linux_old1/arch/mips/kernel/octeon_switch.S

554 lines
14 KiB
ArmAsm

/*
* This file is subject to the terms and conditions of the GNU General Public
* License. See the file "COPYING" in the main directory of this archive
* for more details.
*
* Copyright (C) 1994, 1995, 1996, 1998, 1999, 2002, 2003 Ralf Baechle
* Copyright (C) 1996 David S. Miller (davem@davemloft.net)
* Copyright (C) 1994, 1995, 1996, by Andreas Busse
* Copyright (C) 1999 Silicon Graphics, Inc.
* Copyright (C) 2000 MIPS Technologies, Inc.
* written by Carsten Langgaard, carstenl@mips.com
*/
#define USE_ALTERNATE_RESUME_IMPL 1
.set push
.set arch=mips64r2
#include "r4k_switch.S"
.set pop
/*
* task_struct *resume(task_struct *prev, task_struct *next,
* struct thread_info *next_ti)
*/
.align 7
LEAF(resume)
.set arch=octeon
mfc0 t1, CP0_STATUS
LONG_S t1, THREAD_STATUS(a0)
cpu_save_nonscratch a0
LONG_S ra, THREAD_REG31(a0)
#if CONFIG_CAVIUM_OCTEON_CVMSEG_SIZE > 0
/* Check if we need to store CVMSEG state */
dmfc0 t0, $11,7 /* CvmMemCtl */
bbit0 t0, 6, 3f /* Is user access enabled? */
/* Store the CVMSEG state */
/* Extract the size of CVMSEG */
andi t0, 0x3f
/* Multiply * (cache line size/sizeof(long)/2) */
sll t0, 7-LONGLOG-1
li t1, -32768 /* Base address of CVMSEG */
LONG_ADDI t2, a0, THREAD_CVMSEG /* Where to store CVMSEG to */
synciobdma
2:
.set noreorder
LONG_L t8, 0(t1) /* Load from CVMSEG */
subu t0, 1 /* Decrement loop var */
LONG_L t9, LONGSIZE(t1)/* Load from CVMSEG */
LONG_ADDU t1, LONGSIZE*2 /* Increment loc in CVMSEG */
LONG_S t8, 0(t2) /* Store CVMSEG to thread storage */
LONG_ADDU t2, LONGSIZE*2 /* Increment loc in thread storage */
bnez t0, 2b /* Loop until we've copied it all */
LONG_S t9, -LONGSIZE(t2)/* Store CVMSEG to thread storage */
.set reorder
/* Disable access to CVMSEG */
dmfc0 t0, $11,7 /* CvmMemCtl */
xori t0, t0, 0x40 /* Bit 6 is CVMSEG user enable */
dmtc0 t0, $11,7 /* CvmMemCtl */
#endif
3:
#if defined(CONFIG_CC_STACKPROTECTOR) && !defined(CONFIG_SMP)
PTR_LA t8, __stack_chk_guard
LONG_L t9, TASK_STACK_CANARY(a1)
LONG_S t9, 0(t8)
#endif
/*
* The order of restoring the registers takes care of the race
* updating $28, $29 and kernelsp without disabling ints.
