ncnn模型在预处理copy_make_border时出错，使用NCNN_ARM82预编译指令修改无效。

[jdskddd]

nihui，你好：
我在使用ncnn并将其部署到android过程中时，发现其编译正常，但在运行到copy_make_border时，会报错，报错时堆栈信息为：

此时其具体的堆栈信息为：

ncnn::copy_make_border(ncnn::Mat const&, ncnn::Mat&, int, int, int, int, int, float, ncnn::Option const&):
sub sp, sp, #0x90
str d8, [sp, #0x20]
stp x29, x30, [sp, #0x30]
add x29, sp, #0x30
stp x28, x27, [sp, #0x40]
stp x26, x25, [sp, #0x50]
stp x24, x23, [sp, #0x60]
stp x22, x21, [sp, #0x70]
stp x20, x19, [sp, #0x80]
mrs x28, TPIDR_EL0
mov x22, x0
ldr x8, [x28, #0x28]
mov w0, #0x2b
mov x19, x7
fmov s8, s0
mov w23, w6
mov w24, w5
mov w25, w4
mov w26, w3
mov w27, w2
mov x21, x1
str x8, [sp, #0x18]
bl 0x3989e8 ; ncnn::create_layer(int)
mov x20, x0
add x0, sp, #0x8
bl 0x3b039c ; ncnn::ParamDict::ParamDict()
add x0, sp, #0x8
mov w1, wzr
mov w2, w27
bl 0x3b1294 ; ncnn::ParamDict::set(int, int)
add x0, sp, #0x8
mov w1, #0x1
mov w2, w26
bl 0x3b1294 ; ncnn::ParamDict::set(int, int)
add x0, sp, #0x8
mov w1, #0x2
mov w2, w25
bl 0x3b1294 ; ncnn::ParamDict::set(int, int)
add x0, sp, #0x8
mov w1, #0x3
mov w2, w24
bl 0x3b1294 ; ncnn::ParamDict::set(int, int)
add x0, sp, #0x8
mov w1, #0x4
mov w2, w23
bl 0x3b1294 ; ncnn::ParamDict::set(int, int)
fmov s0, s8
add x0, sp, #0x8
mov w1, #0x5
bl 0x3b12ac ; ncnn::ParamDict::set(int, float)
ldr x8, [x20]
add x1, sp, #0x8
mov x0, x20
ldr x8, [x8, #0x10]
blr x8
ldr x8, [x20]
mov x0, x20
mov x1, x19
ldr x8, [x8, #0x20]
blr x8
ldr x8, [x20]
mov x0, x20
mov x1, x22
mov x2, x21
mov x3, x19
ldr x8, [x8, #0x38]
blr x8
ldr x8, [x20]
mov x0, x20
mov x1, x19
ldr x8, [x8, #0x28]
blr x8
ldr x8, [x20]
mov x0, x20
ldr x8, [x8, #0x8]
blr x8
add x0, sp, #0x8
bl 0x3b09b0 ; ncnn::ParamDict::~ParamDict()
ldr x8, [x28, #0x28]
ldr x9, [sp, #0x18]
cmp x8, x9
b.ne 0x39d5a8 ; <+364>
ldp x20, x19, [sp, #0x80]
ldp x22, x21, [sp, #0x70]
ldp x24, x23, [sp, #0x60]
ldp x26, x25, [sp, #0x50]
ldp x28, x27, [sp, #0x40]
ldp x29, x30, [sp, #0x30]
ldr d8, [sp, #0x20]
add sp, sp, #0x90
ret
bl 0x98a0e0 ; symbol stub for: __stack_chk_fail

ncnn::cast_float32_to_float16(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&):
sub sp, sp, #0x60
stp x29, x30, [sp, #0x20]
add x29, sp, #0x20
str x23, [sp, #0x30]
stp x22, x21, [sp, #0x40]
stp x20, x19, [sp, #0x50]
mrs x23, TPIDR_EL0
mov x21, x0
ldr x8, [x23, #0x28]
mov w0, #0x40
mov x19, x2
mov x20, x1
stur x8, [x29, #-0x8]
bl 0x3989e8 ; ncnn::create_layer(int)
mov x22, x0
add x0, sp, #0x8
bl 0x3b039c ; ncnn::ParamDict::ParamDict()
add x0, sp, #0x8
mov w1, wzr
mov w2, #0x1
bl 0x3b1294 ; ncnn::ParamDict::set(int, int)
add x0, sp, #0x8
mov w1, #0x1
mov w2, #0x2
bl 0x3b1294 ; ncnn::ParamDict::set(int, int)
ldr x8, [x22]
add x1, sp, #0x8
mov x0, x22
ldr x8, [x8, #0x10]
blr x8
ldr x8, [x22]
mov x0, x22
mov x1, x19
ldr x8, [x8, #0x20]
blr x8
ldr x8, [x22]
mov x0, x22
mov x1, x21
mov x2, x20
mov x3, x19
ldr x8, [x8, #0x38]
blr x8
ldr x8, [x22]
mov x0, x22
mov x1, x19
ldr x8, [x8, #0x28]
blr x8
ldr x8, [x22]
mov x0, x22
ldr x8, [x8, #0x8]
blr x8
add x0, sp, #0x8
bl 0x3b09b0 ; ncnn::ParamDict::~ParamDict()
ldr x8, [x23, #0x28]
ldur x9, [x29, #-0x8]
cmp x8, x9
b.ne 0x39e1e0 ; <+252>
ldp x20, x19, [sp, #0x50]
ldp x22, x21, [sp, #0x40]
ldp x29, x30, [sp, #0x20]
ldr x23, [sp, #0x30]
add sp, sp, #0x60
ret
bl 0x98a0e0 ; symbol stub for: __stack_chk_fail

