Make sure all gain steps are applied with the SSE and Neon mixers

2 files changed, 22 insertions, 44 deletions

diff --git a/Alc/mixer_neon.c b/Alc/mixer_neon.c
index 5a598f5c..1b695d81 100644
--- a/Alc/mixer_neon.c
+++ b/Alc/mixer_neon.c
@@ -93,25 +93,14 @@ void MixDirect_Neon(DirectParams *params, const ALfloat *restrict data, ALuint s
         if(Step != 1.0f && Counter > 0)
         {
             DrySend = params->Mix.Gains.Current[srcchan][c];
-            for(;BufferSize-pos > 3 && Counter-pos > 3;pos+=4)
+            for(;pos < BufferSize && pos < Counter;pos++)
             {
-                OutBuffer[c][OutPos+pos  ] += data[pos  ]*DrySend;
+                OutBuffer[c][OutPos+pos] += data[pos]*DrySend;
                 DrySend *= Step;
-                OutBuffer[c][OutPos+pos+1] += data[pos+1]*DrySend;
-                DrySend *= Step;
-                OutBuffer[c][OutPos+pos+2] += data[pos+2]*DrySend;
-                DrySend *= Step;
-                OutBuffer[c][OutPos+pos+4] += data[pos+3]*DrySend;
-                DrySend *= Step;
-            }
-            if(!(BufferSize-pos > 3))
-            {
-                for(;pos < BufferSize && pos < Counter;pos++)
-                {
-                    OutBuffer[c][OutPos+pos] += data[pos]*DrySend;
-                    DrySend *= Step;
-                }
             }
+            /* Mix until pos is aligned with 4 or the mix is done. */
+            for(;pos < BufferSize && (pos&3) != 0;pos++)
+                OutBuffer[c][OutPos+pos] += data[pos]*DrySend;
             params->Mix.Gains.Current[srcchan][c] = DrySend;
         }
 
@@ -146,25 +135,13 @@ void MixSend_Neon(SendParams *params, const ALfloat *restrict data,
         if(Step != 1.0f && Counter > 0)
         {
             WetGain = params->Gain.Current;
-            for(;BufferSize-pos > 3 && Counter-pos > 3;pos+=4)
+            for(;pos < BufferSize && pos < Counter;pos++)
             {
-                OutBuffer[0][OutPos+pos  ] += data[pos  ]*WetGain;
+                OutBuffer[0][OutPos+pos] += data[pos]*WetGain;
                 WetGain *= Step;
-                OutBuffer[0][OutPos+pos+1] += data[pos+1]*WetGain;
-                WetGain *= Step;
-                OutBuffer[0][OutPos+pos+2] += data[pos+2]*WetGain;
-                WetGain *= Step;
-                OutBuffer[0][OutPos+pos+4] += data[pos+3]*WetGain;
-                WetGain *= Step;
-            }
-            if(!(BufferSize-pos > 3))
-            {
-                for(;pos < BufferSize && pos < Counter;pos++)
-                {
-                    OutBuffer[0][OutPos+pos] += data[pos]*WetGain;
-                    WetGain *= Step;
-                }
             }
+            for(;pos < BufferSize && (pos&3) != 0;pos++)
+                OutBuffer[0][OutPos+pos] += data[pos]*WetGain;
             params->Gain.Current = WetGain;
         }
 
diff --git a/Alc/mixer_sse.c b/Alc/mixer_sse.c
index c17a7e08..a2b8461d 100644
--- a/Alc/mixer_sse.c
+++ b/Alc/mixer_sse.c
@@ -156,6 +156,7 @@ void MixDirect_SSE(DirectParams *params, const ALfloat *restrict data, ALuint sr
         if(Step != 1.0f && Counter > 0)
         {
             DrySend = params->Mix.Gains.Current[srcchan][c];
+            /* Mix with applying gain steps in aligned multiples of 4. */
             if(BufferSize-pos > 3 && Counter-pos > 3)
             {
                 gain = _mm_set_ps(
@@ -175,14 +176,15 @@ void MixDirect_SSE(DirectParams *params, const ALfloat *restrict data, ALuint sr
                 } while(BufferSize-pos > 3 && Counter-pos > 3);
                 DrySend = _mm_cvtss_f32(_mm_shuffle_ps(gain, gain, _MM_SHUFFLE(3, 3, 3, 3)));
             }
-            if(!(BufferSize-pos > 3))
+            /* Mix with applying left over gain steps that aren't aligned multiples of 4. */
+            for(;pos < BufferSize && pos < Counter;pos++)
             {
-                for(;pos < BufferSize && pos < Counter;pos++)
-                {
-                    OutBuffer[c][OutPos+pos] += data[pos]*DrySend;
-                    DrySend *= Step;
-                }
+                OutBuffer[c][OutPos+pos] += data[pos]*DrySend;
+                DrySend *= Step;
             }
+            /* Mix until pos is aligned with 4 or the mix is done. */
+            for(;pos < BufferSize && (pos&3) != 0;pos++)
+                OutBuffer[c][OutPos+pos] += data[pos]*DrySend;
             params->Mix.Gains.Current[srcchan][c] = DrySend;
         }
 
@@ -237,14 +239,13 @@ void MixSend_SSE(SendParams *params, const ALfloat *restrict data,
                 } while(BufferSize-pos > 3 && Counter-pos > 3);
                 WetGain = _mm_cvtss_f32(_mm_shuffle_ps(gain, gain, _MM_SHUFFLE(3, 3, 3, 3)));
             }
-            if(!(BufferSize-pos > 3))
+            for(;pos < BufferSize && pos < Counter;pos++)
             {
-                for(;pos < BufferSize && pos < Counter;pos++)
-                {
-                    OutBuffer[0][OutPos+pos] += data[pos]*WetGain;
-                    WetGain *= Step;
-                }
+                OutBuffer[0][OutPos+pos] += data[pos]*WetGain;
+                WetGain *= Step;
             }
+            for(;pos < BufferSize && (pos&3) != 0;pos++)
+                OutBuffer[0][OutPos+pos] += data[pos]*WetGain;
             params->Gain.Current = WetGain;
         }