When I write custom assembly code using labels, the disassembly is really difficult to read since it starts a new block for each label, for example this code:
lui t2, %hi(_edata)
addi t2, t2, %lo(_edata)
.copy_loop:
bgeu t1, t2, .copy_end
lw t3, 0(t0)
sw t3, 0(t1)
addi t0, t0, 4
addi t1, t1, 4
j .copy_loop
.copy_end:
Turns into the following mess in objdump:
168: 00030313 mv t1,t1
16c: 000033b7 lui t2,0x3
170: 01838393 addi t2,t2,24 # 3018 <_edata>
00000174 <.copy_loop>:
174: /-> 00737c63 bgeu t1,t2,18c <.copy_end>
178: | 0002ae03 lw t3,0(t0)
17c: | 01c32023 sw t3,0(t1)
180: | 00428293 addi t0,t0,4
184: | 00430313 addi t1,t1,4
188: \-- fedff06f j 174 <.copy_loop>
0000018c <.copy_end>:
18c: 000032b7 lui t0,0x3
190: 02028293 addi t0,t0,32 # 3020 <_irq_reg_save>
But when I disassemble C code generated by GCC, the labels aren't "separated" like that, and the jump arrows are drawn correctly:
470: 01076713 ori a4,a4,16
474: 00e7a023 sw a4,0(a5)
478: /-- 01c0006f j 494 <lcd_puts+0x4c>
47c: /--|-> fec42783 lw a5,-20(s0)
480: | | 00178713 addi a4,a5,1
484: | | fee42623 sw a4,-20(s0)
488: | | 0007c783 lbu a5,0(a5)
48c: | | 00078513 mv a0,a5
490: | | dfdff0ef jal ra,28c <lcd_write_raw>
494: | \-> fec42783 lw a5,-20(s0)
498: | 0007c783 lbu a5,0(a5)
49c: \----- fe0790e3 bnez a5,47c <lcd_puts+0x34>
4a0: 00000013 nop
4a4: 00000013 nop
Is there a way to obtain the same result when writing custom assembly code in GNU AS? Am I using labels wrong?
