From: "H.J. Lu" <hjl.tools@gmail.com>
To: Noah Goldstein <goldstein.w.n@gmail.com>
Cc: GNU C Library <libc-alpha@sourceware.org>
Subject: Re: [PATCH v2 6/7] x86: Optimize strcmp-evex.S
Date: Sun, 9 Jan 2022 16:41:48 -0800 [thread overview]
Message-ID: <CAMe9rOpQkJGLcdK1za2-+staDD3VAb-rFUGezJRuQsROpUDBBQ@mail.gmail.com> (raw)
In-Reply-To: <20220110002746.3653516-6-goldstein.w.n@gmail.com>
On Sun, Jan 9, 2022 at 4:32 PM Noah Goldstein via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> Optimization are primarily to the loop logic and how the page cross
> logic interacts with the loop.
>
> The page cross logic is at times more expensive for short strings near
> the end of a page but not crossing the page. This is done to retest
> the page cross conditions with a non-faulty check and to improve the
> logic for entering the loop afterwards. This is only particular cases,
> however, and is general made up for by more than 10x improvements on
> the transition from the page cross -> loop case.
>
> The non-page cross cases as well are nearly universally improved.
>
> test-strcmp, test-strncmp, test-wcscmp, and test-wcsncmp all pass.
>
> Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com>
> ---
> sysdeps/x86_64/multiarch/strcmp-evex.S | 1712 +++++++++++++-----------
> 1 file changed, 919 insertions(+), 793 deletions(-)
>
> diff --git a/sysdeps/x86_64/multiarch/strcmp-evex.S b/sysdeps/x86_64/multiarch/strcmp-evex.S
> index 0cd939d5af..e5070f3d53 100644
> --- a/sysdeps/x86_64/multiarch/strcmp-evex.S
> +++ b/sysdeps/x86_64/multiarch/strcmp-evex.S
> @@ -26,54 +26,69 @@
>
> # define PAGE_SIZE 4096
>
> -/* VEC_SIZE = Number of bytes in a ymm register */
> + /* VEC_SIZE = Number of bytes in a ymm register. */
> # define VEC_SIZE 32
> +# define CHAR_PER_VEC (VEC_SIZE / SIZE_OF_CHAR)
>
> -/* Shift for dividing by (VEC_SIZE * 4). */
> -# define DIVIDE_BY_VEC_4_SHIFT 7
> -# if (VEC_SIZE * 4) != (1 << DIVIDE_BY_VEC_4_SHIFT)
> -# error (VEC_SIZE * 4) != (1 << DIVIDE_BY_VEC_4_SHIFT)
> -# endif
> -
> -# define VMOVU vmovdqu64
> -# define VMOVA vmovdqa64
> +# define VMOVU vmovdqu64
> +# define VMOVA vmovdqa64
>
> # ifdef USE_AS_WCSCMP
> -/* Compare packed dwords. */
> -# define VPCMP vpcmpd
> +# define TESTEQ subl $0xff,
> + /* Compare packed dwords. */
> +# define VPCMP vpcmpd
> # define VPMINU vpminud
> # define VPTESTM vptestmd
> -# define SHIFT_REG32 r8d
> -# define SHIFT_REG64 r8
> -/* 1 dword char == 4 bytes. */
> + /* 1 dword char == 4 bytes. */
> # define SIZE_OF_CHAR 4
> # else
> -/* Compare packed bytes. */
> -# define VPCMP vpcmpb
> +# define TESTEQ incl
> + /* Compare packed bytes. */
> +# define VPCMP vpcmpb
> # define VPMINU vpminub
> # define VPTESTM vptestmb
> -# define SHIFT_REG32 ecx
> -# define SHIFT_REG64 rcx
> -/* 1 byte char == 1 byte. */
> + /* 1 byte char == 1 byte. */
> # define SIZE_OF_CHAR 1
> # endif
>
> +# ifdef USE_AS_STRNCMP
> +# define LOOP_REG r9d
> +# define LOOP_REG64 r9
> +
> +# define OFFSET_REG8 r9b
> +# define OFFSET_REG r9d
> +# define OFFSET_REG64 r9
> +# else
> +# define LOOP_REG edx
> +# define LOOP_REG64 rdx
> +
> +# define OFFSET_REG8 dl
> +# define OFFSET_REG edx
> +# define OFFSET_REG64 rdx
> +# endif
> +
> +# if defined USE_AS_STRNCMP || defined USE_AS_WCSCMP
> +# define VEC_OFFSET 0
> +# else
> +# define VEC_OFFSET (-VEC_SIZE)
> +# endif
> +
> # define XMMZERO xmm16
> -# define XMM0 xmm17
> -# define XMM1 xmm18
> +# define XMM0 xmm17
> +# define XMM1 xmm18
>
> # define YMMZERO ymm16
> -# define YMM0 ymm17
> -# define YMM1 ymm18
> -# define YMM2 ymm19
> -# define YMM3 ymm20
> -# define YMM4 ymm21
> -# define YMM5 ymm22
> -# define YMM6 ymm23
> -# define YMM7 ymm24
> -# define YMM8 ymm25
> -# define YMM9 ymm26
> -# define YMM10 ymm27
> +# define YMM0 ymm17
> +# define YMM1 ymm18
> +# define YMM2 ymm19
> +# define YMM3 ymm20
> +# define YMM4 ymm21
> +# define YMM5 ymm22
> +# define YMM6 ymm23
> +# define YMM7 ymm24
> +# define YMM8 ymm25
> +# define YMM9 ymm26
> +# define YMM10 ymm27
>
> /* Warning!
> wcscmp/wcsncmp have to use SIGNED comparison for elements.
> @@ -96,985 +111,1096 @@
> the maximum offset is reached before a difference is found, zero is
> returned. */
>
> - .section .text.evex,"ax",@progbits
> -ENTRY (STRCMP)
> + .section .text.evex, "ax", @progbits
> +ENTRY(STRCMP)
> # ifdef USE_AS_STRNCMP
> - /* Check for simple cases (0 or 1) in offset. */
> - cmp $1, %RDX_LP
> - je L(char0)
> - jb L(zero)
> -# ifdef USE_AS_WCSCMP
> -# ifndef __ILP32__
> - movq %rdx, %rcx
> - /* Check if length could overflow when multiplied by
> - sizeof(wchar_t). Checking top 8 bits will cover all potential
> - overflow cases as well as redirect cases where its impossible to
> - length to bound a valid memory region. In these cases just use
> - 'wcscmp'. */
> - shrq $56, %rcx
> - jnz __wcscmp_evex
> -# endif
> - /* Convert units: from wide to byte char. */
> - shl $2, %RDX_LP
> +# ifdef __ILP32__
> + /* Clear the upper 32 bits. */
> + movl %edx, %rdx
> # endif
> - /* Register %r11 tracks the maximum offset. */
> - mov %RDX_LP, %R11_LP
> + cmp $1, %RDX_LP
> + /* Signed comparison intentional. We use this branch to also
> + test cases where length >= 2^63. These very large sizes can be
> + handled with strcmp as there is no way for that length to
> + actually bound the buffer. */
> + jle L(one_or_less)
> # endif
> movl %edi, %eax
> - xorl %edx, %edx
> - /* Make %XMMZERO (%YMMZERO) all zeros in this function. */
> - vpxorq %XMMZERO, %XMMZERO, %XMMZERO
> orl %esi, %eax
> - andl $(PAGE_SIZE - 1), %eax
> - cmpl $(PAGE_SIZE - (VEC_SIZE * 4)), %eax
> - jg L(cross_page)
> - /* Start comparing 4 vectors. */
> + /* Shift out the bits irrelivant to page boundary ([63:12]). */
> + sall $20, %eax
> + /* Check if s1 or s2 may cross a page in next 4x VEC loads. */
> + cmpl $((PAGE_SIZE -(VEC_SIZE * 4)) << 20), %eax
> + ja L(page_cross)
> +
> +L(no_page_cross):
> + /* Safe to compare 4x vectors. */
> VMOVU (%rdi), %YMM0
> -
> - /* Each bit set in K2 represents a non-null CHAR in YMM0. */
> VPTESTM %YMM0, %YMM0, %k2
> -
> /* Each bit cleared in K1 represents a mismatch or a null CHAR
> in YMM0 and 32 bytes at (%rsi). */
> VPCMP $0, (%rsi), %YMM0, %k1{%k2}
> -
> kmovd %k1, %ecx
> -# ifdef USE_AS_WCSCMP
> - subl $0xff, %ecx
> -# else
> - incl %ecx
> -# endif
> - je L(next_3_vectors)
> - tzcntl %ecx, %edx
> -# ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %edx
> -# endif
> # ifdef USE_AS_STRNCMP
> - /* Return 0 if the mismatched index (%rdx) is after the maximum
> - offset (%r11). */
> - cmpq %r11, %rdx
> - jae L(zero)
> + cmpq $CHAR_PER_VEC, %rdx
> + jbe L(vec_0_test_len)
> # endif
> +
> + /* TESTEQ is `incl` for strcmp/strncmp and `subl $0xff` for
> + wcscmp/wcsncmp. */
> +
> + /* All 1s represents all equals. TESTEQ will overflow to zero in
> + all equals case. Otherwise 1s will carry until position of first
> + mismatch. */
> + TESTEQ %ecx
> + jz L(more_3x_vec)
> +
> + .p2align 4,, 4
> +L(return_vec_0):
> + tzcntl %ecx, %ecx
> # ifdef USE_AS_WCSCMP
> + movl (%rdi, %rcx, SIZE_OF_CHAR), %edx
> xorl %eax, %eax
> - movl (%rdi, %rdx), %ecx
> - cmpl (%rsi, %rdx), %ecx
> - je L(return)
> -L(wcscmp_return):
> + cmpl (%rsi, %rcx, SIZE_OF_CHAR), %edx
> + je L(ret0)
> setl %al
> negl %eax
> orl $1, %eax
> -L(return):
> # else
> - movzbl (%rdi, %rdx), %eax
> - movzbl (%rsi, %rdx), %edx
> - subl %edx, %eax
> + movzbl (%rdi, %rcx), %eax
