diff options
Diffstat (limited to 'Alc/mixer')
-rw-r--r-- | Alc/mixer/mixer_neon.c | 12 | ||||
-rw-r--r-- | Alc/mixer/mixer_sse2.c | 15 | ||||
-rw-r--r-- | Alc/mixer/mixer_sse41.c | 15 |
3 files changed, 21 insertions, 21 deletions
diff --git a/Alc/mixer/mixer_neon.c b/Alc/mixer/mixer_neon.c index 9aa279a2..4feb431d 100644 --- a/Alc/mixer/mixer_neon.c +++ b/Alc/mixer/mixer_neon.c @@ -32,8 +32,12 @@ const ALfloat *Resample_lerp_Neon(const InterpState* UNUSED(state), for(i = 0;numsamples-i > 3;i += 4) { - const float32x4_t val1 = (float32x4_t){src[pos_[0]], src[pos_[1]], src[pos_[2]], src[pos_[3]]}; - const float32x4_t val2 = (float32x4_t){src[pos_[0]+1], src[pos_[1]+1], src[pos_[2]+1], src[pos_[3]+1]}; + const int pos0 = vgetq_lane_s32(pos4, 0); + const int pos1 = vgetq_lane_s32(pos4, 1); + const int pos2 = vgetq_lane_s32(pos4, 2); + const int pos3 = vgetq_lane_s32(pos4, 3); + const float32x4_t val1 = (float32x4_t){src[pos0], src[pos1], src[pos2], src[pos3]}; + const float32x4_t val2 = (float32x4_t){src[pos0+1], src[pos1+1], src[pos2+1], src[pos3+1]}; /* val1 + (val2-val1)*mu */ const float32x4_t r0 = vsubq_f32(val2, val1); @@ -45,8 +49,6 @@ const ALfloat *Resample_lerp_Neon(const InterpState* UNUSED(state), frac4 = vaddq_s32(frac4, increment4); pos4 = vaddq_s32(pos4, vshrq_n_s32(frac4, FRACTIONBITS)); frac4 = vandq_s32(frac4, fracMask4); - - vst1q_s32(pos_, pos4); } if(i < numsamples) @@ -55,7 +57,7 @@ const ALfloat *Resample_lerp_Neon(const InterpState* UNUSED(state), * four samples, so the lowest element is the next position to * resample. */ - ALint pos = pos_[0]; + int pos = vgetq_lane_s32(pos4, 0); frac = vgetq_lane_s32(frac4, 0); do { dst[i] = lerp(src[pos], src[pos+1], frac * (1.0f/FRACTIONONE)); diff --git a/Alc/mixer/mixer_sse2.c b/Alc/mixer/mixer_sse2.c index e0198022..83aaf7f2 100644 --- a/Alc/mixer/mixer_sse2.c +++ b/Alc/mixer/mixer_sse2.c @@ -49,8 +49,12 @@ const ALfloat *Resample_lerp_SSE2(const InterpState* UNUSED(state), for(i = 0;numsamples-i > 3;i += 4) { - const __m128 val1 = _mm_setr_ps(src[pos_.i[0] ], src[pos_.i[1] ], src[pos_.i[2] ], src[pos_.i[3] ]); - const __m128 val2 = _mm_setr_ps(src[pos_.i[0]+1], src[pos_.i[1]+1], src[pos_.i[2]+1], src[pos_.i[3]+1]); + const int pos0 = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(0, 0, 0, 0))); + const int pos1 = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(1, 1, 1, 1))); + const int pos2 = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(2, 2, 2, 2))); + const int pos3 = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(3, 3, 3, 3))); + const __m128 val1 = _mm_setr_ps(src[pos0 ], src[pos1 ], src[pos2 ], src[pos3 ]); + const __m128 val2 = _mm_setr_ps(src[pos0+1], src[pos1+1], src[pos2+1], src[pos3+1]); /* val1 + (val2-val1)*mu */ const __m128 r0 = _mm_sub_ps(val2, val1); @@ -62,17 +66,12 @@ const ALfloat *Resample_lerp_SSE2(const InterpState* UNUSED(state), frac4 = _mm_add_epi32(frac4, increment4); pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, FRACTIONBITS)); frac4 = _mm_and_si128(frac4, fracMask4); - - pos_.i[0] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(0, 0, 0, 0))); - pos_.i[1] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(1, 1, 1, 1))); - pos_.i[2] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(2, 2, 2, 2))); - pos_.i[3] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(3, 3, 3, 3))); } /* NOTE: These four elements represent the position *after* the last four * samples, so the lowest element is the next position to resample. */ - pos = pos_.i[0]; + pos = _mm_cvtsi128_si32(pos4); frac = _mm_cvtsi128_si32(frac4); for(;i < numsamples;i++) diff --git a/Alc/mixer/mixer_sse41.c b/Alc/mixer/mixer_sse41.c index 98a3ef74..396b7c7c 100644 --- a/Alc/mixer/mixer_sse41.c +++ b/Alc/mixer/mixer_sse41.c @@ -50,8 +50,12 @@ const ALfloat *Resample_lerp_SSE41(const InterpState* UNUSED(state), for(i = 0;numsamples-i > 3;i += 4) { - const __m128 val1 = _mm_setr_ps(src[pos_.i[0]], src[pos_.i[1]], src[pos_.i[2]], src[pos_.i[3]]); - const __m128 val2 = _mm_setr_ps(src[pos_.i[0]+1], src[pos_.i[1]+1], src[pos_.i[2]+1], src[pos_.i[3]+1]); + const int pos0 = _mm_extract_epi32(pos4, 0); + const int pos1 = _mm_extract_epi32(pos4, 1); + const int pos2 = _mm_extract_epi32(pos4, 2); + const int pos3 = _mm_extract_epi32(pos4, 3); + const __m128 val1 = _mm_setr_ps(src[pos0 ], src[pos1 ], src[pos2 ], src[pos3 ]); + const __m128 val2 = _mm_setr_ps(src[pos0+1], src[pos1+1], src[pos2+1], src[pos3+1]); /* val1 + (val2-val1)*mu */ const __m128 r0 = _mm_sub_ps(val2, val1); @@ -63,17 +67,12 @@ const ALfloat *Resample_lerp_SSE41(const InterpState* UNUSED(state), frac4 = _mm_add_epi32(frac4, increment4); pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, FRACTIONBITS)); frac4 = _mm_and_si128(frac4, fracMask4); - - pos_.i[0] = _mm_extract_epi32(pos4, 0); - pos_.i[1] = _mm_extract_epi32(pos4, 1); - pos_.i[2] = _mm_extract_epi32(pos4, 2); - pos_.i[3] = _mm_extract_epi32(pos4, 3); } /* NOTE: These four elements represent the position *after* the last four * samples, so the lowest element is the next position to resample. */ - pos = pos_.i[0]; + pos = _mm_cvtsi128_si32(pos4); frac = _mm_cvtsi128_si32(frac4); for(;i < numsamples;i++) |