RecoTracker/MkFitCore/src/MultHelixPropTranspEndcap.ah - master - githubmirror/cmssw

RecoTracker/MkFitCore/src/MultHelixPropTranspEndcap.ah
219 строк6 KB
Shahzad Malik Muzaffar
[RECONSTRUCTION] [LLVM14] Apply clang-tidy and clang-format changes
26 апр 2023, 17:50
26 апр 2023, 17:50cef8b06
#ifdef MPLEX_INTRINSICS

for (int n = 0; n < N; n += MPLEX_INTRINSICS_WIDTH_BYTES / sizeof(T)) {
#ifdef AVX512_INTRINSICS
  IntrVec_t all_zeros = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
#else
  IntrVec_t all_zeros = {0, 0, 0, 0, 0, 0, 0, 0};
#endif

  IntrVec_t b_0 = LD(b, 0);
  IntrVec_t c_0 = b_0;

  IntrVec_t b_2 = LD(b, 2);
  IntrVec_t a_2 = LD(a, 2);
  c_0 = FMA(b_2, a_2, c_0);

  IntrVec_t b_3 = LD(b, 3);
  IntrVec_t a_3 = LD(a, 3);
  c_0 = FMA(b_3, a_3, c_0);

  IntrVec_t b_4 = LD(b, 4);
  IntrVec_t a_4 = LD(a, 4);
  c_0 = FMA(b_4, a_4, c_0);

  IntrVec_t b_5 = LD(b, 5);
  IntrVec_t a_5 = LD(a, 5);
  c_0 = FMA(b_5, a_5, c_0);

  IntrVec_t b_6 = LD(b, 6);
  IntrVec_t c_1 = b_6;

  IntrVec_t b_7 = LD(b, 7);
  IntrVec_t c_2 = b_7;
  ST(c, 0, c_0);

  IntrVec_t b_8 = LD(b, 8);
  c_1 = FMA(b_8, a_2, c_1);
  IntrVec_t a_8 = LD(a, 8);
  c_2 = FMA(b_8, a_8, c_2);

  IntrVec_t b_9 = LD(b, 9);
  c_1 = FMA(b_9, a_3, c_1);
  IntrVec_t a_9 = LD(a, 9);
  c_2 = FMA(b_9, a_9, c_2);

  IntrVec_t b_10 = LD(b, 10);
  c_1 = FMA(b_10, a_4, c_1);
  IntrVec_t a_10 = LD(a, 10);
  c_2 = FMA(b_10, a_10, c_2);

  IntrVec_t b_11 = LD(b, 11);
  c_1 = FMA(b_11, a_5, c_1);
  IntrVec_t a_11 = LD(a, 11);
  c_2 = FMA(b_11, a_11, c_2);

  IntrVec_t b_12 = LD(b, 12);
  IntrVec_t c_3 = b_12;
  ST(c, 1, c_1);

  IntrVec_t b_13 = LD(b, 13);
  IntrVec_t c_4 = b_13;
  ST(c, 2, c_2);

  IntrVec_t b_14 = LD(b, 14);
  c_3 = FMA(b_14, a_2, c_3);
  c_4 = FMA(b_14, a_8, c_4);

  IntrVec_t b_15 = LD(b, 15);
  c_3 = FMA(b_15, a_3, c_3);
  c_4 = FMA(b_15, a_9, c_4);

  IntrVec_t b_16 = LD(b, 16);
  c_3 = FMA(b_16, a_4, c_3);
  c_4 = FMA(b_16, a_10, c_4);

  IntrVec_t b_17 = LD(b, 17);
  c_3 = FMA(b_17, a_5, c_3);
  c_4 = FMA(b_17, a_11, c_4);
  ST(c, 5, all_zeros);

  IntrVec_t b_18 = LD(b, 18);
  IntrVec_t c_6 = b_18;

  IntrVec_t b_19 = LD(b, 19);
  IntrVec_t c_7 = b_19;
  ST(c, 3, c_3);
  ST(c, 4, c_4);

  IntrVec_t b_20 = LD(b, 20);
  c_6 = FMA(b_20, a_2, c_6);
  c_7 = FMA(b_20, a_8, c_7);

  IntrVec_t b_21 = LD(b, 21);
  c_6 = FMA(b_21, a_3, c_6);
  c_7 = FMA(b_21, a_9, c_7);
  IntrVec_t c_9 = b_21;

  IntrVec_t b_22 = LD(b, 22);
  c_6 = FMA(b_22, a_4, c_6);
  c_7 = FMA(b_22, a_10, c_7);

  IntrVec_t b_23 = LD(b, 23);
  c_6 = FMA(b_23, a_5, c_6);
  c_7 = FMA(b_23, a_11, c_7);
  ST(c, 8, all_zeros);

  IntrVec_t b_24 = LD(b, 24);
  IntrVec_t c_10 = b_24;

  IntrVec_t b_25 = LD(b, 25);
  IntrVec_t c_11 = b_25;
  ST(c, 6, c_6);
  ST(c, 7, c_7);
  ST(c, 9, c_9);

  IntrVec_t b_26 = LD(b, 26);
  c_10 = FMA(b_26, a_2, c_10);
  c_11 = FMA(b_26, a_8, c_11);
  IntrVec_t a_26 = LD(a, 26);
  IntrVec_t c_14 = MUL(b_26, a_26);

  IntrVec_t b_27 = LD(b, 27);
  c_10 = FMA(b_27, a_3, c_10);
  c_11 = FMA(b_27, a_9, c_11);
  IntrVec_t c_13 = b_27;
  IntrVec_t a_27 = LD(a, 27);
  c_14 = FMA(b_27, a_27, c_14);

