dnl AMD64 mpn_sqr_basecase optimised for Intel Haswell.
dnl Contributed to the GNU project by Torbjörn Granlund.
dnl Copyright 2008, 2009, 2011-2013 Free Software Foundation, Inc.
dnl This file is part of the GNU MP Library.
dnl
dnl The GNU MP Library is free software; you can redistribute it and/or modify
dnl it under the terms of either:
dnl
dnl * the GNU Lesser General Public License as published by the Free
dnl Software Foundation; either version 3 of the License, or (at your
dnl option) any later version.
dnl
dnl or
dnl
dnl * the GNU General Public License as published by the Free Software
dnl Foundation; either version 2 of the License, or (at your option) any
dnl later version.
dnl
dnl or both in parallel, as here.
dnl
dnl The GNU MP Library is distributed in the hope that it will be useful, but
dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
dnl for more details.
dnl
dnl You should have received copies of the GNU General Public License and the
dnl GNU Lesser General Public License along with the GNU MP Library. If not,
dnl see https://www.gnu.org/licenses/.
include(`../config.m4')
C cycles/limb mul_2 addmul_2 sqr_diag_addlsh1
C AMD K8,K9 n/a n/a n/a
C AMD K10 n/a n/a n/a
C AMD bull n/a n/a n/a
C AMD pile n/a n/a n/a
C AMD steam ? ? ?
C AMD bobcat n/a n/a n/a
C AMD jaguar ? ? ?
C Intel P4 n/a n/a n/a
C Intel core n/a n/a n/a
C Intel NHM n/a n/a n/a
C Intel SBR n/a n/a n/a
C Intel IBR n/a n/a n/a
C Intel HWL 1.86 2.15 ~2.5
C Intel BWL ? ? ?
C Intel atom n/a n/a n/a
C VIA nano n/a n/a n/a
C The inner loops of this code are the result of running a code generation and
C optimisation tool suite written by David Harvey and Torbjörn Granlund, except
C that the sqr_diag_addlsh1 loop was manually written.
C TODO
C * Replace current unoptimised sqr_diag_addlsh1 loop; 1.75 c/l might be
C possible.
C * Consider splitting outer loop into 2, one for n = 1 (mod 2) and one for
C n = 0 (mod 2). These loops could fall into specific "corner" code.
C * Consider splitting outer loop into 4.
C * Streamline pointer updates.
C * Perhaps suppress a few more xor insns in feed-in code.
C * Make sure we write no dead registers in feed-in code.
C * We might use 32-bit size ops, since n >= 2^32 is non-terminating. Watch
C out for negative sizes being zero-extended, though.
C * Provide straight-line code for n = 4; then look for simplifications in
C main code.
define(`rp', `%rdi')
define(`up', `%rsi')
define(`un_param',`%rdx')
ABI_SUPPORT(DOS64)
ABI_SUPPORT(STD64)
ASM_START()
TEXT
ALIGN(32)
PROLOGUE(mpn_sqr_basecase)
FUNC_ENTRY(3)
cmp $2, un_param
jae L(gt1)
mov (up), %rdx
mulx( %rdx, %rax, %rdx)
mov %rax, (rp)
mov %rdx, 8(rp)
FUNC_EXIT()
ret
L(gt1): jne L(gt2)
mov (up), %rdx
mov 8(up), %rcx
mulx( %rcx, %r9, %r10) C v0 * v1 W 1 2
mulx( %rdx, %rax, %r8) C v0 * v0 W 0 1
mov %rcx, %rdx
mulx( %rdx, %r11, %rdx) C v1 * v1 W 2 3
add %r9, %r9 C W 1
adc %r10, %r10 C W 2
adc $0, %rdx C W 3
add %r9, %r8 C W 1
adc %r11, %r10 C W 2
adc $0, %rdx C W 3
mov %rax, (rp)
mov %r8, 8(rp)
mov %r10, 16(rp)
mov %rdx, 24(rp)
FUNC_EXIT()
ret
L(gt2): cmp $4, un_param
jae L(gt3)
define(`v0', `%r8')
define(`v1', `%r9')
define(`w0', `%r10')
define(`w2', `%r11')
mov (up), v0
mov 8(up), %rdx
mov %rdx, v1
mulx( v0, w2, %rax)
mov 16(up), %rdx
mulx( v0, w0, %rcx)
mov w2, %r8
add %rax, w0
adc $0, %rcx
mulx( v1, %rdx, %rax)
add %rcx, %rdx
mov %rdx, 24(rp)
adc $0, %rax
mov %rax, 32(rp)
xor R32(%rcx), R32(%rcx)
mov (up), %rdx
mulx( %rdx, %rax, w2)
mov %rax, (rp)
add %r8, %r8
adc w0, w0
setc R8(%rcx)
mov 8(up), %rdx
mulx( %rdx, %rax, %rdx)
add w2, %r8
adc %rax, w0
mov %r8, 8(rp)
mov w0, 16(rp)
mov 24(rp), %r8
mov 32(rp), w0
lea (%rdx,%rcx), w2
adc %r8, %r8
adc w0, w0
setc R8(%rcx)
mov 16(up), %rdx
mulx( %rdx, %rax, %rdx)
add w2, %r8
adc %rax, w0
mov %r8, 24(rp)
mov w0, 32(rp)
adc %rcx, %rdx
mov %rdx, 40(rp)
FUNC_EXIT()
ret
L(gt3):
define(`v0', `%r8')
define(`v1', `%r9')
define(`w0', `%r10')
define(`w1', `%r11')
define(`w2', `%rbx')
define(`w3', `%rbp')
define(`un', `%r12')
define(`n', `%rcx')
define(`X0', `%r13')
define(`X1', `%r14')
L(do_mul_2):
push %rbx
push %rbp
push %r12
push %r13
push %r14
mov $0, R32(un)
sub un_param, un C free up rdx
push un
mov (up), v0
mov 8(up), %rdx
lea 2(un), n
sar $2, n C FIXME: suppress, change loop?