ncnn::Padding_arm::create_pipeline(ncnn::Option const&):
stp x29, x30, [sp, #-0x20]!
stp x20, x19, [sp, #0x10]
mov x29, sp
ldr x8, [x0]
mov x20, x0
ldrb w10, [x1, #0x42]
mov x19, x1
ldur x8, [x8, #-0x18]
add x8, x0, x8
ldrb w9, [x8, #0xd]
cmp w9, #0x0
ccmp w10, #0x0, #0x4, ne
b.eq 0x6554d8 ; <+92>
ldr s0, [x8, #0xdc]
bl 0x39d3cc ; ncnn::float32_to_float16(float)
ldr x8, [x20]
strh w0, [x20, #0x58]
add x1, x20, #0x60
mov x2, x19
ldur x8, [x8, #-0x18]
add x8, x20, x8
add x0, x8, #0xf0
bl 0x39e0e4 ; ncnn::cast_float32_to_float16(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&)
ldrb w8, [x19, #0x40]
cbz w8, 0x65550c ; <+144>
ldr x8, [x20]
add x1, x20, #0x10
mov x2, x19
ldur x9, [x8, #-0x18]
add x9, x20, x9
ldrh w9, [x9, #0xde]
strh w9, [x20, #0x8]
ldur x8, [x8, #-0x18]
add x8, x20, x8
add x0, x8, #0xf0
bl 0x39e3e4 ; ncnn::cast_float32_to_bfloat16(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&)
ldp x20, x19, [sp, #0x10]
mov w0, wzr
ldp x29, x30, [sp], #0x20
ret