> + movzbl (%rsi, %rcx), %ecx
> + subl %ecx, %eax
> # endif
> +L(ret0):
> ret
>
> -L(return_vec_size):
> - tzcntl %ecx, %edx
> -# ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %edx
> -# endif
> # ifdef USE_AS_STRNCMP
> - /* Return 0 if the mismatched index (%rdx + VEC_SIZE) is after
> - the maximum offset (%r11). */
> - addq $VEC_SIZE, %rdx
> - cmpq %r11, %rdx
> - jae L(zero)
> -# ifdef USE_AS_WCSCMP
> + .p2align 4,, 4
> +L(vec_0_test_len):
> + notl %ecx
> + bzhil %edx, %ecx, %eax
> + jnz L(return_vec_0)
> + /* Align if will cross fetch block. */
> + .p2align 4,, 2
> +L(ret_zero):
> xorl %eax, %eax
> - movl (%rdi, %rdx), %ecx
> - cmpl (%rsi, %rdx), %ecx
> - jne L(wcscmp_return)
> -# else
> - movzbl (%rdi, %rdx), %eax
> - movzbl (%rsi, %rdx), %edx
> - subl %edx, %eax
> -# endif
> -# else
> + ret
> +
> + .p2align 4,, 5
> +L(one_or_less):
> + jb L(ret_zero)
> # ifdef USE_AS_WCSCMP
> + /* 'nbe' covers the case where length is negative (large
> + unsigned). */
> + jnbe __wcscmp_evex
> + movl (%rdi), %edx
> xorl %eax, %eax
> - movl VEC_SIZE(%rdi, %rdx), %ecx
> - cmpl VEC_SIZE(%rsi, %rdx), %ecx
> - jne L(wcscmp_return)
> + cmpl (%rsi), %edx
> + je L(ret1)
> + setl %al
> + negl %eax
> + orl $1, %eax
> # else
> - movzbl VEC_SIZE(%rdi, %rdx), %eax
> - movzbl VEC_SIZE(%rsi, %rdx), %edx
> - subl %edx, %eax
> + /* 'nbe' covers the case where length is negative (large
> + unsigned). */
> + jnbe __strcmp_evex
> + movzbl (%rdi), %eax
> + movzbl (%rsi), %ecx
> + subl %ecx, %eax
> # endif
> -# endif
> +L(ret1):
> ret
> +# endif
>
> -L(return_2_vec_size):
> - tzcntl %ecx, %edx
> + .p2align 4,, 10
> +L(return_vec_1):
> + tzcntl %ecx, %ecx
> +# ifdef USE_AS_STRNCMP
> + /* rdx must be > CHAR_PER_VEC so its safe to subtract without
> + worrying about underflow. */
> + addq $-CHAR_PER_VEC, %rdx
> + cmpq %rcx, %rdx
> + jbe L(ret_zero)
> +# endif
> # ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %edx
> + movl VEC_SIZE(%rdi, %rcx, SIZE_OF_CHAR), %edx
> + xorl %eax, %eax
> + cmpl VEC_SIZE(%rsi, %rcx, SIZE_OF_CHAR), %edx
> + je L(ret2)
> + setl %al
> + negl %eax
> + orl $1, %eax
> +# else
> + movzbl VEC_SIZE(%rdi, %rcx), %eax
> + movzbl VEC_SIZE(%rsi, %rcx), %ecx
> + subl %ecx, %eax
> # endif
> +L(ret2):
> + ret
> +
> + .p2align 4,, 10
> # ifdef USE_AS_STRNCMP
> - /* Return 0 if the mismatched index (%rdx + 2 * VEC_SIZE) is
> - after the maximum offset (%r11). */
> - addq $(VEC_SIZE * 2), %rdx
> - cmpq %r11, %rdx
> - jae L(zero)
> -# ifdef USE_AS_WCSCMP
> - xorl %eax, %eax
> - movl (%rdi, %rdx), %ecx
> - cmpl (%rsi, %rdx), %ecx
> - jne L(wcscmp_return)
> +L(return_vec_3):
> +# if CHAR_PER_VEC <= 16
> + sall $CHAR_PER_VEC, %ecx
> # else
> - movzbl (%rdi, %rdx), %eax
> - movzbl (%rsi, %rdx), %edx
> - subl %edx, %eax
> + salq $CHAR_PER_VEC, %rcx
> # endif
> +# endif
> +L(return_vec_2):
> +# if (CHAR_PER_VEC <= 16) || !(defined USE_AS_STRNCMP)
> + tzcntl %ecx, %ecx
> # else
> -# ifdef USE_AS_WCSCMP
> - xorl %eax, %eax
> - movl (VEC_SIZE * 2)(%rdi, %rdx), %ecx
> - cmpl (VEC_SIZE * 2)(%rsi, %rdx), %ecx
> - jne L(wcscmp_return)
> -# else
> - movzbl (VEC_SIZE * 2)(%rdi, %rdx), %eax
> - movzbl (VEC_SIZE * 2)(%rsi, %rdx), %edx
> - subl %edx, %eax
> -# endif
> + tzcntq %rcx, %rcx
> # endif
> - ret
>
> -L(return_3_vec_size):
> - tzcntl %ecx, %edx
> -# ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %edx
> -# endif
> # ifdef USE_AS_STRNCMP
> - /* Return 0 if the mismatched index (%rdx + 3 * VEC_SIZE) is
> - after the maximum offset (%r11). */
> - addq $(VEC_SIZE * 3), %rdx
> - cmpq %r11, %rdx
> - jae L(zero)
> -# ifdef USE_AS_WCSCMP
> + cmpq %rcx, %rdx
> + jbe L(ret_zero)
> +# endif
> +
> +# ifdef USE_AS_WCSCMP
> + movl (VEC_SIZE * 2)(%rdi, %rcx, SIZE_OF_CHAR), %edx
> xorl %eax, %eax
> - movl (%rdi, %rdx), %ecx
> - cmpl (%rsi, %rdx), %ecx
> - jne L(wcscmp_return)
> -# else
> - movzbl (%rdi, %rdx), %eax
> - movzbl (%rsi, %rdx), %edx
> - subl %edx, %eax
> -# endif
> + cmpl (VEC_SIZE * 2)(%rsi, %rcx, SIZE_OF_CHAR), %edx
> + je L(ret3)
> + setl %al
> + negl %eax
> + orl $1, %eax
> # else
> + movzbl (VEC_SIZE * 2)(%rdi, %rcx), %eax
> + movzbl (VEC_SIZE * 2)(%rsi, %rcx), %ecx
> + subl %ecx, %eax
> +# endif
> +L(ret3):
> + ret
> +
> +# ifndef USE_AS_STRNCMP
> + .p2align 4,, 10
> +L(return_vec_3):
> + tzcntl %ecx, %ecx
> # ifdef USE_AS_WCSCMP
> + movl (VEC_SIZE * 3)(%rdi, %rcx, SIZE_OF_CHAR), %edx
> xorl %eax, %eax
> - movl (VEC_SIZE * 3)(%rdi, %rdx), %ecx
> - cmpl (VEC_SIZE * 3)(%rsi, %rdx), %ecx
> - jne L(wcscmp_return)
> + cmpl (VEC_SIZE * 3)(%rsi, %rcx, SIZE_OF_CHAR), %edx
> + je L(ret4)
> + setl %al
> + negl %eax
> + orl $1, %eax
> # else
> - movzbl (VEC_SIZE * 3)(%rdi, %rdx), %eax
> - movzbl (VEC_SIZE * 3)(%rsi, %rdx), %edx
> - subl %edx, %eax
> + movzbl (VEC_SIZE * 3)(%rdi, %rcx), %eax
> + movzbl (VEC_SIZE * 3)(%rsi, %rcx), %ecx
> + subl %ecx, %eax
> # endif
> -# endif
> +L(ret4):
> ret
> +# endif
>
> - .p2align 4
> -L(next_3_vectors):
> - VMOVU VEC_SIZE(%rdi), %YMM0
> - /* Each bit set in K2 represents a non-null CHAR in YMM0. */
> + /* 32 byte align here ensures the main loop is ideally aligned
> + for DSB. */
> + .p2align 5
> +L(more_3x_vec):
> + /* Safe to compare 4x vectors. */
> + VMOVU (VEC_SIZE)(%rdi), %YMM0
> VPTESTM %YMM0, %YMM0, %k2
> - /* Each bit cleared in K1 represents a mismatch or a null CHAR
> - in YMM0 and 32 bytes at VEC_SIZE(%rsi). */
> - VPCMP $0, VEC_SIZE(%rsi), %YMM0, %k1{%k2}
> + VPCMP $0, (VEC_SIZE)(%rsi), %YMM0, %k1{%k2}
> kmovd %k1, %ecx
> -# ifdef USE_AS_WCSCMP
> - subl $0xff, %ecx
> -# else
> - incl %ecx
> + TESTEQ %ecx
> + jnz L(return_vec_1)
> +
> +# ifdef USE_AS_STRNCMP
> + subq $(CHAR_PER_VEC * 2), %rdx
> + jbe L(ret_zero)
> # endif
> - jne L(return_vec_size)
>
> VMOVU (VEC_SIZE * 2)(%rdi), %YMM0
> - /* Each bit set in K2 represents a non-null CHAR in YMM0. */
> VPTESTM %YMM0, %YMM0, %k2
> - /* Each bit cleared in K1 represents a mismatch or a null CHAR
> - in YMM0 and 32 bytes at (VEC_SIZE * 2)(%rsi). */
> VPCMP $0, (VEC_SIZE * 2)(%rsi), %YMM0, %k1{%k2}
> kmovd %k1, %ecx
> -# ifdef USE_AS_WCSCMP
> - subl $0xff, %ecx
> -# else
> - incl %ecx
> -# endif
> - jne L(return_2_vec_size)
> + TESTEQ %ecx
> + jnz L(return_vec_2)
>
> VMOVU (VEC_SIZE * 3)(%rdi), %YMM0
> - /* Each bit set in K2 represents a non-null CHAR in YMM0. */
> VPTESTM %YMM0, %YMM0, %k2
> - /* Each bit cleared in K1 represents a mismatch or a null CHAR
> - in YMM0 and 32 bytes at (VEC_SIZE * 2)(%rsi). */
> VPCMP $0, (VEC_SIZE * 3)(%rsi), %YMM0, %k1{%k2}
> kmovd %k1, %ecx
> + TESTEQ %ecx
> + jnz L(return_vec_3)
> +
> +# ifdef USE_AS_STRNCMP
> + cmpq $(CHAR_PER_VEC * 2), %rdx
> + jbe L(ret_zero)
> +# endif
> +
> +
> # ifdef USE_AS_WCSCMP
> - subl $0xff, %ecx
> + /* any non-zero positive value that doesn't inference with 0x1.
> + */
> + movl $2, %r8d
> +
> # else
> - incl %ecx
> + xorl %r8d, %r8d
> # endif
> - jne L(return_3_vec_size)
> -L(main_loop_header):
> - leaq (VEC_SIZE * 4)(%rdi), %rdx
> - movl $PAGE_SIZE, %ecx
> - /* Align load via RAX. */
> - andq $-(VEC_SIZE * 4), %rdx
> - subq %rdi, %rdx
> - leaq (%rdi, %rdx), %rax
> +
> + /* The prepare labels are various entry points from the page
> + cross logic. */
> +L(prepare_loop):
> +
> # ifdef USE_AS_STRNCMP
> - /* Starting from this point, the maximum offset, or simply the
> - 'offset', DECREASES by the same amount when base pointers are
> - moved forward. Return 0 when:
> - 1) On match: offset <= the matched vector index.
