We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.
hi: I have a code below:
char pop_value = 0X80; if(pop_value == 0X80) { pop_value = pop_value;//point 1 } pop_value = pop_value;//point 2
I would have thought the code would jump to point 1 after logical judge, but the simulating result exceed my expectations that the code jumped to point 2. I watch the Disassembly code :
82: pop_value = 0X80; C:0x0A68 7853 MOV R0,#pop_value(0x53) C:0x0A6A 7680 MOV @R0,#P0(0x80) 83: if(pop_value == 0X80) C:0x0A6C E6 MOV A,@R0 C:0x0A6D FF MOV R7,A C:0x0A6E FD MOV R5,A C:0x0A6F 33 RLC A C:0x0A70 95E0 SUBB A,ACC(0xE0) C:0x0A72 FC MOV R4,A C:0x0A73 ED MOV A,R5 C:0x0A74 6480 XRL A,#P0(0x80) C:0x0A76 4C ORL A,R4 C:0x0A77 7002 JNZ C:0A7B 84: { 85: pop_value = pop_value; 86: C:0x0A79 A607 MOV @R0,0x07 87: } C:0x0A7B 22 RET PUTCHAR:
I can not find the error happened.so I also compile this C code in another project. the code can jump to point 1 after logical judge ,nor jump to point 2.I watch the the Disassembly code at once:
61: item = 0X80; C:0x0A94 752400 MOV 0x24,#malloc_mempool(0x00) C:0x0A97 752580 MOV 0x25,#P0(0x80) 62: if( item== 0x80) C:0x0A9A E525 MOV A,0x25 C:0x0A9C 6480 XRL A,#P0(0x80) C:0x0A9E 4524 ORL A,0x24 C:0x0AA0 7006 JNZ C:0AA8 63: { 64: item =item; C:0x0AA2 852424 MOV 0x24,0x24 C:0x0AA5 852525 MOV 0x25,0x25 65: }
my keiC51 version is V8.16. I don't know why the same C code has different assemble code in different project. The important thing is that the compiler make a wrong logical judge.
Yes, but still not as smart as it should be. It should have optimized away everything except the first line. maybe not, but Keil, in their infinite wisdom, do all optimizing in the linker.
Erik
maybe not, but Keil, in their infinite wisdom, do all optimizing in the linker
Not all --- not by a rather wide margin. I'm pretty sure none of the linker-based optimazation are relevant to the example under study. This should be optimized to nothing before the linker is even run.
I still wonder why the making a switch an indirect jump instead of a combersome calculation (when the state values are 'organized') is not done in the compiler but in the linker