ncnn::Cast_arm::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const:
sub sp, sp, #0x80
stp x29, x30, [sp, #0x20]
add x29, sp, #0x20
stp x28, x27, [sp, #0x30]
stp x26, x25, [sp, #0x40]
stp x24, x23, [sp, #0x50]
stp x22, x21, [sp, #0x60]
stp x20, x19, [sp, #0x70]
mrs x22, TPIDR_EL0
mov x23, x0
adrp x0, 719
ldr x8, [x22, #0x28]
add x0, x0, #0xcc0
mov x25, x3
mov x19, x2
mov x20, x1
stur x8, [x29, #-0x8]
bl 0x98ae20 ; symbol stub for: __kmpc_global_thread_num
ldr x8, [x23]
ldur x8, [x8, #-0x18]
add x8, x23, x8
ldp w9, w10, [x8, #0xc8]
cmp w9, w10
b.ne 0x6e53f4 ; <+184>
cmp x19, x20
b.eq 0x6e58e0 ; <+1444>
ldr x8, [x20, #0x8]
cbz x8, 0x6e53bc ; <+128>
ldaxr w9, [x8]
add w9, w9, #0x1
stlxr w10, w9, [x8]
cbnz w10, 0x6e53ac ; <+112>
ldr x8, [x19, #0x8]
cbz x8, 0x6e57b4 ; <+1144>
ldaxr w9, [x8]
subs w9, w9, #0x1
stlxr w10, w9, [x8]
cbnz w10, 0x6e53c4 ; <+136>
b.ne 0x6e57b4 ; <+1144>
ldr x0, [x19, #0x20]
ldr x1, [x19]
cbz x0, 0x6e57a8 ; <+1132>
ldr x8, [x0]
ldr x8, [x8, #0x18]
blr x8
b 0x6e57b4 ; <+1144>
ldp w24, w11, [x20, #0x30]
ldp w21, w26, [x20, #0x28]
ldr w28, [x20, #0x38]
sub w10, w10, #0x1
ldrsw x27, [x20, #0x18]
cmp w10, #0x3
stp w0, w11, [sp]
str w28, [sp, #0xc]
b.hi 0x6e5498 ; <+348>
adrp x11, -1060
add x11, x11, #0xc94
adr x12, #0x10 ; <+244>
ldrb w13, [x11, x10]
add x12, x12, x13, lsl #2
br x12
lsl w8, w27, #1
sxtw x8, w8
sub w9, w21, #0x1
mov x21, x25
cmp w9, #0x3
b.ls 0x6e54ac ; <+368>
b 0x6e5540 ; <+516>
cmp w9, #0x3
b.ne 0x6e5468 ; <+300>
mov x0, x8
mov x1, x20
mov x2, x19
mov x3, x25
bl 0x6e45b8 ; ncnn::Cast::forward(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&) const
lsl x8, x27, #2
sub w9, w21, #0x1
mov x21, x25
cmp w9, #0x3
b.ls 0x6e54ac ; <+368>
b 0x6e5540 ; <+516>
mov x8, x27
sub w9, w21, #0x1
mov x21, x25
cmp w9, #0x3
b.ls 0x6e54ac ; <+368>
b 0x6e5540 ; <+516>
ldr x8, [x20, #0x10]
sub w9, w21, #0x1
mov x21, x25
cmp w9, #0x3
b.hi 0x6e5540 ; <+516>
adrp x10, -1060
add x10, x10, #0xc98
adr x11, #0x10 ; <+392>
ldrb w12, [x10, x9]
add x11, x11, x12, lsl #2
br x11
ldr x4, [x21, #0x8]
mov x0, x19
mov w1, w26
mov x2, x8
mov w3, w27
bl 0x398d10 ; ncnn::Mat::create(int, unsigned long, int, ncnn::Allocator*)
ldr x8, [x19]
mov w0, #-0x64
cbnz x8, 0x6e554c ; <+528>
b 0x6e58e4 ; <+1448>
ldr x6, [x21, #0x8]
mov x0, x19
mov w1, w26
mov w2, w24
mov w3, w28
mov x4, x8
mov w5, w27
bl 0x399024 ; ncnn::Mat::create(int, int, int, unsigned long, int, ncnn::Allocator*)
ldr x8, [x19]
mov w0, #-0x64
cbnz x8, 0x6e554c ; <+528>
b 0x6e58e4 ; <+1448>
ldr x7, [x21, #0x8]
mov x0, x19
mov w1, w26
mov w2, w24
ldr w3, [sp, #0x4]
mov w4, w28