> - 2) On mistmach, offset is before the mistmatched index.
> - */
> - subq %rdx, %r11
> - jbe L(zero)
> +# ifdef USE_AS_WCSCMP
> +L(prepare_loop_no_len):
> + movl %edi, %ecx
> + andl $(VEC_SIZE * 4 - 1), %ecx
> + shrl $2, %ecx
> + leaq (CHAR_PER_VEC * 2)(%rdx, %rcx), %rdx
> +# else
> + /* Store N + (VEC_SIZE * 4) and place check at the begining of
> + the loop. */
> + leaq (VEC_SIZE * 2)(%rdi, %rdx), %rdx
> +L(prepare_loop_no_len):
> +# endif
> +# else
> +L(prepare_loop_no_len):
> # endif
> - addq %rsi, %rdx
> - movq %rdx, %rsi
> - andl $(PAGE_SIZE - 1), %esi
> - /* Number of bytes before page crossing. */
> - subq %rsi, %rcx
> - /* Number of VEC_SIZE * 4 blocks before page crossing. */
> - shrq $DIVIDE_BY_VEC_4_SHIFT, %rcx
> - /* ESI: Number of VEC_SIZE * 4 blocks before page crossing. */
> - movl %ecx, %esi
> - jmp L(loop_start)
>
> + /* Align s1 and adjust s2 accordingly. */
> + subq %rdi, %rsi
> + andq $-(VEC_SIZE * 4), %rdi
> +L(prepare_loop_readj):
> + addq %rdi, %rsi
> +# if (defined USE_AS_STRNCMP) && !(defined USE_AS_WCSCMP)
> + subq %rdi, %rdx
> +# endif
> +
> +L(prepare_loop_aligned):
> + /* eax stores distance from rsi to next page cross. These cases
> + need to be handled specially as the 4x loop could potentially
> + read memory past the length of s1 or s2 and across a page
> + boundary. */
> + movl $-(VEC_SIZE * 4), %eax
> + subl %esi, %eax
> + andl $(PAGE_SIZE - 1), %eax
> +
> + vpxorq %YMMZERO, %YMMZERO, %YMMZERO
> +
> + /* Loop 4x comparisons at a time. */
> .p2align 4
> L(loop):
> +
> + /* End condition for strncmp. */
> # ifdef USE_AS_STRNCMP
> - /* Base pointers are moved forward by 4 * VEC_SIZE. Decrease
> - the maximum offset (%r11) by the same amount. */
> - subq $(VEC_SIZE * 4), %r11
> - jbe L(zero)
> + subq $(CHAR_PER_VEC * 4), %rdx
> + jbe L(ret_zero)
> # endif
> - addq $(VEC_SIZE * 4), %rax
> - addq $(VEC_SIZE * 4), %rdx
> -L(loop_start):
> - testl %esi, %esi
> - leal -1(%esi), %esi
> - je L(loop_cross_page)
> -L(back_to_loop):
> - /* Main loop, comparing 4 vectors are a time. */
> - VMOVA (%rax), %YMM0
> - VMOVA VEC_SIZE(%rax), %YMM2
> - VMOVA (VEC_SIZE * 2)(%rax), %YMM4
> - VMOVA (VEC_SIZE * 3)(%rax), %YMM6
> +
> + subq $-(VEC_SIZE * 4), %rdi
> + subq $-(VEC_SIZE * 4), %rsi
> +
> + /* Check if rsi loads will cross a page boundary. */
> + addl $-(VEC_SIZE * 4), %eax
> + jnb L(page_cross_during_loop)
> +
> + /* Loop entry after handling page cross during loop. */
> +L(loop_skip_page_cross_check):
> + VMOVA (VEC_SIZE * 0)(%rdi), %YMM0
> + VMOVA (VEC_SIZE * 1)(%rdi), %YMM2
> + VMOVA (VEC_SIZE * 2)(%rdi), %YMM4
> + VMOVA (VEC_SIZE * 3)(%rdi), %YMM6
>
> VPMINU %YMM0, %YMM2, %YMM8
> VPMINU %YMM4, %YMM6, %YMM9
>
> - /* A zero CHAR in YMM8 means that there is a null CHAR. */
> - VPMINU %YMM8, %YMM9, %YMM8
> + /* A zero CHAR in YMM9 means that there is a null CHAR. */
> + VPMINU %YMM8, %YMM9, %YMM9
>
> /* Each bit set in K1 represents a non-null CHAR in YMM8. */
> - VPTESTM %YMM8, %YMM8, %k1
> + VPTESTM %YMM9, %YMM9, %k1
>
> - /* (YMM ^ YMM): A non-zero CHAR represents a mismatch. */
> - vpxorq (%rdx), %YMM0, %YMM1
> - vpxorq VEC_SIZE(%rdx), %YMM2, %YMM3
> - vpxorq (VEC_SIZE * 2)(%rdx), %YMM4, %YMM5
> - vpxorq (VEC_SIZE * 3)(%rdx), %YMM6, %YMM7
> + vpxorq (VEC_SIZE * 0)(%rsi), %YMM0, %YMM1
> + vpxorq (VEC_SIZE * 1)(%rsi), %YMM2, %YMM3
> + vpxorq (VEC_SIZE * 2)(%rsi), %YMM4, %YMM5
> + /* Ternary logic to xor (VEC_SIZE * 3)(%rsi) with YMM6 while
> + oring with YMM1. Result is stored in YMM6. */
> + vpternlogd $0xde, (VEC_SIZE * 3)(%rsi), %YMM1, %YMM6
>
> - vporq %YMM1, %YMM3, %YMM9
> - vporq %YMM5, %YMM7, %YMM10
> + /* Or together YMM3, YMM5, and YMM6. */
> + vpternlogd $0xfe, %YMM3, %YMM5, %YMM6
>
> - /* A non-zero CHAR in YMM9 represents a mismatch. */
> - vporq %YMM9, %YMM10, %YMM9
>
> - /* Each bit cleared in K0 represents a mismatch or a null CHAR. */
> - VPCMP $0, %YMMZERO, %YMM9, %k0{%k1}
> - kmovd %k0, %ecx
> -# ifdef USE_AS_WCSCMP
> - subl $0xff, %ecx
> -# else
> - incl %ecx
> -# endif
> - je L(loop)
> + /* A non-zero CHAR in YMM6 represents a mismatch. */
> + VPCMP $0, %YMMZERO, %YMM6, %k0{%k1}
> + kmovd %k0, %LOOP_REG
>
> - /* Each bit set in K1 represents a non-null CHAR in YMM0. */
> + TESTEQ %LOOP_REG
> + jz L(loop)
> +
> +
> + /* Find which VEC has the mismatch of end of string. */
> VPTESTM %YMM0, %YMM0, %k1
> - /* Each bit cleared in K0 represents a mismatch or a null CHAR
> - in YMM0 and (%rdx). */
> VPCMP $0, %YMMZERO, %YMM1, %k0{%k1}
> kmovd %k0, %ecx
> -# ifdef USE_AS_WCSCMP
> - subl $0xff, %ecx
> -# else
> - incl %ecx
> -# endif
> - je L(test_vec)
> - tzcntl %ecx, %ecx
> -# ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %ecx
> -# endif
> -# ifdef USE_AS_STRNCMP
> - cmpq %rcx, %r11
> - jbe L(zero)
> -# ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> - xorl %eax, %eax
> - movl (%rsi, %rcx), %edi
> - cmpl (%rdx, %rcx), %edi
> - jne L(wcscmp_return)
> -# else
> - movzbl (%rax, %rcx), %eax
> - movzbl (%rdx, %rcx), %edx
> - subl %edx, %eax
> -# endif
> -# else
> -# ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> - xorl %eax, %eax
> - movl (%rsi, %rcx), %edi
> - cmpl (%rdx, %rcx), %edi
> - jne L(wcscmp_return)
> -# else
> - movzbl (%rax, %rcx), %eax
> - movzbl (%rdx, %rcx), %edx
> - subl %edx, %eax
> -# endif
> -# endif
> - ret
> + TESTEQ %ecx
> + jnz L(return_vec_0_end)
>
> - .p2align 4
> -L(test_vec):
> -# ifdef USE_AS_STRNCMP
> - /* The first vector matched. Return 0 if the maximum offset
> - (%r11) <= VEC_SIZE. */
> - cmpq $VEC_SIZE, %r11
> - jbe L(zero)
> -# endif
> - /* Each bit set in K1 represents a non-null CHAR in YMM2. */
> VPTESTM %YMM2, %YMM2, %k1
> - /* Each bit cleared in K0 represents a mismatch or a null CHAR
> - in YMM2 and VEC_SIZE(%rdx). */
> VPCMP $0, %YMMZERO, %YMM3, %k0{%k1}
> kmovd %k0, %ecx
> -# ifdef USE_AS_WCSCMP
> - subl $0xff, %ecx
> -# else
> - incl %ecx
> -# endif
> - je L(test_2_vec)
> - tzcntl %ecx, %edi
> -# ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %edi
> -# endif
> -# ifdef USE_AS_STRNCMP
> - addq $VEC_SIZE, %rdi
> - cmpq %rdi, %r11
> - jbe L(zero)
> -# ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> - xorl %eax, %eax
> - movl (%rsi, %rdi), %ecx
> - cmpl (%rdx, %rdi), %ecx
> - jne L(wcscmp_return)
> -# else
> - movzbl (%rax, %rdi), %eax
> - movzbl (%rdx, %rdi), %edx
> - subl %edx, %eax
> -# endif
> -# else
> -# ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> - xorl %eax, %eax
> - movl VEC_SIZE(%rsi, %rdi), %ecx
> - cmpl VEC_SIZE(%rdx, %rdi), %ecx
> - jne L(wcscmp_return)
> -# else
> - movzbl VEC_SIZE(%rax, %rdi), %eax
> - movzbl VEC_SIZE(%rdx, %rdi), %edx
> - subl %edx, %eax
> -# endif
> -# endif
> - ret
> + TESTEQ %ecx
> + jnz L(return_vec_1_end)
>
> - .p2align 4
> -L(test_2_vec):
> +
> + /* Handle VEC 2 and 3 without branches. */
