aes neon: Tweak register usage.
- Call r12 by its usual name, ip. - No need for r7 or r11=fp at the moment.
This commit is contained in:
riastradh
parent
b5c99049d3
commit
db39c37e7d
|
|
@ -1,4 +1,4 @@
|
|||
/* $NetBSD: aes_neon_32.S,v 1.8 2020/09/10 11:29:43 riastradh Exp $ */
|
||||
/* $NetBSD: aes_neon_32.S,v 1.9 2020/09/10 11:30:08 riastradh Exp $ */
|
||||
|
||||
/*-
|
||||
* Copyright (c) 2020 The NetBSD Foundation, Inc.
|
||||
|
|
@ -28,7 +28,7 @@
|
|||
|
||||
#include <arm/asm.h>
|
||||
|
||||
RCSID("$NetBSD: aes_neon_32.S,v 1.8 2020/09/10 11:29:43 riastradh Exp $")
|
||||
RCSID("$NetBSD: aes_neon_32.S,v 1.9 2020/09/10 11:30:08 riastradh Exp $")
|
||||
|
||||
.fpu neon
|
||||
|
||||
|
|
@ -205,14 +205,14 @@ ENTRY(aes_neon_enc1)
|
|||
vldr d1, [sp] /* d1 := x hi */
|
||||
ldr r1, [sp, #8] /* r1 := nrounds */
|
||||
#endif
|
||||
push {r4, r5, r6, r7, r8, r10, r11, lr}
|
||||
push {r4, r5, r6, r8, r10, lr}
|
||||
vpush {d8-d15}
|
||||
|
||||
/*
|
||||
* r3: rmod4
|
||||
* r4: mc_forward
|
||||
* r5: mc_backward
|
||||
* r6,r7,r8,r10,r11,r12: temporaries
|
||||
* r6,r8,r10,ip: temporaries
|
||||
* q0={d0-d1}: x/ak/A
|
||||
* q1={d2-d3}: 0x0f0f...
|
||||
* q2={d4-d5}: lo/k/j/io
|
||||
|
|
@ -231,32 +231,32 @@ ENTRY(aes_neon_enc1)
|
|||
* q15={d30-d31}: A2_B/sr[rmod4]
|
||||
*/
|
||||
|
||||
/* r12 := .Lconstants - .Lconstants_addr, r11 := .Lconstants_addr */
|
||||
ldr r12, .Lconstants_addr
|
||||
adr r11, .Lconstants_addr
|
||||
/* ip := .Lconstants - .Lconstants_addr, r10 := .Lconstants_addr */
|
||||
ldr ip, .Lconstants_addr
|
||||
adr r10, .Lconstants_addr
|
||||
|
||||
vld1.8 {q14}, [r0 :128]! /* q14 = *rk++ */
|
||||
movw r3, #0
|
||||
vmov.i8 q1, #0x0f
|
||||
|
||||
/* r12 := .Lconstants */
|
||||
add r12, r12, r11
|
||||
/* ip := .Lconstants */
|
||||
add ip, ip, r10
|
||||
|
||||
/* (q4, q5) := (iptlo, ipthi) */
|
||||
add r6, r12, #(ipt - .Lconstants)
|
||||
add r6, ip, #(ipt - .Lconstants)
|
||||
vld1.8 {q4-q5}, [r6 :256]
|
||||
|
||||
/* load the rest of the constants */
|
||||
add r4, r12, #(sb1 - .Lconstants)
|
||||
add r6, r12, #(sb2 - .Lconstants)
|
||||
add r8, r12, #(.Linv_inva - .Lconstants)
|
||||
add r4, ip, #(sb1 - .Lconstants)
|
||||
add r6, ip, #(sb2 - .Lconstants)
|
||||
add r8, ip, #(.Linv_inva - .Lconstants)
|
||||
vld1.8 {q6-q7}, [r4 :256] /* q6 = sb1[0], q7 = sb1[1] */
|
||||