mov x5, x8
mov w6, w27
bl 0x3991e8 ; ncnn::Mat::create(int, int, int, int, unsigned long, int, ncnn::Allocator*)
ldr x8, [x19]
mov w0, #-0x64
cbz x8, 0x6e58e4 ; <+1448>
ldr x8, [x19, #0x40]
ldrsw x9, [x19, #0x38]
mul x8, x8, x9
cbz x8, 0x6e58e4 ; <+1448>
ldr x8, [x23]
mul w9, w24, w26
ldr w10, [sp, #0x4]
ldur x8, [x8, #-0x18]
mul w9, w9, w10
add x10, x23, x8
mul w11, w9, w27
ldr w9, [x10, #0xc8]
str w11, [sp, #0x8]
cmp w9, #0x1
b.ne 0x6e5620 ; <+740>
ldr w9, [x10, #0xcc]
ldr w25, [sp]
cmp w9, #0x2
b.ne 0x6e5714 ; <+984>
ldp w8, w9, [x20, #0x2c]
ldr w11, [x20, #0x18]
adrp x24, 719
add x24, x24, #0xcc0
ldr w2, [x21, #0x4]
mov x0, x24
mov w1, w25
mul w8, w9, w8
ldp w9, w10, [x20, #0x34]
mul w8, w8, w9
stur w10, [x29, #-0xc]
mul w8, w8, w11
str w8, [sp, #0x10]
bl 0x98ae40 ; symbol stub for: __kmpc_push_num_threads
adrp x2, 0
sub x3, x29, #0xc
add x2, x2, #0xc98
add x6, sp, #0x10
mov x0, x24
mov w1, #0x4
mov x4, x20
mov x5, x19
bl 0x98ae50 ; symbol stub for: __kmpc_fork_call
ldr x8, [x23]
ldur x8, [x8, #-0x18]
add x9, x23, x8
ldr w9, [x9, #0xc8]
cmp w9, #0x2
b.eq 0x6e562c ; <+752>
cmp w9, #0x3
b.eq 0x6e56b0 ; <+884>
cmp w9, #0x1
b.eq 0x6e5714 ; <+984>
b 0x6e573c ; <+1024>
ldr w25, [sp]
cmp w9, #0x2
b.ne 0x6e560c ; <+720>
add x8, x23, x8
ldr w8, [x8, #0xcc]
cmp w8, #0x1
b.ne 0x6e573c ; <+1024>
ldp w8, w9, [x20, #0x2c]
ldr w11, [x20, #0x18]
adrp x24, 719
add x24, x24, #0xcc0
ldr w2, [x21, #0x4]
mov x0, x24
mov w1, w25
mul w8, w9, w8
ldp w9, w10, [x20, #0x34]
mul w8, w8, w9
stur w10, [x29, #-0xc]
mul w8, w8, w11
str w8, [sp, #0x10]
bl 0x98ae40 ; symbol stub for: __kmpc_push_num_threads
adrp x2, 0
sub x3, x29, #0xc
add x2, x2, #0xe70
add x6, sp, #0x10
mov x0, x24
mov w1, #0x4
mov x4, x20
mov x5, x19
bl 0x98ae50 ; symbol stub for: __kmpc_fork_call
ldr x8, [x23]
ldur x8, [x8, #-0x18]
add x9, x23, x8
ldr w9, [x9, #0xc8]
cmp w9, #0x3
b.ne 0x6e5614 ; <+728>
add x8, x23, x8
ldr w8, [x8, #0xcc]
cmp w8, #0x1
b.ne 0x6e573c ; <+1024>
adrp x24, 719
ldr w2, [x21, #0x4]
add x24, x24, #0xcc0
mov w1, w25
mov x0, x24
bl 0x98ae40 ; symbol stub for: __kmpc_push_num_threads
adrp x2, 0
add x3, sp, #0xc
add x2, x2, #0x9f8
add x6, sp, #0x8
mov x0, x24
mov w1, #0x4
mov x4, x20
mov x5, x19
bl 0x98ae50 ; symbol stub for: __kmpc_fork_call
ldr x8, [x23]
ldur x8, [x8, #-0x18]
add x9, x23, x8
ldr w9, [x9, #0xc8]
cmp w9, #0x1
b.ne 0x6e573c ; <+1024>
add x8, x23, x8
ldr w8, [x8, #0xcc]
cmp w8, #0x4
b.ne 0x6e573c ; <+1024>
bl 0x3a10cc ; ncnn::cpu_support_arm_bf16()
cbz w0, 0x6e586c ; <+1328>
mov x0, x20
mov x1, x19
mov x2, x21