> +L(return_vec_2_3_end):
> # ifdef USE_AS_STRNCMP
> - /* The first 2 vectors matched. Return 0 if the maximum offset
> - (%r11) <= 2 * VEC_SIZE. */
> - cmpq $(VEC_SIZE * 2), %r11
> - jbe L(zero)
> + subq $(CHAR_PER_VEC * 2), %rdx
> + jbe L(ret_zero_end)
> # endif
> - /* Each bit set in K1 represents a non-null CHAR in YMM4. */
> +
> VPTESTM %YMM4, %YMM4, %k1
> - /* Each bit cleared in K0 represents a mismatch or a null CHAR
> - in YMM4 and (VEC_SIZE * 2)(%rdx). */
> VPCMP $0, %YMMZERO, %YMM5, %k0{%k1}
> kmovd %k0, %ecx
> -# ifdef USE_AS_WCSCMP
> - subl $0xff, %ecx
> + TESTEQ %ecx
> +# if CHAR_PER_VEC <= 16
> + sall $CHAR_PER_VEC, %LOOP_REG
> + orl %ecx, %LOOP_REG
> # else
> - incl %ecx
> + salq $CHAR_PER_VEC, %LOOP_REG64
> + orq %rcx, %LOOP_REG64
> +# endif
> +L(return_vec_3_end):
> + /* LOOP_REG contains matches for null/mismatch from the loop. If
> + VEC 0,1,and 2 all have no null and no mismatches then mismatch
> + must entirely be from VEC 3 which is fully represented by
> + LOOP_REG. */
> +# if CHAR_PER_VEC <= 16
> + tzcntl %LOOP_REG, %LOOP_REG
> +# else
> + tzcntq %LOOP_REG64, %LOOP_REG64
> +# endif
> +# ifdef USE_AS_STRNCMP
> + cmpq %LOOP_REG64, %rdx
> + jbe L(ret_zero_end)
> # endif
> - je L(test_3_vec)
> - tzcntl %ecx, %edi
> +
> # ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %edi
> + movl (VEC_SIZE * 2)(%rdi, %LOOP_REG64, SIZE_OF_CHAR), %ecx
> + xorl %eax, %eax
> + cmpl (VEC_SIZE * 2)(%rsi, %LOOP_REG64, SIZE_OF_CHAR), %ecx
> + je L(ret5)
> + setl %al
> + negl %eax
> + xorl %r8d, %eax
> +# else
> + movzbl (VEC_SIZE * 2)(%rdi, %LOOP_REG64), %eax
> + movzbl (VEC_SIZE * 2)(%rsi, %LOOP_REG64), %ecx
> + subl %ecx, %eax
> + xorl %r8d, %eax
> + subl %r8d, %eax
> # endif
> +L(ret5):
> + ret
> +
> # ifdef USE_AS_STRNCMP
> - addq $(VEC_SIZE * 2), %rdi
> - cmpq %rdi, %r11
> - jbe L(zero)
> -# ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> + .p2align 4,, 2
> +L(ret_zero_end):
> xorl %eax, %eax
> - movl (%rsi, %rdi), %ecx
> - cmpl (%rdx, %rdi), %ecx
> - jne L(wcscmp_return)
> + ret
> +# endif
> +
> +
> + /* The L(return_vec_N_end) differ from L(return_vec_N) in that
> + they use the value of `r8` to negate the return value. This is
> + because the page cross logic can swap `rdi` and `rsi`. */
> + .p2align 4,, 10
> +# ifdef USE_AS_STRNCMP
> +L(return_vec_1_end):
> +# if CHAR_PER_VEC <= 16
> + sall $CHAR_PER_VEC, %ecx
> # else
> - movzbl (%rax, %rdi), %eax
> - movzbl (%rdx, %rdi), %edx
> - subl %edx, %eax
> + salq $CHAR_PER_VEC, %rcx
> # endif
> +# endif
> +L(return_vec_0_end):
> +# if (CHAR_PER_VEC <= 16) || !(defined USE_AS_STRNCMP)
> + tzcntl %ecx, %ecx
> # else
> -# ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> - xorl %eax, %eax
> - movl (VEC_SIZE * 2)(%rsi, %rdi), %ecx
> - cmpl (VEC_SIZE * 2)(%rdx, %rdi), %ecx
> - jne L(wcscmp_return)
> -# else
> - movzbl (VEC_SIZE * 2)(%rax, %rdi), %eax
> - movzbl (VEC_SIZE * 2)(%rdx, %rdi), %edx
> - subl %edx, %eax
> -# endif
> + tzcntq %rcx, %rcx
> # endif
> - ret
>
> - .p2align 4
> -L(test_3_vec):
> # ifdef USE_AS_STRNCMP
> - /* The first 3 vectors matched. Return 0 if the maximum offset
> - (%r11) <= 3 * VEC_SIZE. */
> - cmpq $(VEC_SIZE * 3), %r11
> - jbe L(zero)
> + cmpq %rcx, %rdx
> + jbe L(ret_zero_end)
> # endif
> - /* Each bit set in K1 represents a non-null CHAR in YMM6. */
> - VPTESTM %YMM6, %YMM6, %k1
> - /* Each bit cleared in K0 represents a mismatch or a null CHAR
> - in YMM6 and (VEC_SIZE * 3)(%rdx). */
> - VPCMP $0, %YMMZERO, %YMM7, %k0{%k1}
> - kmovd %k0, %ecx
> +
> # ifdef USE_AS_WCSCMP
> - subl $0xff, %ecx
> + movl (%rdi, %rcx, SIZE_OF_CHAR), %edx
> + xorl %eax, %eax
> + cmpl (%rsi, %rcx, SIZE_OF_CHAR), %edx
> + je L(ret6)
> + setl %al
> + negl %eax
> + /* This is the non-zero case for `eax` so just xorl with `r8d`
> + flip is `rdi` and `rsi` where swapped. */
> + xorl %r8d, %eax
> # else
> - incl %ecx
> + movzbl (%rdi, %rcx), %eax
> + movzbl (%rsi, %rcx), %ecx
> + subl %ecx, %eax
> + /* Flip `eax` if `rdi` and `rsi` where swapped in page cross
> + logic. Subtract `r8d` after xor for zero case. */
> + xorl %r8d, %eax
> + subl %r8d, %eax
> # endif
> +L(ret6):
> + ret
> +
> +# ifndef USE_AS_STRNCMP
> + .p2align 4,, 10
> +L(return_vec_1_end):
> tzcntl %ecx, %ecx
> -# ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %ecx
> -# endif
> -# ifdef USE_AS_STRNCMP
> - addq $(VEC_SIZE * 3), %rcx
> - cmpq %rcx, %r11
> - jbe L(zero)
> # ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> + movl VEC_SIZE(%rdi, %rcx, SIZE_OF_CHAR), %edx
> xorl %eax, %eax
> - movl (%rsi, %rcx), %esi
> - cmpl (%rdx, %rcx), %esi
> - jne L(wcscmp_return)
> -# else
> - movzbl (%rax, %rcx), %eax
> - movzbl (%rdx, %rcx), %edx
> - subl %edx, %eax
> -# endif
> -# else
> -# ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> - xorl %eax, %eax
> - movl (VEC_SIZE * 3)(%rsi, %rcx), %esi
> - cmpl (VEC_SIZE * 3)(%rdx, %rcx), %esi
> - jne L(wcscmp_return)
> + cmpl VEC_SIZE(%rsi, %rcx, SIZE_OF_CHAR), %edx
> + je L(ret7)
> + setl %al
> + negl %eax
> + xorl %r8d, %eax
> # else
> - movzbl (VEC_SIZE * 3)(%rax, %rcx), %eax
> - movzbl (VEC_SIZE * 3)(%rdx, %rcx), %edx
> - subl %edx, %eax
> + movzbl VEC_SIZE(%rdi, %rcx), %eax
> + movzbl VEC_SIZE(%rsi, %rcx), %ecx
> + subl %ecx, %eax
> + xorl %r8d, %eax
> + subl %r8d, %eax
> # endif
> -# endif
> +L(ret7):
> ret
> -
> - .p2align 4
> -L(loop_cross_page):
> - xorl %r10d, %r10d
> - movq %rdx, %rcx
> - /* Align load via RDX. We load the extra ECX bytes which should
> - be ignored. */
> - andl $((VEC_SIZE * 4) - 1), %ecx
> - /* R10 is -RCX. */
> - subq %rcx, %r10
> -
> - /* This works only if VEC_SIZE * 2 == 64. */
> -# if (VEC_SIZE * 2) != 64
> -# error (VEC_SIZE * 2) != 64
> # endif
>
> - /* Check if the first VEC_SIZE * 2 bytes should be ignored. */
> - cmpl $(VEC_SIZE * 2), %ecx
> - jge L(loop_cross_page_2_vec)
>
> - VMOVU (%rax, %r10), %YMM2
> - VMOVU VEC_SIZE(%rax, %r10), %YMM3
> + /* Page cross in rsi in next 4x VEC. */
>
> - /* Each bit set in K2 represents a non-null CHAR in YMM2. */
> - VPTESTM %YMM2, %YMM2, %k2
> - /* Each bit cleared in K1 represents a mismatch or a null CHAR
> - in YMM2 and 32 bytes at (%rdx, %r10). */
> - VPCMP $0, (%rdx, %r10), %YMM2, %k1{%k2}
> - kmovd %k1, %r9d
> - /* Don't use subl since it is the lower 16/32 bits of RDI
> - below. */
> - notl %r9d
> -# ifdef USE_AS_WCSCMP
> - /* Only last 8 bits are valid. */
> - andl $0xff, %r9d
> -# endif
> + /* TODO: Improve logic here. */
> + .p2align 4,, 10
> +L(page_cross_during_loop):
> + /* eax contains [distance_from_page - (VEC_SIZE * 4)]. */
>
> - /* Each bit set in K4 represents a non-null CHAR in YMM3. */
> - VPTESTM %YMM3, %YMM3, %k4
> - /* Each bit cleared in K3 represents a mismatch or a null CHAR
> - in YMM3 and 32 bytes at VEC_SIZE(%rdx, %r10). */
> - VPCMP $0, VEC_SIZE(%rdx, %r10), %YMM3, %k3{%k4}