*/
move $28, a2
cpu_restore_nonscratch a1
PTR_ADDU t0, $28, _THREAD_SIZE - 32
set_saved_sp t0, t1, t2
mfc0 t1, CP0_STATUS /* Do we really need this? */
li a3, 0xff01
and t1, a3
LONG_L a2, THREAD_STATUS(a1)
nor a3, $0, a3
and a2, a3
or a2, t1
mtc0 a2, CP0_STATUS
move v0, a0
jr ra
END(resume)
/*
* void octeon_cop2_save(struct octeon_cop2_state *a0)
*/
.align 7
.set push
.set noreorder
LEAF(octeon_cop2_save)
dmfc0 t9, $9,7 /* CvmCtl register. */
/* Save the COP2 CRC state */
dmfc2 t0, 0x0201
dmfc2 t1, 0x0202
dmfc2 t2, 0x0200
sd t0, OCTEON_CP2_CRC_IV(a0)
sd t1, OCTEON_CP2_CRC_LENGTH(a0)
/* Skip next instructions if CvmCtl[NODFA_CP2] set */
bbit1 t9, 28, 1f
sd t2, OCTEON_CP2_CRC_POLY(a0)
/* Save the LLM state */
dmfc2 t0, 0x0402
dmfc2 t1, 0x040A
sd t0, OCTEON_CP2_LLM_DAT(a0)
1: bbit1 t9, 26, 3f /* done if CvmCtl[NOCRYPTO] set */
sd t1, OCTEON_CP2_LLM_DAT+8(a0)
/* Save the COP2 crypto state */
/* this part is mostly common to both pass 1 and later revisions */
dmfc2 t0, 0x0084
dmfc2 t1, 0x0080
dmfc2 t2, 0x0081
dmfc2 t3, 0x0082
sd t0, OCTEON_CP2_3DES_IV(a0)
dmfc2 t0, 0x0088
sd t1, OCTEON_CP2_3DES_KEY(a0)
dmfc2 t1, 0x0111 /* only necessary for pass 1 */
sd t2, OCTEON_CP2_3DES_KEY+8(a0)
dmfc2 t2, 0x0102
sd t3, OCTEON_CP2_3DES_KEY+16(a0)
dmfc2 t3, 0x0103
sd t0, OCTEON_CP2_3DES_RESULT(a0)
dmfc2 t0, 0x0104
sd t1, OCTEON_CP2_AES_INP0(a0) /* only necessary for pass 1 */
dmfc2 t1, 0x0105
sd t2, OCTEON_CP2_AES_IV(a0)
dmfc2 t2, 0x0106
sd t3, OCTEON_CP2_AES_IV+8(a0)
dmfc2 t3, 0x0107
sd t0, OCTEON_CP2_AES_KEY(a0)
dmfc2 t0, 0x0110
sd t1, OCTEON_CP2_AES_KEY+8(a0)
dmfc2 t1, 0x0100
sd t2, OCTEON_CP2_AES_KEY+16(a0)
dmfc2 t2, 0x0101
sd t3, OCTEON_CP2_AES_KEY+24(a0)
mfc0 v0, $15,0 /* Get the processor ID register */
sd t0, OCTEON_CP2_AES_KEYLEN(a0)
li v1, 0x000d0000 /* This is the processor ID of Octeon Pass1 */
sd t1, OCTEON_CP2_AES_RESULT(a0)
/* Skip to the Pass1 version of the remainder of the COP2 state */
beq v0, v1, 2f
sd t2, OCTEON_CP2_AES_RESULT+8(a0)
/* the non-pass1 state when !CvmCtl[NOCRYPTO] */
dmfc2 t1, 0x0240
dmfc2 t2, 0x0241
ori v1, v1, 0x9500 /* lowest OCTEON III PrId*/
dmfc2 t3, 0x0242
subu v1, v0, v1 /* prid - lowest OCTEON III PrId */
dmfc2 t0, 0x0243
sd t1, OCTEON_CP2_HSH_DATW(a0)
dmfc2 t1, 0x0244
sd t2, OCTEON_CP2_HSH_DATW+8(a0)
dmfc2 t2, 0x0245
sd t3, OCTEON_CP2_HSH_DATW+16(a0)
dmfc2 t3, 0x0246
sd t0, OCTEON_CP2_HSH_DATW+24(a0)
dmfc2 t0, 0x0247
sd t1, OCTEON_CP2_HSH_DATW+32(a0)