bl 0x6e7ad4 ; ncnn::cast_fp32_to_bf16_neon_bf16(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&)
ldr x8, [x23]
ldur x8, [x8, #-0x18]
add x8, x23, x8
ldr w9, [x8, #0xc8]
cmp w9, #0x4
b.ne 0x6e58e0 ; <+1444>
ldr w8, [x8, #0xcc]
cmp w8, #0x1
b.ne 0x6e58e0 ; <+1444>
bl 0x3a10cc ; ncnn::cpu_support_arm_bf16()
cbz w0, 0x6e580c ; <+1232>
mov x0, x20
mov x1, x19
mov x2, x21
bl 0x6e7d6c ; ncnn::cast_bf16_to_fp32_neon_bf16(ncnn::Mat const&, ncnn::Mat&, ncnn::Option const&)
b 0x6e58e0 ; <+1444>
ldr x5, [x21, #0x8]
mov x0, x19
mov w1, w26
mov w2, w24
mov x3, x8
mov w4, w27
bl 0x398e8c ; ncnn::Mat::create(int, int, unsigned long, int, ncnn::Allocator*)
ldr x8, [x19]
mov w0, #-0x64
cbnz x8, 0x6e554c ; <+528>
b 0x6e58e4 ; <+1448>
cbz x1, 0x6e57b4 ; <+1144>
mov x0, x1
bl 0x98a120 ; symbol stub for: free
str xzr, [x19, #0x40]
mov w0, wzr
stp xzr, xzr, [x19, #0x8]
str xzr, [x19]
str wzr, [x19, #0x18]
str wzr, [x19, #0x38]
stp xzr, xzr, [x19, #0x28]
ldr q0, [x20]
str q0, [x19]
ldr x8, [x20, #0x10]
str x8, [x19, #0x10]
ldr w8, [x20, #0x18]
str w8, [x19, #0x18]
ldr x8, [x20, #0x20]
str x8, [x19, #0x20]
ldur q0, [x20, #0x28]
stur q0, [x19, #0x28]
ldr w8, [x20, #0x38]
str w8, [x19, #0x38]
ldr x8, [x20, #0x40]
str x8, [x19, #0x40]
b 0x6e58e4 ; <+1448>
ldp w8, w9, [x20, #0x2c]
ldr w11, [x20, #0x18]
mov w1, w25
ldr w2, [x21, #0x4]
adrp x21, 719
add x21, x21, #0xcc0
mul w8, w9, w8
mov x0, x21
ldp w9, w10, [x20, #0x34]
mul w8, w8, w9
stur w10, [x29, #-0xc]
mul w8, w8, w11
str w8, [sp, #0x10]
bl 0x98ae40 ; symbol stub for: __kmpc_push_num_threads
adrp x2, 1
sub x3, x29, #0xc
add x2, x2, #0x298
add x6, sp, #0x10
mov x0, x21
mov w1, #0x4
mov x4, x20
mov x5, x19
bl 0x98ae50 ; symbol stub for: __kmpc_fork_call
b 0x6e58e0 ; <+1444>
ldp w8, w9, [x20, #0x2c]
ldr w11, [x20, #0x18]
adrp x24, 719
add x24, x24, #0xcc0
ldr w2, [x21, #0x4]
mov x0, x24
mov w1, w25
mul w8, w9, w8
ldp w9, w10, [x20, #0x34]
mul w8, w8, w9
stur w10, [x29, #-0xc]
mul w8, w8, w11
str w8, [sp, #0x10]
bl 0x98ae40 ; symbol stub for: __kmpc_push_num_threads
adrp x2, 1
sub x3, x29, #0xc
add x2, x2, #0x48
add x6, sp, #0x10
mov x0, x24
mov w1, #0x4
mov x4, x20
mov x5, x19
bl 0x98ae50 ; symbol stub for: __kmpc_fork_call
ldr x8, [x23]
ldur x8, [x8, #-0x18]
add x8, x23, x8
ldr w9, [x8, #0xc8]
cmp w9, #0x4
b.eq 0x6e5754 ; <+1048>
mov w0, wzr
ldr x8, [x22, #0x28]
ldur x9, [x29, #-0x8]
cmp x8, x9
b.ne 0x6e5914 ; <+1496>
ldp x20, x19, [sp, #0x70]
ldp x22, x21, [sp, #0x60]
ldp x24, x23, [sp, #0x50]
ldp x26, x25, [sp, #0x40]
ldp x28, x27, [sp, #0x30]
ldp x29, x30, [sp, #0x20]
add sp, sp, #0x80
ret
bl 0x98a0e0 ; symbol stub for: __stack_chk_fail

