dnl Power9 mpn_mul_basecase.
dnl Copyright 1999-2001, 2003-2006, 2008, 2017-2018 Free Software Foundation,
dnl 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
C POWER3/PPC630 -
C POWER4/PPC970 -
C POWER5 -
C POWER6 -
C POWER7 -
C POWER8 -
C POWER9 1.62
C TODO
C * Check if (inner) loop alignment affects performance.
C * Could we schedule loads less in addmul_2/mul_2? That would save some regs
C and make the tail code more manageable.
C * Postpone some register saves to main loop.
C * Perhaps write more small operands (3x1, 3x2, 3x3) code.
C * Consider restoring rp,up after loop using arithmetic, eliminating rp2, up2.
C On the other hand, the current rp,up restore register are useful for OSP.
C * Do OSP. This should save a lot with the current deep addmul_2 pipeline.
C INPUT PARAMETERS
define(`rp', `r3')
define(`up', `r4')
define(`un', `r5')
define(`vp', `r6')
define(`vn', `r7')
define(`v0', `r0')
define(`v1', `r7')
define(`rp2', `r24')
define(`up2', `r25')
ASM_START()
PROLOGUE(mpn_mul_basecase)
cmpdi cr0, un, 2
bgt cr0, L(un_gt2)
cmpdi cr6, vn, 1
ld r7, 0(vp)
ld r5, 0(up)
mulld r8, r5, r7 C weight 0
mulhdu r9, r5, r7 C weight 1
std r8, 0(rp)
beq cr0, L(2x)
std r9, 8(rp)
blr
ALIGN(16)
L(2x): ld r0, 8(up)
mulld r8, r0, r7 C weight 1
mulhdu r10, r0, r7 C weight 2
addc r9, r9, r8
addze r10, r10
bne cr6, L(2x2)
std r9, 8(rp)
std r10, 16(rp)
blr
ALIGN(16)
L(2x2): ld r6, 8(vp)
mulld r8, r5, r6 C weight 1
mulhdu r11, r5, r6 C weight 2
addc r9, r9, r8
std r9, 8(rp)
adde r11, r11, r10
mulld r12, r0, r6 C weight 2
mulhdu r0, r0, r6 C weight 3
addze r0, r0
addc r11, r11, r12
addze r0, r0
std r11, 16(rp)
std r0, 24(rp)
blr
L(un_gt2):
std r22, -80(r1)
std r23, -72(r1)
std r24, -64(r1)
std r25, -56(r1)
std r26, -48(r1)
std r27, -40(r1)
std r28, -32(r1)
std r29, -24(r1)
std r30, -16(r1)
std r31, -8(r1)
mr rp2, r3 C rp
mr up2, r4 C up
srdi r22, r5, 2 C un
subfic r23, r7, 0 C -vn, clear CA
subfo r0, r0, r0 C clear OV (and r0)
cmpdi cr6, un, 3
rldicl r0, un, 0, 63 C r0 = un & 1
cmpdi cr7, r0, 0
rldicl r0, un, 63, 63 C FIXME: unused for vn = 1
cmpdi cr5, r0, 0 C FIXME: unused for vn = 1
ld v0, 0(vp)
rldicl. r9, vn, 0, 63
beq cr0, L(vn_evn)
L(vn_odd):
addi r10, un, -2
ld r5, 0(up)
srdi r10, r10, 1
mtctr r10
bne cr7, L(m1_b1)
L(m1_b0):
ld r10, 8(up)
mulld r9, r5, v0
mulhdu r11, r5, v0
ld r12, 16(up)
mulld r8, r10, v0
mulhdu r5, r10, v0
addi rp, rp, -8
b L(m1_mid)
L(m1_b1):
ld r12, 8(up)
mulld r8, r5, v0
mulhdu r5, r5, v0
ld r10, 16(up)
mulld r9, r12, v0
mulhdu r11, r12, v0
addi up, up, 8
beq cr6, L(m1_end) C jump taken means un = 3, vn = {1,3}
ALIGN(16)
L(m1_top):
ld r12, 16(up)
std r8, 0(rp)
