773 JREQ ?CASE_0115
002C A108 CP A,#008H
002E 2777 JREQ ?NXT_0604
0030 A10A CP A,#00AH
0032 2673 JRNE ?NXT_0604
0034 2071 JRA ?NXT_0604
0036 ?CASE_0109:
...
...
...
I don´t know why it uses the Accumulator the second time instead of X like the first. It seems to me that could be some kind of types thing or signs thing but i don´t know how to fix it, could you help me please? thank you very much. I´m using STVD with the Discovery (STM8S105C6) and rkit RKit-STM8_2.26.09.0317 with the patch. Regards Alberto |
Hi Alberto,
The compiler makes a lot of optimizations. Amonst them, it follows the constants, and in your second example, it knows that even if codigo_funcion_recibido is an int, it is actually loaded by a u8, so its high byte is 0. This is why the switch is made on A.
In the first exmple, the contents comes from the guarda_en_la_eeprom() function return, so it is an actual 16-bit value.
So I do not think there is an issue in here.
Try to run your program with disabled optimizations. If it works without opti but does not when optimized, then the code can be suspected. In such a case, please send us the full listing files (with/without opti) and if possible a whole project that shows up the problem, so that we can reproduce it in our labs.
Let us know what happens.
Regards,
Hi bruno!
thank you for the response. If I disable the peephole optimization it works fine, the .lst looks similar to the first example:
; FUNCTION ?processing_required_action (BEGIN) ; SOURCE LINE # 28 0000 5206 SUB SP,#006H ; SOURCE LINE # 34 0002 5F CLRW X 0003 B600 F LD A,buffer_recepcion_trama + 01H 0005 97 LD XL,A 0006 1F01 F LDW (001H,SP),X ; [ codigo_funcion_recibido ] ; SOURCE LINE # 39 0008 1E01 F LDW X,(001H,SP) ; [ codigo_funcion_recibido ] 000A A30000 CPW X,#00000H 000D 2755 JREQ ?CASE_0108 000F A30001 CPW X,#00001H 0012 2735 JREQ ?CASE_0109 0014 A30002 CPW X,#00002H 0017 2767 JREQ ?CASE_0110 0019 A30003 CPW X,#00003H 001C 2770 JREQ ?CASE_0111 001E A30004 CPW X,#00004H 0021 2603 JRNE ?LAB_0171 0023 CC0000 F JP ?CASE_0112 0026 ?LAB_0171: 0026 A30005 CPW X,#00005H 0029 277F JREQ ?CASE_0113 002B A30006 CPW X,#00006H 002E 277C JREQ ?CASE_0114 0030 A30007 CPW X,#00007H 0033 2603 JRNE ?LAB_0172 0035 CC0000 F JP ?CASE_0115 0038 ?LAB_0172: 0038 A30008 CPW X,#00008H 003B 2603 JRNE ?LAB_0173 003D CC0000 F JP ?CASE_0116 0040 ?LAB_0173: 0040 A3000A CPW X,#0000AH 0043 2602 JRNE ?LAB_0161 0045 207D JRA ?CASE_0117 0047 ?LAB_0161: 0047 207D JRA ?NXT_0604 0049 ?CASE_0109:I don´t understand a lot of assembler but it seems to me that this sentences:
are wrong because at first A is compared with 0, secondly with 1 but it's decremented so in the third comparison it will be CP A,#001H or also DEC A, it isn´t?
I have a lot of enum/switch in the project but this strange thing only happens here and in other with the same switch.
tell me what you think please,
regards,
Alberto.
Hi Alberto!
I was able to reproduce your issue and indeed, it is a peephole bug.
Unfortunately, we are not able to provide a fix in the next weeks.
The workaround is to disable the peephole optimization, at least for that function. But a better
way will be to use char for enums (if possible). As there is a bug in the STVD interface (compiler
always uses int for enums), you must set it by yourself with the command-line option/pragma
ENUMTYPE(CHAR)
Stéphane
PS: enumtype(char) is a good choice anyhow, as generated code is (usually) smaller.
Ok thank you!!
I´m working with STVD, then I have the two bugs... I'll try ENUMTYPE(CHAR), thank you!!
regards
Alberto