vld1.8 {q8-q9}, [r6 :256] /* q8 = sb2[0], q9 = sb2[1] */
|
||||
vld1.8 {q10-q11}, [r8 :256] /* q10 = inv, q11 = inva */
|
||||
|
||||
/* (r4, r5) := (&mc_forward[0], &mc_backward[0]) */
|
||||
add r4, r12, #(mc_forward - .Lconstants)
|
||||
add r5, r12, #(mc_backward - .Lconstants)
|
||||
add r4, ip, #(mc_forward - .Lconstants)
|
||||
add r5, ip, #(mc_backward - .Lconstants)
|
||||
|
||||
/* (q2, q3) := (lo, hi) */
|
||||
vshr.u8 q3, q0, #4
|
||||
|
|
@ -295,9 +295,9 @@ ENTRY(aes_neon_enc1)
|
|||
|
||||
/* (q12, q13) := (mc_forward[rmod4], mc_backward[rmod4]) */
|
||||
add r6, r4, r3, lsl #4
|
||||
add r7, r5, r3, lsl #4
|
||||
add r8, r5, r3, lsl #4
|
||||
vld1.8 {q12}, [r6 :128]
|
||||
vld1.8 {q13}, [r7 :128]
|
||||
vld1.8 {q13}, [r8 :128]
|
||||
|
||||
/* q15 := A2_B = A2 + A(mcf) */
|
||||
vtbl.8 d30, {q0}, d24
|
||||
|
|
@ -365,8 +365,8 @@ ENTRY(aes_neon_enc1)
|
|||
bne 1b
|
||||
|
||||
/* (q6, q7, q15) := (sbo[0], sbo[1], sr[rmod4]) */
|
||||
add r8, r12, #(sr - .Lconstants)
|
||||
add r6, r12, #(sbo - .Lconstants)
|
||||
add r8, ip, #(sr - .Lconstants)
|
||||
add r6, ip, #(sbo - .Lconstants)
|
||||
add r8, r8, r3, lsl #4
|
||||
vld1.8 {q6-q7}, [r6 :256]
|
||||
vld1.8 {q15}, [r8 :128]
|
||||
|
|
@ -388,7 +388,7 @@ ENTRY(aes_neon_enc1)
|
|||
vtbl.8 d1, {q2}, d31
|
||||
|
||||
vpop {d8-d15}
|
||||
pop {r4, r5, r6, r7, r8, r10, r11, lr}
|
||||
pop {r4, r5, r6, r8, r10, lr}
|
||||
#ifdef __SOFTFP__
|
||||
#ifdef __ARM_BIG_ENDIAN
|
||||
vmov r1, r0, d0
|
||||
|
|
@ -426,7 +426,7 @@ ENTRY(aes_neon_dec1)
|
|||
vldr d1, [sp] /* d1 := x hi */
|
||||
ldr r1, [sp, #8] /* r1 := nrounds */
|
||||
#endif
|
||||
push {r4, r5, r6, r7, r8, r10, r11, lr}
|
||||
push {r4, r5, r6, r8, r10, lr}
|
||||
vpush {d8-d15}
|
||||
|
||||
/*
|
||||
|
|
@ -449,26 +449,26 @@ ENTRY(aes_neon_dec1)
|
|||
* q15={d30-d31}: mc/sr[3 & ~(nrounds - 1)]
|
||||
*/
|
||||
|
||||
/* r12 := .Lconstants - .Lconstants_addr, r11 := .Lconstants_addr */
|
||||
ldr r12, .Lconstants_addr
|
||||
adr r11, .Lconstants_addr
|
||||
/* ip := .Lconstants - .Lconstants_addr, r10 := .Lconstants_addr */
|
||||
ldr ip, .Lconstants_addr
|
||||
adr r10, .Lconstants_addr
|
||||
|
||||
vld1.8 {q14}, [r0 :128]! /* q14 = *rk++ */
|
||||
rsb r3, r1, #0 /* r3 := ~(x - 1) = -x */
|
||||
vmov.i8 q1, #0x0f
|
||||
and r3, r3, #3 /* r3 := 3 & ~(x - 1) */
|
||||
|
||||
/* r12 := .Lconstants */
|
||||
add r12, r12, r11
|
||||
