Question: I am attempting to loop unroll some y86 code. xorq %rax,%rax # count = 0; andq %rdx,%rdx # len
I am attempting to loop unroll some y86 code.
xorq %rax,%rax # count = 0;
andq %rdx,%rdx # len <= 0?
jle Done # if so, goto Done:
Loop:
mrmovq (%rdi), %r10 # read val from src...
rmmovq %r10, (%rsi) # ...and store it to dst
andq %r10, %r10 # val <= 0?
jle Npos # if so, goto Npos:
#irmovq $1, %r10
#addq %r10, %rax
iaddq $1, %rax # count++
Npos:
irmovq $1, %r10
subq %r10, %rdx # len--
#irmovq $8, %r10
#addq %r10, %rdi
#addq %r10, %rsi
iaddq $8, %rdi # src++
iaddq $8, %rsi # dst++
andq %rdx,%rdx # len > 0?
jg Loop # if so, goto Loop:
Done:
ret
The above is the original code. And Below is the unrolled code i made.
xorq %rax,%rax # count = 0;
andq %rdx,%rdx # len <= 0?
jle Done # if so, goto Done:
iaddq $-2, %rdx
jl len_cleanup
Loop: mrmovq (%rdi), %r10 # read val from src mrmovq 8(%rdi), %r11 # <- from class get second value rmmovq %r10, (%rsi) # ...and store it to dst rmmovq %r11, 8(%rsi) # store second val to dst andq %r10, %r10 # val <= 0? jle Npos # if so, goto Npos: iaddq $1, %rax
Npos: andq %r11, %r11 # check if src[1] <= 0 jle Npos2 # if it is, dont increase count iaddq $1, %rax
Npos2: irmvoq %2, %r10 iaddq $16, %rdi # increase stack or base pointer to get next 2 vals iaddq $16, %rsi # increase stack or base pointer to store next 2 vals subq %r10, %rdx # decrease length by 2 jge Loop # go back into loop if length >= 2
len_cleanup: iaddq $2, %rdx
cleanup: irmovq $1, %r10 subq %r10, %rdx jl Done # if length < 0, jmp to Done, no cleanup needed mrmovq (%rdi), %r10 # get next val rmmovq %r10, (%rsi) # move val onto stack andq %r10, %r10 # check if val <= 0 jle Done # skip count if val < 0 iaddq $1, %rax # same as iaddq $1, %rax
Done: ret
Can someone help me unroll it further?
Step by Step Solution
There are 3 Steps involved in it
Get step-by-step solutions from verified subject matter experts
