Make utf8_length_from_utf32 branchless. #489
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This helps GCC to autovectorize its implementation. https://compiler-explorer.com/z/61xhGE784 so instead of ``` utf8_length_from_utf32(char32_t const*, unsigned long): test rsi, rsi je .L8 lea rcx, [rdi+rsi*4] xor edx, edx jmp .L7 .L11: add rdx, 1 .L4: add rdi, 4 cmp rcx, rdi je .L1 .L7: mov eax, DWORD PTR [rdi] cmp eax, 127 jbe .L11 cmp eax, 2047 ja .L5 add rdi, 4 add rdx, 2 cmp rcx, rdi jne .L7 .L1: mov rax, rdx ret .L5: mov rsi, rdx xor edx, edx cmp eax, 65535 seta dl lea rdx, [rdx+3+rsi] jmp .L4 .L8: xor edx, edx mov rax, rdx ret ``` it generates ``` utf8_length_from_utf32(char32_t const*, unsigned long): mov rcx, rsi test rsi, rsi je .L7 lea rax, [rsi-1] cmp rax, 2 jbe .L8 mov rdx, rsi pxor xmm5, xmm5 mov rax, rdi movdqa xmm4, XMMWORD PTR .LC1[rip] shr rdx, 2 movdqa xmm3, xmm5 movdqa xmm10, XMMWORD PTR .LC2[rip] movdqa xmm6, XMMWORD PTR .LC5[rip] sal rdx, 4 movdqa xmm7, XMMWORD PTR .LC8[rip] pxor xmm2, xmm2 movdqa xmm9, XMMWORD PTR .LC6[rip] movdqa xmm8, XMMWORD PTR .LC7[rip] add rdx, rdi .L4: movdqu xmm0, XMMWORD PTR [rax] add rax, 16 psubd xmm0, xmm10 movdqa xmm1, xmm0 movdqa xmm11, xmm0 pcmpgtd xmm1, xmm9 pcmpgtd xmm11, xmm8 pcmpgtd xmm0, xmm7 pand xmm1, xmm4 pand xmm11, xmm4 movdqa xmm13, xmm1 movdqa xmm12, xmm11 punpckhdq xmm1, xmm2 pand xmm0, xmm4 punpckhdq xmm11, xmm2 punpckldq xmm13, xmm2 paddq xmm11, xmm1 movdqa xmm1, xmm0 punpckldq xmm12, xmm2 punpckldq xmm1, xmm2 punpckhdq xmm0, xmm2 paddq xmm12, xmm13 paddq xmm1, xmm6 paddq xmm0, xmm6 paddq xmm1, xmm12 paddq xmm0, xmm11 paddq xmm3, xmm1 paddq xmm5, xmm0 cmp rax, rdx jne .L4 paddq xmm3, xmm5 movdqa xmm0, xmm3 psrldq xmm0, 8 paddq xmm3, xmm0 movq rax, xmm3 test cl, 3 je .L1 mov rdx, rcx and rdx, -4 .L3: mov r8d, DWORD PTR [rdi+rdx*4] xor esi, esi lea r9, [0+rdx*4] cmp r8d, 127 seta sil lea rax, [rax+1+rsi] xor esi, esi cmp r8d, 2047 seta sil add rsi, rax xor eax, eax cmp r8d, 65535 seta al add rax, rsi lea rsi, [rdx+1] cmp rsi, rcx jnb .L1 mov r8d, DWORD PTR [rdi+4+r9] xor r10d, r10d cmp r8d, 127 seta r10b xor esi, esi cmp r8d, 2047 seta sil cmp r8d, 65535 seta r8b lea rsi, [r10+1+rsi] add rdx, 2 movzx r8d, r8b add rsi, r8 add rax, rsi cmp rdx, rcx jnb .L1 mov ecx, DWORD PTR [rdi+8+r9] xor esi, esi cmp ecx, 2047 seta sil xor edx, edx cmp ecx, 127 seta dl cmp ecx, 65535 seta cl lea rdx, [rsi+1+rdx] movzx ecx, cl add rdx, rcx add rax, rdx ret .L7: xor eax, eax .L1: ret .L8: xor edx, edx xor eax, eax jmp .L3 .LC1: .long 1 .long 1 .long 1 .long 1 .LC2: .long -2147483648 .long -2147483648 .long -2147483648 .long -2147483648 .LC5: .quad 1 .quad 1 .LC6: .long -2147483521 .long -2147483521 .long -2147483521 .long -2147483521 .LC7: .long -2147481601 .long -2147481601 .long -2147481601 .long -2147481601 .LC8: .long -2147418113 .long -2147418113 .long -2147418113 .long -2147418113 ```
ttsugriy
force-pushed
the
utf8_length_from_utf32
branch
from
4b35b67
to
03945fd
Compare
|
ada_idna.cpp file is originally implemented in idna repository. Can you update it directly? |
|
oh, sorry, I didn't realize it's from a different repo. I'll send a PR to it. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This helps GCC to autovectorize its implementation. https://compiler-explorer.com/z/M6Tjd43nz
so instead of
it generates