  IntrVec_t b_28 = LD(b, 28);
  c_10 = FMA(b_28, a_4, c_10);
  c_11 = FMA(b_28, a_10, c_11);
  c_14 = ADD(b_28, c_14);

  IntrVec_t b_29 = LD(b, 29);
  c_10 = FMA(b_29, a_5, c_10);
  c_11 = FMA(b_29, a_11, c_11);
  ST(c, 12, all_zeros);
  IntrVec_t a_29 = LD(a, 29);
  c_14 = FMA(b_29, a_29, c_14);

  IntrVec_t b_30 = LD(b, 30);
  IntrVec_t c_15 = b_30;
  ST(c, 10, c_10);
  ST(c, 11, c_11);
  ST(c, 13, c_13);
  ST(c, 14, c_14);

  IntrVec_t b_31 = LD(b, 31);
  IntrVec_t c_16 = b_31;

  IntrVec_t b_32 = LD(b, 32);
  c_15 = FMA(b_32, a_2, c_15);
  c_16 = FMA(b_32, a_8, c_16);
  IntrVec_t c_19 = MUL(b_32, a_26);

  IntrVec_t b_33 = LD(b, 33);
  c_15 = FMA(b_33, a_3, c_15);
  c_16 = FMA(b_33, a_9, c_16);
  IntrVec_t c_18 = b_33;
  c_19 = FMA(b_33, a_27, c_19);

  IntrVec_t b_34 = LD(b, 34);
  c_15 = FMA(b_34, a_4, c_15);
  c_16 = FMA(b_34, a_10, c_16);
  c_19 = ADD(b_34, c_19);

  IntrVec_t b_35 = LD(b, 35);
  c_15 = FMA(b_35, a_5, c_15);
  c_16 = FMA(b_35, a_11, c_16);
  ST(c, 17, all_zeros);
  c_19 = FMA(b_35, a_29, c_19);
  IntrVec_t c_20 = b_35;
  ST(c, 15, c_15);
  ST(c, 16, c_16);
  ST(c, 18, c_18);
  ST(c, 19, c_19);
  ST(c, 20, c_20);
}

#else

#pragma omp simd
for (int n = 0; n < N; ++n) {
  c[0 * N + n] = b[0 * N + n] + b[2 * N + n] * a[2 * N + n] + b[3 * N + n] * a[3 * N + n] +
                 b[4 * N + n] * a[4 * N + n] + b[5 * N + n] * a[5 * N + n];
  c[1 * N + n] = b[6 * N + n] + b[8 * N + n] * a[2 * N + n] + b[9 * N + n] * a[3 * N + n] +
                 b[10 * N + n] * a[4 * N + n] + b[11 * N + n] * a[5 * N + n];
  c[2 * N + n] = b[7 * N + n] + b[8 * N + n] * a[8 * N + n] + b[9 * N + n] * a[9 * N + n] +
                 b[10 * N + n] * a[10 * N + n] + b[11 * N + n] * a[11 * N + n];
  c[3 * N + n] = b[12 * N + n] + b[14 * N + n] * a[2 * N + n] + b[15 * N + n] * a[3 * N + n] +
                 b[16 * N + n] * a[4 * N + n] + b[17 * N + n] * a[5 * N + n];
  c[4 * N + n] = b[13 * N + n] + b[14 * N + n] * a[8 * N + n] + b[15 * N + n] * a[9 * N + n] +
                 b[16 * N + n] * a[10 * N + n] + b[17 * N + n] * a[11 * N + n];
  c[5 * N + n] = 0;
  c[6 * N + n] = b[18 * N + n] + b[20 * N + n] * a[2 * N + n] + b[21 * N + n] * a[3 * N + n] +
                 b[22 * N + n] * a[4 * N + n] + b[23 * N + n] * a[5 * N + n];
  c[7 * N + n] = b[19 * N + n] + b[20 * N + n] * a[8 * N + n] + b[21 * N + n] * a[9 * N + n] +
                 b[22 * N + n] * a[10 * N + n] + b[23 * N + n] * a[11 * N + n];
  c[8 * N + n] = 0;
  c[9 * N + n] = b[21 * N + n];
  c[10 * N + n] = b[24 * N + n] + b[26 * N + n] * a[2 * N + n] + b[27 * N + n] * a[3 * N + n] +
                  b[28 * N + n] * a[4 * N + n] + b[29 * N + n] * a[5 * N + n];
  c[11 * N + n] = b[25 * N + n] + b[26 * N + n] * a[8 * N + n] + b[27 * N + n] * a[9 * N + n] +
                  b[28 * N + n] * a[10 * N + n] + b[29 * N + n] * a[11 * N + n];
  c[12 * N + n] = 0;
  c[13 * N + n] = b[27 * N + n];
  c[14 * N + n] =
      b[26 * N + n] * a[26 * N + n] + b[27 * N + n] * a[27 * N + n] + b[28 * N + n] + b[29 * N + n] * a[29 * N + n];
  c[15 * N + n] = b[30 * N + n] + b[32 * N + n] * a[2 * N + n] + b[33 * N + n] * a[3 * N + n] +
                  b[34 * N + n] * a[4 * N + n] + b[35 * N + n] * a[5 * N + n];
  c[16 * N + n] = b[31 * N + n] + b[32 * N + n] * a[8 * N + n] + b[33 * N + n] * a[9 * N + n] +
                  b[34 * N + n] * a[10 * N + n] + b[35 * N + n] * a[11 * N + n];
  c[17 * N + n] = 0;
  c[18 * N + n] = b[33 * N + n];
  c[19 * N + n] =
      b[32 * N + n] * a[26 * N + n] + b[33 * N + n] * a[27 * N + n] + b[34 * N + n] + b[35 * N + n] * a[29 * N + n];
  c[20 * N + n] = b[35 * N + n];
}
#endif