/* ip := .Lconstants */
|
||||
add ip, ip, r10
|
||||
|
||||
/* (q4, q5) := (diptlo, dipthi) */
|
||||
add r6, r12, #(dipt - .Lconstants)
|
||||
add r6, ip, #(dipt - .Lconstants)
|
||||
vld1.8 {q4-q5}, [r6 :256]
|
||||
|
||||
/* load the rest of the constants */
|
||||
add r4, r12, #(dsbb - .Lconstants)
|
||||
add r6, r12, #(.Linv_inva - .Lconstants)
|
||||
add r8, r12, #(.Lmc_forward_3 - .Lconstants)
|
||||
add r4, ip, #(dsbb - .Lconstants)
|
||||
add r6, ip, #(.Linv_inva - .Lconstants)
|
||||
add r8, ip, #(.Lmc_forward_3 - .Lconstants)
|
||||
vld1.8 {q6-q7}, [r4 :256] /* q6 := dsbb[0], q7 := dsbb[1] */
|
||||
vld1.8 {q10-q11}, [r6 :256] /* q10 := inv, q11 := inva */
|
||||
vld1.8 {q15}, [r8 :128] /* q15 := mc_forward[3] */
|
||||
|
|
@ -485,7 +485,7 @@ ENTRY(aes_neon_dec1)
|
|||
vtbl.8 d7, {q5}, d7
|
||||
|
||||
/* load dsb9 */
|
||||
add r4, r12, #(dsb9 - .Lconstants)
|
||||
add r4, ip, #(dsb9 - .Lconstants)
|
||||
vld1.8 {q4-q5}, [r4 :256] /* q4 := dsb9[0], q5 := dsb9[1] */
|
||||
|
||||
/* q0 := rk[0] + diptlo(lo) + dipthi(hi) */
|
||||
|
|
@ -496,7 +496,7 @@ ENTRY(aes_neon_dec1)
|
|||
|
||||
_ALIGN_TEXT
|
||||
1: /* load dsbd */
|
||||
add r4, r12, #(dsbd - .Lconstants)
|
||||
add r4, ip, #(dsbd - .Lconstants)
|
||||
vld1.8 {q8-q9}, [r4 :256] /* q8 := dsbd[0], q9 := dsbd[1] */
|
||||
|
||||
vld1.8 {q14}, [r0 :128]! /* q14 = *rk++ */
|
||||
|
|
@ -522,7 +522,7 @@ ENTRY(aes_neon_dec1)
|
|||
veor q0, q0, q13
|
||||
|
||||
/* load dsbe */
|
||||
add r4, r12, #(dsbe - .Lconstants)
|
||||
add r4, ip, #(dsbe - .Lconstants)
|
||||
vld1.8 {q8-q9}, [r4 :256]! /* q8 := dsbe[0], q9 := dsbe[1] */
|
||||
|
||||
/* q0 := x(mc) + dsbb_0(io) + dsbb_1(jo) */
|
||||
|
|
@ -597,8 +597,8 @@ ENTRY(aes_neon_dec1)
|
|||
bne 1b
|
||||
|
||||
/* (q6, q7, q15) := (dsbo[0], dsbo[1], sr[i]) */
|
||||
add r8, r12, #(sr - .Lconstants)
|
||||
add r6, r12, #(dsbo - .Lconstants)
|
||||
add r8, ip, #(sr - .Lconstants)
|
||||
add r6, ip, #(dsbo - .Lconstants)
|
||||
add r8, r8, r3, lsl #4
|
||||
vld1.8 {q6-q7}, [r6 :256]
|
||||
vld1.8 {q15}, [r8 :128]
|
||||
|
|
@ -620,7 +620,7 @@ ENTRY(aes_neon_dec1)
|
|||
vtbl.8 d1, {q2}, d31
|
||||
|
||||
vpop {d8-d15}
|
||||
pop {r4, r5, r6, r7, r8, r10, r11, lr}
|
||||
pop {r4, r5, r6, r8, r10, lr}
|
||||
#ifdef __SOFTFP__
|
||||
#ifdef __ARM_BIG_ENDIAN
|
||||
vmov r1, r0, d0
|
||||
|
|
|
|||
Loading…
Reference in New Issue