Warning, /RecoTracker/MkFitCore/src/JacCCS2Loc.ah is written in an unsupported language. File is not indexed.
0001 #ifdef MPLEX_INTRINSICS
0002
0003 for (int n = 0; n < N; n += MPLEX_INTRINSICS_WIDTH_BYTES / sizeof(T)) {
0004 #ifdef AVX512_INTRINSICS
0005 IntrVec_t all_zeros = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
0006 #else
0007 IntrVec_t all_zeros = {0, 0, 0, 0, 0, 0, 0, 0};
0008 #endif
0009
0010 IntrVec_t b_3 = LD(b, 3);
0011 IntrVec_t c_3 = b_3;
0012 IntrVec_t b_5 = LD(b, 5);
0013 IntrVec_t c_5 = b_5;
0014
0015 ST(c, 0, all_zeros);
0016 ST(c, 1, all_zeros);
0017 ST(c, 2, all_zeros);
0018 ST(c, 4, all_zeros);
0019
0020 IntrVec_t a_6 = LD(a, 6);
0021 IntrVec_t b_11 = LD(b, 11);
0022 IntrVec_t c_11 = MUL(a_6, b_11);
0023
0024 IntrVec_t a_7 = LD(a, 7);
0025 IntrVec_t c_10 = a_7;
0026 ST(c, 3, c_3);
0027 ST(c, 5, c_5);
0028
0029 IntrVec_t a_8 = LD(a, 8);
0030 IntrVec_t b_18 = LD(b, 18);
0031 IntrVec_t c_6 = MUL(a_8, b_18);
0032 IntrVec_t b_19 = LD(b, 19);
0033 IntrVec_t c_7 = MUL(a_8, b_19);
0034
0035 IntrVec_t a_9 = LD(a, 9);
0036 IntrVec_t b_24 = LD(b, 24);
0037 c_6 = FMA(a_9, b_24, c_6);
0038 IntrVec_t b_25 = LD(b, 25);
0039 c_7 = FMA(a_9, b_25, c_7);
0040 IntrVec_t b_26 = LD(b, 26);
0041 IntrVec_t c_8 = MUL(a_9, b_26);
0042 ST(c, 6, c_6);
0043 ST(c, 9, all_zeros);
0044
0045 IntrVec_t a_11 = LD(a, 11);
0046 IntrVec_t c_17 = MUL(a_11, b_11);
0047 ST(c, 7, c_7);
0048 ST(c, 8, c_8);
0049 ST(c, 10, c_10);
0050 ST(c, 11, c_11);
0051
0052 IntrVec_t a_12 = LD(a, 12);
0053 IntrVec_t c_16 = a_12;
0054
0055 IntrVec_t a_13 = LD(a, 13);
0056 IntrVec_t c_12 = MUL(a_13, b_18);
0057 IntrVec_t c_13 = MUL(a_13, b_19);
0058
0059 IntrVec_t a_14 = LD(a, 14);
0060 c_12 = FMA(a_14, b_24, c_12);
0061 c_13 = FMA(a_14, b_25, c_13);
0062 IntrVec_t c_14 = MUL(a_14, b_26);
0063 ST(c, 15, all_zeros);
0064
0065 IntrVec_t a_18 = LD(a, 18);
0066 IntrVec_t c_18 = MUL(a_18, b_18);
0067 ST(c, 12, c_12);
0068 ST(c, 13, c_13);
0069 ST(c, 14, c_14);
0070 ST(c, 16, c_16);
0071 ST(c, 17, c_17);
0072 IntrVec_t c_19 = MUL(a_18, b_19);
0073
0074 IntrVec_t a_19 = LD(a, 19);
0075 c_18 = FMA(a_19, b_24, c_18);
0076 c_19 = FMA(a_19, b_25, c_19);
0077 IntrVec_t c_20 = MUL(a_19, b_26);
0078 ST(c, 21, all_zeros);
0079 ST(c, 22, all_zeros);
0080 ST(c, 23, all_zeros);
0081
0082 IntrVec_t a_23 = LD(a, 23);
0083 IntrVec_t c_24 = MUL(a_23, b_18);
0084 ST(c, 18, c_18);
0085 ST(c, 19, c_19);
0086 ST(c, 20, c_20);
0087 IntrVec_t c_25 = MUL(a_23, b_19);
0088
0089 IntrVec_t a_24 = LD(a, 24);
0090 c_24 = FMA(a_24, b_24, c_24);
0091 c_25 = FMA(a_24, b_25, c_25);
0092 IntrVec_t c_26 = MUL(a_24, b_26);
0093 ST(c, 27, all_zeros);
0094 ST(c, 28, all_zeros);
0095 ST(c, 29, all_zeros);
0096 ST(c, 24, c_24);
0097 ST(c, 25, c_25);
0098 ST(c, 26, c_26);
0099 }
0100
0101 #else
0102
0103 #pragma omp simd
0104 for (int n = 0; n < N; ++n) {
0105 c[0 * N + n] = 0;
0106 c[1 * N + n] = 0;
0107 c[2 * N + n] = 0;
0108 c[3 * N + n] = b[3 * N + n];
0109 c[4 * N + n] = 0;
0110 c[5 * N + n] = b[5 * N + n];
0111 c[6 * N + n] = a[8 * N + n] * b[18 * N + n] + a[9 * N + n] * b[24 * N + n];
0112 c[7 * N + n] = a[8 * N + n] * b[19 * N + n] + a[9 * N + n] * b[25 * N + n];
0113 c[8 * N + n] = a[9 * N + n] * b[26 * N + n];
0114 c[9 * N + n] = 0;
0115 c[10 * N + n] = a[7 * N + n];
0116 c[11 * N + n] = a[6 * N + n] * b[11 * N + n];
0117 c[12 * N + n] = a[13 * N + n] * b[18 * N + n] + a[14 * N + n] * b[24 * N + n];
0118 c[13 * N + n] = a[13 * N + n] * b[19 * N + n] + a[14 * N + n] * b[25 * N + n];
0119 c[14 * N + n] = a[14 * N + n] * b[26 * N + n];
0120 c[15 * N + n] = 0;
0121 c[16 * N + n] = a[12 * N + n];
0122 c[17 * N + n] = a[11 * N + n] * b[11 * N + n];
0123 c[18 * N + n] = a[18 * N + n] * b[18 * N + n] + a[19 * N + n] * b[24 * N + n];
0124 c[19 * N + n] = a[18 * N + n] * b[19 * N + n] + a[19 * N + n] * b[25 * N + n];
0125 c[20 * N + n] = a[19 * N + n] * b[26 * N + n];
0126 c[21 * N + n] = 0;
0127 c[22 * N + n] = 0;
0128 c[23 * N + n] = 0;
0129 c[24 * N + n] = a[23 * N + n] * b[18 * N + n] + a[24 * N + n] * b[24 * N + n];
0130 c[25 * N + n] = a[23 * N + n] * b[19 * N + n] + a[24 * N + n] * b[25 * N + n];
0131 c[26 * N + n] = a[24 * N + n] * b[26 * N + n];
0132 c[27 * N + n] = 0;
0133 c[28 * N + n] = 0;
0134 c[29 * N + n] = 0;
0135 }
0136 #endif