dmfc2 t1, 0x0248
sd t2, OCTEON_CP2_HSH_DATW+40(a0)
dmfc2 t2, 0x0249
sd t3, OCTEON_CP2_HSH_DATW+48(a0)
dmfc2 t3, 0x024A
sd t0, OCTEON_CP2_HSH_DATW+56(a0)
dmfc2 t0, 0x024B
sd t1, OCTEON_CP2_HSH_DATW+64(a0)
dmfc2 t1, 0x024C
sd t2, OCTEON_CP2_HSH_DATW+72(a0)
dmfc2 t2, 0x024D
sd t3, OCTEON_CP2_HSH_DATW+80(a0)
dmfc2 t3, 0x024E
sd t0, OCTEON_CP2_HSH_DATW+88(a0)
dmfc2 t0, 0x0250
sd t1, OCTEON_CP2_HSH_DATW+96(a0)
dmfc2 t1, 0x0251
sd t2, OCTEON_CP2_HSH_DATW+104(a0)
dmfc2 t2, 0x0252
sd t3, OCTEON_CP2_HSH_DATW+112(a0)
dmfc2 t3, 0x0253
sd t0, OCTEON_CP2_HSH_IVW(a0)
dmfc2 t0, 0x0254
sd t1, OCTEON_CP2_HSH_IVW+8(a0)
dmfc2 t1, 0x0255
sd t2, OCTEON_CP2_HSH_IVW+16(a0)
dmfc2 t2, 0x0256
sd t3, OCTEON_CP2_HSH_IVW+24(a0)
dmfc2 t3, 0x0257
sd t0, OCTEON_CP2_HSH_IVW+32(a0)
dmfc2 t0, 0x0258
sd t1, OCTEON_CP2_HSH_IVW+40(a0)
dmfc2 t1, 0x0259
sd t2, OCTEON_CP2_HSH_IVW+48(a0)
dmfc2 t2, 0x025E
sd t3, OCTEON_CP2_HSH_IVW+56(a0)
dmfc2 t3, 0x025A
sd t0, OCTEON_CP2_GFM_MULT(a0)
dmfc2 t0, 0x025B
sd t1, OCTEON_CP2_GFM_MULT+8(a0)
sd t2, OCTEON_CP2_GFM_POLY(a0)
sd t3, OCTEON_CP2_GFM_RESULT(a0)
bltz v1, 4f
sd t0, OCTEON_CP2_GFM_RESULT+8(a0)
/* OCTEON III things*/
dmfc2 t0, 0x024F
dmfc2 t1, 0x0050
sd t0, OCTEON_CP2_SHA3(a0)
sd t1, OCTEON_CP2_SHA3+8(a0)
4:
jr ra
nop
2: /* pass 1 special stuff when !CvmCtl[NOCRYPTO] */
dmfc2 t3, 0x0040
dmfc2 t0, 0x0041
dmfc2 t1, 0x0042
dmfc2 t2, 0x0043
sd t3, OCTEON_CP2_HSH_DATW(a0)
dmfc2 t3, 0x0044
sd t0, OCTEON_CP2_HSH_DATW+8(a0)
dmfc2 t0, 0x0045
sd t1, OCTEON_CP2_HSH_DATW+16(a0)
dmfc2 t1, 0x0046
sd t2, OCTEON_CP2_HSH_DATW+24(a0)
dmfc2 t2, 0x0048
sd t3, OCTEON_CP2_HSH_DATW+32(a0)
dmfc2 t3, 0x0049
sd t0, OCTEON_CP2_HSH_DATW+40(a0)
dmfc2 t0, 0x004A
sd t1, OCTEON_CP2_HSH_DATW+48(a0)
sd t2, OCTEON_CP2_HSH_IVW(a0)
sd t3, OCTEON_CP2_HSH_IVW+8(a0)
sd t0, OCTEON_CP2_HSH_IVW+16(a0)
3: /* pass 1 or CvmCtl[NOCRYPTO] set */
jr ra
nop
END(octeon_cop2_save)
.set pop
/*
* void octeon_cop2_restore(struct octeon_cop2_state *a0)
*/
.align 7
.set push
.set noreorder
LEAF(octeon_cop2_restore)
/* First cache line was prefetched before the call */
pref 4, 128(a0)
dmfc0 t9, $9,7 /* CvmCtl register. */
pref 4, 256(a0)
ld t0, OCTEON_CP2_CRC_IV(a0)
pref 4, 384(a0)
ld t1, OCTEON_CP2_CRC_LENGTH(a0)
ld t2, OCTEON_CP2_CRC_POLY(a0)
/* Restore the COP2 CRC state */
dmtc2 t0, 0x0201
dmtc2 t1, 0x1202
bbit1 t9, 28, 2f /* Skip LLM if CvmCtl[NODFA_CP2] is set */
dmtc2 t2, 0x4200
/* Restore the LLM state */
