From 2b04e1ece6def1bd836dee9c223f5c1be59c403e Mon Sep 17 00:00:00 2001
From: Themaister <maister@archlinux.us>
Date: Mon, 3 Dec 2012 00:22:07 +0100
Subject: [PATCH 1/7] Start adding NEON sinc.

---
 audio/sinc.c      | 14 +++++++++++++
 audio/sinc_neon.S | 52 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 66 insertions(+)
 create mode 100644 audio/sinc_neon.S

diff --git a/audio/sinc.c b/audio/sinc.c
index 4041e8caed..e4ab461755 100644
--- a/audio/sinc.c
+++ b/audio/sinc.c
@@ -123,6 +123,8 @@ rarch_resampler_t *resampler_new(void)
    RARCH_LOG("Sinc resampler [AVX]\n");
 #elif defined(__SSE__)
    RARCH_LOG("Sinc resampler [SSE]\n");
+#elif defined(HAVE_NEON)
+   RARCH_LOG("Sinc resampler [NEON]\n");
 #else
    RARCH_LOG("Sinc resampler [C]\n");
 #endif
@@ -208,6 +210,18 @@ static void process_sinc(rarch_resampler_t *resamp, float *out_buffer)
    // movehl { X, R, X, L } == { X, R, X, R }
    _mm_store_ss(out_buffer + 1, _mm_movehl_ps(sum, sum));
 }
+#elif defined(HAVE_NEON)
+void process_sinc_neon_asm(float *out, const float *left, const float *right, const float *coeff);
+static void process_sinc(rarch_resampler_t *resamp, float *out_buffer)
+{
+   const float *buffer_l = resamp->buffer_l + resamp->ptr;
+   const float *buffer_r = resamp->buffer_r + resamp->ptr;
+
+   unsigned phase = resamp->time >> SUBPHASE_BITS;
+   const float *phase_table = resamp->phase_table[phase];
+
+   process_sinc_neon_asm(out_buffer, buffer_l, buffer_r, phase_table);
+}
 #else // Plain ol' C99
 static void process_sinc(rarch_resampler_t *resamp, float *out_buffer)
 {
diff --git a/audio/sinc_neon.S b/audio/sinc_neon.S
new file mode 100644
index 0000000000..f6c78ce122
--- /dev/null
+++ b/audio/sinc_neon.S
@@ -0,0 +1,52 @@
+/*  RetroArch - A frontend for libretro.
+ *  Copyright (C) 2010-2012 - Hans-Kristian Arntzen
+ * 
+ *  RetroArch is free software: you can redistribute it and/or modify it under the terms
+ *  of the GNU General Public License as published by the Free Software Found-
+ *  ation, either version 3 of the License, or (at your option) any later version.
+ *
+ *  RetroArch is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY;
+ *  without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
+ *  PURPOSE.  See the GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License along with RetroArch.
+ *  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+.arm
+.align 4
+.global process_sinc_neon_asm
+# void process_sinc_neon(float *out, const float *left, const float *right, const float *coeff)
+# Hardcoded to 16 taps.
+process_sinc_neon_asm:
+   # Left
+   vld1.f32 {q0-q1},   [r1]!
+   vld1.f32 {q2-q3},   [r1]!
+   # Right
+   vld1.f32 {q4-q5},   [r2]!
+   vld1.f32 {q6-q7},   [r2]!
+   # Coeff
+   vld1.f32 {q8-q9},   [r3, :1024]!
+   vld1.f32 {q10-q11}, [r3, :1024]!
+
+   # Left
+   vmul.f32 q0, q0, q8
+   vmul.f32 q1, q1, q9
+   vmla.f32 q0, q2, q10
+   vmla.f32 q1, q3, q11
+
+   # Right
+   vmul.f32 q4, q4, q8
+   vmul.f32 q5, q5, q9
+   vmla.f32 q4, q6, q10
+   vmla.f32 q5, q7, q11
+
+   # Add everything together
+   vadd.f32 q0, q0, q1
+   vadd.f32 q4, q4, q5
+   vadd.f32 d0, d0, d1
+   vadd.f32 d8, d8, d9
+   vpadd.f32 d0, d0, d8
+   vst1.f32 d0, [r0]
+   
+   bx lr

From 1c9dec56a8b8a5b3c6ec173a33fd6e9daea56b11 Mon Sep 17 00:00:00 2001
From: twinaphex <autechre1024@hotmail.com>
Date: Sun, 2 Dec 2012 23:50:32 +0100
Subject: [PATCH 2/7] (NEON/Android) Some fixes

