minio/pkg/encoding/erasure/gf-3vect-dot-prod-avx2.asm

306 lines
7.8 KiB
NASM

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2014 Intel 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 Intel 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.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;
;;; gf_3vect_dot_prod_avx2(len, vec, *g_tbls, **buffs, **dests);
;;;
;;; Author: Gregory Tucker
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_3VECT_DOT_PROD_AVX2 _gf_3vect_dot_prod_avx2
%else
%define GF_3VECT_DOT_PROD_AVX2 gf_3vect_dot_prod_avx2
%endif
%ifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg1 rsi
%define arg2 rdx
%define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define func(x) x:
%macro FUNC_SAVE 0
push r12
push r13
%endmacro
%macro FUNC_RESTORE 0
pop r13
pop r12
%endmacro
%endif
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg1 rdx
%define arg2 r8
%define arg3 r9
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 6*16 + 5*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
alloc_stack stack_size
vmovdqa [rsp + 0*16], xmm6
vmovdqa [rsp + 1*16], xmm7
vmovdqa [rsp + 2*16], xmm8
vmovdqa [rsp + 3*16], xmm9
vmovdqa [rsp + 4*16], xmm10
vmovdqa [rsp + 5*16], xmm11
save_reg r12, 6*16 + 0*8
save_reg r13, 6*16 + 1*8
save_reg r14, 6*16 + 2*8
save_reg r15, 6*16 + 3*8
end_prolog
mov arg4, arg(4)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
mov r12, [rsp + 6*16 + 0*8]
mov r13, [rsp + 6*16 + 1*8]
mov r14, [rsp + 6*16 + 2*8]
mov r15, [rsp + 6*16 + 3*8]
add rsp, stack_size
%endmacro
%endif
%define len arg0
%define vec arg1
%define mul_array arg2
%define src arg3
%define dest1 arg4
%define ptr arg5
%define vec_i tmp2
%define dest2 tmp3
%define dest3 tmp4
%define pos return
%ifndef EC_ALIGNED_ADDR
;;; Use Un-aligned load/store
%define XLDR vmovdqu
%define XSTR vmovdqu
%else
;;; Use Non-temporal load/stor
%ifdef NO_NT_LDST
%define XLDR vmovdqa
%define XSTR vmovdqa
%else
%define XLDR vmovntdqa
%define XSTR vmovntdq
%endif
%endif
default rel
[bits 64]
section .text
%define xmask0f ymm11
%define xmask0fx xmm11
%define xgft1_lo ymm10
%define xgft1_hi ymm9
%define xgft2_lo ymm8
%define xgft2_hi ymm7
%define xgft3_lo ymm6
%define xgft3_hi ymm5
%define x0 ymm0
%define xtmpa ymm1
%define xp1 ymm2
%define xp2 ymm3
%define xp3 ymm4
align 16
global GF_3VECT_DOT_PROD_AVX2:function
func(GF_3VECT_DOT_PROD_AVX2)
FUNC_SAVE
sub len, 32
jl .return_fail
xor pos, pos
mov tmp.b, 0x0f
vpinsrb xmask0fx, xmask0fx, tmp.w, 0
vpbroadcastb xmask0f, xmask0fx ;Construct mask 0x0f0f0f...
sal vec, LOG_PS ;vec *= PS. Make vec_i count by PS
mov dest2, [dest1+PS]
mov dest3, [dest1+2*PS]
mov dest1, [dest1]
.loop32:
vpxor xp1, xp1
vpxor xp2, xp2
vpxor xp3, xp3
mov tmp, mul_array
xor vec_i, vec_i
.next_vect:
mov ptr, [src+vec_i]
vmovdqu xgft1_lo, [tmp] ;Load array Ax{00}, Ax{01}, ..., Ax{0f}
; " Ax{00}, Ax{10}, ..., Ax{f0}
vperm2i128 xgft1_hi, xgft1_lo, xgft1_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft1_lo, xgft1_lo, xgft1_lo, 0x00 ; swapped to lo | lo
vmovdqu xgft2_lo, [tmp+vec*(32/PS)] ;Load array Bx{00}, Bx{01}, ..., Bx{0f}
; " Bx{00}, Bx{10}, ..., Bx{f0}
vperm2i128 xgft2_hi, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
vmovdqu xgft3_lo, [tmp+vec*(64/PS)] ;Load array Cx{00}, Cx{01}, ..., Cx{0f}
; " Cx{00}, Cx{10}, ..., Cx{f0}
vperm2i128 xgft3_hi, xgft3_lo, xgft3_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft3_lo, xgft3_lo, xgft3_lo, 0x00 ; swapped to lo | lo
add tmp, 32
add vec_i, PS
XLDR x0, [ptr+pos] ;Get next source vector
vpand xtmpa, x0, xmask0f ;Mask low src nibble in bits 4-0
vpsraw x0, x0, 4 ;Shift to put high nibble into bits 4-0
vpand x0, x0, xmask0f ;Mask high src nibble in bits 4-0
vpshufb xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft1_hi, xgft1_lo ;GF add high and low partials
vpxor xp1, xgft1_hi ;xp1 += partial
vpshufb xgft2_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft2_hi, xgft2_lo ;GF add high and low partials
vpxor xp2, xgft2_hi ;xp2 += partial
vpshufb xgft3_hi, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xgft3_hi, xgft3_lo ;GF add high and low partials
vpxor xp3, xgft3_hi ;xp3 += partial
cmp vec_i, vec
jl .next_vect
XSTR [dest1+pos], xp1
XSTR [dest2+pos], xp2
XSTR [dest3+pos], xp3
add pos, 32 ;Loop on 32 bytes at a time
cmp pos, len
jle .loop32
lea tmp, [len + 32]
cmp pos, tmp
je .return_pass
;; Tail len
mov pos, len ;Overlapped offset length-16
jmp .loop32 ;Do one more overlap pass
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
%macro slversion 4
global %1_slver_%2%3%4
global %1_slver
%1_slver:
%1_slver_%2%3%4:
dw 0x%4
db 0x%3, 0x%2
%endmacro
;;; func core, ver, snum
slversion GF_3VECT_DOT_PROD_AVX2, 04, 03, 0197