[pcsx_rearmed.git] / plugins / dfsound / arm_utils.s

/*
 * (C) Gražvydas "notaz" Ignotas, 2011
 *
 * This work is licensed under the terms of any of these licenses
 * (at your option):
 *  - GNU GPL, version 2 or later.
 *  - GNU LGPL, version 2.1 or later.
 * See the COPYING file in the top-level directory.
 */


.text
.align 2

@ XXX: should be HAVE_NEON
.if HAVE_ARMV7

.global mix_chan @ (int start, int count, int lv, int rv)
mix_chan:
    vmov.32     d14[0], r2
    vmov.32     d14[1], r3             @ multipliers
    mov         r12, r0
    movw        r0, #:lower16:ChanBuf
    movw        r2, #:lower16:SSumLR
    movt        r0, #:upper16:ChanBuf
    movt        r2, #:upper16:SSumLR
    add         r0, r12, lsl #2
    add         r2, r12, lsl #3
0:
    vldmia      r0!, {d0-d1}
    vldmia      r2, {d2-d5}
    vmul.s32    d10, d14, d0[0]
    vmul.s32    d11, d14, d0[1]
    vmul.s32    d12, d14, d1[0]
    vmul.s32    d13, d14, d1[1]
    vsra.s32    q1, q5, #14
    vsra.s32    q2, q6, #14
    subs        r1, #4
    blt         mc_finish
    vstmia      r2!, {d2-d5}
    bgt         0b
    nop
    bxeq        lr

mc_finish:
    vstmia      r2!, {d2}
    cmp         r1, #-2
    vstmiage    r2!, {d3}
    cmp         r1, #-1
    vstmiage    r2!, {d4}
    bx          lr


.global mix_chan_rvb @ (int start, int count, int lv, int rv)
mix_chan_rvb:
    vmov.32     d14[0], r2
    vmov.32     d14[1], r3             @ multipliers
    mov         r12, r0
    movw        r0, #:lower16:ChanBuf
    movw        r3, #:lower16:sRVBStart
    movw        r2, #:lower16:SSumLR
    movt        r0, #:upper16:ChanBuf
    movt        r3, #:upper16:sRVBStart
    movt        r2, #:upper16:SSumLR
    ldr         r3, [r3]
    add         r0, r12, lsl #2
    add         r2, r12, lsl #3
    add         r3, r12, lsl #3
0:
    vldmia      r0!, {d0-d1}
    vldmia      r2, {d2-d5}
    vldmia      r3, {d6-d9}
    vmul.s32    d10, d14, d0[0]
    vmul.s32    d11, d14, d0[1]
    vmul.s32    d12, d14, d1[0]
    vmul.s32    d13, d14, d1[1]
    vsra.s32    q1, q5, #14
    vsra.s32    q2, q6, #14
    vsra.s32    q3, q5, #14
    vsra.s32    q4, q6, #14
    subs        r1, #4
    blt         mcr_finish
    vstmia      r2!, {d2-d5}
    vstmia      r3!, {d6-d9}
    bgt         0b
    nop
    bxeq        lr

mcr_finish:
    vstmia      r2!, {d2}
    vstmia      r3!, {d6}
    cmp         r1, #-2
    vstmiage    r2!, {d3}
    vstmiage    r3!, {d7}
    cmp         r1, #-1
    vstmiage    r2!, {d4}
    vstmiage    r3!, {d8}
    bx          lr

.else

.global mix_chan @ (int start, int count, int lv, int rv)
mix_chan:
    stmfd       sp!, {r4-r8,lr}
    orr         r3, r2, r3, lsl #16
    lsl         r3, #1                 @ packed multipliers << 1
    mov         r12, r0
    ldr         r0, =ChanBuf
    ldr         r2, =SSumLR
    add         r0, r12, lsl #2
    add         r2, r12, lsl #3
0:
    ldmia       r0!, {r4,r5}
    ldmia       r2, {r6-r8,lr}
    lsl         r4, #1                 @ adjust for mul
    lsl         r5, #1
    smlawb      r6, r4, r3, r6
    smlawt      r7, r4, r3, r7
    smlawb      r8, r5, r3, r8
    smlawt      lr, r5, r3, lr
    subs        r1, #2
    blt         mc_finish
    stmia       r2!, {r6-r8,lr}
    bgt         0b
    ldmeqfd     sp!, {r4-r8,pc}