---
 android/native/jni/Android.mk | 9 +++++----
 audio/sinc_neon.S             | 4 ++--
 console/griffin/griffin.c     | 4 ++++
 gfx/scaler/filter.c           | 4 ++--
 4 files changed, 13 insertions(+), 8 deletions(-)

diff --git a/android/native/jni/Android.mk b/android/native/jni/Android.mk
index ca27487b94..a84b349509 100644
--- a/android/native/jni/Android.mk
+++ b/android/native/jni/Android.mk
@@ -1,6 +1,6 @@
 RARCH_VERSION		= "0.9.8-beta3"
 LOCAL_PATH := $(call my-dir)
-PERF_TEST := 0
+PERF_TEST := 1
 HAVE_OPENSL     := 1
 
 include $(CLEAR_VARS)
@@ -15,8 +15,9 @@ LOCAL_CFLAGS += -DANDROID_X86 -DHAVE_SSSE3
 endif
 
 ifeq ($(TARGET_ARCH_ABI),armeabi-v7a)
-#LOCAL_CFLAGS += -mfpu=neon
-LOCAL_CFLAGS += -DANDROID_ARM_V7
+LOCAL_CFLAGS += -mfpu=neon
+LOCAL_CFLAGS += -DANDROID_ARM_V7 -DHAVE_NEON
+LOCAL_SRC_FILES += ../../../audio/sinc_neon.S.neon
 endif
 
 ifeq ($(TARGET_ARCH),mips)
@@ -27,7 +28,7 @@ LOCAL_MODULE    := retroarch-activity
 
 RARCH_PATH  := ../../..
 LIBXML_PATH := ../../../deps/libxml2
-LOCAL_SRC_FILES    =	$(RARCH_PATH)/console/griffin/griffin.c
+LOCAL_SRC_FILES  +=	$(RARCH_PATH)/console/griffin/griffin.c
 
 
 ifeq ($(PERF_TEST), 1)
diff --git a/audio/sinc_neon.S b/audio/sinc_neon.S
index f6c78ce122..2b67215214 100644
--- a/audio/sinc_neon.S
+++ b/audio/sinc_neon.S
@@ -26,8 +26,8 @@ process_sinc_neon_asm:
    vld1.f32 {q4-q5},   [r2]!
    vld1.f32 {q6-q7},   [r2]!
    # Coeff
-   vld1.f32 {q8-q9},   [r3, :1024]!
-   vld1.f32 {q10-q11}, [r3, :1024]!
+   vld1.f32 {q8-q9},   [r3, :128]!
+   vld1.f32 {q10-q11}, [r3, :128]!
 
    # Left
    vmul.f32 q0, q0, q8
diff --git a/console/griffin/griffin.c b/console/griffin/griffin.c
index cf397a897c..e5e7cdd580 100644
--- a/console/griffin/griffin.c
+++ b/console/griffin/griffin.c
@@ -247,7 +247,11 @@ FIFO BUFFER
 /*============================================================
 AUDIO HERMITE
 ============================================================ */
+#ifdef HAVE_NEON
+#include "../../audio/sinc.c"
+#else
 #include "../../audio/hermite.c"
+#endif
 
 /*============================================================
 RSOUND
diff --git a/gfx/scaler/filter.c b/gfx/scaler/filter.c
index c4bb12f606..df469eca98 100644
--- a/gfx/scaler/filter.c
+++ b/gfx/scaler/filter.c
@@ -95,7 +95,7 @@ static bool gen_filter_bilinear(struct scaler_ctx *ctx)
    return true;
 }
 
-static inline double sinc(double phase)
+static inline double filter_sinc(double phase)
 {
    if (fabs(phase) < 0.0001)
       return 1.0;
@@ -116,7 +116,7 @@ static void gen_filter_sinc_sub(struct scaler_filter *filter, int len, int pos,
       {
          double sinc_phase    = M_PI * ((double)((sinc_size << 15) + (pos & 0xffff)) / 0x10000 - j);
          double lanczos_phase = sinc_phase / ((sinc_size >> 1));
-         int16_t sinc_val     = FILTER_UNITY * sinc(sinc_phase * phase_mul) * sinc(lanczos_phase) * phase_mul;
+         int16_t sinc_val     = FILTER_UNITY * filter_sinc(sinc_phase * phase_mul) * filter_sinc(lanczos_phase) * phase_mul;
          //sinc_sum += sinc_val;
 
          filter->filter[i * sinc_size + j] = sinc_val;

