148 lines
5.2 KiB
ArmAsm
148 lines
5.2 KiB
ArmAsm
/*-
|
|
* Copyright (c) 2011 The NetBSD Foundation, Inc.
|
|
* All rights reserved.
|
|
*
|
|
* This code is derived from software contributed to The NetBSD Foundation
|
|
* by Matt Thomas of 3am Software Foundry.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. 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.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. 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 FOUNDATION 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.
|
|
*/
|
|
|
|
#include <machine/asm.h>
|
|
#include "assym.h"
|
|
|
|
RCSID("$NetBSD: spe_subr.S,v 1.4 2014/08/10 18:08:33 joerg Exp $")
|
|
|
|
.text
|
|
#ifndef __clang__
|
|
.machine e500x2
|
|
#endif
|
|
/*
|
|
* Lintstub: void vec_load_from_vreg(const struct vreg *);
|
|
*/
|
|
ENTRY(vec_load_from_vreg)
|
|
/*
|
|
* Load and initialize the accumulator. Don't care about trashing
|
|
* r0's high half since we are about to load it.
|
|
*/
|
|
evldd %r0,(8 << 4)(%r3)
|
|
evmra %r0,%r0
|
|
|
|
/*
|
|
* SPEFSCR is in the trapframe.
|
|
*/
|
|
|
|
/*
|
|
* Since the high part of the register is saved in vreg (because we
|
|
* are loading it), and we know SPE is on (to load it), just load two
|
|
* high parts at one time and merge appropriately. For each even-odd
|
|
* register pair, the evldw will load high part for the even register
|
|
* in the high 32 bits and the high part for the odd regiter in the
|
|
* low 32 bits.
|
|
*
|
|
* The evmergelo will move the lower half of r0 to the high half of the
|
|
* destination register and evmergehilo will merge the high half of r0
|
|
* and lower half of the other register.
|
|
*/
|
|
evldw %r0,(30 << 2)(%r3)
|
|
evmergelo %r31,%r0,%r31; evmergehilo %r30,%r0,%r30
|
|
evldw %r0,(28 << 2)(%r3)
|
|
evmergelo %r29,%r0,%r29; evmergehilo %r28,%r0,%r28
|
|
evldw %r0,(26 << 2)(%r3)
|
|
evmergelo %r27,%r0,%r27; evmergehilo %r26,%r0,%r26
|
|
evldw %r0,(24 << 2)(%r3)
|
|
evmergelo %r25,%r0,%r25; evmergehilo %r24,%r0,%r24
|
|
evldw %r0,(22 << 2)(%r3)
|
|
evmergelo %r23,%r0,%r23; evmergehilo %r22,%r0,%r22
|
|
evldw %r0,(20 << 2)(%r3)
|
|
evmergelo %r21,%r0,%r21; evmergehilo %r20,%r0,%r20
|
|
evldw %r0,(18 << 2)(%r3)
|
|
evmergelo %r19,%r0,%r19; evmergehilo %r18,%r0,%r18
|
|
evldw %r0,(16 << 2)(%r3)
|
|
evmergelo %r17,%r0,%r17; evmergehilo %r16,%r0,%r16
|
|
evldw %r0,(14 << 2)(%r3)
|
|
evmergelo %r15,%r0,%r15; evmergehilo %r14,%r0,%r14
|
|
evldw %r0,(12 << 2)(%r3)
|
|
evmergelo %r13,%r0,%r13; evmergehilo %r12,%r0,%r12
|
|
/*
|
|
* Done with callee-saved registers. For caller-saved we can just
|
|
* trash the register contents. So load the two words with the high
|
|
* halves into the appropriate register and merge the half for the
|
|
* next register into it.
|
|
*/
|
|
evldw %r10,(10 << 2)(%r3)
|
|
evmergelo %r11,%r10,%r11
|
|
evldw %r8,(8 << 2)(%r3)
|
|
evmergelo %r9,%r8,%r9
|
|
evldw %r6,(6 << 2)(%r3)
|
|
evmergelo %r7,%r6,%r7
|
|
evldw %r4,(4 << 2)(%r3)
|
|
evmergelo %r5,%r4,%r5
|
|
|
|
/*
|
|
* R2 isn't a callee-saved, so load into r0 because we still need r3
|
|
*/
|
|
evldw %r0,(2 << 2)(%r3)
|
|
evmergelo %r3,%r0,%r3; evmergehilo %r2,%r0,%r2
|
|
evldd %r0,(0 << 2)(%r3)
|
|
evmergelo %r1,%r0,%r1 /* why bother? */
|
|
|
|
blr
|
|
END(vec_load_from_vreg)
|
|
|
|
/*
|
|
* Lintstub: void vec_unload_to_vreg(struct vreg *);
|
|
*/
|
|
ENTRY(vec_unload_to_vreg)
|
|
evmergehi %r4,%r4,%r0; /* save r0's high part in lo r4 */
|
|
evmergehi %r0,%r0,%r1; evstdw %r0,(0 << 2)(%r3)
|
|
evmergehi %r0,%r2,%r3; evstdw %r0,(2 << 2)(%r3)
|
|
evmergehi %r0,%r4,%r5; evstdw %r0,(4 << 2)(%r3)
|
|
evmergehi %r0,%r6,%r7; evstdw %r0,(6 << 2)(%r3)
|
|
evmergehi %r0,%r8,%r9; evstdw %r0,(8 << 2)(%r3)
|
|
evmergehi %r0,%r10,%r11; evstdw %r0,(10 << 2)(%r3)
|
|
evmergehi %r0,%r12,%r13; evstdw %r0,(12 << 2)(%r3)
|
|
evmergehi %r0,%r14,%r15; evstdw %r0,(14 << 2)(%r3)
|
|
evmergehi %r0,%r16,%r17; evstdw %r0,(16 << 2)(%r3)
|
|
evmergehi %r0,%r18,%r19; evstdw %r0,(18 << 2)(%r3)
|
|
evmergehi %r0,%r20,%r21; evstdw %r0,(20 << 2)(%r3)
|
|
evmergehi %r0,%r22,%r23; evstdw %r0,(22 << 2)(%r3)
|
|
evmergehi %r0,%r24,%r25; evstdw %r0,(24 << 2)(%r3)
|
|
evmergehi %r0,%r26,%r27; evstdw %r0,(26 << 2)(%r3)
|
|
evmergehi %r0,%r28,%r29; evstdw %r0,(28 << 2)(%r3)
|
|
evmergehi %r0,%r30,%r31; evstdw %r0,(30 << 2)(%r3)
|
|
|
|
/*
|
|
* Now save the accumulator.
|
|
*/
|
|
evxor %r0,%r0,%r0 /* zero r0 */
|
|
evaddumiaaw %r0,%r0 /* r0 = accum + r0 */
|
|
evstdd %r0,(8 << 4)(%r3) /* store it */
|
|
|
|
evmergelo %r0,%r4,%r0 /* retore r0's high half */
|
|
|
|
/*
|
|
* The SPEFSCR will be restored when the exception returns.
|
|
*/
|
|
blr
|
|
END(vec_unload_to_vreg)
|