> - kmovd %k3, %edi
> - /* Must use notl %edi here as lower bits are for CHAR
> - comparisons potentially out of range thus can be 0 without
> - indicating mismatch. */
> - notl %edi
> -# ifdef USE_AS_WCSCMP
> - /* Don't use subl since it is the upper 8 bits of EDI below. */
> - andl $0xff, %edi
> -# endif
> + /* Optimistically rsi and rdi and both aligned in which case we
> + don't need any logic here. */
> + cmpl $-(VEC_SIZE * 4), %eax
> + /* Don't adjust eax before jumping back to loop and we will
> + never hit page cross case again. */
> + je L(loop_skip_page_cross_check)
>
> -# ifdef USE_AS_WCSCMP
> - /* NB: Each bit in EDI/R9D represents 4-byte element. */
> - sall $8, %edi
> - /* NB: Divide shift count by 4 since each bit in K1 represent 4
> - bytes. */
> - movl %ecx, %SHIFT_REG32
> - sarl $2, %SHIFT_REG32
> -
> - /* Each bit in EDI represents a null CHAR or a mismatch. */
> - orl %r9d, %edi
> -# else
> - salq $32, %rdi
> + /* Check if we can safely load a VEC. */
> + cmpl $-(VEC_SIZE * 3), %eax
> + jle L(less_1x_vec_till_page_cross)
>
> - /* Each bit in RDI represents a null CHAR or a mismatch. */
> - orq %r9, %rdi
> -# endif
> + VMOVA (%rdi), %YMM0
> + VPTESTM %YMM0, %YMM0, %k2
> + VPCMP $0, (%rsi), %YMM0, %k1{%k2}
> + kmovd %k1, %ecx
> + TESTEQ %ecx
> + jnz L(return_vec_0_end)
> +
> + /* if distance >= 2x VEC then eax > -(VEC_SIZE * 2). */
> + cmpl $-(VEC_SIZE * 2), %eax
> + jg L(more_2x_vec_till_page_cross)
> +
> + .p2align 4,, 4
> +L(less_1x_vec_till_page_cross):
> + subl $-(VEC_SIZE * 4), %eax
> + /* Guranteed safe to read from rdi - VEC_SIZE here. The only
> + concerning case is first iteration if incoming s1 was near start
> + of a page and s2 near end. If s1 was near the start of the page
> + we already aligned up to nearest VEC_SIZE * 4 so gurnateed safe
> + to read back -VEC_SIZE. If rdi is truly at the start of a page
> + here, it means the previous page (rdi - VEC_SIZE) has already
> + been loaded earlier so must be valid. */
> + VMOVU -VEC_SIZE(%rdi, %rax), %YMM0
> + VPTESTM %YMM0, %YMM0, %k2
> + VPCMP $0, -VEC_SIZE(%rsi, %rax), %YMM0, %k1{%k2}
> +
> + /* Mask of potentially valid bits. The lower bits can be out of
> + range comparisons (but safe regarding page crosses). */
>
> - /* Since ECX < VEC_SIZE * 2, simply skip the first ECX bytes. */
> - shrxq %SHIFT_REG64, %rdi, %rdi
> - testq %rdi, %rdi
> - je L(loop_cross_page_2_vec)
> - tzcntq %rdi, %rcx
> # ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %ecx
> + movl $-1, %r10d
> + movl %esi, %ecx
> + andl $(VEC_SIZE - 1), %ecx
> + shrl $2, %ecx
> + shlxl %ecx, %r10d, %ecx
> + movzbl %cl, %r10d
> +# else
> + movl $-1, %ecx
> + shlxl %esi, %ecx, %r10d
> # endif
> +
> + kmovd %k1, %ecx
> + notl %ecx
> +
> +
> # ifdef USE_AS_STRNCMP
> - cmpq %rcx, %r11
> - jbe L(zero)
> # ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> - xorl %eax, %eax
> - movl (%rsi, %rcx), %edi
> - cmpl (%rdx, %rcx), %edi
> - jne L(wcscmp_return)
> + movl %eax, %r11d
> + shrl $2, %r11d
> + cmpq %r11, %rdx
> # else
> - movzbl (%rax, %rcx), %eax
> - movzbl (%rdx, %rcx), %edx
> - subl %edx, %eax
> + cmpq %rax, %rdx
> # endif
> + jbe L(return_page_cross_end_check)
> +# endif
> + movl %eax, %OFFSET_REG
> +
> + /* Readjust eax before potentially returning to the loop. */
> + addl $(PAGE_SIZE - VEC_SIZE * 4), %eax
> +
> + andl %r10d, %ecx
> + jz L(loop_skip_page_cross_check)
> +
> + .p2align 4,, 3
> +L(return_page_cross_end):
> + tzcntl %ecx, %ecx
> +
> +# if (defined USE_AS_STRNCMP) || (defined USE_AS_WCSCMP)
> + leal -VEC_SIZE(%OFFSET_REG64, %rcx, SIZE_OF_CHAR), %ecx
> +L(return_page_cross_cmp_mem):
> # else
> -# ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> + addl %OFFSET_REG, %ecx
> +# endif
> +# ifdef USE_AS_WCSCMP
> + movl VEC_OFFSET(%rdi, %rcx), %edx
> xorl %eax, %eax
> - movl (%rsi, %rcx), %edi
> - cmpl (%rdx, %rcx), %edi
> - jne L(wcscmp_return)
> -# else
> - movzbl (%rax, %rcx), %eax
> - movzbl (%rdx, %rcx), %edx
> - subl %edx, %eax
> -# endif
> + cmpl VEC_OFFSET(%rsi, %rcx), %edx
> + je L(ret8)
> + setl %al
> + negl %eax
> + xorl %r8d, %eax
> +# else
> + movzbl VEC_OFFSET(%rdi, %rcx), %eax
> + movzbl VEC_OFFSET(%rsi, %rcx), %ecx
> + subl %ecx, %eax
> + xorl %r8d, %eax
> + subl %r8d, %eax
> # endif
> +L(ret8):
> ret
>
> - .p2align 4
> -L(loop_cross_page_2_vec):
> - /* The first VEC_SIZE * 2 bytes match or are ignored. */
> - VMOVU (VEC_SIZE * 2)(%rax, %r10), %YMM0
> - VMOVU (VEC_SIZE * 3)(%rax, %r10), %YMM1
> +# ifdef USE_AS_STRNCMP
> + .p2align 4,, 10
> +L(return_page_cross_end_check):
> + tzcntl %ecx, %ecx
> + leal -VEC_SIZE(%rax, %rcx, SIZE_OF_CHAR), %ecx
> +# ifdef USE_AS_WCSCMP
> + sall $2, %edx
> +# endif
> + cmpl %ecx, %edx
> + ja L(return_page_cross_cmp_mem)
> + xorl %eax, %eax
> + ret
> +# endif
> +
>
> + .p2align 4,, 10
> +L(more_2x_vec_till_page_cross):
> + /* If more 2x vec till cross we will complete a full loop
> + iteration here. */
> +
> + VMOVA VEC_SIZE(%rdi), %YMM0
> VPTESTM %YMM0, %YMM0, %k2
> - /* Each bit cleared in K1 represents a mismatch or a null CHAR
> - in YMM0 and 32 bytes at (VEC_SIZE * 2)(%rdx, %r10). */
> - VPCMP $0, (VEC_SIZE * 2)(%rdx, %r10), %YMM0, %k1{%k2}
> - kmovd %k1, %r9d
> - /* Don't use subl since it is the lower 16/32 bits of RDI
> - below. */
> - notl %r9d
> -# ifdef USE_AS_WCSCMP
> - /* Only last 8 bits are valid. */
> - andl $0xff, %r9d
> -# endif
> + VPCMP $0, VEC_SIZE(%rsi), %YMM0, %k1{%k2}
> + kmovd %k1, %ecx
> + TESTEQ %ecx
> + jnz L(return_vec_1_end)
>
> - VPTESTM %YMM1, %YMM1, %k4
> - /* Each bit cleared in K3 represents a mismatch or a null CHAR
> - in YMM1 and 32 bytes at (VEC_SIZE * 3)(%rdx, %r10). */
> - VPCMP $0, (VEC_SIZE * 3)(%rdx, %r10), %YMM1, %k3{%k4}
> - kmovd %k3, %edi
> - /* Must use notl %edi here as lower bits are for CHAR
> - comparisons potentially out of range thus can be 0 without
> - indicating mismatch. */
> - notl %edi
> -# ifdef USE_AS_WCSCMP
> - /* Don't use subl since it is the upper 8 bits of EDI below. */
> - andl $0xff, %edi
> +# ifdef USE_AS_STRNCMP
> + cmpq $(CHAR_PER_VEC * 2), %rdx
> + jbe L(ret_zero_in_loop_page_cross)
> # endif
>
> -# ifdef USE_AS_WCSCMP
> - /* NB: Each bit in EDI/R9D represents 4-byte element. */
> - sall $8, %edi
> + subl $-(VEC_SIZE * 4), %eax
>
> - /* Each bit in EDI represents a null CHAR or a mismatch. */
> - orl %r9d, %edi
> -# else
> - salq $32, %rdi
> + /* Safe to include comparisons from lower bytes. */
> + VMOVU -(VEC_SIZE * 2)(%rdi, %rax), %YMM0
> + VPTESTM %YMM0, %YMM0, %k2
> + VPCMP $0, -(VEC_SIZE * 2)(%rsi, %rax), %YMM0, %k1{%k2}
> + kmovd %k1, %ecx
> + TESTEQ %ecx
> + jnz L(return_vec_page_cross_0)
> +
> + VMOVU -(VEC_SIZE * 1)(%rdi, %rax), %YMM0
> + VPTESTM %YMM0, %YMM0, %k2
> + VPCMP $0, -(VEC_SIZE * 1)(%rsi, %rax), %YMM0, %k1{%k2}
> + kmovd %k1, %ecx
> + TESTEQ %ecx