ld t0, OCTEON_CP2_LLM_DAT(a0)
ld t1, OCTEON_CP2_LLM_DAT+8(a0)
dmtc2 t0, 0x0402
dmtc2 t1, 0x040A
2:
bbit1 t9, 26, done_restore /* done if CvmCtl[NOCRYPTO] set */
nop
/* Restore the COP2 crypto state common to pass 1 and pass 2 */
ld t0, OCTEON_CP2_3DES_IV(a0)
ld t1, OCTEON_CP2_3DES_KEY(a0)
ld t2, OCTEON_CP2_3DES_KEY+8(a0)
dmtc2 t0, 0x0084
ld t0, OCTEON_CP2_3DES_KEY+16(a0)
dmtc2 t1, 0x0080
ld t1, OCTEON_CP2_3DES_RESULT(a0)
dmtc2 t2, 0x0081
ld t2, OCTEON_CP2_AES_INP0(a0) /* only really needed for pass 1 */
dmtc2 t0, 0x0082
ld t0, OCTEON_CP2_AES_IV(a0)
dmtc2 t1, 0x0098
ld t1, OCTEON_CP2_AES_IV+8(a0)
dmtc2 t2, 0x010A /* only really needed for pass 1 */
ld t2, OCTEON_CP2_AES_KEY(a0)
dmtc2 t0, 0x0102
ld t0, OCTEON_CP2_AES_KEY+8(a0)
dmtc2 t1, 0x0103
ld t1, OCTEON_CP2_AES_KEY+16(a0)
dmtc2 t2, 0x0104
ld t2, OCTEON_CP2_AES_KEY+24(a0)
dmtc2 t0, 0x0105
ld t0, OCTEON_CP2_AES_KEYLEN(a0)
dmtc2 t1, 0x0106
ld t1, OCTEON_CP2_AES_RESULT(a0)
dmtc2 t2, 0x0107
ld t2, OCTEON_CP2_AES_RESULT+8(a0)
mfc0 t3, $15,0 /* Get the processor ID register */
dmtc2 t0, 0x0110
li v0, 0x000d0000 /* This is the processor ID of Octeon Pass1 */
dmtc2 t1, 0x0100
bne v0, t3, 3f /* Skip the next stuff for non-pass1 */
dmtc2 t2, 0x0101
/* this code is specific for pass 1 */
ld t0, OCTEON_CP2_HSH_DATW(a0)
ld t1, OCTEON_CP2_HSH_DATW+8(a0)
ld t2, OCTEON_CP2_HSH_DATW+16(a0)
dmtc2 t0, 0x0040
ld t0, OCTEON_CP2_HSH_DATW+24(a0)
dmtc2 t1, 0x0041
ld t1, OCTEON_CP2_HSH_DATW+32(a0)
dmtc2 t2, 0x0042
ld t2, OCTEON_CP2_HSH_DATW+40(a0)
dmtc2 t0, 0x0043
ld t0, OCTEON_CP2_HSH_DATW+48(a0)
dmtc2 t1, 0x0044
ld t1, OCTEON_CP2_HSH_IVW(a0)
dmtc2 t2, 0x0045
ld t2, OCTEON_CP2_HSH_IVW+8(a0)
dmtc2 t0, 0x0046
ld t0, OCTEON_CP2_HSH_IVW+16(a0)
dmtc2 t1, 0x0048
dmtc2 t2, 0x0049
b done_restore /* unconditional branch */
dmtc2 t0, 0x004A
3: /* this is post-pass1 code */
ld t2, OCTEON_CP2_HSH_DATW(a0)
ori v0, v0, 0x9500 /* lowest OCTEON III PrId*/
ld t0, OCTEON_CP2_HSH_DATW+8(a0)
ld t1, OCTEON_CP2_HSH_DATW+16(a0)
dmtc2 t2, 0x0240
ld t2, OCTEON_CP2_HSH_DATW+24(a0)
dmtc2 t0, 0x0241
ld t0, OCTEON_CP2_HSH_DATW+32(a0)
dmtc2 t1, 0x0242
ld t1, OCTEON_CP2_HSH_DATW+40(a0)
dmtc2 t2, 0x0243
ld t2, OCTEON_CP2_HSH_DATW+48(a0)
dmtc2 t0, 0x0244
ld t0, OCTEON_CP2_HSH_DATW+56(a0)
dmtc2 t1, 0x0245
ld t1, OCTEON_CP2_HSH_DATW+64(a0)
dmtc2 t2, 0x0246
ld t2, OCTEON_CP2_HSH_DATW+72(a0)
dmtc2 t0, 0x0247
ld t0, OCTEON_CP2_HSH_DATW+80(a0)
dmtc2 t1, 0x0248
ld t1, OCTEON_CP2_HSH_DATW+88(a0)