__kmp_env_get:
sub sp, sp, #0x50
stp x29, x30, [sp, #0x30]
stp x20, x19, [sp, #0x40]
add x29, sp, #0x30
bl 0x9904b0 ; symbol stub for: getenv
cbz x0, 0x92f324 ; <+60>
mov x19, x0
bl 0x98aa20 ; symbol stub for: strlen
add x20, x0, #0x1
mov x0, x20
bl 0x98a810 ; symbol stub for: malloc
cbz x0, 0x92f334 ; <+76>
mov x1, x19
mov x2, x20
bl 0x9907d0 ; symbol stub for: strncpy
ldp x20, x19, [sp, #0x40]
ldp x29, x30, [sp, #0x30]
add sp, sp, #0x50
ret
mov w0, #0x6e
add x8, sp, #0x18
movk w0, #0x4, lsl #16
bl 0x990640 ; symbol stub for: __kmp_msg_format
adrp x8, 151
ldr x8, [x8, #0xcd8]
add x0, sp, #0x18
mov x1, sp
ldr q0, [x8]
ldr x8, [x8, #0x10]
str q0, [sp]
str x8, [sp, #0x10]
bl 0x990890 ; symbol stub for: __kmp_fatal

__kmp_get_global_thread_id_reg:
stp x29, x30, [sp, #-0x20]!
stp x20, x19, [sp, #0x10]
adrp x20, 151
ldr x20, [x20, #0xd30]
mov x29, sp
ldr w8, [x20]
cbz w8, 0x92fd18 ; <+132>
adrp x8, 151
ldr x8, [x8, #0xe30]
ldr w8, [x8]
cmp w8, #0x3
b.lt 0x92fcec ; <+88>
adrp x0, 151
ldr x0, [x0, #0xe38]
bl 0x989e58 ; __emutls_get_address at emutls.c:173
ldr w19, [x0]
cmn w19, #0x2
b.eq 0x92fd18 ; <+132>
mov w0, w19
ldp x20, x19, [sp, #0x10]
ldp x29, x30, [sp], #0x20
ret
cmp w8, #0x2
b.ne 0x92fd08 ; <+116>
bl 0x990ae0 ; symbol stub for: __kmp_gtid_get_specific
mov w19, w0
cmn w19, #0x2
b.ne 0x92fcdc ; <+72>
b 0x92fd18 ; <+132>
bl 0x9907e0 ; symbol stub for: __kmp_get_global_thread_id
mov w19, w0
cmn w19, #0x2
b.ne 0x92fcdc ; <+72>
adrp x0, 151
ldr x0, [x0, #0xe50]
mov w1, #-0x2
bl 0x990970 ; symbol stub for: __kmp_acquire_ticket_lock
ldr w8, [x20]
cbz w8, 0x92fd3c ; <+168>
mov w0, wzr
bl 0x990b00 ; symbol stub for: __kmp_register_root
b 0x92fd44 ; <+176>
bl 0x92fd68 ; __kmp_do_serial_initialize()
bl 0x990ae0 ; symbol stub for: __kmp_gtid_get_specific
mov w19, w0
adrp x0, 151
ldr x0, [x0, #0xe50]
mov w1, #-0x2
bl 0x990980 ; symbol stub for: __kmp_release_ticket_lock
mov w0, w19
ldp x20, x19, [sp, #0x10]
ldp x29, x30, [sp], #0x20
ret