inc un C decrement |un|
mov %rdx, v1
test $1, R8(un)
jnz L(mx1)
L(mx0): mulx( v0, w2, w1)
mov 16(up), %rdx
mov w2, 8(rp)
xor w2, w2
mulx( v0, w0, w3)
test $2, R8(un)
jz L(m00)
L(m10): lea -8(rp), rp
lea -8(up), up
jmp L(mlo2)
L(m00): lea 8(up), up
lea 8(rp), rp
jmp L(mlo0)
L(mx1): mulx( v0, w0, w3)
mov 16(up), %rdx
mov w0, 8(rp)
xor w0, w0
mulx( v0, w2, w1)
test $2, R8(un)
jz L(mlo3)
L(m01): lea 16(rp), rp
lea 16(up), up
jmp L(mlo1)
ALIGN(32)
L(mtop):mulx( v1, %rax, w0)
add %rax, w2 C 0
mov (up), %rdx
mulx( v0, %rax, w1)
adc $0, w0 C 1
add %rax, w2 C 0
L(mlo1):adc $0, w1 C 1
add w3, w2 C 0
mov w2, (rp) C 0
adc $0, w1 C 1
mulx( v1, %rax, w2)
add %rax, w0 C 1
mov 8(up), %rdx
adc $0, w2 C 2
mulx( v0, %rax, w3)
add %rax, w0 C 1
adc $0, w3 C 2
L(mlo0):add w1, w0 C 1
mov w0, 8(rp) C 1
adc $0, w3 C 2
mulx( v1, %rax, w0)
add %rax, w2 C 2
mov 16(up), %rdx
mulx( v0, %rax, w1)
adc $0, w0 C 3
add %rax, w2 C 2
adc $0, w1 C 3
L(mlo3):add w3, w2 C 2
mov w2, 16(rp) C 2
adc $0, w1 C 3
mulx( v1, %rax, w2)
add %rax, w0 C 3
mov 24(up), %rdx
adc $0, w2 C 4
mulx( v0, %rax, w3)
add %rax, w0 C 3
adc $0, w3 C 4
L(mlo2):add w1, w0 C 3
lea 32(up), up
mov w0, 24(rp) C 3
adc $0, w3 C 4
inc n
lea 32(rp), rp
jnz L(mtop)
L(mend):mulx( v1, %rdx, %rax)
add %rdx, w2
adc $0, %rax
add w3, w2
mov w2, (rp)
adc $0, %rax
mov %rax, 8(rp)
lea 16(up), up
lea -16(rp), rp
L(do_addmul_2):
L(outer):
lea (up,un,8), up C put back up to 2 positions above last time
lea 48(rp,un,8), rp C put back rp to 4 positions above last time
mov -8(up), v0 C shared between addmul_2 and corner
add $2, un C decrease |un|
cmp $-2, un
jge L(corner)
mov (up), v1
lea 1(un), n
sar $2, n C FIXME: suppress, change loop?
