minio/pkg/erasure/gf_3vect_mad_avx2.asm

348 lines
9.5 KiB
NASM

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2015 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_mad_avx2(len, vec, vec_i, mul_array, src, dest);
;;;
%ifidn __OUTPUT_FORMAT__, macho64
%define GF_3VECT_MAD_AVX2 _gf_3vect_mad_avx2
%else
%define GF_3VECT_MAD_AVX2 gf_3vect_mad_avx2
%endif
%define PS 8
%ifidn __OUTPUT_FORMAT__, win64
%define arg0 rcx
%define arg0.w ecx
%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 return rax
%define return.w eax
%define stack_size 16*10 + 3*8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
%macro FUNC_SAVE 0
sub rsp, stack_size
vmovdqa [rsp+16*0],xmm6
vmovdqa [rsp+16*1],xmm7
vmovdqa [rsp+16*2],xmm8
vmovdqa [rsp+16*3],xmm9
vmovdqa [rsp+16*4],xmm10
vmovdqa [rsp+16*5],xmm11
vmovdqa [rsp+16*6],xmm12
vmovdqa [rsp+16*7],xmm13
vmovdqa [rsp+16*8],xmm14
vmovdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r15, 10*16 + 1*8
end_prolog
mov arg4, arg(4)
mov arg5, arg(5)
%endmacro
%macro FUNC_RESTORE 0
vmovdqa xmm6, [rsp+16*0]
vmovdqa xmm7, [rsp+16*1]
vmovdqa xmm8, [rsp+16*2]
vmovdqa xmm9, [rsp+16*3]
vmovdqa xmm10, [rsp+16*4]
vmovdqa xmm11, [rsp+16*5]
vmovdqa xmm12, [rsp+16*6]
vmovdqa xmm13, [rsp+16*7]
vmovdqa xmm14, [rsp+16*8]
vmovdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r15, [rsp + 10*16 + 1*8]
add rsp, stack_size
%endmacro
%elifidn __OUTPUT_FORMAT__, elf64
%define arg0 rdi
%define arg0.w edi
%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 return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%elifidn __OUTPUT_FORMAT__, macho64
%define arg0 rdi
%define arg0.w edi
%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 return rax
%define return.w eax
%define func(x) x:
%define FUNC_SAVE
%define FUNC_RESTORE
%endif
;;; gf_3vect_mad_avx2(len, vec, vec_i, mul_array, src, dest)
%define len arg0
%define len.w arg0.w
%define vec arg1
%define vec_i arg2
%define mul_array arg3
%define src arg4
%define dest1 arg5
%define pos return
%define pos.w return.w
%define dest2 mul_array
%define dest3 vec_i
%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 ymm15
%define xmask0fx xmm15
%define xgft1_lo ymm14
%define xgft1_hi ymm13
%define xgft2_lo ymm12
%define xgft3_lo ymm11
%define x0 ymm0
%define xtmpa ymm1
%define xtmph1 ymm2
%define xtmpl1 ymm3
%define xtmph2 ymm4
%define xtmpl2 ymm5
%define xtmpl2x xmm5
%define xtmph3 ymm6
%define xtmpl3 ymm7
%define xtmpl3x xmm7
%define xd1 ymm8
%define xd2 ymm9
%define xd3 ymm10
align 16
global GF_3VECT_MAD_AVX2:function
func(GF_3VECT_MAD_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_i, 5 ;Multiply by 32
sal vec, 5
lea tmp, [mul_array + 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] ;Load array Bx{00}, Bx{01}, Bx{02}, ...