dmtc2 t2, 0x0249
ld t2, OCTEON_CP2_HSH_DATW+96(a0)
dmtc2 t0, 0x024A
ld t0, OCTEON_CP2_HSH_DATW+104(a0)
dmtc2 t1, 0x024B
ld t1, OCTEON_CP2_HSH_DATW+112(a0)
dmtc2 t2, 0x024C
ld t2, OCTEON_CP2_HSH_IVW(a0)
dmtc2 t0, 0x024D
ld t0, OCTEON_CP2_HSH_IVW+8(a0)
dmtc2 t1, 0x024E
ld t1, OCTEON_CP2_HSH_IVW+16(a0)
dmtc2 t2, 0x0250
ld t2, OCTEON_CP2_HSH_IVW+24(a0)
dmtc2 t0, 0x0251
ld t0, OCTEON_CP2_HSH_IVW+32(a0)
dmtc2 t1, 0x0252
ld t1, OCTEON_CP2_HSH_IVW+40(a0)
dmtc2 t2, 0x0253
ld t2, OCTEON_CP2_HSH_IVW+48(a0)
dmtc2 t0, 0x0254
ld t0, OCTEON_CP2_HSH_IVW+56(a0)
dmtc2 t1, 0x0255
ld t1, OCTEON_CP2_GFM_MULT(a0)
dmtc2 t2, 0x0256
ld t2, OCTEON_CP2_GFM_MULT+8(a0)
dmtc2 t0, 0x0257
ld t0, OCTEON_CP2_GFM_POLY(a0)
dmtc2 t1, 0x0258
ld t1, OCTEON_CP2_GFM_RESULT(a0)
dmtc2 t2, 0x0259
ld t2, OCTEON_CP2_GFM_RESULT+8(a0)
dmtc2 t0, 0x025E
subu v0, t3, v0 /* prid - lowest OCTEON III PrId */
dmtc2 t1, 0x025A
bltz v0, done_restore
dmtc2 t2, 0x025B
/* OCTEON III things*/
ld t0, OCTEON_CP2_SHA3(a0)
ld t1, OCTEON_CP2_SHA3+8(a0)
dmtc2 t0, 0x0051
dmtc2 t1, 0x0050
done_restore:
jr ra
nop
END(octeon_cop2_restore)
.set pop
/*
* void octeon_mult_save()
* sp is assumed to point to a struct pt_regs
*
* NOTE: This is called in SAVE_TEMP in stackframe.h. It can
* safely modify v1,k0, k1,$10-$15, and $24. It will
* be overwritten with a processor specific version of the code.
*/
.p2align 7
.set push
.set noreorder
LEAF(octeon_mult_save)
jr ra
nop
.space 30 * 4, 0
octeon_mult_save_end:
EXPORT(octeon_mult_save_end)
END(octeon_mult_save)
LEAF(octeon_mult_save2)
/* Save the multiplier state OCTEON II and earlier*/
v3mulu k0, $0, $0
v3mulu k1, $0, $0
sd k0, PT_MTP(sp) /* PT_MTP has P0 */
v3mulu k0, $0, $0
sd k1, PT_MTP+8(sp) /* PT_MTP+8 has P1 */
ori k1, $0, 1
v3mulu k1, k1, $0
sd k0, PT_MTP+16(sp) /* PT_MTP+16 has P2 */
v3mulu k0, $0, $0
sd k1, PT_MPL(sp) /* PT_MPL has MPL0 */
v3mulu k1, $0, $0
sd k0, PT_MPL+8(sp) /* PT_MPL+8 has MPL1 */
jr ra
sd k1, PT_MPL+16(sp) /* PT_MPL+16 has MPL2 */
octeon_mult_save2_end:
EXPORT(octeon_mult_save2_end)
END(octeon_mult_save2)
LEAF(octeon_mult_save3)
/* Save the multiplier state OCTEON III */
v3mulu $10, $0, $0 /* read P0 */
v3mulu $11, $0, $0 /* read P1 */
v3mulu $12, $0, $0 /* read P2 */
sd $10, PT_MTP+(0*8)(sp) /* store P0 */
v3mulu $10, $0, $0 /* read P3 */
sd $11, PT_MTP+(1*8)(sp) /* store P1 */
v3mulu $11, $0, $0 /* read P4 */
sd $12, PT_MTP+(2*8)(sp) /* store P2 */
ori $13, $0, 1