> + jnz L(return_vec_page_cross_1)
>
> - /* Each bit in RDI represents a null CHAR or a mismatch. */
> - orq %r9, %rdi
> +# ifdef USE_AS_STRNCMP
> + /* Must check length here as length might proclude reading next
> + page. */
> +# ifdef USE_AS_WCSCMP
> + movl %eax, %r11d
> + shrl $2, %r11d
> + cmpq %r11, %rdx
> +# else
> + cmpq %rax, %rdx
> +# endif
> + jbe L(ret_zero_in_loop_page_cross)
> # endif
>
> - xorl %r8d, %r8d
> - /* If ECX > VEC_SIZE * 2, skip ECX - (VEC_SIZE * 2) bytes. */
> - subl $(VEC_SIZE * 2), %ecx
> - jle 1f
> - /* R8 has number of bytes skipped. */
> - movl %ecx, %r8d
> -# ifdef USE_AS_WCSCMP
> - /* NB: Divide shift count by 4 since each bit in RDI represent 4
> - bytes. */
> - sarl $2, %ecx
> - /* Skip ECX bytes. */
> - shrl %cl, %edi
> + /* Finish the loop. */
> + VMOVA (VEC_SIZE * 2)(%rdi), %YMM4
> + VMOVA (VEC_SIZE * 3)(%rdi), %YMM6
> + VPMINU %YMM4, %YMM6, %YMM9
> + VPTESTM %YMM9, %YMM9, %k1
> +
> + vpxorq (VEC_SIZE * 2)(%rsi), %YMM4, %YMM5
> + /* YMM6 = YMM5 | ((VEC_SIZE * 3)(%rsi) ^ YMM6). */
> + vpternlogd $0xde, (VEC_SIZE * 3)(%rsi), %YMM5, %YMM6
> +
> + VPCMP $0, %YMMZERO, %YMM6, %k0{%k1}
> + kmovd %k0, %LOOP_REG
> + TESTEQ %LOOP_REG
> + jnz L(return_vec_2_3_end)
> +
> + /* Best for code size to include ucond-jmp here. Would be faster
> + if this case is hot to duplicate the L(return_vec_2_3_end) code
> + as fall-through and have jump back to loop on mismatch
> + comparison. */
> + subq $-(VEC_SIZE * 4), %rdi
> + subq $-(VEC_SIZE * 4), %rsi
> + addl $(PAGE_SIZE - VEC_SIZE * 8), %eax
> +# ifdef USE_AS_STRNCMP
> + subq $(CHAR_PER_VEC * 4), %rdx
> + ja L(loop_skip_page_cross_check)
> +L(ret_zero_in_loop_page_cross):
> + xorl %eax, %eax
> + ret
> # else
> - /* Skip ECX bytes. */
> - shrq %cl, %rdi
> + jmp L(loop_skip_page_cross_check)
> # endif
> -1:
> - /* Before jumping back to the loop, set ESI to the number of
> - VEC_SIZE * 4 blocks before page crossing. */
> - movl $(PAGE_SIZE / (VEC_SIZE * 4) - 1), %esi
>
> - testq %rdi, %rdi
> -# ifdef USE_AS_STRNCMP
> - /* At this point, if %rdi value is 0, it already tested
> - VEC_SIZE*4+%r10 byte starting from %rax. This label
> - checks whether strncmp maximum offset reached or not. */
> - je L(string_nbyte_offset_check)
> +
> + .p2align 4,, 10
> +L(return_vec_page_cross_0):
> + addl $-VEC_SIZE, %eax
> +L(return_vec_page_cross_1):
> + tzcntl %ecx, %ecx
> +# if defined USE_AS_STRNCMP || defined USE_AS_WCSCMP
> + leal -VEC_SIZE(%rax, %rcx, SIZE_OF_CHAR), %ecx
> +# ifdef USE_AS_STRNCMP
> +# ifdef USE_AS_WCSCMP
> + /* Must divide ecx instead of multiply rdx due to overflow. */
> + movl %ecx, %eax
> + shrl $2, %eax
> + cmpq %rax, %rdx
> +# else
> + cmpq %rcx, %rdx
> +# endif
> + jbe L(ret_zero_in_loop_page_cross)
> +# endif
> # else
> - je L(back_to_loop)
> + addl %eax, %ecx
> # endif
> - tzcntq %rdi, %rcx
> +
> # ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %ecx
> -# endif
> - addq %r10, %rcx
> - /* Adjust for number of bytes skipped. */
> - addq %r8, %rcx
> -# ifdef USE_AS_STRNCMP
> - addq $(VEC_SIZE * 2), %rcx
> - subq %rcx, %r11
> - jbe L(zero)
> -# ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> + movl VEC_OFFSET(%rdi, %rcx), %edx
> xorl %eax, %eax
> - movl (%rsi, %rcx), %edi
> - cmpl (%rdx, %rcx), %edi
> - jne L(wcscmp_return)
> -# else
> - movzbl (%rax, %rcx), %eax
> - movzbl (%rdx, %rcx), %edx
> - subl %edx, %eax
> -# endif
> + cmpl VEC_OFFSET(%rsi, %rcx), %edx
> + je L(ret9)
> + setl %al
> + negl %eax
> + xorl %r8d, %eax
> # else
> -# ifdef USE_AS_WCSCMP
> - movq %rax, %rsi
> - xorl %eax, %eax
> - movl (VEC_SIZE * 2)(%rsi, %rcx), %edi
> - cmpl (VEC_SIZE * 2)(%rdx, %rcx), %edi
> - jne L(wcscmp_return)
> -# else
> - movzbl (VEC_SIZE * 2)(%rax, %rcx), %eax
> - movzbl (VEC_SIZE * 2)(%rdx, %rcx), %edx
> - subl %edx, %eax
> -# endif
> + movzbl VEC_OFFSET(%rdi, %rcx), %eax
> + movzbl VEC_OFFSET(%rsi, %rcx), %ecx
> + subl %ecx, %eax
> + xorl %r8d, %eax
> + subl %r8d, %eax
> # endif
> +L(ret9):
> ret
>
> -# ifdef USE_AS_STRNCMP
> -L(string_nbyte_offset_check):
> - leaq (VEC_SIZE * 4)(%r10), %r10
> - cmpq %r10, %r11
> - jbe L(zero)
> - jmp L(back_to_loop)
> +
> + .p2align 4,, 10
> +L(page_cross):
> +# ifndef USE_AS_STRNCMP
> + /* If both are VEC aligned we don't need any special logic here.
> + Only valid for strcmp where stop condition is guranteed to be
> + reachable by just reading memory. */
> + testl $((VEC_SIZE - 1) << 20), %eax
> + jz L(no_page_cross)
> # endif
>
> - .p2align 4
> -L(cross_page_loop):
> - /* Check one byte/dword at a time. */
> + movl %edi, %eax
> + movl %esi, %ecx
> + andl $(PAGE_SIZE - 1), %eax
> + andl $(PAGE_SIZE - 1), %ecx
> +
> + xorl %OFFSET_REG, %OFFSET_REG
> +
> + /* Check which is closer to page cross, s1 or s2. */
> + cmpl %eax, %ecx
> + jg L(page_cross_s2)
> +
> + /* The previous page cross check has false positives. Check for
> + true positive as page cross logic is very expensive. */
> + subl $(PAGE_SIZE - VEC_SIZE * 4), %eax
> + jbe L(no_page_cross)
> +
> +
> + /* Set r8 to not interfere with normal return value (rdi and rsi
> + did not swap). */
> # ifdef USE_AS_WCSCMP
> - cmpl %ecx, %eax
> + /* any non-zero positive value that doesn't inference with 0x1.
> + */
> + movl $2, %r8d
> # else
> - subl %ecx, %eax
> + xorl %r8d, %r8d
> # endif
> - jne L(different)
> - addl $SIZE_OF_CHAR, %edx
> - cmpl $(VEC_SIZE * 4), %edx
> - je L(main_loop_header)
> +
> + /* Check if less than 1x VEC till page cross. */
> + subl $(VEC_SIZE * 3), %eax
> + jg L(less_1x_vec_till_page)
> +
> +
> + /* If more than 1x VEC till page cross, loop throuh safely
> + loadable memory until within 1x VEC of page cross. */
> + .p2align 4,, 8
> +L(page_cross_loop):
> + VMOVU (%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %YMM0
> + VPTESTM %YMM0, %YMM0, %k2
> + VPCMP $0, (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %YMM0, %k1{%k2}
> + kmovd %k1, %ecx
> + TESTEQ %ecx
> + jnz L(check_ret_vec_page_cross)
> + addl $CHAR_PER_VEC, %OFFSET_REG
> # ifdef USE_AS_STRNCMP
> - cmpq %r11, %rdx
> - jae L(zero)
> + cmpq %OFFSET_REG64, %rdx
> + jbe L(ret_zero_page_cross)
> # endif
> + addl $VEC_SIZE, %eax
> + jl L(page_cross_loop)
> +
> # ifdef USE_AS_WCSCMP
> - movl (%rdi, %rdx), %eax
> - movl (%rsi, %rdx), %ecx
> -# else
> - movzbl (%rdi, %rdx), %eax
> - movzbl (%rsi, %rdx), %ecx
> + shrl $2, %eax
> # endif
> - /* Check null CHAR. */
> - testl %eax, %eax
> - jne L(cross_page_loop)
> - /* Since %eax == 0, subtract is OK for both SIGNED and UNSIGNED
> - comparisons. */
> - subl %ecx, %eax
> -# ifndef USE_AS_WCSCMP
> -L(different):
> +
> +
> + subl %eax, %OFFSET_REG
> + /* OFFSET_REG has distance to page cross - VEC_SIZE. Guranteed
> + to not cross page so is safe to load. Since we have already
> + loaded at least 1 VEC from rsi it is also guranteed to be safe.