mc_finish:
    stmia       r2!, {r6,r7}
    ldmfd       sp!, {r4-r8,pc}


.global mix_chan_rvb @ (int start, int count, int lv, int rv)
mix_chan_rvb:
    stmfd       sp!, {r4-r8,lr}
    orr         lr, r2, r3, lsl #16
    lsl         lr, #1
    ldr         r3, =sRVBStart
    ldr         r2, =SSumLR
    ldr         r4, =ChanBuf
    ldr         r3, [r3]
    add         r2, r2, r0, lsl #3
    add         r3, r3, r0, lsl #3
    add         r0, r4, r0, lsl #2
0:
    ldr         r4, [r0], #4
    ldmia       r2, {r6,r7}
    ldmia       r3, {r8,r12}
    lsl         r4, #1
    smlawb      r6, r4, lr, r6         @ supposedly takes single cycle?
    smlawt      r7, r4, lr, r7
    smlawb      r8, r4, lr, r8
    smlawt      r12,r4, lr, r12
    subs        r1, #1
    stmia       r2!, {r6,r7}
    stmia       r3!, {r8,r12}
    bgt         0b
    ldmfd       sp!, {r4-r8,pc}

.endif

@ vim:filetype=armasm
Commit	Line	Data
b17618c0	1	/*
	2	* (C) Gražvydas "notaz" Ignotas, 2011
	3	*
	4	* This work is licensed under the terms of any of these licenses
	5	* (at your option):
	6	* - GNU GPL, version 2 or later.
	7	* - GNU LGPL, version 2.1 or later.
	8	* See the COPYING file in the top-level directory.
	9	*/
	10
	11
	12	.text
	13	.align 2
	14
3a721c1f	15	@ XXX: should be HAVE_NEON
3a721c1f	16	.if HAVE_ARMV7
b17618c0	17
	18	.global mix_chan @ (int start, int count, int lv, int rv)
	19	mix_chan:
	20	vmov.32 d14[0], r2
	21	vmov.32 d14[1], r3 @ multipliers
	22	mov r12, r0
	23	movw r0, #:lower16:ChanBuf
	24	movw r2, #:lower16:SSumLR
	25	movt r0, #:upper16:ChanBuf
	26	movt r2, #:upper16:SSumLR
	27	add r0, r12, lsl #2
	28	add r2, r12, lsl #3
	29	0:
	30	vldmia r0!, {d0-d1}
	31	vldmia r2, {d2-d5}
	32	vmul.s32 d10, d14, d0[0]
	33	vmul.s32 d11, d14, d0[1]
	34	vmul.s32 d12, d14, d1[0]
	35	vmul.s32 d13, d14, d1[1]
	36	vsra.s32 q1, q5, #14
	37	vsra.s32 q2, q6, #14
	38	subs r1, #4
	39	blt mc_finish
	40	vstmia r2!, {d2-d5}
	41	bgt 0b
	42	nop
	43	bxeq lr
	44
	45	mc_finish:
	46	vstmia r2!, {d2}
587fa7de	47	cmp r1, #-2
b17618c0	48	vstmiage r2!, {d3}
587fa7de	49	cmp r1, #-1
b17618c0	50	vstmiage r2!, {d4}
	51	bx lr
	52
	53
	54	.global mix_chan_rvb @ (int start, int count, int lv, int rv)
	55	mix_chan_rvb:
	56	vmov.32 d14[0], r2
	57	vmov.32 d14[1], r3 @ multipliers
	58	mov r12, r0
	59	movw r0, #:lower16:ChanBuf
