diff options
Diffstat (limited to 'alc')
| -rw-r--r-- | alc/alu.h | 4 | ||||
| -rw-r--r-- | alc/mixer/mixer_c.cpp | 12 | ||||
| -rw-r--r-- | alc/mixer/mixer_neon.cpp | 11 | ||||
| -rw-r--r-- | alc/mixer/mixer_sse.cpp | 6 |
4 files changed, 14 insertions, 19 deletions
@@ -67,8 +67,8 @@ extern Resampler ResamplerDefault; */ struct BsincState { ALfloat sf; /* Scale interpolation factor. */ - ALsizei m; /* Coefficient count. */ - ALsizei l; /* Left coefficient offset. */ + ALuint m; /* Coefficient count. */ + ALuint l; /* Left coefficient offset. */ /* Filter coefficients, followed by the scale, phase, and scale-phase * delta coefficients. Starting at phase index 0, each subsequent phase * index follows contiguously. diff --git a/alc/mixer/mixer_c.cpp b/alc/mixer/mixer_c.cpp index 6e96e54e..42d515ae 100644 --- a/alc/mixer/mixer_c.cpp +++ b/alc/mixer/mixer_c.cpp @@ -21,7 +21,7 @@ inline ALfloat do_cubic(const InterpState&, const ALfloat *RESTRICT vals, const { return cubic(vals[0], vals[1], vals[2], vals[3], static_cast<float>(frac)*(1.0f/FRACTIONONE)); } inline ALfloat do_bsinc(const InterpState &istate, const ALfloat *RESTRICT vals, const ALuint frac) { - ASSUME(istate.bsinc.m > 0); + const size_t m{istate.bsinc.m}; // Calculate the phase index and factor. #define FRAC_PHASE_BITDIFF (FRACTIONBITS-BSINC_PHASE_BITS) @@ -30,14 +30,14 @@ inline ALfloat do_bsinc(const InterpState &istate, const ALfloat *RESTRICT vals, (1.0f/(1<<FRAC_PHASE_BITDIFF))}; #undef FRAC_PHASE_BITDIFF - const ALfloat *fil{istate.bsinc.filter + istate.bsinc.m*static_cast<ptrdiff_t>(pi*4)}; - const ALfloat *scd{fil + istate.bsinc.m}; - const ALfloat *phd{scd + istate.bsinc.m}; - const ALfloat *spd{phd + istate.bsinc.m}; + const ALfloat *fil{istate.bsinc.filter + m*pi*4}; + const ALfloat *scd{fil + m}; + const ALfloat *phd{scd + m}; + const ALfloat *spd{phd + m}; // Apply the scale and phase interpolated filter. ALfloat r{0.0f}; - for(ALsizei j_f{0};j_f < istate.bsinc.m;j_f++) + for(size_t j_f{0};j_f < m;j_f++) r += (fil[j_f] + istate.bsinc.sf*scd[j_f] + pf*(phd[j_f] + istate.bsinc.sf*spd[j_f])) * vals[j_f]; return r; } diff --git a/alc/mixer/mixer_neon.cpp b/alc/mixer/mixer_neon.cpp index 8a98f5de..cd2b0ebc 100644 --- a/alc/mixer/mixer_neon.cpp +++ b/alc/mixer/mixer_neon.cpp @@ -74,9 +74,7 @@ const ALfloat *Resample_<BSincTag,NEONTag>(const InterpState *state, const ALflo { 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); + const size_t m{state->bsinc.m}; src -= state->bsinc.l; for(float &out_sample : dst) @@ -92,11 +90,11 @@ const ALfloat *Resample_<BSincTag,NEONTag>(const InterpState *state, const ALflo float32x4_t r4{vdupq_n_f32(0.0f)}; { const float32x4_t pf4{vdupq_n_f32(pf)}; - const float *fil{filter + m*static_cast<ptrdiff_t>(pi*4)}; + const float *fil{filter + m*pi*4}; const float *scd{fil + m}; const float *phd{scd + m}; const float *spd{phd + m}; - ptrdiff_t td{m >> 2}; + size_t td{m >> 2}; size_t j{0u}; do { @@ -110,8 +108,7 @@ const ALfloat *Resample_<BSincTag,NEONTag>(const InterpState *state, const ALflo j += 4; } while(--td); } - r4 = vaddq_f32(r4, vcombine_f32(vrev64_f32(vget_high_f32(r4)), - vrev64_f32(vget_low_f32(r4)))); + r4 = vaddq_f32(r4, vrev64q_f32(r4)); out_sample = vget_lane_f32(vadd_f32(vget_low_f32(r4), vget_high_f32(r4)), 0); frac += increment; diff --git a/alc/mixer/mixer_sse.cpp b/alc/mixer/mixer_sse.cpp index d47a0e66..9bb3bb8a 100644 --- a/alc/mixer/mixer_sse.cpp +++ b/alc/mixer/mixer_sse.cpp @@ -19,9 +19,7 @@ const ALfloat *Resample_<BSincTag,SSETag>(const InterpState *state, const ALfloa { 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); + const size_t m{state->bsinc.m}; src -= state->bsinc.l; for(float &out_sample : dst) @@ -41,7 +39,7 @@ const ALfloat *Resample_<BSincTag,SSETag>(const InterpState *state, const ALfloa const float *scd{fil + m}; const float *phd{scd + m}; const float *spd{phd + m}; - ptrdiff_t td{m >> 2}; + size_t td{m >> 2}; size_t j{0u}; #define MLA4(x, y, z) _mm_add_ps(x, _mm_mul_ps(y, z)) |