v3mulu $12, $0, $0 /* read P5 */
sd $10, PT_MTP+(3*8)(sp) /* store P3 */
v3mulu $13, $13, $0 /* P4-P0 = MPL5-MPL1, $13 = MPL0 */
sd $11, PT_MTP+(4*8)(sp) /* store P4 */
v3mulu $10, $0, $0 /* read MPL1 */
sd $12, PT_MTP+(5*8)(sp) /* store P5 */
v3mulu $11, $0, $0 /* read MPL2 */
sd $13, PT_MPL+(0*8)(sp) /* store MPL0 */
v3mulu $12, $0, $0 /* read MPL3 */
sd $10, PT_MPL+(1*8)(sp) /* store MPL1 */
v3mulu $10, $0, $0 /* read MPL4 */
sd $11, PT_MPL+(2*8)(sp) /* store MPL2 */
v3mulu $11, $0, $0 /* read MPL5 */
sd $12, PT_MPL+(3*8)(sp) /* store MPL3 */
sd $10, PT_MPL+(4*8)(sp) /* store MPL4 */
jr ra
sd $11, PT_MPL+(5*8)(sp) /* store MPL5 */
octeon_mult_save3_end:
EXPORT(octeon_mult_save3_end)
END(octeon_mult_save3)
.set pop
/*
* void octeon_mult_restore()
* sp is assumed to point to a struct pt_regs
*
* NOTE: This is called in RESTORE_TEMP in stackframe.h.
*/
.p2align 7
.set push
.set noreorder
LEAF(octeon_mult_restore)
jr ra
nop
.space 30 * 4, 0
octeon_mult_restore_end:
EXPORT(octeon_mult_restore_end)
END(octeon_mult_restore)
LEAF(octeon_mult_restore2)
ld v0, PT_MPL(sp) /* MPL0 */
ld v1, PT_MPL+8(sp) /* MPL1 */
ld k0, PT_MPL+16(sp) /* MPL2 */
/* Restore the multiplier state */
ld k1, PT_MTP+16(sp) /* P2 */
mtm0 v0 /* MPL0 */
ld v0, PT_MTP+8(sp) /* P1 */
mtm1 v1 /* MPL1 */
ld v1, PT_MTP(sp) /* P0 */
mtm2 k0 /* MPL2 */
mtp2 k1 /* P2 */
mtp1 v0 /* P1 */
jr ra
mtp0 v1 /* P0 */
octeon_mult_restore2_end:
EXPORT(octeon_mult_restore2_end)
END(octeon_mult_restore2)
LEAF(octeon_mult_restore3)
ld $12, PT_MPL+(0*8)(sp) /* read MPL0 */
ld $13, PT_MPL+(3*8)(sp) /* read MPL3 */
ld $10, PT_MPL+(1*8)(sp) /* read MPL1 */
ld $11, PT_MPL+(4*8)(sp) /* read MPL4 */
.word 0x718d0008
/* mtm0 $12, $13 restore MPL0 and MPL3 */
ld $12, PT_MPL+(2*8)(sp) /* read MPL2 */
.word 0x714b000c
/* mtm1 $10, $11 restore MPL1 and MPL4 */
ld $13, PT_MPL+(5*8)(sp) /* read MPL5 */
ld $10, PT_MTP+(0*8)(sp) /* read P0 */
ld $11, PT_MTP+(3*8)(sp) /* read P3 */
.word 0x718d000d
/* mtm2 $12, $13 restore MPL2 and MPL5 */
ld $12, PT_MTP+(1*8)(sp) /* read P1 */
.word 0x714b0009
/* mtp0 $10, $11 restore P0 and P3 */
ld $13, PT_MTP+(4*8)(sp) /* read P4 */
ld $10, PT_MTP+(2*8)(sp) /* read P2 */
ld $11, PT_MTP+(5*8)(sp) /* read P5 */
.word 0x718d000a
/* mtp1 $12, $13 restore P1 and P4 */
jr ra
.word 0x714b000b
/* mtp2 $10, $11 restore P2 and P5 */
octeon_mult_restore3_end:
EXPORT(octeon_mult_restore3_end)
END(octeon_mult_restore3)
.set pop