isa-l/crc/aarch64/crc32_refl_common_pmull.h
Zhiyuan Zhu a46da529d9 crc: optimize crc with arm64 assembly
Change-Id: I49166ee06b3ad24babb90aeb0b834d8aacfc2d03
Signed-off-by: Zhiyuan Zhu <zhiyuan.zhu@arm.com>
2019-06-21 17:02:16 +08:00

281 lines
6.0 KiB
C

########################################################################
# Copyright(c) 2019 Arm Corporation All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions
# are met:
# * Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in
# the documentation and/or other materials provided with the
# distribution.
# * Neither the name of Arm Corporation nor the names of its
# contributors may be used to endorse or promote products derived
# from this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#########################################################################
.macro crc32_refl_func name:req
.arch armv8-a+crc+crypto
.text
.align 3
.global \name
.type \name, %function
/* crc32_refl_func(uint32_t seed, uint8_t * buf, uint64_t len) */
// constant
.equ FOLD_SIZE, 1024
// paramter
w_seed .req w0
x_seed .req x0
x_buf .req x1
x_len .req x2
x_buf_tmp .req x0
// crc32 refl function entry
\name\():
mvn w_seed, w_seed
mov x3, 0
mov w4, 0
cmp x_len, (FOLD_SIZE - 1)
bhi .crc32_clmul_pre
.crc32_refl_tab_pre:
cmp x_len, x3
bls .done
sxtw x4, w4
adrp x5, .LANCHOR0
sub x_buf, x_buf, x4
add x5, x5, :lo12:.LANCHOR0
.align 3
.loop_crc32_refl_tab:
ldrb w3, [x_buf, x4]
add x4, x4, 1
cmp x_len, x4
eor x3, x_seed, x3
and x3, x3, 255
ldr w3, [x5, x3, lsl 2]
eor w_seed, w3, w_seed, lsr 8
bhi .loop_crc32_refl_tab
.done:
mvn w_seed, w_seed
ret
d_y0_tmp .req d0
v_y0_tmp .req v0
q_x0_tmp .req q3
v_x0_tmp .req v3
v_x0 .req v0
q_x1 .req q2
q_x2 .req q4
q_x3 .req q1
d_p4_low .req d17
d_p4_high .req d19
x_buf_end .req x3
.align 2
.crc32_clmul_pre:
and x4, x_len, -64
uxtw x_seed, w_seed
cmp x4, 63
bls .clmul_end
fmov d_y0_tmp, x_seed
ins v_y0_tmp.d[1], x3
ldr q_x0_tmp, [x_buf]
ldr q_x1, [x_buf, 16]
ldr q_x2, [x_buf, 32]
ldr q_x3, [x_buf, 48]
eor v_x0.16b, v_y0_tmp.16b, v_x0_tmp.16b
sub x5, x4, #64
cmp x5, 63
add x_buf_tmp, x_buf, 64
bls .clmul_loop_end
mov x4, p4_high_b0
movk x4, p4_high_b1, lsl 16
fmov d_p4_high, x4
mov x4, p4_low_b0
movk x4, p4_low_b1, lsl 16
fmov d_p4_low, x4
add x_buf_end, x_buf_tmp, x5
v_p4_low .req v17
v_p4_high .req v19
// v_x0 .req v0
v_x1 .req v2
v_x2 .req v4
v_x3 .req v1
q_y0 .req q7
q_y1 .req q5
q_y2 .req q3
q_y3 .req q21
v_y0 .req v7
v_y1 .req v5
v_y2 .req v3
v_y3 .req v21
d_x0_h .req d22
d_x1_h .req d20
d_x2_h .req d18
