Make the resampler increment unsigned

6 files changed, 74 insertions, 78 deletions

diff --git a/alc/mixer/defs.h b/alc/mixer/defs.h
index 4aa3c6b6..a6e07bd4 100644
--- a/alc/mixer/defs.h
+++ b/alc/mixer/defs.h
@@ -28,7 +28,7 @@ enum ResampleType {
 
 template<ResampleType TypeTag, InstSetType InstTag>
 const ALfloat *Resample_(const InterpState *state, const ALfloat *RESTRICT src, ALuint frac,
-    ALint increment, const al::span<float> dst);
+    ALuint increment, const al::span<float> dst);
 
 template<InstSetType InstTag>
 void Mix_(const al::span<const float> InSamples, const al::span<FloatBufferLine> OutBuffer,
@@ -45,13 +45,14 @@ template<InstSetType InstTag>
 void MixDirectHrtf_(FloatBufferLine &LeftOut, FloatBufferLine &RightOut, const al::span<const FloatBufferLine> InSamples, float2 *AccumSamples, DirectHrtfState *State, const size_t BufferSize);
 
 /* Vectorized resampler helpers */
-inline void InitiatePositionArrays(ALsizei frac, ALint increment, ALsizei *RESTRICT frac_arr, ALsizei *RESTRICT pos_arr, ALsizei size)
+inline void InitPosArrays(ALuint frac, ALuint increment, ALuint *frac_arr, ALuint *pos_arr,
+    size_t size)
 {
     pos_arr[0] = 0;
     frac_arr[0] = frac;
-    for(ALsizei i{1};i < size;i++)
+    for(size_t i{1};i < size;i++)
     {
-        ALint frac_tmp = frac_arr[i-1] + increment;
+        const ALuint frac_tmp{frac_arr[i-1] + increment};
         pos_arr[i] = pos_arr[i-1] + (frac_tmp>>FRACTIONBITS);
         frac_arr[i] = frac_tmp&FRACTIONMASK;
     }
diff --git a/alc/mixer/mixer_c.cpp b/alc/mixer/mixer_c.cpp
index a8fb9a19..c5e712f9 100644
--- a/alc/mixer/mixer_c.cpp
+++ b/alc/mixer/mixer_c.cpp
@@ -44,10 +44,8 @@ inline ALfloat do_bsinc(const InterpState &istate, const ALfloat *RESTRICT vals,
 using SamplerT = ALfloat(const InterpState&, const ALfloat*RESTRICT, const ALuint);
 template<SamplerT &Sampler>
 const ALfloat *DoResample(const InterpState *state, const ALfloat *RESTRICT src,
-    ALuint frac, ALint increment, const al::span<float> dst)
+    ALuint frac, ALuint increment, const al::span<float> dst)
 {
-    ASSUME(increment > 0);
-
     const InterpState istate{*state};
     auto proc_sample = [&src,&frac,istate,increment]() -> ALfloat
     {
@@ -68,7 +66,7 @@ const ALfloat *DoResample(const InterpState *state, const ALfloat *RESTRICT src,
 
 template<>
 const ALfloat *Resample_<CopyTag,CTag>(const InterpState*, const ALfloat *RESTRICT src, ALuint,
-    ALint, const al::span<float> dst)
+    ALuint, const al::span<float> dst)
 {
 #if defined(HAVE_SSE) || defined(HAVE_NEON)
     /* Avoid copying the source data if it's aligned like the destination. */
@@ -81,22 +79,22 @@ const ALfloat *Resample_<CopyTag,CTag>(const InterpState*, const ALfloat *RESTRI
 
 template<>
 const ALfloat *Resample_<PointTag,CTag>(const InterpState *state, const ALfloat *RESTRICT src,
-    ALuint frac, ALint increment, const al::span<float> dst)
+    ALuint frac, ALuint increment, const al::span<float> dst)
 { return DoResample<do_point>(state, src, frac, increment, dst); }
 
 template<>
 const ALfloat *Resample_<LerpTag,CTag>(const InterpState *state, const ALfloat *RESTRICT src,
-    ALuint frac, ALint increment, const al::span<float> dst)
+    ALuint frac, ALuint increment, const al::span<float> dst)
 { return DoResample<do_lerp>(state, src, frac, increment, dst); }
 
 template<>
 const ALfloat *Resample_<CubicTag,CTag>(const InterpState *state, const ALfloat *RESTRICT src,
-    ALuint frac, ALint increment, const al::span<float> dst)
+    ALuint frac, ALuint increment, const al::span<float> dst)
 { return DoResample<do_cubic>(state, src-1, frac, increment, dst); }
 