	60	movw r3, #:lower16:sRVBStart
	61	movw r2, #:lower16:SSumLR
	62	movt r0, #:upper16:ChanBuf
	63	movt r3, #:upper16:sRVBStart
	64	movt r2, #:upper16:SSumLR
	65	ldr r3, [r3]
	66	add r0, r12, lsl #2
	67	add r2, r12, lsl #3
	68	add r3, r12, lsl #3
	69	0:
	70	vldmia r0!, {d0-d1}
	71	vldmia r2, {d2-d5}
	72	vldmia r3, {d6-d9}
	73	vmul.s32 d10, d14, d0[0]
	74	vmul.s32 d11, d14, d0[1]
	75	vmul.s32 d12, d14, d1[0]
	76	vmul.s32 d13, d14, d1[1]
	77	vsra.s32 q1, q5, #14
	78	vsra.s32 q2, q6, #14
	79	vsra.s32 q3, q5, #14
	80	vsra.s32 q4, q6, #14
	81	subs r1, #4
	82	blt mcr_finish
	83	vstmia r2!, {d2-d5}
	84	vstmia r3!, {d6-d9}
	85	bgt 0b
	86	nop
	87	bxeq lr
	88
	89	mcr_finish:
	90	vstmia r2!, {d2}
	91	vstmia r3!, {d6}
587fa7de	92	cmp r1, #-2
b17618c0	93	vstmiage r2!, {d3}
b17618c0	94	vstmiage r3!, {d7}
587fa7de	95	cmp r1, #-1
b17618c0	96	vstmiage r2!, {d4}
	97	vstmiage r3!, {d8}
	98	bx lr
	99
3a721c1f	100	.else
	101
	102	.global mix_chan @ (int start, int count, int lv, int rv)
	103	mix_chan:
	104	stmfd sp!, {r4-r8,lr}
	105	orr r3, r2, r3, lsl #16
	106	lsl r3, #1 @ packed multipliers << 1
	107	mov r12, r0
	108	ldr r0, =ChanBuf
	109	ldr r2, =SSumLR
	110	add r0, r12, lsl #2
	111	add r2, r12, lsl #3
	112	0:
	113	ldmia r0!, {r4,r5}
	114	ldmia r2, {r6-r8,lr}
	115	lsl r4, #1 @ adjust for mul
	116	lsl r5, #1
	117	smlawb r6, r4, r3, r6
	118	smlawt r7, r4, r3, r7
	119	smlawb r8, r5, r3, r8
	120	smlawt lr, r5, r3, lr
	121	subs r1, #2
	122	blt mc_finish
	123	stmia r2!, {r6-r8,lr}
	124	bgt 0b
	125	ldmeqfd sp!, {r4-r8,pc}
	126
	127	mc_finish:
	128	stmia r2!, {r6,r7}
	129	ldmfd sp!, {r4-r8,pc}
	130
	131
	132	.global mix_chan_rvb @ (int start, int count, int lv, int rv)
	133	mix_chan_rvb:
	134	stmfd sp!, {r4-r8,lr}
	135	orr lr, r2, r3, lsl #16
	136	lsl lr, #1
	137	ldr r3, =sRVBStart
	138	ldr r2, =SSumLR
	139	ldr r4, =ChanBuf
	140	ldr r3, [r3]
	141	add r2, r2, r0, lsl #3
	142	add r3, r3, r0, lsl #3
	143	add r0, r4, r0, lsl #2
	144	0:
	145	ldr r4, [r0], #4
	146	ldmia r2, {r6,r7}
	147	ldmia r3, {r8,r12}
	148	lsl r4, #1
	149	smlawb r6, r4, lr, r6 @ supposedly takes single cycle?
	150	smlawt r7, r4, lr, r7
	151	smlawb r8, r4, lr, r8
	152	smlawt r12,r4, lr, r12
	153	subs r1, #1
	154	stmia r2!, {r6,r7}
	155	stmia r3!, {r8,r12}
	156	bgt 0b
	157	ldmfd sp!, {r4-r8,pc}
	158
	159	.endif
	160
b17618c0	161	@ vim:filetype=armasm