Add some optimization hints

3 files changed, 24 insertions, 18 deletions

diff --git a/core/mixer/mixer_c.cpp b/core/mixer/mixer_c.cpp
index 6b9c4344..8c2c60c5 100644
--- a/core/mixer/mixer_c.cpp
+++ b/core/mixer/mixer_c.cpp
@@ -32,15 +32,16 @@ inline float do_cubic(const InterpState&, const float *RESTRICT vals, const uint
 inline float do_bsinc(const InterpState &istate, const float *RESTRICT vals, const uint frac)
 {
     const size_t m{istate.bsinc.m};
+    ASSUME(m > 0);
 
     // Calculate the phase index and factor.
     const uint pi{frac >> FracPhaseBitDiff};
     const float pf{static_cast<float>(frac & (FracPhaseDiffOne-1)) * (1.0f/FracPhaseDiffOne)};
 
-    const float *fil{istate.bsinc.filter + m*pi*4};
-    const float *phd{fil + m};
-    const float *scd{phd + m};
-    const float *spd{scd + m};
+    const float *RESTRICT fil{istate.bsinc.filter + m*pi*4};
+    const float *RESTRICT phd{fil + m};
+    const float *RESTRICT scd{phd + m};
+    const float *RESTRICT spd{scd + m};
 
     // Apply the scale and phase interpolated filter.
     float r{0.0f};
@@ -51,13 +52,14 @@ inline float do_bsinc(const InterpState &istate, const float *RESTRICT vals, con
 inline float do_fastbsinc(const InterpState &istate, const float *RESTRICT vals, const uint frac)
 {
     const size_t m{istate.bsinc.m};
+    ASSUME(m > 0);
 
     // Calculate the phase index and factor.
     const uint pi{frac >> FracPhaseBitDiff};
     const float pf{static_cast<float>(frac & (FracPhaseDiffOne-1)) * (1.0f/FracPhaseDiffOne)};
 
-    const float *fil{istate.bsinc.filter + m*pi*4};
-    const float *phd{fil + m};
+    const float *RESTRICT fil{istate.bsinc.filter + m*pi*4};
+    const float *RESTRICT phd{fil + m};
 
     // Apply the phase interpolated filter.
     float r{0.0f};
diff --git a/core/mixer/mixer_neon.cpp b/core/mixer/mixer_neon.cpp
index 220d412e..8cd7e610 100644
--- a/core/mixer/mixer_neon.cpp
+++ b/core/mixer/mixer_neon.cpp
@@ -118,6 +118,7 @@ float *Resample_<BSincTag,NEONTag>(const InterpState *state, float *RESTRICT src
     const float *const filter{state->bsinc.filter};
     const float32x4_t sf4{vdupq_n_f32(state->bsinc.sf)};
     const size_t m{state->bsinc.m};
+    ASSUME(m > 0);
 
     src -= state->bsinc.l;
     for(float &out_sample : dst)
@@ -130,10 +131,10 @@ float *Resample_<BSincTag,NEONTag>(const InterpState *state, float *RESTRICT src
         float32x4_t r4{vdupq_n_f32(0.0f)};
         {
             const float32x4_t pf4{vdupq_n_f32(pf)};
-            const float *fil{filter + m*pi*4};
-            const float *phd{fil + m};
-            const float *scd{phd + m};
-            const float *spd{scd + m};
+            const float *RESTRICT fil{filter + m*pi*4};
+            const float *RESTRICT phd{fil + m};
+            const float *RESTRICT scd{phd + m};
+            const float *RESTRICT spd{scd + m};
             size_t td{m >> 2};
             size_t j{0u};
 
@@ -163,6 +164,7 @@ float *Resample_<FastBSincTag,NEONTag>(const InterpState *state, float *RESTRICT
 {
     const float *const filter{state->bsinc.filter};
     const size_t m{state->bsinc.m};
+    ASSUME(m > 0);
 
     src -= state->bsinc.l;
     for(float &out_sample : dst)
@@ -175,8 +177,8 @@ float *Resample_<FastBSincTag,NEONTag>(const InterpState *state, float *RESTRICT
         float32x4_t r4{vdupq_n_f32(0.0f)};
         {
             const float32x4_t pf4{vdupq_n_f32(pf)};
-            const float *fil{filter + m*pi*4};
-            const float *phd{fil + m};
+            const float *RESTRICT fil{filter + m*pi*4};
+            const float *RESTRICT phd{fil + m};
             size_t td{m >> 2};
             size_t j{0u};
 
diff --git a/core/mixer/mixer_sse.cpp b/core/mixer/mixer_sse.cpp
index 8c748a57..ba53d0a5 100644
--- a/core/mixer/mixer_sse.cpp
+++ b/core/mixer/mixer_sse.cpp
@@ -83,6 +83,7 @@ float *Resample_<BSincTag,SSETag>(const InterpState *state, float *RESTRICT src,
     const float *const filter{state->bsinc.filter};
     const __m128 sf4{_mm_set1_ps(state->bsinc.sf)};
     const size_t m{state->bsinc.m};
+    ASSUME(m > 0);
 
     src -= state->bsinc.l;
     for(float &out_sample : dst)
@@ -95,10 +96,10 @@ float *Resample_<BSincTag,SSETag>(const InterpState *state, float *RESTRICT src,
         __m128 r4{_mm_setzero_ps()};
         {
             const __m128 pf4{_mm_set1_ps(pf)};
-            const float *fil{filter + m*pi*4};
-            const float *phd{fil + m};
-            const float *scd{phd + m};
-            const float *spd{scd + m};
+            const float *RESTRICT fil{filter + m*pi*4};
+            const float *RESTRICT phd{fil + m};
+            const float *RESTRICT scd{phd + m};
+            const float *RESTRICT spd{scd + m};
             size_t td{m >> 2};
             size_t j{0u};
 
@@ -129,6 +130,7 @@ float *Resample_<FastBSincTag,SSETag>(const InterpState *state, float *RESTRICT
 {
     const float *const filter{state->bsinc.filter};
     const size_t m{state->bsinc.m};
+    ASSUME(m > 0);
 
     src -= state->bsinc.l;
     for(float &out_sample : dst)
@@ -141,8 +143,8 @@ float *Resample_<FastBSincTag,SSETag>(const InterpState *state, float *RESTRICT
         __m128 r4{_mm_setzero_ps()};
         {
             const __m128 pf4{_mm_set1_ps(pf)};
-            const float *fil{filter + m*pi*4};
-            const float *phd{fil + m};
+            const float *RESTRICT fil{filter + m*pi*4};
+            const float *RESTRICT phd{fil + m};
             size_t td{m >> 2};
             size_t j{0u};