 template<>
 const ALfloat *Resample_<BSincTag,CTag>(const InterpState *state, const ALfloat *RESTRICT src,
-    ALuint frac, ALint increment, const al::span<float> dst)
+    ALuint frac, ALuint increment, const al::span<float> dst)
 { return DoResample<do_bsinc>(state, src-state->bsinc.l, frac, increment, dst); }
 
 
diff --git a/alc/mixer/mixer_neon.cpp b/alc/mixer/mixer_neon.cpp
index d5b1658f..c0a8e628 100644
--- a/alc/mixer/mixer_neon.cpp
+++ b/alc/mixer/mixer_neon.cpp
@@ -16,22 +16,20 @@
 
 template<>
 const ALfloat *Resample_<LerpTag,NEONTag>(const InterpState*, const ALfloat *RESTRICT src,
-    ALuint frac, ALint increment, const al::span<float> dst)
+    ALuint frac, ALuint increment, const al::span<float> dst)
 {
-    const int32x4_t increment4 = vdupq_n_s32(increment*4);
+    const int32x4_t increment4 = vdupq_n_s32(static_cast<int>(increment*4));
     const float32x4_t fracOne4 = vdupq_n_f32(1.0f/FRACTIONONE);
     const int32x4_t fracMask4 = vdupq_n_s32(FRACTIONMASK);
-    alignas(16) ALsizei pos_[4], frac_[4];
+    alignas(16) ALuint pos_[4], frac_[4];
     int32x4_t pos4, frac4;
 
-    ASSUME(increment > 0);
-
-    InitiatePositionArrays(frac, increment, frac_, pos_, 4);
-    frac4 = vld1q_s32(frac_);
-    pos4 = vld1q_s32(pos_);
+    InitPosArrays(frac, increment, frac_, pos_, 4);
+    frac4 = vld1q_s32(reinterpret_cast<int*>(frac_));
+    pos4 = vld1q_s32(reinterpret_cast<int*>(pos_));
 
     auto dst_iter = dst.begin();
-    const auto aligned_end = (dst.size()&~3) + dst_iter;
+    const auto aligned_end = (dst.size()&~3u) + dst_iter;
     while(dst_iter != aligned_end)
     {
         const int pos0{vgetq_lane_s32(pos4, 0)};
@@ -54,33 +52,31 @@ const ALfloat *Resample_<LerpTag,NEONTag>(const InterpState*, const ALfloat *RES
         frac4 = vandq_s32(frac4, fracMask4);
     }
 
-    /* NOTE: These four elements represent the position *after* the last four
-     * samples, so the lowest element is the next position to resample.
-     */
-    src += static_cast<ALuint>(vgetq_lane_s32(pos4, 0));
-    frac = vgetq_lane_s32(frac4, 0);
-
-    while(dst_iter != dst.end())
+    if(dst_iter != dst.end())
     {
-        *(dst_iter++) = lerp(src[0], src[1], frac * (1.0f/FRACTIONONE));
+        src += static_cast<ALuint>(vgetq_lane_s32(pos4, 0));
+        frac = vgetq_lane_s32(frac4, 0);
 
-        frac += increment;
-        src  += frac>>FRACTIONBITS;
-        frac &= FRACTIONMASK;
+        do {
+            *(dst_iter++) = lerp(src[0], src[1], frac * (1.0f/FRACTIONONE));
+
+            frac += increment;
+            src  += frac>>FRACTIONBITS;
+            frac &= FRACTIONMASK;
+        } while(dst_iter != dst.end());
     }
     return dst.begin();
 }
 
 template<>
 const ALfloat *Resample_<BSincTag,NEONTag>(const InterpState *state, const ALfloat *RESTRICT src,
-    ALuint frac, ALint increment, const al::span<float> dst)
+    ALuint frac, ALuint increment, const al::span<float> dst)
 {
     const ALfloat *const filter{state->bsinc.filter};
     const float32x4_t sf4{vdupq_n_f32(state->bsinc.sf)};
     const ptrdiff_t m{state->bsinc.m};
 