adde r9, r5, r9
mulld r8, r10, v0
mulhdu r5, r10, v0
L(m1_mid):
ld r10, 24(up)
std r9, 8(rp)
adde r8, r11, r8
mulld r9, r12, v0
mulhdu r11, r12, v0
addi rp, rp, 16
addi up, up, 16
bdnz L(m1_top)
L(m1_end):
std r8, 0(rp)
mulld r8, r10, v0
adde r9, r5, r9
mulhdu r5, r10, v0
std r9, 8(rp)
adde r8, r11, r8
std r8, 16(rp)
addze r10, r5
std r10, 24(rp)
addi rp2, rp2, 8
addi vp, vp, 8
addic. r23, r23, 1
b L(do_outer)
L(vn_evn):
ld v1, 8(vp)
addi r23, r23, 2
mtctr r22
bne cr7, L(m2_bx1)
L(m2_bx0):
ld r8, 0(up)
ld r9, 8(up)
li r11, 0
mulld r28, r8, v0
mulhdu r31, r8, v0
mulld r5, r8, v1
mulhdu r10, r8, v1
li r12, 0
bne cr5, L(m2_b10)
L(m2_b00):
addi up, up, -8
addi rp, rp, -24
b L(m2_lo0)
L(m2_b10):
addi up, up, 8
addi rp, rp, -8
b L(m2_lo2)
L(m2_bx1):
ld r9, 0(up)
ld r8, 8(up)
li r10, 0
mulld r29, r9, v0
mulhdu r30, r9, v0
mulld r12, r9, v1
mulhdu r11, r9, v1
li r5, 0
bne cr5, L(m2_b11)
L(m2_b01):
addi rp, rp, -16
b L(m2_lo1)
L(m2_b11):
addi up, up, 16
beq cr6, L(m2_end) C taken means un = 3, vn = 2. We're done.
L(m2_top):
ld r9, 0(up)
maddld( r28, r8, v0, r10)
maddhdu(r31, r8, v0, r10)
adde r5, r29, r5
std r5, 0(rp)
mulld r5, r8, v1
mulhdu r10, r8, v1
addex( r12, r12, r30, 0)
L(m2_lo2):
ld r8, 8(up)
maddld( r29, r9, v0, r11)
maddhdu(r30, r9, v0, r11)
adde r12, r28, r12
std r12, 8(rp)
mulld r12, r9, v1
mulhdu r11, r9, v1
addex( r5, r5, r31, 0)
L(m2_lo1):
ld r9, 16(up)
maddld( r28, r8, v0, r10)
maddhdu(r31, r8, v0, r10)
adde r5, r29, r5
std r5, 16(rp)
mulld r5, r8, v1
mulhdu r10, r8, v1
addex( r12, r12, r30, 0)
L(m2_lo0):
ld r8, 24(up)
maddld( r29, r9, v0, r11)
maddhdu(r30, r9, v0, r11)
adde r12, r28, r12
std r12, 24(rp)
mulld r12, r9, v1
mulhdu r11, r9, v1
addex( r5, r5, r31, 0)
addi up, up, 32
addi rp, rp, 32
bdnz L(m2_top)
L(m2_end):
ld r9, 0(up)
maddld( r28, r8, v0, r10)
maddhdu(r31, r8, v0, r10)
adde r5, r29, r5
std r5, 0(rp)
mulld r5, r8, v1
mulhdu r10, r8, v1
b L(cj)
L(outer):
ld v0, 0(vp)
ld v1, 8(vp)
addi r23, r23, 2
mtctr r22
bne cr7, L(bx1)
L(bx0): ld r26, 0(rp2)
ld r8, 0(up2)
ld r11, 8(rp2)
ld r9, 8(up2)
maddld( r28, r8, v0, r26)
maddhdu(r31, r8, v0, r26)
ld r26, 16(rp2)
mulld r5, r8, v1
mulhdu r10, r8, v1
li r12, 0
bne cr5, L(b10)
L(b00): addi up, up2, -8
addi rp, rp2, -24
b L(lo0)
L(b10): addi up, up2, 8
addi rp, rp2, -8
b L(lo2)
L(bx1): ld r27, 0(rp2)
ld r9, 0(up2)
ld r10, 8(rp2)
ld r8, 8(up2)
maddld( r29, r9, v0, r27)
maddhdu(r30, r9, v0, r27)
ld r27, 16(rp2)
mulld r12, r9, v1
mulhdu r11, r9, v1
li r5, 0
bne cr5, L(b11)
L(b01): addi up, up2, 0
addi rp, rp2, -16
b L(lo1)
L(b11): addi up, up2, 16
addi rp, rp2, 0
beq cr6, L(end) C taken means un = 3, vn = 3. We're done.
