webrtc/modules/audio_coding/codecs/isac/fix/source/lattice_armv7.S
Mirko Bonadei 08973eed36 Using fully qualified #include paths in isac code.
WebRTC internal code should always used include paths that starts
from the root of the project and that clearly identify the header file.

This allows 'gn check' to actually keep dependencies under control
because 'gn check' cannot enforce anything if the include path
is not fully qualified (starting from the root of the project).

Bug: webrtc:8815
Change-Id: I23fb4fed0c27a4d98bea360315b959af843587bc
Reviewed-on: https://webrtc-review.googlesource.com/46101
Commit-Queue: Mirko Bonadei <mbonadei@webrtc.org>
Reviewed-by: Karl Wiberg <kwiberg@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21847}
2018-02-01 14:57:44 +00:00

77 lines
2.6 KiB
ArmAsm

@
@ Copyright (c) 2011 The WebRTC project authors. All Rights Reserved.
@
@ Use of this source code is governed by a BSD-style license
@ that can be found in the LICENSE file in the root of the source
@ tree. An additional intellectual property rights grant can be found
@ in the file PATENTS. All contributing project authors may
@ be found in the AUTHORS file in the root of the source tree.
@
@ Contains a function for the core loop in the normalized lattice AR
@ filter routine for iSAC codec, optimized for ARMv7 platforms.
@
@ Output is bit-exact with the reference C code in lattic_c.c
@
@ Register usage:
@
@ r0: &ar_g_Q0
@ r1: &ar_f_Q0
@ r2: &cth_Q15
@ r3: &sth_Q15
@ r4: out loop counter
@ r5: tmpAR
@ r9: inner loop counter
@ r12: constant #16384
@ r6, r7, r8, r10, r11: scratch
#include "system_wrappers/include/asm_defines.h"
#include "modules/audio_coding/codecs/isac/fix/source/settings.h"
GLOBAL_FUNCTION WebRtcIsacfix_FilterArLoop
.align 2
DEFINE_FUNCTION WebRtcIsacfix_FilterArLoop
push {r4-r11}
add r1, #2 @ &ar_f_Q0[1]
mov r12, #16384
mov r4, #HALF_SUBFRAMELEN
sub r4, #1 @ Outer loop counter = HALF_SUBFRAMELEN - 1
HALF_SUBFRAME_LOOP: @ for (n = 0; n < HALF_SUBFRAMELEN - 1; n++)
ldr r9, [sp, #32] @ Restore the inner loop counter to order_coef
ldrh r5, [r1] @ tmpAR = ar_f_Q0[n+1]
add r0, r9, asl #1 @ Restore r0 to &ar_g_Q0[order_coef]
add r2, r9, asl #1 @ Restore r2 to &cth_Q15[order_coef]
add r3, r9, asl #1 @ Restore r3 to &sth_Q15[order_coef]
ORDER_COEF_LOOP: @ for (k = order_coef; k > 0; k--)
ldrh r7, [r3, #-2]! @ sth_Q15[k - 1]
ldrh r6, [r2, #-2]! @ cth_Q15[k - 1]
ldrh r8, [r0, #-2] @ ar_g_Q0[k - 1]
smlabb r11, r7, r5, r12 @ sth_Q15[k - 1] * tmpAR + 16384
smlabb r10, r6, r5, r12 @ cth_Q15[k - 1] * tmpAR + 16384
smulbb r7, r7, r8 @ sth_Q15[k - 1] * ar_g_Q0[k - 1]
smlabb r11, r6, r8, r11 @ cth_Q15[k - 1] * ar_g_Q0[k - 1] +
@ (sth_Q15[k - 1] * tmpAR + 16384)
sub r10, r10, r7 @ cth_Q15[k - 1] * tmpAR + 16384 -
@ (sth_Q15[k - 1] * ar_g_Q0[k - 1])
ssat r11, #16, r11, asr #15
ssat r5, #16, r10, asr #15
strh r11, [r0], #-2 @ Output: ar_g_Q0[k]
subs r9, #1
bgt ORDER_COEF_LOOP
strh r5, [r0] @ Output: ar_g_Q0[0] = tmpAR;
strh r5, [r1], #2 @ Output: ar_f_Q0[n+1] = tmpAR;
subs r4, #1
bne HALF_SUBFRAME_LOOP
pop {r4-r11}
bx lr