From 63cf9247f1f2c8481e4a0c2ad72c18df0608d0b4 Mon Sep 17 00:00:00 2001
From: twinaphex <autechre1024@hotmail.com>
Date: Sun, 2 Dec 2012 23:53:08 +0100
Subject: [PATCH 3/7] (Hermite) Add log

---
 audio/hermite.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/audio/hermite.c b/audio/hermite.c
index c96008b8a3..97fd75b125 100644
--- a/audio/hermite.c
+++ b/audio/hermite.c
@@ -52,6 +52,7 @@ static inline float hermite_kernel(float mu1, float a, float b, float c, float d
 
 rarch_resampler_t *resampler_new(void)
 {
+   RARCH_LOG("Hermite resampler [C]\n");
    return (rarch_resampler_t*)calloc(1, sizeof(rarch_resampler_t));
 }
 

From 7403faa6083325d1c31debb2f031754346f8c576 Mon Sep 17 00:00:00 2001
From: twinaphex <autechre1024@hotmail.com>
Date: Mon, 3 Dec 2012 00:20:21 +0100
Subject: [PATCH 4/7] HAVE_NEON define seems to give problems - WANT_NEON seems
 to work - so quick-fix hack for now

---
 android/native/jni/Android.mk | 4 ++--
 audio/sinc.c                  | 4 ++--
 console/griffin/griffin.c     | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/android/native/jni/Android.mk b/android/native/jni/Android.mk
index a84b349509..b32b8a6fbe 100644
--- a/android/native/jni/Android.mk
+++ b/android/native/jni/Android.mk
@@ -15,8 +15,8 @@ LOCAL_CFLAGS += -DANDROID_X86 -DHAVE_SSSE3
 endif
 
 ifeq ($(TARGET_ARCH_ABI),armeabi-v7a)
-LOCAL_CFLAGS += -mfpu=neon
-LOCAL_CFLAGS += -DANDROID_ARM_V7 -DHAVE_NEON
+LOCAL_CFLAGS += -DANDROID_ARM_V7
+LOCAL_CFLAGS += -DWANT_NEON
 LOCAL_SRC_FILES += ../../../audio/sinc_neon.S.neon
 endif
 
diff --git a/audio/sinc.c b/audio/sinc.c
index e4ab461755..519254d116 100644
--- a/audio/sinc.c
+++ b/audio/sinc.c
@@ -123,7 +123,7 @@ rarch_resampler_t *resampler_new(void)
    RARCH_LOG("Sinc resampler [AVX]\n");
 #elif defined(__SSE__)
    RARCH_LOG("Sinc resampler [SSE]\n");
-#elif defined(HAVE_NEON)
+#elif defined(WANT_NEON)
    RARCH_LOG("Sinc resampler [NEON]\n");
 #else
    RARCH_LOG("Sinc resampler [C]\n");
@@ -210,7 +210,7 @@ static void process_sinc(rarch_resampler_t *resamp, float *out_buffer)
    // movehl { X, R, X, L } == { X, R, X, R }
    _mm_store_ss(out_buffer + 1, _mm_movehl_ps(sum, sum));
 }
-#elif defined(HAVE_NEON)
+#elif defined(WANT_NEON)
 void process_sinc_neon_asm(float *out, const float *left, const float *right, const float *coeff);
 static void process_sinc(rarch_resampler_t *resamp, float *out_buffer)
 {
diff --git a/console/griffin/griffin.c b/console/griffin/griffin.c
index e5e7cdd580..c355d5fa64 100644
--- a/console/griffin/griffin.c
+++ b/console/griffin/griffin.c
@@ -247,7 +247,7 @@ FIFO BUFFER
 /*============================================================
 AUDIO HERMITE
 ============================================================ */
-#ifdef HAVE_NEON
+#ifdef WANT_NEON
 #include "../../audio/sinc.c"
 #else
 #include "../../audio/hermite.c"

From 5eaaaacaf6cd60e1ca0599e35d218650a2e54198 Mon Sep 17 00:00:00 2001
From: twinaphex <autechre1024@hotmail.com>
Date: Mon, 3 Dec 2012 01:29:04 +0100
Subject: [PATCH 5/7] (Android) Make Neon compilation optional with a switch

---
 android/native/jni/Android.mk | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/android/native/jni/Android.mk b/android/native/jni/Android.mk
index b32b8a6fbe..ac8f268e00 100644
--- a/android/native/jni/Android.mk
+++ b/android/native/jni/Android.mk
@@ -2,6 +2,7 @@ RARCH_VERSION		= "0.9.8-beta3"
 LOCAL_PATH := $(call my-dir)
 PERF_TEST := 1
 HAVE_OPENSL     := 1
+WANT_NEON := 0
 
 include $(CLEAR_VARS)
 