; " Bx{00}, Bx{10}, Bx{20}, ... , Bx{f0}
vmovdqu xgft3_lo, [tmp+2*vec] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
; " Cx{00}, Cx{10}, Cx{20}, ... , Cx{f0}
mov dest2, [dest1+PS] ; reuse mul_array
mov dest3, [dest1+2*PS] ; reuse vec_i
mov dest1, [dest1]
.loop32:
XLDR x0, [src+pos] ;Get next source vector
XLDR xd1, [dest1+pos] ;Get next dest vector
XLDR xd2, [dest2+pos] ;Get next dest vector
XLDR xd3, [dest3+pos] ;Get next dest vector
vperm2i128 xtmph2, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xtmpl2, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
vperm2i128 xtmph3, xgft3_lo, xgft3_lo, 0x11 ; swapped to hi | hi
vperm2i128 xtmpl3, xgft3_lo, xgft3_lo, 0x00 ; swapped to lo | lo
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
; dest1
vpshufb xtmph1, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpxor xd1, xd1, xtmph1 ;xd1 += partial
; dest2
vpshufb xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xtmpl2, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmpl2 ;GF add high and low partials
vpxor xd2, xtmph2 ;xd2 += partial
; dest3
vpshufb xtmph3, x0 ;Lookup mul table of high nibble
vpshufb xtmpl3, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmpl3 ;GF add high and low partials
vpxor xd3, xtmph3 ;xd3 += partial
XSTR [dest1+pos], xd1
XSTR [dest2+pos], xd2
XSTR [dest3+pos], xd3
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
.lessthan32:
;; Tail len
;; Do one more overlap pass
mov tmp.b, 0x1f
vpinsrb xtmpl2x, xtmpl2x, tmp.w, 0
vpbroadcastb xtmpl2, xtmpl2x ;Construct mask 0x1f1f1f...
mov tmp, len ;Overlapped offset length-32
XLDR x0, [src+tmp] ;Get next source vector
XLDR xd1, [dest1+tmp] ;Get next dest vector
XLDR xd2, [dest2+tmp] ;Get next dest vector
XLDR xd3, [dest3+tmp] ;Get next dest vector
sub len, pos
vmovdqa xtmph3, [constip32] ;Load const of i + 32
vpinsrb xtmpl3x, xtmpl3x, len.w, 15
vinserti128 xtmpl3, xtmpl3, xtmpl3x, 1 ;swapped to xtmpl3x | xtmpl3x
vpshufb xtmpl3, xtmpl3, xtmpl2 ;Broadcast len to all bytes. xtmpl2=0x1f1f1f...
vpcmpgtb xtmpl3, xtmpl3, xtmph3
vperm2i128 xtmph2, xgft2_lo, xgft2_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft2_lo, xgft2_lo, xgft2_lo, 0x00 ; swapped to lo | lo
vperm2i128 xtmph3, xgft3_lo, xgft3_lo, 0x11 ; swapped to hi | hi
vperm2i128 xgft3_lo, xgft3_lo, xgft3_lo, 0x00 ; swapped to lo | lo
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
; dest1
vpshufb xtmph1, xgft1_hi, x0 ;Lookup mul table of high nibble
vpshufb xtmpl1, xgft1_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph1, xtmph1, xtmpl1 ;GF add high and low partials
vpand xtmph1, xtmph1, xtmpl3
vpxor xd1, xd1, xtmph1 ;xd1 += partial
; dest2
vpshufb xtmph2, xtmph2, x0 ;Lookup mul table of high nibble
vpshufb xgft2_lo, xgft2_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph2, xtmph2, xgft2_lo ;GF add high and low partials
vpand xtmph2, xtmph2, xtmpl3
vpxor xd2, xd2, xtmph2 ;xd2 += partial
; dest3
vpshufb xtmph3, xtmph3, x0 ;Lookup mul table of high nibble
vpshufb xgft3_lo, xgft3_lo, xtmpa ;Lookup mul table of low nibble
vpxor xtmph3, xtmph3, xgft3_lo ;GF add high and low partials
vpand xtmph3, xtmph3, xtmpl3
vpxor xd3, xd3, xtmph3 ;xd3 += partial
XSTR [dest1+tmp], xd1
XSTR [dest2+tmp], xd2
XSTR [dest3+tmp], xd3
.return_pass:
mov return, 0
FUNC_RESTORE
ret
.return_fail:
mov return, 1
FUNC_RESTORE
ret
endproc_frame
section .data
align 32
constip32:
ddq 0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
ddq 0xe0e1e2e3e4e5e6e7e8e9eaebecedeeef
%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_MAD_AVX2, 04, 00, 0208