From 8a4e9922141cd6bf3471e386cd52978bc4064ede Mon Sep 17 00:00:00 2001
From: itsmattkc <34096995+itsmattkc@users.noreply.github.com>
Date: Wed, 13 Oct 2021 17:00:43 -0700
Subject: [PATCH] minor formatting

---
 app/audio/audiovisualwaveform.cpp | 28 ++++++++++++++--------------
 app/node/math/math/mathbase.cpp   | 27 +++++++++++++++++++++++++++
 2 files changed, 41 insertions(+), 14 deletions(-)

diff --git a/app/audio/audiovisualwaveform.cpp b/app/audio/audiovisualwaveform.cpp
index 22e169a2f..1018c3e76 100644
--- a/app/audio/audiovisualwaveform.cpp
+++ b/app/audio/audiovisualwaveform.cpp
@@ -17,15 +17,15 @@
   along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
 ***/
-#include <QtGlobal>
-
-#ifdef Q_PROCESSOR_X86_64
-#include <xmmintrin.h>
-#endif
 
 #include "audiovisualwaveform.h"
 
 #include <QDebug>
+#include <QtGlobal>
+
+#ifdef Q_PROCESSOR_X86
+#include <xmmintrin.h>
+#endif
 
 #include "config/config.h"
 #include "common/functiontimer.h"
@@ -291,15 +291,15 @@ AudioVisualWaveform::Sample AudioVisualWaveform::SumSamples(const float *samples
   return summed_samples;
 }
 
-#ifdef Q_PROCESSOR_X86_64
-void min_max_sse(float *a, int start, int end, float &min_val, float &max_val) 
+#ifdef Q_PROCESSOR_X86
+void ExpandMinMaxSSE(float *a, int start, int end, float &min_val, float &max_val)
 {
   // load the first 4 elements of 'a' into min and max (they are 4 * 32 = 128 bits)
-  __m128 max = _mm_loadu_ps(a + start); 
+  __m128 max = _mm_loadu_ps(a + start);
   __m128 min = _mm_loadu_ps(a + start);
 
-  // loop over 'a' and compare current elements with min and max 4 by 4. 
-  // we need to make sure we don't read out of boundaries should 'a' lenght be not mod. 4 
+  // loop over 'a' and compare current elements with min and max 4 by 4.
+  // we need to make sure we don't read out of boundaries should 'a' lenght be not mod. 4
   for(int i = 4; i < end-4; i+=4) {
     __m128 cur = _mm_loadu_ps(a + start + i);
     max = _mm_max_ps(max, cur);
@@ -311,7 +311,7 @@ void min_max_sse(float *a, int start, int end, float &min_val, float &max_val)
   min = _mm_min_ps(min, cur);
   // this potentially overlaps up to the last 3 elements but it's not an issue.
 
-  // min and max will contain 4 min and max. To get the absolute min and max 
+  // min and max will contain 4 min and max. To get the absolute min and max
   // we need to compare the 4 values over themselves by shuffling each time.
   for (int i = 0; i < 3; i++) {
     max = _mm_max_ps(max, _mm_shuffle_ps(max, max, 0x93));
@@ -331,11 +331,11 @@ AudioVisualWaveform::Sample AudioVisualWaveform::SumSamples(SampleBufferPtr samp
   int channels = samples->audio_params().channel_count();
   AudioVisualWaveform::Sample summed_samples(channels);
 
-  #ifdef Q_PROCESSOR_X86_64
+  #ifdef Q_PROCESSOR_X86
     for (int channel=0; channel<samples->audio_params().channel_count(); channel++) {
-      min_max_sse(samples->data(channel), start_index, length, summed_samples[channel].min, summed_samples[channel].max);
+      ExpandMinMaxSSE(samples->data(channel), start_index, length, summed_samples[channel].min, summed_samples[channel].max);
     }
-  #else 
+  #else
     int end_index = start_index + length;
     for (int channel=0; channel<samples->audio_params().channel_count(); channel++) {
       for (int i=start_index; i<end_index; i++) {
diff --git a/app/node/math/math/mathbase.cpp b/app/node/math/math/mathbase.cpp
index f19ede011..3633afe2f 100644
--- a/app/node/math/math/mathbase.cpp
+++ b/app/node/math/math/mathbase.cpp
@@ -23,6 +23,10 @@
 #include <QMatrix4x4>
 #include <QVector2D>
 
+#ifdef Q_PROCESSOR_X86
+#include <xmmintrin.h>
+#endif
+
 #include "common/tohex.h"
 #include "node/distort/transform/transformdistortnode.h"
 #include "render/color.h"
@@ -349,9 +353,32 @@ void MathNodeBase::ValueInternal(Operation operation, Pairing pairing, const QSt
       if (IsInputStatic(number_param)) {
         if (!NumberIsNoOp(operation, number)) {
           for (int i=0;i<job.samples()->audio_params().channel_count();i++) {
+#ifdef Q_PROCESSOR_X86
+            // Use SSE instructions for optimization
+
+            float *a = job.samples()->data(i);
+            int end = job.samples()->sample_count();
+
+            int end_divisible_4 = (end / 4) * 4;
+
+            // loop over 'a' and compare current elements with min and max 4 by 4.
+            // we need to make sure we don't read out of boundaries should 'a' lenght be not mod. 4
+            __m128 mult = _mm_load1_ps(&number);
+            for(int j = 0; j < end_divisible_4; j+=4) {
+              __m128 cur = _mm_loadu_ps(a + j);
+              __m128 res = _mm_mul_ps(cur, mult);
+              _mm_storeu_ps(a + j, res);
+            }
+
+            // Do last three numbers, if non-divisible by 4
+            for (int j=end_divisible_4; j<end; j++) {
+              a[j] *= number;
+            }
+#else
             for (int j=0;j<job.samples()->sample_count();j++) {
               job.samples()->data(i)[j] = PerformAll(operation, job.samples()->data(i)[j], number);
             }
+#endif
           }
         }