> + */
> + VMOVU (%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %YMM0
> + VPTESTM %YMM0, %YMM0, %k2
> + VPCMP $0, (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %YMM0, %k1{%k2}
> +
> + kmovd %k1, %ecx
> +# ifdef USE_AS_STRNCMP
> + leal CHAR_PER_VEC(%OFFSET_REG64), %eax
> + cmpq %rax, %rdx
> + jbe L(check_ret_vec_page_cross2)
> +# ifdef USE_AS_WCSCMP
> + addq $-(CHAR_PER_VEC * 2), %rdx
> +# else
> + addq %rdi, %rdx
> +# endif
> # endif
> - ret
> + TESTEQ %ecx
> + jz L(prepare_loop_no_len)
>
> + .p2align 4,, 4
> +L(ret_vec_page_cross):
> +# ifndef USE_AS_STRNCMP
> +L(check_ret_vec_page_cross):
> +# endif
> + tzcntl %ecx, %ecx
> + addl %OFFSET_REG, %ecx
> +L(ret_vec_page_cross_cont):
> # ifdef USE_AS_WCSCMP
> - .p2align 4
> -L(different):
> - /* Use movl to avoid modifying EFLAGS. */
> - movl $0, %eax
> + movl (%rdi, %rcx, SIZE_OF_CHAR), %edx
> + xorl %eax, %eax
> + cmpl (%rsi, %rcx, SIZE_OF_CHAR), %edx
> + je L(ret12)
> setl %al
> negl %eax
> - orl $1, %eax
> - ret
> + xorl %r8d, %eax
> +# else
> + movzbl (%rdi, %rcx, SIZE_OF_CHAR), %eax
> + movzbl (%rsi, %rcx, SIZE_OF_CHAR), %ecx
> + subl %ecx, %eax
> + xorl %r8d, %eax
> + subl %r8d, %eax
> # endif
> +L(ret12):
> + ret
> +
>
> # ifdef USE_AS_STRNCMP
> - .p2align 4
> -L(zero):
> + .p2align 4,, 10
> +L(check_ret_vec_page_cross2):
> + TESTEQ %ecx
> +L(check_ret_vec_page_cross):
> + tzcntl %ecx, %ecx
> + addl %OFFSET_REG, %ecx
> + cmpq %rcx, %rdx
> + ja L(ret_vec_page_cross_cont)
> + .p2align 4,, 2
> +L(ret_zero_page_cross):
> xorl %eax, %eax
> ret
> +# endif
>
> - .p2align 4
> -L(char0):
> -# ifdef USE_AS_WCSCMP
> - xorl %eax, %eax
> - movl (%rdi), %ecx
> - cmpl (%rsi), %ecx
> - jne L(wcscmp_return)
> -# else
> - movzbl (%rsi), %ecx
> - movzbl (%rdi), %eax
> - subl %ecx, %eax
> -# endif
> - ret
> + .p2align 4,, 4
> +L(page_cross_s2):
> + /* Ensure this is a true page cross. */
> + subl $(PAGE_SIZE - VEC_SIZE * 4), %ecx
> + jbe L(no_page_cross)
> +
> +
> + movl %ecx, %eax
> + movq %rdi, %rcx
> + movq %rsi, %rdi
> + movq %rcx, %rsi
> +
> + /* set r8 to negate return value as rdi and rsi swapped. */
> +# ifdef USE_AS_WCSCMP
> + movl $-4, %r8d
> +# else
> + movl $-1, %r8d
> # endif
> + xorl %OFFSET_REG, %OFFSET_REG
>
> - .p2align 4
> -L(last_vector):
> - addq %rdx, %rdi
> - addq %rdx, %rsi
> -# ifdef USE_AS_STRNCMP
> - subq %rdx, %r11
> + /* Check if more than 1x VEC till page cross. */
> + subl $(VEC_SIZE * 3), %eax
> + jle L(page_cross_loop)
> +
> + .p2align 4,, 6
> +L(less_1x_vec_till_page):
> +# ifdef USE_AS_WCSCMP
> + shrl $2, %eax
> # endif
> - tzcntl %ecx, %edx
> + /* Find largest load size we can use. */
> + cmpl $(16 / SIZE_OF_CHAR), %eax
> + ja L(less_16_till_page)
> +
> + /* Use 16 byte comparison. */
> + vmovdqu (%rdi), %xmm0
> + VPTESTM %xmm0, %xmm0, %k2
> + VPCMP $0, (%rsi), %xmm0, %k1{%k2}
> + kmovd %k1, %ecx
> # ifdef USE_AS_WCSCMP
> - /* NB: Multiply wchar_t count by 4 to get the number of bytes. */
> - sall $2, %edx
> + subl $0xf, %ecx
> +# else
> + incw %cx
> # endif
> + jnz L(check_ret_vec_page_cross)
> + movl $(16 / SIZE_OF_CHAR), %OFFSET_REG
> # ifdef USE_AS_STRNCMP
> - cmpq %r11, %rdx
> - jae L(zero)
> + cmpq %OFFSET_REG64, %rdx
> + jbe L(ret_zero_page_cross_slow_case0)
> + subl %eax, %OFFSET_REG
> +# else
> + /* Explicit check for 16 byte alignment. */
> + subl %eax, %OFFSET_REG
> + jz L(prepare_loop)
> # endif
> + vmovdqu (%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm0
> + VPTESTM %xmm0, %xmm0, %k2
> + VPCMP $0, (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm0, %k1{%k2}
> + kmovd %k1, %ecx
> # ifdef USE_AS_WCSCMP
> - xorl %eax, %eax
> - movl (%rdi, %rdx), %ecx
> - cmpl (%rsi, %rdx), %ecx
> - jne L(wcscmp_return)
> + subl $0xf, %ecx
> # else
> - movzbl (%rdi, %rdx), %eax
> - movzbl (%rsi, %rdx), %edx
> - subl %edx, %eax
> + incw %cx
> # endif
> + jnz L(check_ret_vec_page_cross)
> +# ifdef USE_AS_STRNCMP
> + addl $(16 / SIZE_OF_CHAR), %OFFSET_REG
> + subq %OFFSET_REG64, %rdx
> + jbe L(ret_zero_page_cross_slow_case0)
> + subq $-(CHAR_PER_VEC * 4), %rdx
> +
> + leaq -(VEC_SIZE * 4)(%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %rdi
> + leaq -(VEC_SIZE * 4)(%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %rsi
> +# else
> + leaq (16 - VEC_SIZE * 4)(%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %rdi
> + leaq (16 - VEC_SIZE * 4)(%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %rsi
> +# endif
> + jmp L(prepare_loop_aligned)
> +
> +# ifdef USE_AS_STRNCMP
> + .p2align 4,, 2
> +L(ret_zero_page_cross_slow_case0):
> + xorl %eax, %eax
> ret
> +# endif
>
> - /* Comparing on page boundary region requires special treatment:
> - It must done one vector at the time, starting with the wider
> - ymm vector if possible, if not, with xmm. If fetching 16 bytes
> - (xmm) still passes the boundary, byte comparison must be done.
> - */
> - .p2align 4
> -L(cross_page):
> - /* Try one ymm vector at a time. */
> - cmpl $(PAGE_SIZE - VEC_SIZE), %eax
> - jg L(cross_page_1_vector)
> -L(loop_1_vector):
> - VMOVU (%rdi, %rdx), %YMM0
>
> - VPTESTM %YMM0, %YMM0, %k2
> - /* Each bit cleared in K1 represents a mismatch or a null CHAR
> - in YMM0 and 32 bytes at (%rsi, %rdx). */
> - VPCMP $0, (%rsi, %rdx), %YMM0, %k1{%k2}
> + .p2align 4,, 10
> +L(less_16_till_page):
> + cmpl $(24 / SIZE_OF_CHAR), %eax
> + ja L(less_8_till_page)
> +
> + /* Use 8 byte comparison. */
> + vmovq (%rdi), %xmm0
> + vmovq (%rsi), %xmm1
> + VPTESTM %xmm0, %xmm0, %k2
> + VPCMP $0, %xmm1, %xmm0, %k1{%k2}
> kmovd %k1, %ecx
> # ifdef USE_AS_WCSCMP
> - subl $0xff, %ecx
> + subl $0x3, %ecx
> # else
> - incl %ecx
> + incb %cl
> # endif
> - jne L(last_vector)
> + jnz L(check_ret_vec_page_cross)
>
> - addl $VEC_SIZE, %edx
>
> - addl $VEC_SIZE, %eax
> # ifdef USE_AS_STRNCMP
> - /* Return 0 if the current offset (%rdx) >= the maximum offset
> - (%r11). */
> - cmpq %r11, %rdx
> - jae L(zero)
> + cmpq $(8 / SIZE_OF_CHAR), %rdx
> + jbe L(ret_zero_page_cross_slow_case0)
> # endif
> - cmpl $(PAGE_SIZE - VEC_SIZE), %eax
> - jle L(loop_1_vector)
> -L(cross_page_1_vector):
> - /* Less than 32 bytes to check, try one xmm vector. */
> - cmpl $(PAGE_SIZE - 16), %eax
> - jg L(cross_page_1_xmm)
> - VMOVU (%rdi, %rdx), %XMM0
> + movl $(24 / SIZE_OF_CHAR), %OFFSET_REG
> + subl %eax, %OFFSET_REG
>
> - VPTESTM %YMM0, %YMM0, %k2
> - /* Each bit cleared in K1 represents a mismatch or a null CHAR
> - in XMM0 and 16 bytes at (%rsi, %rdx). */
> - VPCMP $0, (%rsi, %rdx), %XMM0, %k1{%k2}
> + vmovq (%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm0
> + vmovq (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm1
> + VPTESTM %xmm0, %xmm0, %k2
> + VPCMP $0, %xmm1, %xmm0, %k1{%k2}
> kmovd %k1, %ecx
> # ifdef USE_AS_WCSCMP
> - subl $0xf, %ecx
> + subl $0x3, %ecx
> # else
> - subl $0xffff, %ecx
> + incb %cl
> # endif
> - jne L(last_vector)
> + jnz L(check_ret_vec_page_cross)
> +
>
> - addl $16, %edx
> -# ifndef USE_AS_WCSCMP
> - addl $16, %eax
> -# endif
> # ifdef USE_AS_STRNCMP
> - /* Return 0 if the current offset (%rdx) >= the maximum offset
> - (%r11). */
> - cmpq %r11, %rdx
> - jae L(zero)
> + addl $(8 / SIZE_OF_CHAR), %OFFSET_REG
> + subq %OFFSET_REG64, %rdx
> + jbe L(ret_zero_page_cross_slow_case0)
> + subq $-(CHAR_PER_VEC * 4), %rdx
> +
> + leaq -(VEC_SIZE * 4)(%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %rdi
> + leaq -(VEC_SIZE * 4)(%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %rsi
> +# else
> + leaq (8 - VEC_SIZE * 4)(%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %rdi
> + leaq (8 - VEC_SIZE * 4)(%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %rsi
> # endif
> + jmp L(prepare_loop_aligned)
>
> -L(cross_page_1_xmm):
> -# ifndef USE_AS_WCSCMP
> - /* Less than 16 bytes to check, try 8 byte vector. NB: No need
> - for wcscmp nor wcsncmp since wide char is 4 bytes. */
> - cmpl $(PAGE_SIZE - 8), %eax
> - jg L(cross_page_8bytes)
> - vmovq (%rdi, %rdx), %XMM0
> - vmovq (%rsi, %rdx), %XMM1
>
> - VPTESTM %YMM0, %YMM0, %k2
> - /* Each bit cleared in K1 represents a mismatch or a null CHAR
> - in XMM0 and XMM1. */
> - VPCMP $0, %XMM1, %XMM0, %k1{%k2}
> - kmovb %k1, %ecx
> +
> +
> + .p2align 4,, 10
> +L(less_8_till_page):
> # ifdef USE_AS_WCSCMP
> - subl $0x3, %ecx
> + /* If using wchar then this is the only check before we reach
> + the page boundary. */
> + movl (%rdi), %eax
> + movl (%rsi), %ecx
> + cmpl %ecx, %eax
> + jnz L(ret_less_8_wcs)
> +# ifdef USE_AS_STRNCMP
> + addq $-(CHAR_PER_VEC * 2), %rdx
> + /* We already checked for len <= 1 so cannot hit that case here.