@@ -15,10 +16,13 @@ LOCAL_CFLAGS += -DANDROID_X86 -DHAVE_SSSE3
 endif
 
 ifeq ($(TARGET_ARCH_ABI),armeabi-v7a)
-LOCAL_CFLAGS += -DANDROID_ARM_V7
+ifeq ($(WANT_NEON),1)
 LOCAL_CFLAGS += -DWANT_NEON
 LOCAL_SRC_FILES += ../../../audio/sinc_neon.S.neon
 endif
+LOCAL_CFLAGS += -DANDROID_ARM_V7
+endif
+
 
 ifeq ($(TARGET_ARCH),mips)
 LOCAL_CFLAGS += -DANDROID_MIPS -D__mips__ -D__MIPSEL__

From 6a7e03aadb54bb5598664d205059e8fa850179f1 Mon Sep 17 00:00:00 2001
From: twinaphex <autechre1024@hotmail.com>
Date: Mon, 3 Dec 2012 03:21:10 +0100
Subject: [PATCH 6/7] (Android) Fix sound - don't fix samplerate at 44Khz

---
 audio/opensl.c | 2 --
 gfx/gl.c       | 2 +-
 2 files changed, 1 insertion(+), 3 deletions(-)

diff --git a/audio/opensl.c b/audio/opensl.c
index 8483e815d6..f672b99e89 100644
--- a/audio/opensl.c
+++ b/audio/opensl.c
@@ -109,8 +109,6 @@ static void *sl_init(const char *device, unsigned rate, unsigned latency)
 {
    (void)device;
 
-   g_settings.audio.out_rate = 44100;
-
    SLDataFormat_PCM fmt_pcm = {0};
    SLDataSource audio_src   = {0};
    SLDataSink audio_sink    = {0};
diff --git a/gfx/gl.c b/gfx/gl.c
index 2606410ea3..bccfd634fe 100644
--- a/gfx/gl.c
+++ b/gfx/gl.c
@@ -1175,7 +1175,7 @@ static bool gl_frame(void *data, const void *frame, unsigned width, unsigned hei
    if (msg)
       gl_render_msg(gl, msg);
 
-#if !defined(RARCH_CONSOLE) && !defined(ANDROID)
+#if !defined(RARCH_CONSOLE)
    context_update_window_title_func(false);
 #endif
 

From ddc9c94884f061b6e92782921483cea0a0b1ca39 Mon Sep 17 00:00:00 2001
From: twinaphex <autechre1024@hotmail.com>
Date: Mon, 3 Dec 2012 04:58:44 +0100
Subject: [PATCH 7/7] (Android) Refactor Android.mk - better defines

---
 android/native/jni/Android.mk | 11 ++++++++++-
 console/griffin/griffin.c     |  2 +-
 2 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/android/native/jni/Android.mk b/android/native/jni/Android.mk
index ac8f268e00..d318c9e179 100644
--- a/android/native/jni/Android.mk
+++ b/android/native/jni/Android.mk
@@ -2,7 +2,8 @@ RARCH_VERSION		= "0.9.8-beta3"
 LOCAL_PATH := $(call my-dir)
 PERF_TEST := 1
 HAVE_OPENSL     := 1
-WANT_NEON := 0
+WANT_NEON := 1
+HAVE_SINC := 1
 
 include $(CLEAR_VARS)
 
@@ -16,10 +17,18 @@ LOCAL_CFLAGS += -DANDROID_X86 -DHAVE_SSSE3
 endif
 
 ifeq ($(TARGET_ARCH_ABI),armeabi-v7a)
+
 ifeq ($(WANT_NEON),1)
 LOCAL_CFLAGS += -DWANT_NEON
+endif
+
+ifeq ($(HAVE_SINC),1)
+ifeq ($(WANT_NEON),1)
 LOCAL_SRC_FILES += ../../../audio/sinc_neon.S.neon
 endif
+LOCAL_CFLAGS += -DHAVE_SINC
+endif
+
 LOCAL_CFLAGS += -DANDROID_ARM_V7
 endif
 
diff --git a/console/griffin/griffin.c b/console/griffin/griffin.c
index c355d5fa64..d0bd986416 100644
--- a/console/griffin/griffin.c
+++ b/console/griffin/griffin.c
@@ -247,7 +247,7 @@ FIFO BUFFER
 /*============================================================
 AUDIO HERMITE
 ============================================================ */
-#ifdef WANT_NEON
+#ifdef HAVE_SINC
 #include "../../audio/sinc.c"
 #else
 #include "../../audio/hermite.c"