mov v1, %rdx
test $1, R8(un)
jnz L(bx1)
L(bx0): mov (rp), X0
mov 8(rp), X1
mulx( v0, %rax, w1)
add %rax, X0
adc $0, w1
mov X0, (rp)
xor w2, w2
test $2, R8(un)
jnz L(b10)
L(b00): mov 8(up), %rdx
lea 16(rp), rp
lea 16(up), up
jmp L(lo0)
L(b10): mov 8(up), %rdx
mov 16(rp), X0
lea 32(up), up
inc n
mulx( v0, %rax, w3)
jz L(ex)
jmp L(lo2)
L(bx1): mov (rp), X1
mov 8(rp), X0
mulx( v0, %rax, w3)
mov 8(up), %rdx
add %rax, X1
adc $0, w3
xor w0, w0
mov X1, (rp)
mulx( v0, %rax, w1)
test $2, R8(un)
jz L(b11)
L(b01): mov 16(rp), X1
lea 24(rp), rp
lea 24(up), up
jmp L(lo1)
L(b11): lea 8(rp), rp
lea 8(up), up
jmp L(lo3)
ALIGN(32)
L(top): mulx( v0, %rax, w3)
add w0, X1
adc $0, w2
L(lo2): add %rax, X1
adc $0, w3
mulx( v1, %rax, w0)
add %rax, X0
adc $0, w0
lea 32(rp), rp
add w1, X1
mov -16(up), %rdx
mov X1, -24(rp)
adc $0, w3
add w2, X0
mov -8(rp), X1
mulx( v0, %rax, w1)
adc $0, w0
L(lo1): add %rax, X0
mulx( v1, %rax, w2)
adc $0, w1
add w3, X0
mov X0, -16(rp)
adc $0, w1
add %rax, X1
adc $0, w2
add w0, X1
mov -8(up), %rdx
adc $0, w2
L(lo0): mulx( v0, %rax, w3)
add %rax, X1
adc $0, w3
mov (rp), X0
mulx( v1, %rax, w0)
add %rax, X0
adc $0, w0
add w1, X1
mov X1, -8(rp)
adc $0, w3
mov (up), %rdx
add w2, X0
mulx( v0, %rax, w1)
adc $0, w0
L(lo3): add %rax, X0
adc $0, w1
mulx( v1, %rax, w2)
add w3, X0
mov 8(rp), X1
mov X0, (rp)
mov 16(rp), X0
adc $0, w1
add %rax, X1
adc $0, w2
mov 8(up), %rdx
lea 32(up), up
inc n
jnz L(top)
L(end): mulx( v0, %rax, w3)
add w0, X1
adc $0, w2
L(ex): add %rax, X1
adc $0, w3
mulx( v1, %rdx, %rax)
add w1, X1
mov X1, 8(rp)
adc $0, w3
add w2, %rdx
adc $0, %rax
add %rdx, w3
mov w3, 16(rp)
adc $0, %rax
mov %rax, 24(rp)
jmp L(outer) C loop until a small corner remains
L(corner):
pop un
mov (up), %rdx
jg L(small_corner)
mov %rdx, v1
mov (rp), X0
mov %rax, X1 C Tricky rax reuse of last iteration
mulx( v0, %rax, w1)
add %rax, X0
adc $0, w1
mov X0, (rp)
mov 8(up), %rdx
mulx( v0, %rax, w3)
add %rax, X1
adc $0, w3
mulx( v1, %rdx, %rax)
add w1, X1
mov X1, 8(rp)
adc $0, w3
add w3, %rdx
mov %rdx, 16(rp)
adc $0, %rax
mov %rax, 24(rp)
lea 32(rp), rp
lea 16(up), up
jmp L(com)
L(small_corner):
mulx( v0, X1, w3)
add %rax, X1 C Tricky rax reuse of last iteration
adc $0, w3
mov X1, (rp)
mov w3, 8(rp)
lea 16(rp), rp
lea 8(up), up
L(com):
L(sqr_diag_addlsh1):
lea 8(up,un,8), up C put back up at its very beginning
lea (rp,un,8), rp
lea (rp,un,8), rp C put back rp at its very beginning
inc un
mov -8(up), %rdx
xor R32(%rbx), R32(%rbx) C clear CF as side effect
mulx( %rdx, %rax, %r10)
mov %rax, 8(rp)
mov 16(rp), %r8
mov 24(rp), %r9
jmp L(dm)
ALIGN(16)
L(dtop):mov 32(rp), %r8
mov 40(rp), %r9
lea 16(rp), rp
lea (%rdx,%rbx), %r10
L(dm): adc %r8, %r8
adc %r9, %r9
setc R8(%rbx)
mov (up), %rdx
lea 8(up), up
mulx( %rdx, %rax, %rdx)
add %r10, %r8
adc %rax, %r9
mov %r8, 16(rp)
mov %r9, 24(rp)
inc un
jnz L(dtop)
L(dend):adc %rbx, %rdx
mov %rdx, 32(rp)
pop %r14
pop %r13
pop %r12
pop %rbp
pop %rbx
FUNC_EXIT()
ret
EPILOGUE()
CF_PROT