CODEWITHSUNDEEP
cgccassemblyarmconstants
Lawrencium
Lawrencium

Reputation: 13

GCC: Why is const data filled inside my function and not at the beginning?

I want to benchmark the number of cycles used by different machine instructions on my system (for this example it is ARM Cortex-M4). So i use a macro, which repeats the target instruction a number of times and before and after this, i read the cycle counter of my controller. In the asm-dump, i see that at some position the const data (address of my cycle-counter register) is filled in (position 8003140 to 8003150, marked with ">"):

    08002d58 <testThis>:
    8002d58:    48fa        ldr r0, [pc, #1000] ; (8003144 <testThis+0x3ec>)
    8002d5a:    49fb        ldr r1, [pc, #1004] ; (8003148 <testThis+0x3f0>)
    8002d5c:    4bfb        ldr r3, [pc, #1004] ; (800314c <testThis+0x3f4>)
    8002d5e:    4afc        ldr r2, [pc, #1008] ; (8003150 <testThis+0x3f8>)
    8002d60:    6800        ldr r0, [r0, #0]
    8002d62:    6008        str r0, [r1, #0]
    8002d64:    681b        ldr r3, [r3, #0]
    8002d66:    6812        ldr r2, [r2, #0]
    8002d68:    fa82 f183   qadd    r1, r3, r2
    8002d6c:    fa82 f183   qadd    r1, r3, r2
    ..
    8003138:    fa82 f183   qadd    r1, r3, r2
    800313c:    fa82 f183   qadd    r1, r3, r2
    > 8003140:  e008        b.n 8003154 <testThis+0x3fc>
    > 8003142:  bf00        nop
    > 8003144:  e0001004    .word   0xe0001004
    > 8003148:  20000598    .word   0x20000598
    > 800314c:  20000594    .word   0x20000594
    > 8003150:  200002e4    .word   0x200002e4
    8003154:    fa82 f183   qadd    r1, r3, r2
    8003158:    fa82 f183   qadd    r1, r3, r2
    ..
    8003b84:    fa82 f183   qadd    r1, r3, r2
    8003b88:    fa82 f383   qadd    r3, r3, r2
    8003b8c:    4803        ldr r0, [pc, #12]   ; (8003b9c <testThis+0xe44>)
    8003b8e:    4904        ldr r1, [pc, #16]   ; (8003ba0 <testThis+0xe48>)
    8003b90:    6003        str r3, [r0, #0]
    8003b92:    4b04        ldr r3, [pc, #16]   ; (8003ba4 <testThis+0xe4c>)
    8003b94:    680a        ldr r2, [r1, #0]
    8003b96:    601a        str r2, [r3, #0]
    8003b98:    4770        bx  lr
    8003b9a:    bf00        nop
    8003b9c:    2000058c    .word   0x2000058c
    8003ba0:    e0001004    .word   0xe0001004
    8003ba4:    20000338    .word   0x20000338

Why is this not filled in at the beginning? Am i able to control this?

GCC version:

gcc version 4.8.3 20140228 (release) [ARM/embedded-4_8-branch revision 208322]

C-Code:

    #define READCYCCNT() *((volatile unsigned int *)0xE0001004)

    uint32_t cyc_begin, cyc_end;
    int c, a, b;
    void testThis(void *obj)
    {
        cyc_begin = READCYCCNT();
        REP(9,0,0, c, __QADD, a, b);
        cyc_end = READCYCCNT();
    }

The REP-macro is a bit lengthy. It just adds 900 calls to 

    c = __QADD(a,b)

Compiler-call:

    arm-atollic-eabi-gcc -c -mthumb -mcpu=cortex-m4 -mfloat-abi=softfp -mfpu=fpv4-sp-d16 -std=gnu90 -DDEBUG=1 -I../Inc -I../CoreSupport -I../DeviceSupport -Ofast -ffunction-sections -fdata-sections -g -Wall -o Application\Main.o ..\Application\Main.c

Upvotes: 1

Views: 145

Answers (1)

Dmitry Grigoryev
Dmitry Grigoryev

Reputation: 3203

The compiler have generated your code using ldr instructions with addressing relative to PC. Those instructions have only 5 bits to store the relative address, so they can only access data within range 0-124 words from the current program counter position. This is why the compiler have placed your data in the middle of the code. Here's the quick reference card for thumb instructions.

There are several ways to avoid this. You could replace the macro by a hand-written assembly which uses different addressing mode. You could replace variables by constants and avoid addressing altogether. You could reduce the number of times your macro is called. You could remove the -mthumb flag to generate 32-bit instructions which have more bits for addressing. It really depends on what you want to evaluate with your test.

Upvotes: 1

Related Questions