L(top): ld r9, 0(up)
maddld( r28, r8, v0, r10)
maddhdu(r31, r8, v0, r10)
adde r5, r29, r5
ld r26, 24(rp)
std r5, 0(rp)
maddld( r5, r8, v1, r27)
maddhdu(r10, r8, v1, r27)
addex( r12, r12, r30, 0)
L(lo2): ld r8, 8(up)
maddld( r29, r9, v0, r11)
maddhdu(r30, r9, v0, r11)
adde r12, r28, r12
ld r27, 32(rp)
std r12, 8(rp)
maddld( r12, r9, v1, r26)
maddhdu(r11, r9, v1, r26)
addex( r5, r5, r31, 0)
L(lo1): ld r9, 16(up)
maddld( r28, r8, v0, r10)
maddhdu(r31, r8, v0, r10)
adde r5, r29, r5
ld r26, 40(rp)
std r5, 16(rp)
maddld( r5, r8, v1, r27)
maddhdu(r10, r8, v1, r27)
addex( r12, r12, r30, 0)
L(lo0): ld r8, 24(up)
maddld( r29, r9, v0, r11)
maddhdu(r30, r9, v0, r11)
adde r12, r28, r12
ld r27, 48(rp)
std r12, 24(rp)
maddld( r12, r9, v1, r26)
maddhdu(r11, r9, v1, r26)
addex( r5, r5, r31, 0)
addi up, up, 32
addi rp, rp, 32
bdnz L(top)
L(end): ld r9, 0(up)
maddld( r28, r8, v0, r10)
maddhdu(r31, r8, v0, r10)
adde r5, r29, r5
std r5, 0(rp)
maddld( r5, r8, v1, r27)
maddhdu(r10, r8, v1, r27)
L(cj): addex( r12, r12, r30, 0)
maddld( r29, r9, v0, r11)
maddhdu(r30, r9, v0, r11)
adde r12, r28, r12
std r12, 8(rp)
mulld r12, r9, v1
mulhdu r11, r9, v1
addex( r5, r5, r31, 0)
adde r5, r29, r5
std r5, 16(rp)
addex( r12, r12, r30, 0)
adde r12, r12, r10
std r12, 24(rp)
li r4, 0
addze r5, r11
addex( r5, r5, r4, 0)
std r5, 32(rp)
cmpdi cr0, r23, 0
addi rp2, rp2, 16
addi vp, vp, 16
L(do_outer):
bne cr0, L(outer)
L(ret):
ld r22, -80(r1)
ld r23, -72(r1)
ld r24, -64(r1)
ld r25, -56(r1)
ld r26, -48(r1)
ld r27, -40(r1)
ld r28, -32(r1)
ld r29, -24(r1)
ld r30, -16(r1)
ld r31, -8(r1)
blr
EPILOGUE()
ASM_END()