mpir/mpn/x86_64/atom/lshift.as

175 lines
3.9 KiB
ActionScript

; AMD64 mpn_lshift
; Copyright 2009 Jason Moxham
; This file is part of the MPIR Library.
; The MPIR Library is free software; you can redistribute it and/or modify
; it under the terms of the GNU Lesser General Public License as published
; by the Free Software Foundation; either version 2.1 of the License, or (at
; your option) any later version.
; The MPIR Library is distributed in the hope that it will be useful, but
; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
; License for more details.
; You should have received a copy of the GNU Lesser General Public License
; along with the MPIR Library; see the file COPYING.LIB. If not, write
; to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
; Boston, MA 02110-1301, USA.
; (rdi,rdx) = (rsi,rdx)<<rcx
; rax = carry
; decent assmeblers understand what movq means ,except
; microsofts/apple masm (what a suprise there) so for the broken old masm
; assembler. Needed for movq reg64,mediareg and movq mediareg,reg64
; only , where mediareg is xmm or mm
%include 'yasm_mac.inc'
%define MOVQ movd
BITS 64
GLOBAL_FUNC mpn_lshift
cmp rdx, 2
ja threeormore
jz two
one:
mov rdx, [rsi]
mov rax, rdx
shl rdx, cl
neg rcx
shr rax, cl
mov [rdi], rdx
ret
two:
mov r8, [rsi]
mov r9, [rsi+8]
mov r11, r8
mov rax, r9
shl r8, cl
shl r9, cl
neg rcx
shr r11, cl
shr rax, cl
or r9, r11
mov [rdi], r8
mov [rdi+8], r9
ret
threeormore:
mov eax, 64
sub rax, rcx
MOVQ xmm0, rcx
MOVQ xmm1, rax
mov r8, rdx
lea r9, [rsi+r8*8-16]
mov r10, r9
and r9, -16
movdqa xmm3, [r9]
movdqa xmm5, xmm3
psrlq xmm3, xmm1
pshufd xmm3, xmm3, 0x4E
MOVQ rax, xmm3
cmp r10, r9
je aligned
movq xmm2, [rsi+r8*8-8]
movq xmm4, xmm2
psrlq xmm2, xmm1
MOVQ rax, xmm2
psllq xmm4, xmm0
por xmm4, xmm3
movq [rdi+r8*8-8], xmm4
dec r8
aligned:
sub r8, 5
jle skiploop
align 16
loop1:
movdqa xmm2, [rsi+r8*8+8]
movdqa xmm4, xmm2
psllq xmm5, xmm0
psrlq xmm2, xmm1
movhlps xmm3, xmm2
por xmm5, xmm3
movq [rdi+r8*8+24], xmm5
pshufd xmm2, xmm2, 0x4E
movhpd [rdi+r8*8+32], xmm5
movdqa xmm3, [rsi+r8*8-8]
movdqa xmm5, xmm3
psrlq xmm3, xmm1
movhlps xmm2, xmm3
psllq xmm4, xmm0
pshufd xmm3, xmm3, 0x4E
por xmm4, xmm2
movq [rdi+r8*8+8], xmm4
movhpd [rdi+r8*8+16], xmm4
sub r8, 4
jg loop1
skiploop:
cmp r8, -1
je left2
jg left3
jp left1
left0:
; may be easier to bswap xmm5 first , same with other cases
pxor xmm2, xmm2
psllq xmm5, xmm0
movhlps xmm3, xmm2
por xmm5, xmm3
movq [rdi+r8*8+24], xmm5
movhpd [rdi+r8*8+32], xmm5
ret
align 16
left3:
movdqa xmm2, [rsi+r8*8+8]
movdqa xmm4, xmm2
psllq xmm5, xmm0
psrlq xmm2, xmm1
movhlps xmm3, xmm2
por xmm5, xmm3
movq [rdi+r8*8+24], xmm5
pshufd xmm2, xmm2, 0x4E
movhpd [rdi+r8*8+32], xmm5
movq xmm3, [rsi+r8*8]
pshufd xmm3, xmm3, 0x4E
movdqa xmm5, xmm3
psrlq xmm3, xmm1
movhlps xmm2, xmm3
psllq xmm4, xmm0
por xmm4, xmm2
movq [rdi+r8*8+8], xmm4
movhpd [rdi+r8*8+16], xmm4
psllq xmm5, xmm0
movhpd [rdi+r8*8], xmm5
ret
align 16
left2:
movdqa xmm2, [rsi+r8*8+8]
movdqa xmm4, xmm2
psllq xmm5, xmm0
psrlq xmm2, xmm1
movhlps xmm3, xmm2
por xmm5, xmm3
movq [rdi+r8*8+24], xmm5
pshufd xmm2, xmm2, 0x4E
movhpd [rdi+r8*8+32], xmm5
pxor xmm3, xmm3
movhlps xmm2, xmm3
psllq xmm4, xmm0
por xmm4, xmm2
movq [rdi+r8*8+8], xmm4
movhpd [rdi+r8*8+16], xmm4
ret
align 16
left1:
movq xmm2, [rsi+r8*8+16]
pshufd xmm2, xmm2, 0x4E
movdqa xmm4, xmm2
psllq xmm5, xmm0
psrlq xmm2, xmm1
movhlps xmm3, xmm2
por xmm5, xmm3
movq [rdi+r8*8+24], xmm5
movhpd [rdi+r8*8+32], xmm5
psllq xmm4, xmm0
movhpd [rdi+r8*8+16], xmm4
ret