d_x3_h .req d6
v_x0_h .req v22
v_x1_h .req v20
v_x2_h .req v18
v_x3_h .req v6
.align 3
.clmul_loop:
add x_buf_tmp, x_buf_tmp, 64
cmp x_buf_tmp, x_buf_end
dup d_x0_h, v_x0.d[1]
dup d_x1_h, v_x1.d[1]
dup d_x2_h, v_x2.d[1]
dup d_x3_h, v_x3.d[1]
ldr q_y0, [x_buf_tmp, -64]
ldr q_y1, [x_buf_tmp, -48]
ldr q_y2, [x_buf_tmp, -32]
ldr q_y3, [x_buf_tmp, -16]
pmull v_x0.1q, v_x0.1d, v_p4_low.1d
pmull v_x1.1q, v_x1.1d, v_p4_low.1d
pmull v_x2.1q, v_x2.1d, v_p4_low.1d
pmull v_x3.1q, v_x3.1d, v_p4_low.1d
pmull v_x0_h.1q, v_x0_h.1d, v_p4_high.1d
pmull v_x1_h.1q, v_x1_h.1d, v_p4_high.1d
pmull v_x2_h.1q, v_x2_h.1d, v_p4_high.1d
pmull v_x3_h.1q, v_x3_h.1d, v_p4_high.1d
eor v_y0.16b, v_y0.16b, v22.16b
eor v_y1.16b, v_y1.16b, v20.16b
eor v_y2.16b, v_y2.16b, v18.16b
eor v_y3.16b, v_y3.16b, v6.16b
eor v_x0.16b, v_y0.16b, v_x0.16b
eor v_x1.16b, v_y1.16b, v_x1.16b
eor v_x2.16b, v_y2.16b, v_x2.16b
eor v_x3.16b, v_y3.16b, v_x3.16b
bne .clmul_loop
// v_x0 .req v0
// v_x1 .req v2
// v_x2 .req v4
// v_x3 .req v1
d_x0 .req d0
d_p1_high .req d7
d_p1_low .req d17
v_p1_high .req v7
v_p1_low .req v17
.clmul_loop_end:
// fold 128b
mov x0, p1_high_b0
movk x0, p1_high_b1, lsl 16
fmov d_p1_high, x0
mov x0, p1_low_b0
movk x0, p1_low_b1, lsl 16
fmov d_p1_low, x0
dup d6, v_x0.d[1]
pmull v_x0.1q, v_x0.1d, v_p1_low.1d
pmull v6.1q, v6.1d, v_p1_high.1d
eor v6.16b, v6.16b, v_x0.16b
eor v_x1.16b, v6.16b, v_x1.16b
dup d6, v_x1.d[1]
pmull v_x1.1q, v_x1.1d, v_p1_low.1d
pmull v6.1q, v6.1d, v_p1_high.1d
eor v6.16b, v6.16b, v_x1.16b
eor v_x2.16b, v6.16b, v_x2.16b
dup d_x0, v_x2.d[1] // d_x0 temparory saved v_x2 high
pmull v_x2.1q, v_x2.1d, v_p1_low.1d
pmull v_x0.1q, v_x0.1d, v_p1_high.1d
eor v_x0.16b, v_x0.16b, v_x2.16b
eor v_x0.16b, v_x0.16b, v_x3.16b
// all
mov x0, 4294967295
fmov d3, x0
movi v5.4s, 0
// fold 64b
mov x4, p0_low_b0
movk x4, p0_low_b1, lsl 16
fmov d1, x4
dup d2, v0.d[0]
ext v0.16b, v0.16b, v5.16b, #8
pmull v2.1q, v2.1d, v7.1d
eor v0.16b, v0.16b, v2.16b
and v2.16b, v3.16b, v0.16b
ext v0.16b, v0.16b, v5.16b, #4
pmull v2.1q, v2.1d, v1.1d
// barrett reduction
mov x3, br_high_b0
movk x3, br_high_b1, lsl 16
movk x3, br_high_b2, lsl 32
fmov d1, x3
eor v0.16b, v0.16b, v2.16b
and v2.16b, v0.16b, v3.16b
pmull v2.1q, v2.1d, v1.1d
mov x0, br_low_b0
movk x0, br_low_b1, lsl 16
movk x0, br_low_b2, lsl 32
fmov d1, x0
and v2.16b, v3.16b, v2.16b
pmull v2.1q, v2.1d, v1.1d
eor v0.16b, v0.16b, v2.16b
umov w_seed, v0.s[1]
uxtw x_seed, w_seed
.clmul_end:
and w4, w2, -64
sxtw x3, w4
add x_buf, x_buf, x3
b .crc32_refl_tab_pre
.size \name, .-\name
.endm