Simplify bsinc filter storage in the filter state

Rather than storing individual pointers to filter, scale delta, phase delta,
and scale  phase delta entries, per phase index, the new table layout makes it
trivial to access the per-phase filter and delta entries given the base offset
and coefficient count.

4 files changed, 18 insertions, 24 deletions

diff --git a/Alc/ALu.c b/Alc/ALu.c
index d559649a..2b83ea00 100644
--- a/Alc/ALu.c
+++ b/Alc/ALu.c
@@ -226,9 +226,9 @@ void aluInit(void)
  */
 ALboolean BsincPrepare(const ALuint increment, BsincState *state)
 {
-    ALfloat sf;
-    ALsizei si, pi;
     ALboolean uncut = AL_TRUE;
+    ALfloat sf;
+    ALsizei si;
 
     if(increment > FRACTIONONE)
     {
@@ -262,16 +262,7 @@ ALboolean BsincPrepare(const ALuint increment, BsincState *state)
     state->sf = sf;
     state->m = bsinc.m[si];
     state->l = -((state->m/2) - 1);
-    /* The CPU cost of this table re-mapping could be traded for the memory
-     * cost of a complete table map (1024 elements large).
-     */
-    for(pi = 0;pi < BSINC_PHASE_COUNT;pi++)
-    {
-        state->coeffs[pi].filter  = &bsinc.Tab[bsinc.filterOffset[si] + bsinc.m[si]*(pi*4 + 0)];
-        state->coeffs[pi].scDelta = &bsinc.Tab[bsinc.filterOffset[si] + bsinc.m[si]*(pi*4 + 1)];
-        state->coeffs[pi].phDelta = &bsinc.Tab[bsinc.filterOffset[si] + bsinc.m[si]*(pi*4 + 2)];
-        state->coeffs[pi].spDelta = &bsinc.Tab[bsinc.filterOffset[si] + bsinc.m[si]*(pi*4 + 3)];
-    }
+    state->filter = bsinc.Tab + bsinc.filterOffset[si];
     return uncut;
 }
 
diff --git a/Alc/mixer_c.c b/Alc/mixer_c.c
index 87f2fe90..6d616cbf 100644
--- a/Alc/mixer_c.c
+++ b/Alc/mixer_c.c
@@ -57,6 +57,7 @@ const ALfloat *Resample_bsinc32_C(const InterpState *state, const ALfloat *restr
                                   ALsizei dstlen)
 {
     const ALfloat *fil, *scd, *phd, *spd;
+    const ALfloat *filter = state->bsinc.filter;
     const ALfloat sf = state->bsinc.sf;
     const ALsizei m = state->bsinc.m;
     ALsizei j_f, pi, i;
@@ -71,10 +72,10 @@ const ALfloat *Resample_bsinc32_C(const InterpState *state, const ALfloat *restr
         pf = (frac & ((1<<FRAC_PHASE_BITDIFF)-1)) * (1.0f/(1<<FRAC_PHASE_BITDIFF));
 #undef FRAC_PHASE_BITDIFF
 
-        fil = ASSUME_ALIGNED(state->bsinc.coeffs[pi].filter, 16);
-        scd = ASSUME_ALIGNED(state->bsinc.coeffs[pi].scDelta, 16);
-        phd = ASSUME_ALIGNED(state->bsinc.coeffs[pi].phDelta, 16);
-        spd = ASSUME_ALIGNED(state->bsinc.coeffs[pi].spDelta, 16);
+        fil = ASSUME_ALIGNED(filter + m*pi*4, 16);
+        scd = ASSUME_ALIGNED(fil + m, 16);
+        phd = ASSUME_ALIGNED(scd + m, 16);
+        spd = ASSUME_ALIGNED(phd + m, 16);
 
         // Apply the scale and phase interpolated filter.
         r = 0.0f;
diff --git a/Alc/mixer_neon.c b/Alc/mixer_neon.c
index 51191783..dd7e4226 100644
--- a/Alc/mixer_neon.c
+++ b/Alc/mixer_neon.c
@@ -140,6 +140,7 @@ const ALfloat *Resample_bsinc32_Neon(const InterpState *state,
   const ALfloat *restrict src, ALsizei frac, ALint increment,
   ALfloat *restrict dst, ALsizei dstlen)
 {
+    const ALfloat *filter = state->bsinc.filter;
     const float32x4_t sf4 = vdupq_n_f32(state->bsinc.sf);
     const ALsizei m = state->bsinc.m;
     const ALfloat *fil, *scd, *phd, *spd;
@@ -156,10 +157,10 @@ const ALfloat *Resample_bsinc32_Neon(const InterpState *state,
         pf = (frac & ((1<<FRAC_PHASE_BITDIFF)-1)) * (1.0f/(1<<FRAC_PHASE_BITDIFF));
 #undef FRAC_PHASE_BITDIFF
 
-        fil = ASSUME_ALIGNED(state->bsinc.coeffs[pi].filter, 16);
-        scd = ASSUME_ALIGNED(state->bsinc.coeffs[pi].scDelta, 16);
-        phd = ASSUME_ALIGNED(state->bsinc.coeffs[pi].phDelta, 16);
-        spd = ASSUME_ALIGNED(state->bsinc.coeffs[pi].spDelta, 16);
+        fil = ASSUME_ALIGNED(filter + m*pi*4, 16);
+        scd = ASSUME_ALIGNED(fil + m, 16);
+        phd = ASSUME_ALIGNED(scd + m, 16);
+        spd = ASSUME_ALIGNED(phd + m, 16);
 
         // Apply the scale and phase interpolated filter.
         r4 = vdupq_n_f32(0.0f);
diff --git a/Alc/mixer_sse.c b/Alc/mixer_sse.c
index 68786573..bd7928e4 100644
--- a/Alc/mixer_sse.c
+++ b/Alc/mixer_sse.c
@@ -16,6 +16,7 @@ const ALfloat *Resample_bsinc32_SSE(const InterpState *state, const ALfloat *res
                                     ALsizei frac, ALint increment, ALfloat *restrict dst,
                                     ALsizei dstlen)
 {
+    const ALfloat *filter = state->bsinc.filter;
     const __m128 sf4 = _mm_set1_ps(state->bsinc.sf);
     const ALsizei m = state->bsinc.m;
     const ALfloat *fil, *scd, *phd, *spd;
@@ -32,10 +33,10 @@ const ALfloat *Resample_bsinc32_SSE(const InterpState *state, const ALfloat *res
         pf = (frac & ((1<<FRAC_PHASE_BITDIFF)-1)) * (1.0f/(1<<FRAC_PHASE_BITDIFF));
 #undef FRAC_PHASE_BITDIFF
 
-        fil = ASSUME_ALIGNED(state->bsinc.coeffs[pi].filter, 16);
-        scd = ASSUME_ALIGNED(state->bsinc.coeffs[pi].scDelta, 16);
-        phd = ASSUME_ALIGNED(state->bsinc.coeffs[pi].phDelta, 16);
-        spd = ASSUME_ALIGNED(state->bsinc.coeffs[pi].spDelta, 16);
+        fil = ASSUME_ALIGNED(filter + m*pi*4, 16);
+        scd = ASSUME_ALIGNED(fil + m, 16);
+        phd = ASSUME_ALIGNED(scd + m, 16);
+        spd = ASSUME_ALIGNED(phd + m, 16);
 
         // Apply the scale and phase interpolated filter.
         r4 = _mm_setzero_ps();