__kmp_do_serial_initialize():
sub sp, sp, #0x80
stp x29, x30, [sp, #0x50]
str x21, [sp, #0x60]
stp x20, x19, [sp, #0x70]
add x29, sp, #0x50
bl 0x990b10 ; symbol stub for: ompt_pre_init
bl 0x990b20 ; symbol stub for: __kmp_validate_locks
bl 0x990b30 ; symbol stub for: __kmp_init_memkind
bl 0x990b40 ; symbol stub for: __kmp_register_library_startup()
adrp x19, 151
ldr x19, [x19, #0xe58]
adrp x0, 151
ldr wzr, [x19, #0x44]
str wzr, [x19, #0x40]
str wzr, [x19, #0x44]
ldr x0, [x0, #0xe60]
bl 0x990b50 ; symbol stub for: __kmp_init_ticket_lock
adrp x0, 151
ldr x0, [x0, #0xe68]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xe70]
bl 0x990b50 ; symbol stub for: __kmp_init_ticket_lock
adrp x0, 151
ldr x0, [x0, #0xe78]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xe80]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xe88]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xe90]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xe98]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xea0]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xea8]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xeb0]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xeb8]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xec0]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xec8]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xed0]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xed8]
bl 0x990b60 ; symbol stub for: __kmp_init_queuing_lock
adrp x0, 151
ldr x0, [x0, #0xee0]
bl 0x990b50 ; symbol stub for: __kmp_init_ticket_lock
adrp x0, 151
ldr x0, [x0, #0xee8]
bl 0x990b50 ; symbol stub for: __kmp_init_ticket_lock
adrp x0, 151
ldr x0, [x0, #0xef0]
bl 0x990b50 ; symbol stub for: __kmp_init_ticket_lock
bl 0x990b70 ; symbol stub for: __kmp_runtime_initialize
adrp x9, 151
adrp x8, 151
adrp x10, 151
adrp x20, 151
adrp x21, 151
adrp x11, 151
adrp x12, 151
adrp x13, 151
ldr x9, [x9, #0xef8]
ldr x8, [x8, #0xd88]
ldr x10, [x10, #0xf00]
ldr x20, [x20, #0xf08]
ldr x21, [x21, #0xf10]
ldr x11, [x11, #0xf18]
ldr x12, [x12, #0xf20]
ldr x13, [x13, #0xf28]
adrp x15, 151
ldr x15, [x15, #0xf30]
mov w14, #0xc8
str w14, [x13]
mov w13, #0x3
adrp x14, 151
ldr x14, [x14, #0xf38]
str w13, [x15]
adrp x15, 151
ldr x15, [x15, #0xf40]
mov w13, #0x29
str w13, [x14]
adrp x13, 151
adrp x14, 151
ldr w15, [x15]
ldr x13, [x13, #0xf48]
ldr x14, [x14, #0xf50]
mov x0, xzr
stp w15, w15, [x14]
mov w15, #0x1
str w15, [x14, #0x8]
adrp x14, 151
ldr x14, [x14, #0xf58]
ldr w13, [x13]
str w15, [x14]
adrp x14, 151
ldr x14, [x14, #0xf60]
str w15, [x14, #0x8]
adrp x15, 151
ldr x15, [x15, #0xf68]
stp w13, w13, [x14]
adrp x13, 151
adrp x14, 151
ldr x13, [x13, #0xf70]
ldr x14, [x14, #0xf78]
str wzr, [x9]
mov w9, #0x2
str w9, [x13, #0x8]
str w9, [x14, #0x8]
adrp x9, 151
ldr x9, [x9, #0xf80]
str wzr, [x9]
adrp x9, 151
ldr x9, [x9, #0xf88]
ldr w8, [x8]
ldr w10, [x10]
ldr w15, [x15]
ldr w9, [x9]
cmp w8, #0x1
str w10, [x21]
str w10, [x11]
stp w9, w9, [x14]
csinc w9, w8, wzr, gt
cmp w9, w10
csel w9, w10, w9, gt
cmp w8, w10
csel w8, w10, w8, gt
stp w15, w15, [x13]
str w9, [x20]
str w8, [x12]
str xzr, [x19, #0x48]
bl 0x990b80 ; symbol stub for: __kmp_env_initialize(char const*)
ldr w0, [x20]
bl 0x990b90 ; symbol stub for: __kmp_initial_threads_capacity(int)
adrp x9, 151
ldr w8, [x20]
ldr w1, [x21]
ldr x9, [x9, #0xf90]
adrp x19, 151
ldr w2, [x9]
ldr x19, [x19, #0xe40]
str w0, [x19]
mov w0, w8
bl 0x990ba0 ; symbol stub for: __kmp_default_tp_capacity(int, int, int)
adrp x8, 151
adrp x9, 151
adrp x10, 151
ldr x8, [x8, #0xf98]
ldr x9, [x9, #0xfa0]
b 0x989fec ; __CortexA53843419_930004
ldr w11, [x19]
str w0, [x8]
str xzr, [x9]
str xzr, [x10]
adrp x8, 150
ldr x8, [x8, #0xfb0]
str xzr, [x8]
lsl w8, w11, #4
add w8, w8, #0x40
sxtw x0, w8
bl 0x9909b0 ; symbol stub for: ___kmp_allocate
adrp x20, 150
ldr x20, [x20, #0xc90]
ldrsw x8, [x19]
adrp x21, 150
adrp x9, 150
adrp x10, 150
ldr x21, [x21, #0xfb8]
ldr x9, [x9, #0xc88]
ldr x10, [x10, #0xd78]
str x0, [x20]
add x8, x0, x8, lsl #3
mov w0, #0x1
str x8, [x21]
str wzr, [x9]
str wzr, [x10]
bl 0x990b00 ; symbol stub for: __kmp_register_root
mov w19, w0
tbnz w0, #0x1f, 0x9300a4 ; <+828>
ldr x8, [x21]
ldr x8, [x8, w19, uxtw #3]
cbz x8, 0x9300a4 ; <+828>
ldr x9, [x20]
mov w10, w19
ldr x9, [x9, x10, lsl #3]
cbz x9, 0x9300a4 ; <+828>
ldr x8, [x8, #0x18]
cmp x9, x8
b.eq 0x9300bc ; <+852>
adrp x0, -2116
adrp x1, -2085
add x0, x0, #0xa1a
add x1, x1, #0x988
mov w2, #0x1a26
bl 0x9905e0 ; symbol stub for: __kmp_debug_assert
cbz w19, 0x9300d8 ; <+880>
adrp x0, -2116
adrp x1, -2085
add x0, x0, #0xa1a
add x1, x1, #0x988
mov w2, #0x1a27
bl 0x9905e0 ; symbol stub for: __kmp_debug_assert
dmb ish
bl 0x990bb0 ; symbol stub for: __kmp_common_initialize
bl 0x990bc0 ; symbol stub for: __kmp_register_atfork
adrp x0, 150
ldr x0, [x0, #0xfc0]
bl 0x34ebf0 ; atexit
cbnz w0, 0x930170 ; <+1032>
bl 0x990bd0 ; symbol stub for: __kmp_install_signals
adrp x8, 150
ldr x8, [x8, #0xfc8]
adrp x10, 150
mov w11, #0x1
ldr w9, [x8]
ldr x10, [x10, #0xd30]
add w9, w9, #0x1
str w9, [x8]
str w11, [x10]
adrp x8, 150
ldr x8, [x8, #0xfd0]
ldr w8, [x8]
cbz w8, 0x930130 ; <+968>
bl 0x990be0 ; symbol stub for: __kmp_env_print()
adrp x8, 150
adrp x9, 150
ldr x8, [x8, #0xfd8]
ldr x9, [x9, #0xfe0]
ldr w8, [x8]
ldr w9, [x9]
orr w8, w9, w8
cbz w8, 0x930154 ; <+1004>
bl 0x990bf0 ; symbol stub for: __kmp_env_print_2()
bl 0x990c00 ; symbol stub for: ompt_post_init
dmb ish
ldp x20, x19, [sp, #0x70]
ldr x21, [sp, #0x60]
ldp x29, x30, [sp, #0x50]
add sp, sp, #0x80
ret
mov w19, w0
adrp x1, -2079
mov w0, #0xb2
add x1, x1, #0xa16
sub x8, x29, #0x18
movk w0, #0x4, lsl #16
bl 0x990640 ; symbol stub for: __kmp_msg_format
add x8, sp, #0x20
mov w0, w19
bl 0x990a10 ; symbol stub for: __kmp_msg_error_code
adrp x8, 150
ldr x8, [x8, #0xcd8]
sub x0, x29, #0x18
add x1, sp, #0x20
mov x2, sp
ldr q0, [x8]
ldr x8, [x8, #0x10]
str q0, [sp]
str x8, [sp, #0x10]
bl 0x990890 ; symbol stub for: __kmp_fatal