     ASSUME(m > 0);
-    ASSUME(increment > 0);
 
     src -= state->bsinc.l;
     for(float &out_sample : dst)
@@ -183,7 +179,7 @@ void Mix_<NEONTag>(const al::span<const float> InSamples, const al::span<FloatBu
     const ALfloat delta{(Counter > 0) ? 1.0f / static_cast<ALfloat>(Counter) : 0.0f};
     const bool reached_target{InSamples.size() >= Counter};
     const auto min_end = reached_target ? InSamples.begin() + Counter : InSamples.end();
-    const auto aligned_end = minz(InSamples.size(), (min_end-InSamples.begin()+3) & ~3) +
+    const auto aligned_end = minz(InSamples.size(), (min_end-InSamples.begin()+3) & ~3u) +
         InSamples.begin();
     for(FloatBufferLine &output : OutBuffer)
     {
diff --git a/alc/mixer/mixer_sse.cpp b/alc/mixer/mixer_sse.cpp
index 1965f3e6..82e2bfbe 100644
--- a/alc/mixer/mixer_sse.cpp
+++ b/alc/mixer/mixer_sse.cpp
@@ -15,14 +15,13 @@
 
 template<>
 const ALfloat *Resample_<BSincTag,SSETag>(const InterpState *state, const ALfloat *RESTRICT src,
-    ALuint frac, ALint increment, const al::span<float> dst)
+    ALuint frac, ALuint increment, const al::span<float> dst)
 {
     const ALfloat *const filter{state->bsinc.filter};
     const __m128 sf4{_mm_set1_ps(state->bsinc.sf)};
     const ptrdiff_t m{state->bsinc.m};
 
     ASSUME(m > 0);
-    ASSUME(increment > 0);
 
     src -= state->bsinc.l;
     for(float &out_sample : dst)
@@ -146,7 +145,7 @@ void Mix_<SSETag>(const al::span<const float> InSamples, const al::span<FloatBuf
     const ALfloat delta{(Counter > 0) ? 1.0f / static_cast<ALfloat>(Counter) : 0.0f};
     const bool reached_target{InSamples.size() >= Counter};
     const auto min_end = reached_target ? InSamples.begin() + Counter : InSamples.end();
-    const auto aligned_end = minz(InSamples.size(), (min_end-InSamples.begin()+3) & ~3) +
+    const auto aligned_end = minz(InSamples.size(), (min_end-InSamples.begin()+3) & ~3u) +
         InSamples.begin();
     for(FloatBufferLine &output : OutBuffer)
     {
diff --git a/alc/mixer/mixer_sse2.cpp b/alc/mixer/mixer_sse2.cpp
index b126cd25..38d77fd9 100644
--- a/alc/mixer/mixer_sse2.cpp
+++ b/alc/mixer/mixer_sse2.cpp
@@ -29,21 +29,21 @@
 
 template<>
 const ALfloat *Resample_<LerpTag,SSE2Tag>(const InterpState*, const ALfloat *RESTRICT src,
-    ALuint frac, ALint increment, const al::span<float> dst)
+    ALuint frac, ALuint increment, const al::span<float> dst)
 {
-    const __m128i increment4{_mm_set1_epi32(increment*4)};
+    const __m128i increment4{_mm_set1_epi32(static_cast<int>(increment*4))};
     const __m128 fracOne4{_mm_set1_ps(1.0f/FRACTIONONE)};
     const __m128i fracMask4{_mm_set1_epi32(FRACTIONMASK)};
 
-    ASSUME(increment > 0);
-
-    alignas(16) ALsizei pos_[4], frac_[4];
-    InitiatePositionArrays(frac, increment, frac_, pos_, 4);
-    __m128i frac4{_mm_setr_epi32(frac_[0], frac_[1], frac_[2], frac_[3])};
-    __m128i pos4{_mm_setr_epi32(pos_[0], pos_[1], pos_[2], pos_[3])};
+    alignas(16) ALuint pos_[4], frac_[4];
+    InitPosArrays(frac, increment, frac_, pos_, 4);
+    __m128i frac4{_mm_setr_epi32(static_cast<int>(frac_[0]), static_cast<int>(frac_[1]),
+        static_cast<int>(frac_[2]), static_cast<int>(frac_[3]))};
+    __m128i pos4{_mm_setr_epi32(static_cast<int>(pos_[0]), static_cast<int>(pos_[1]),
+        static_cast<int>(pos_[2]), static_cast<int>(pos_[3]))};
 
     auto dst_iter = dst.begin();
-    const auto aligned_end = (dst.size()&~3) + dst_iter;
+    const auto aligned_end = (dst.size()&~3u) + dst_iter;
     while(dst_iter != aligned_end)
     {
         const int pos0{_mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(0, 0, 0, 0)))};
@@ -66,19 +66,18 @@ const ALfloat *Resample_<LerpTag,SSE2Tag>(const InterpState*, const ALfloat *RES
         frac4 = _mm_and_si128(frac4, fracMask4);
     }
 
