3 #ifdef HAVE_XMMINTRIN_H
20 const ALfloat (*RESTRICT CoeffStep)[2],
24 __m128
coeffs, deltas, imp0, imp1;
25 __m128 vals = _mm_setzero_ps();
31 const ALuint o1 = (Offset+IrSize-1)&HRIR_MASK;
33 coeffs = _mm_load_ps(&Coeffs[0][0]);
34 deltas = _mm_load_ps(&CoeffStep[0][0]);
35 vals = _mm_loadl_pi(vals, (__m64*)&Values[o0][0]);
36 imp0 = _mm_mul_ps(lrlr, coeffs);
37 coeffs = _mm_add_ps(coeffs, deltas);
38 vals = _mm_add_ps(imp0, vals);
39 _mm_store_ps(&Coeffs[0][0], coeffs);
40 _mm_storel_pi((__m64*)&Values[o0][0], vals);
41 for(i = 1;i < IrSize-1;i += 2)
43 const ALuint o2 = (Offset+
i)&HRIR_MASK;
45 coeffs = _mm_load_ps(&Coeffs[i+1][0]);
46 deltas = _mm_load_ps(&CoeffStep[i+1][0]);
47 vals = _mm_load_ps(&Values[o2][0]);
48 imp1 = _mm_mul_ps(lrlr, coeffs);
49 coeffs = _mm_add_ps(coeffs, deltas);
50 imp0 = _mm_shuffle_ps(imp0, imp1, _MM_SHUFFLE(1, 0, 3, 2));
51 vals = _mm_add_ps(imp0, vals);
52 _mm_store_ps(&Coeffs[i+1][0], coeffs);
53 _mm_store_ps(&Values[o2][0], vals);
56 vals = _mm_loadl_pi(vals, (__m64*)&Values[o1][0]);
57 imp0 = _mm_movehl_ps(imp0, imp0);
58 vals = _mm_add_ps(imp0, vals);
59 _mm_storel_pi((__m64*)&Values[o1][0], vals);
63 for(i = 0;i < IrSize;i += 2)
67 coeffs = _mm_load_ps(&Coeffs[i][0]);
68 deltas = _mm_load_ps(&CoeffStep[i][0]);
69 vals = _mm_load_ps(&Values[o][0]);
70 imp0 = _mm_mul_ps(lrlr, coeffs);
71 coeffs = _mm_add_ps(coeffs, deltas);
72 vals = _mm_add_ps(imp0, vals);
73 _mm_store_ps(&Coeffs[i][0], coeffs);
74 _mm_store_ps(&Values[o][0], vals);
85 __m128 vals = _mm_setzero_ps();
92 const ALuint o1 = (Offset+IrSize-1)&HRIR_MASK;
95 coeffs = _mm_load_ps(&Coeffs[0][0]);
96 vals = _mm_loadl_pi(vals, (__m64*)&Values[o0][0]);
97 imp0 = _mm_mul_ps(lrlr, coeffs);
98 vals = _mm_add_ps(imp0, vals);
99 _mm_storel_pi((__m64*)&Values[o0][0], vals);
100 for(i = 1;i < IrSize-1;i += 2)
102 const ALuint o2 = (Offset+
i)&HRIR_MASK;
104 coeffs = _mm_load_ps(&Coeffs[i+1][0]);
105 vals = _mm_load_ps(&Values[o2][0]);
106 imp1 = _mm_mul_ps(lrlr, coeffs);
107 imp0 = _mm_shuffle_ps(imp0, imp1, _MM_SHUFFLE(1, 0, 3, 2));
108 vals = _mm_add_ps(imp0, vals);
109 _mm_store_ps(&Values[o2][0], vals);
112 vals = _mm_loadl_pi(vals, (__m64*)&Values[o1][0]);
113 imp0 = _mm_movehl_ps(imp0, imp0);
114 vals = _mm_add_ps(imp0, vals);
115 _mm_storel_pi((__m64*)&Values[o1][0], vals);
119 for(i = 0;i < IrSize;i += 2)
123 coeffs = _mm_load_ps(&Coeffs[i][0]);
124 vals = _mm_load_ps(&Values[o][0]);
125 vals = _mm_add_ps(vals, _mm_mul_ps(lrlr, coeffs));
126 _mm_store_ps(&Values[o][0], vals);
150 DrySend = params->
Gains[srcchan][
c];
151 if(DrySend < 0.00001
f)
155 ClickRemoval[
c] -= data[0]*DrySend;
157 gain = _mm_set1_ps(DrySend);
158 for(pos = 0;pos < BufferSize-3;pos += 4)
160 const __m128 val4 = _mm_load_ps(&data[pos]);
161 __m128 dry4 = _mm_load_ps(&DryBuffer[c][OutPos+pos]);
162 dry4 = _mm_add_ps(dry4, _mm_mul_ps(val4, gain));
163 _mm_store_ps(&DryBuffer[c][OutPos+pos], dry4);
165 for(;pos < BufferSize;pos++)
166 DryBuffer[c][OutPos+pos] += data[pos]*DrySend;
168 if(OutPos+pos == SamplesToDo)
169 PendingClicks[
c] += data[pos]*DrySend;
185 if(WetGain < 0.00001
f)
189 WetClickRemoval[0] -= data[0] * WetGain;
191 gain = _mm_set1_ps(WetGain);
192 for(pos = 0;pos < BufferSize-3;pos+=4)
194 const __m128 val4 = _mm_load_ps(&data[pos]);
195 __m128 wet4 = _mm_load_ps(&WetBuffer[0][OutPos+pos]);
196 wet4 = _mm_add_ps(wet4, _mm_mul_ps(val4, gain));
197 _mm_store_ps(&WetBuffer[0][OutPos+pos], wet4);
199 for(;pos < BufferSize;pos++)
200 WetBuffer[0][OutPos+pos] += data[pos] * WetGain;
202 if(OutPos+pos == SamplesToDo)
203 WetPendingClicks[0] += data[pos] * WetGain;
static __inline void ApplyCoeffsStep(ALuint Offset, ALfloat(*RESTRICT Values)[2], const ALuint IrSize, ALfloat(*RESTRICT Coeffs)[2], const ALfloat(*RESTRICT CoeffStep)[2], ALfloat left, ALfloat right)
ALfloat(* OutBuffer)[BUFFERSIZE]
void MixDirect_SSE(const struct DirectParams *, const ALfloat *RESTRICT, ALuint, ALuint, ALuint, ALuint)
void MixSend_SSE(const struct SendParams *, const ALfloat *RESTRICT, ALuint, ALuint, ALuint)
static __inline void ApplyCoeffs(ALuint Offset, ALfloat(*RESTRICT Values)[2], const ALuint IrSize, ALfloat(*RESTRICT Coeffs)[2], ALfloat left, ALfloat right)
struct ALeffectslot * Slot
GLint GLenum GLsizei GLsizei GLsizei GLint GLsizei const GLvoid * data
ALfloat Gains[MaxChannels][MaxChannels]
GLenum GLenum const GLfloat * coeffs