Index
> Compiler Internals > fasm is hanging |
| Author |
|
|
Tomasz Grysztar 27 Oct 2015, 12:03
CandyMan wrote:
|
|||
27 Oct 2015, 12:03 |
|
|
CandyMan 27 Oct 2015, 13:10
Code: use16 vpgatherdd xmm0{k1},[rax+xmm0+64] Code:
check_vsib:
xor ah,ah
check_vsib_base: ;*
test bh,bh
jz check_vsib_index
mov al,bh
shr al,4
cmp al,4
je check_vsib_base_size
cmp [code_type],64
jne swap_vsib_registers ;*
cmp al,8
jne swap_vsib_registers
check_vsib_base_size:
mov ah,[address_size]
and ah,0Fh
jz check_vsib_index
cmp al,ah
jne invalid_address
check_vsib_index:
mov al,bl
and al,0E0h
cmp al,0C0h
jae check_index_scale
cmp al,60h
je check_index_scale
jmp invalid_address
swap_vsib_registers:
xor al,-1
jz invalid_address
cmp cl,1
ja invalid_address
xchg bl,bh
mov cl,1
jmp check_vsib_base ;* closed loop
_________________ smaller is better |
|||
|
27 Oct 2015, 13:10 |
|
|
Tomasz Grysztar 27 Oct 2015, 13:13
No, I meant the second one of your samples. The first one is already fixed.
|
|||
|
27 Oct 2015, 13:13 |
|
|
CandyMan 27 Oct 2015, 13:46
Whether you could what combinations of registers are disallowed?
thanks... _________________ smaller is better |
|||
|
27 Oct 2015, 13:46 |
|
|
Tomasz Grysztar 27 Oct 2015, 14:12
The disallowed combinations are defined in Intel manuals this way:
Intel wrote: The instruction will #UD fault if the destination vector zmm1 is the same as index vector VINDEX. Code: vpgatherdd xmm0{k1},dword [eax+xmm0+0x40] ; error: disallowed combination of registers |
|||
|
27 Oct 2015, 14:12 |
|
< Last Thread | Next Thread > |
Forum Rules:
|