-    /* NOTE: These four elements represent the position *after* the last four
-     * samples, so the lowest element is the next position to resample.
-     */
-    src += static_cast<ALuint>(_mm_cvtsi128_si32(pos4));
-    frac = _mm_cvtsi128_si32(frac4);
-
-    while(dst_iter != dst.end())
+    if(dst_iter != dst.end())
     {
-        *(dst_iter++) = lerp(src[0], src[1], frac * (1.0f/FRACTIONONE));
+        src += static_cast<ALuint>(_mm_cvtsi128_si32(pos4));
+        frac = static_cast<ALuint>(_mm_cvtsi128_si32(frac4));
+
+        do {
+            *(dst_iter++) = lerp(src[0], src[1], frac * (1.0f/FRACTIONONE));
 
-        frac += increment;
-        src  += frac>>FRACTIONBITS;
-        frac &= FRACTIONMASK;
+            frac += increment;
+            src  += frac>>FRACTIONBITS;
+            frac &= FRACTIONMASK;
+        } while(dst_iter != dst.end());
     }
     return dst.begin();
 }
diff --git a/alc/mixer/mixer_sse41.cpp b/alc/mixer/mixer_sse41.cpp
index 06c51e6a..0a87f76f 100644
--- a/alc/mixer/mixer_sse41.cpp
+++ b/alc/mixer/mixer_sse41.cpp
@@ -30,21 +30,21 @@
 
 template<>
 const ALfloat *Resample_<LerpTag,SSE4Tag>(const InterpState*, const ALfloat *RESTRICT src,
-    ALuint frac, ALint increment, const al::span<float> dst)
+    ALuint frac, ALuint increment, const al::span<float> dst)
 {
-    const __m128i increment4{_mm_set1_epi32(increment*4)};
+    const __m128i increment4{_mm_set1_epi32(static_cast<int>(increment*4))};
     const __m128 fracOne4{_mm_set1_ps(1.0f/FRACTIONONE)};
     const __m128i fracMask4{_mm_set1_epi32(FRACTIONMASK)};
 
-    ASSUME(increment > 0);
-
-    alignas(16) ALsizei pos_[4], frac_[4];
-    InitiatePositionArrays(frac, increment, frac_, pos_, 4);
-    __m128i frac4{_mm_setr_epi32(frac_[0], frac_[1], frac_[2], frac_[3])};
-    __m128i pos4{_mm_setr_epi32(pos_[0], pos_[1], pos_[2], pos_[3])};
+    alignas(16) ALuint pos_[4], frac_[4];
+    InitPosArrays(frac, increment, frac_, pos_, 4);
+    __m128i frac4{_mm_setr_epi32(static_cast<int>(frac_[0]), static_cast<int>(frac_[1]),
+        static_cast<int>(frac_[2]), static_cast<int>(frac_[3]))};
+    __m128i pos4{_mm_setr_epi32(static_cast<int>(pos_[0]), static_cast<int>(pos_[1]),
+        static_cast<int>(pos_[2]), static_cast<int>(pos_[3]))};
 
     auto dst_iter = dst.begin();
-    const auto aligned_end = (dst.size()&~3) + dst_iter;
+    const auto aligned_end = (dst.size()&~3u) + dst_iter;
     while(dst_iter != aligned_end)
     {
         const int pos0{_mm_extract_epi32(pos4, 0)};
@@ -67,19 +67,22 @@ const ALfloat *Resample_<LerpTag,SSE4Tag>(const InterpState*, const ALfloat *RES
         frac4 = _mm_and_si128(frac4, fracMask4);
     }
 
-    /* NOTE: These four elements represent the position *after* the last four
-     * samples, so the lowest element is the next position to resample.
-     */
-    src += static_cast<ALuint>(_mm_cvtsi128_si32(pos4));
-    frac = _mm_cvtsi128_si32(frac4);
-
-    while(dst_iter != dst.end())
+    if(dst_iter != dst.end())
     {
-        *(dst_iter++) = lerp(src[0], src[1], frac * (1.0f/FRACTIONONE));
+        /* NOTE: These four elements represent the position *after* the last
+         * four samples, so the lowest element is the next position to
+         * resample.
+         */
+        src += static_cast<ALuint>(_mm_cvtsi128_si32(pos4));
+        frac = static_cast<ALuint>(_mm_cvtsi128_si32(frac4));
+
+        do {
+            *(dst_iter++) = lerp(src[0], src[1], frac * (1.0f/FRACTIONONE));
 
-        frac += increment;
-        src  += frac>>FRACTIONBITS;
-        frac &= FRACTIONMASK;
+            frac += increment;
+            src  += frac>>FRACTIONBITS;
+            frac &= FRACTIONMASK;
+        } while(dst_iter != dst.end());
     }
     return dst.begin();
 }