> + */
> +# endif
> + testl %eax, %eax
> + jnz L(prepare_loop)
> + ret
> +
> + .p2align 4,, 8
> +L(ret_less_8_wcs):
> + setl %OFFSET_REG8
> + negl %OFFSET_REG
> + movl %OFFSET_REG, %eax
> + xorl %r8d, %eax
> + ret
> +
> # else
> - subl $0xff, %ecx
> -# endif
> - jne L(last_vector)
> + cmpl $28, %eax
> + ja L(less_4_till_page)
> +
> + vmovd (%rdi), %xmm0
> + vmovd (%rsi), %xmm1
> + VPTESTM %xmm0, %xmm0, %k2
> + VPCMP $0, %xmm1, %xmm0, %k1{%k2}
> + kmovd %k1, %ecx
> + subl $0xf, %ecx
> + jnz L(check_ret_vec_page_cross)
>
> - addl $8, %edx
> - addl $8, %eax
> # ifdef USE_AS_STRNCMP
> - /* Return 0 if the current offset (%rdx) >= the maximum offset
> - (%r11). */
> - cmpq %r11, %rdx
> - jae L(zero)
> + cmpq $4, %rdx
> + jbe L(ret_zero_page_cross_slow_case1)
> # endif
> + movl $(28 / SIZE_OF_CHAR), %OFFSET_REG
> + subl %eax, %OFFSET_REG
>
> -L(cross_page_8bytes):
> - /* Less than 8 bytes to check, try 4 byte vector. */
> - cmpl $(PAGE_SIZE - 4), %eax
> - jg L(cross_page_4bytes)
> - vmovd (%rdi, %rdx), %XMM0
> - vmovd (%rsi, %rdx), %XMM1
> -
> - VPTESTM %YMM0, %YMM0, %k2
> - /* Each bit cleared in K1 represents a mismatch or a null CHAR
> - in XMM0 and XMM1. */
> - VPCMP $0, %XMM1, %XMM0, %k1{%k2}
> + vmovd (%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm0
> + vmovd (%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %xmm1
> + VPTESTM %xmm0, %xmm0, %k2
> + VPCMP $0, %xmm1, %xmm0, %k1{%k2}
> kmovd %k1, %ecx
> -# ifdef USE_AS_WCSCMP
> - subl $0x1, %ecx
> -# else
> subl $0xf, %ecx
> -# endif
> - jne L(last_vector)
> + jnz L(check_ret_vec_page_cross)
> +# ifdef USE_AS_STRNCMP
> + addl $(4 / SIZE_OF_CHAR), %OFFSET_REG
> + subq %OFFSET_REG64, %rdx
> + jbe L(ret_zero_page_cross_slow_case1)
> + subq $-(CHAR_PER_VEC * 4), %rdx
> +
> + leaq -(VEC_SIZE * 4)(%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %rdi
> + leaq -(VEC_SIZE * 4)(%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %rsi
> +# else
> + leaq (4 - VEC_SIZE * 4)(%rdi, %OFFSET_REG64, SIZE_OF_CHAR), %rdi
> + leaq (4 - VEC_SIZE * 4)(%rsi, %OFFSET_REG64, SIZE_OF_CHAR), %rsi
> +# endif
> + jmp L(prepare_loop_aligned)
> +
>
> - addl $4, %edx
> # ifdef USE_AS_STRNCMP
> - /* Return 0 if the current offset (%rdx) >= the maximum offset
> - (%r11). */
> - cmpq %r11, %rdx
> - jae L(zero)
> + .p2align 4,, 2
> +L(ret_zero_page_cross_slow_case1):
> + xorl %eax, %eax
> + ret
> # endif
>
> -L(cross_page_4bytes):
> -# endif
> - /* Less than 4 bytes to check, try one byte/dword at a time. */
> -# ifdef USE_AS_STRNCMP
> - cmpq %r11, %rdx
> - jae L(zero)
> -# endif
> -# ifdef USE_AS_WCSCMP
> - movl (%rdi, %rdx), %eax
> - movl (%rsi, %rdx), %ecx
> -# else
> - movzbl (%rdi, %rdx), %eax
> - movzbl (%rsi, %rdx), %ecx
> -# endif
> - testl %eax, %eax
> - jne L(cross_page_loop)
> + .p2align 4,, 10
> +L(less_4_till_page):
> + subq %rdi, %rsi
> + /* Extremely slow byte comparison loop. */
> +L(less_4_loop):
> + movzbl (%rdi), %eax
> + movzbl (%rsi, %rdi), %ecx
> subl %ecx, %eax
> + jnz L(ret_less_4_loop)
> + testl %ecx, %ecx
> + jz L(ret_zero_4_loop)
> +# ifdef USE_AS_STRNCMP
> + decq %rdx
> + jz L(ret_zero_4_loop)
> +# endif
> + incq %rdi
> + /* end condition is reach page boundary (rdi is aligned). */
> + testl $31, %edi
> + jnz L(less_4_loop)
> + leaq -(VEC_SIZE * 4)(%rdi, %rsi), %rsi
> + addq $-(VEC_SIZE * 4), %rdi
> +# ifdef USE_AS_STRNCMP
> + subq $-(CHAR_PER_VEC * 4), %rdx
> +# endif
> + jmp L(prepare_loop_aligned)
> +
> +L(ret_zero_4_loop):
> + xorl %eax, %eax
> + ret
> +L(ret_less_4_loop):
> + xorl %r8d, %eax
> + subl %r8d, %eax
> ret
> -END (STRCMP)
> +# endif
> +END(STRCMP)
> #endif
> --
> 2.25.1
>
LGTM.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
Thanks.
--
H.J.
next prev parent reply other threads:[~2022-01-10 0:42 UTC|newest]
Thread overview: 59+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-01-09 12:29 [PATCH v1 1/5] x86: Optimize strcmp-avx2.S and fix for [BZ# 28755] Noah Goldstein
2022-01-09 12:29 ` [PATCH v1 2/5] x86: Optimize strcmp-evex.S " Noah Goldstein
2022-01-09 12:29 ` [PATCH v1 3/5] string: remove stupid_[strcmp, strncmp, wcscmp, wcsncmp] Noah Goldstein
2022-01-09 12:29 ` [PATCH v1 4/5] string: Improve coverage in test-strcmp.c and test-strncmp.c Noah Goldstein
2022-01-09 12:29 ` [PATCH v1 5/5] benchtests: Add more coverage for strcmp and strncmp benchmarks Noah Goldstein
2022-01-09 12:35 ` [PATCH v1 1/5] x86: Optimize strcmp-avx2.S and fix for [BZ# 28755] Noah Goldstein
2022-01-09 14:07 ` H.J. Lu
2022-01-10 0:29 ` Noah Goldstein
2022-01-10 0:27 ` [PATCH v2 1/7] x86: Fix __wcsncmp_avx2 in strcmp-avx2.S " Noah Goldstein
2022-01-10 0:27 ` [PATCH v2 2/7] x86: Fix __wcsncmp_evex in strcmp-evex.S " Noah Goldstein
2022-01-10 0:35 ` H.J. Lu
2022-01-10 0:27 ` [PATCH v2 3/7] string/test-str*cmp: remove stupid_[strcmp, strncmp, wcscmp, wcsncmp] Noah Goldstein
2022-01-10 0:37 ` H.J. Lu
2022-01-10 0:27 ` [PATCH v2 4/7] string: Improve coverage in test-strcmp.c and test-strncmp.c Noah Goldstein
2022-01-10 0:38 ` H.J. Lu
2022-01-10 2:51 ` Noah Goldstein
2022-01-10 0:27 ` [PATCH v2 5/7] x86: Optimize strcmp-avx2.S Noah Goldstein
2022-01-10 0:41 ` H.J. Lu
2022-01-10 1:06 ` Noah Goldstein
2022-01-10 1:58 ` H.J. Lu
2022-01-10 2:54 ` Noah Goldstein
2022-01-10 0:27 ` [PATCH v2 6/7] x86: Optimize strcmp-evex.S Noah Goldstein
2022-01-10 0:41 ` H.J. Lu [this message]
2022-01-10 0:27 ` [PATCH v2 7/7] benchtests: Add more coverage for strcmp and strncmp benchmarks Noah Goldstein
2022-01-10 0:34 ` [PATCH v2 1/7] x86: Fix __wcsncmp_avx2 in strcmp-avx2.S [BZ# 28755] H.J. Lu
2022-01-10 21:35 ` [PATCH v3 " Noah Goldstein
2022-01-10 21:35 ` [PATCH v3 2/7] x86: Fix __wcsncmp_evex in strcmp-evex.S " Noah Goldstein
2022-01-11 2:15 ` H.J. Lu
2022-01-26 22:04 ` H.J. Lu
2022-04-29 22:05 ` Sunil Pandey
2022-01-10 21:35 ` [PATCH v3 3/7] string/test-str*cmp: remove stupid_[strcmp, strncmp, wcscmp, wcsncmp] Noah Goldstein
2022-01-10 21:35 ` [PATCH v3 4/7] string: Improve coverage in test-strcmp.c and test-strncmp.c Noah Goldstein
2022-01-10 21:35 ` [PATCH v3 5/7] x86: Optimize strcmp-avx2.S Noah Goldstein
2022-02-14 14:10 ` Andreas Schwab
2022-02-14 18:23 ` H.J. Lu
2022-02-14 19:16 ` Andreas Schwab
2022-02-14 19:30 ` H.J. Lu
2022-02-14 19:35 ` Andreas Schwab
2022-02-14 20:59 ` H.J. Lu
2022-02-14 21:10 ` H.J. Lu
2022-02-15 11:11 ` Andreas Schwab
2022-02-15 12:55 ` Andreas Schwab
2022-02-15 12:58 ` Noah Goldstein
2022-02-15 13:09 ` Noah Goldstein
2022-02-15 13:32 ` Noah Goldstein
2022-02-15 13:37 ` Noah Goldstein
2022-02-15 16:33 ` Noah Goldstein
2022-02-14 23:42 ` Noah Goldstein
2022-02-15 10:43 ` Andreas Schwab
2022-02-15 11:22 ` Andreas Schwab
2022-02-15 11:28 ` Noah Goldstein
2022-02-15 12:24 ` Andreas Schwab
2022-01-10 21:35 ` [PATCH v3 6/7] x86: Optimize strcmp-evex.S Noah Goldstein
2022-01-10 21:35 ` [PATCH v3 7/7] benchtests: Add more coverage for strcmp and strncmp benchmarks Noah Goldstein
2022-01-11 2:15 ` [PATCH v3 1/7] x86: Fix __wcsncmp_avx2 in strcmp-avx2.S [BZ# 28755] H.J. Lu
2022-01-26 22:05 ` H.J. Lu
2022-01-27 4:29 ` H.J. Lu
2022-01-27 5:10 ` H.J. Lu
2022-01-27 5:52 ` Noah Goldstein
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAMe9rOpQkJGLcdK1za2-+staDD3VAb-rFUGezJRuQsROpUDBBQ@mail.gmail.com \
--to=hjl.tools@gmail.com \
--cc=goldstein.w.n@gmail.com \
--cc=libc-alpha@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).