__kmp_abort_thread:
stp x29, x30, [sp, #-0x10]!
mov x29, sp
bl 0x990c90 ; symbol stub for: __kmp_infinite_loop

__kmp_serial_initialize:
stp x29, x30, [sp, #-0x20]!
str x19, [sp, #0x10]
adrp x19, 143
ldr x19, [x19, #0xd30]
mov x29, sp
ldr w8, [x19]
cbz w8, 0x937444 ; <+40>
ldr x19, [sp, #0x10]
ldp x29, x30, [sp], #0x20
ret
adrp x0, 143
ldr x0, [x0, #0xe50]
mov w1, #-0x2
bl 0x990970 ; symbol stub for: __kmp_acquire_ticket_lock
ldr w8, [x19]
cbnz w8, 0x937460 ; <+68>
bl 0x92fd68 ; __kmp_do_serial_initialize()
adrp x0, 143
ldr x0, [x0, #0xe50]
ldr x19, [sp, #0x10]
mov w1, #-0x2
ldp x29, x30, [sp], #0x20
b 0x990980 ; symbol stub for: __kmp_release_ticket_lock

abort:
sub sp, sp, #0xd0
str x21, [sp, #0xa0]
stp x20, x19, [sp, #0xb0]
stp x29, x30, [sp, #0xc0]
add x29, sp, #0xc0
sub x0, x29, #0x18
bl 0xe4e50 ; symbol stub for: sigfillset64
sub x0, x29, #0x18
orr w1, wzr, #0x6
bl 0xe4a10 ; symbol stub for: sigdelset64
sub x1, x29, #0x18
orr w0, wzr, #0x2
mov x2, xzr
bl 0xe4e60 ; symbol stub for: sigprocmask64
mov w0, #0xac
bl 0xe4bb0 ; symbol stub for: syscall
mov x19, x0
mov w0, #0xb2
bl 0xe4bb0 ; symbol stub for: syscall
mov x20, x0
movi v0.2d, #0000000000000000
mov w21, #-0x1
stp q0, q0, [sp, #0x20]
stp q0, q0, [sp, #0x80]
stp q0, q0, [sp, #0x60]
stp q0, q0, [sp, #0x40]
str w21, [sp, #0x28]
str w19, [sp, #0x30]
bl 0xe4a90 ; symbol stub for: getuid
mov w8, w0
str xzr, [sp, #0x38]
sxtw x0, w19
sxtw x1, w20
add x3, sp, #0x20
adrp x9, -93
add x9, x9, #0x4e0
ldp q1, q0, [x9]
str w8, [sp, #0x34]
orr w2, wzr, #0x6
orr w8, wzr, #0xf0
svc #0
stp q1, q0, [sp]
mov x1, sp
orr w0, wzr, #0x6
mov x2, xzr
bl 0xe4e70 ; symbol stub for: sigaction64
sub x1, x29, #0x18
orr w0, wzr, #0x2
mov x2, xzr
bl 0xe4e60 ; symbol stub for: sigprocmask64
mov w0, #0xac
bl 0xe4bb0 ; symbol stub for: syscall
mov x19, x0
mov w0, #0xb2
bl 0xe4bb0 ; symbol stub for: syscall
mov x20, x0
movi v0.2d, #0000000000000000
stp q0, q0, [sp, #0x20]
stp q0, q0, [sp, #0x80]
stp q0, q0, [sp, #0x60]
stp q0, q0, [sp, #0x40]
str w21, [sp, #0x28]
str w19, [sp, #0x30]
bl 0xe4a90 ; symbol stub for: getuid
mov w8, w0
str xzr, [sp, #0x38]
sxtw x0, w19
sxtw x1, w20
add x3, sp, #0x20
str w8, [sp, #0x34]
orr w2, wzr, #0x6
orr w8, wzr, #0xf0
svc #0
orr w0, wzr, #0x7f
bl 0xe4e80 ; symbol stub for: _exit

我检查后认为，其可能是因为不支持arm82，故在编译时在cmakelist中增加了：
option(NCNN_ARM82 "optimize aarch64 platform with armv8.2 fp16" OFF)
此时编译语句为：
cd ncnn
mkdir -p build-android-armv7
cd build-android-armv7
cmake -DCMAKE_TOOLCHAIN_FILE="$ANDROID_NDK/build/cmake/android.toolchain.cmake" -DANDROID_ABI="armeabi-v7a" -DANDROID_ARM_NEON=ON -DANDROID_PLATFORM=android-28 -DNCNN_VULKAN=ON -DNCNN_ARM82=OFF ..
make -j16
make install

重新编译后，将其导入android项目，报错任然没有任何变化，为何我修改了ARM82编译命令后，其还是会进入cast_float32_to_float16这个函数？为何我的修改没有生效？？