// 
// sub_10449C from 0x10449c to 0x104768 (716 bytes)
// 
// 1 xrefs:
// > <???> @ 0x11f748
// 
// This function is hookable on this platform!
// 
// Using BromaIDA 8.0.0 @ https://github.com/Stazzical/BromaIDA
// Using bindings at commit ba9f177b at Thu Jul 16 22:52:06 2026 from https://github.com/geode-sdk/bindings
// 

__int64 __fastcall sub_10449C(__int64 a1, _QWORD *a2, __int64 a3, __int64 a4, __int64 a5, int a6)
{
  float32x4_t *v8; // x22
  float32x4_t *v9; // x21
  int v10; // w23
  long double v11; // q0
  int v12; // w5
  int v13; // w23
  __int64 j; // x0
  int16x8_t v15; // q0
  int8x16_t v16; // q3
  int32x4_t v17; // q5
  int32x4_t v18; // q4
  int32x4_t v19; // q0
  int32x4_t v20; // q2
  long double v21; // q0
  __int64 v22; // x0
  __int64 v23; // x1
  int8x16_t v24; // q0
  int32x4_t v25; // q2
  int32x4_t v26; // q3
  long double v27; // q0
  int16x8_t v28; // q0
  float32x4_t v29; // q1
  float32x4_t *v30; // x9
  float32x4_t *v31; // x8
  float32x4_t v32; // q4
  int8x16_t v33; // q3
  float32x4_t v34; // q5
  int k; // w23
  int32x4_t v36; // q5
  int32x4_t v37; // q6
  int32x4_t v38; // q4
  int32x4_t v39; // q0
  float32x4_t v40; // q2
  float32x4_t v41; // q0
  int16x8_t v43; // [xsp+0h] [xbp-D0h]
  int16x8_t v44; // [xsp+10h] [xbp-C0h]
  int16x8_t v45; // [xsp+20h] [xbp-B0h]
  int8x16_t v46; // [xsp+20h] [xbp-B0h]
  int16x8_t i; // [xsp+20h] [xbp-B0h]
  float32x4_t v48; // [xsp+30h] [xbp-A0h]
  float32x4_t v49; // [xsp+40h] [xbp-90h]
  int8x16_t v50; // [xsp+50h] [xbp-80h]
  int8x16_t v51; // [xsp+60h] [xbp-70h] BYREF
  __n128 v52[3]; // [xsp+70h] [xbp-60h] BYREF

  v50 = *(int8x16_t *)a1;
  v49 = *(float32x4_t *)(a1 + 16);
  v8 = (float32x4_t *)(*a2 + 4 * a3);
  v9 = (float32x4_t *)(a2[1] + 4 * a3);
  v10 = a6 >> 2;
  v48 = *(float32x4_t *)(a1 + 32);
  switch ( *(_BYTE *)(a1 + 100) )
  {
    case 0:
    case 1:
      while ( v10 )
      {
        v21 = sub_104768(a1);
        v46 = COERCE_INT8X16_T(sub_104774(v21));
        v24.n128_u64[0] = sub_103C94(v22, v23);
        v25 = vmlaq_f32(v49, vcvtq_f32_u32(vandq_s8(v46, v50)), v48);
        v26 = vmlaq_f32(v49, vcvtq_f32_u32(vandq_s8(v24, v50)), v48);
        *v8 = vaddq_f32(vuzp1q_s32(v25, v26), *v8);
        --v10;
        ++v8;
        *v9 = vaddq_f32(vuzp2q_s32(v25, v26), *v9);
        ++v9;
      }
      break;
    case 2:
    case 3:
      for ( i = vdupq_n_s16(-*(unsigned __int8 *)(a1 + 102)); v10; v9 = v30 + 1 )
      {
        v27 = sub_104768(a1);
        --v10;
        *(long double *)&v28 = sub_104804(v27, *(long double *)v8);
        *v8 = v29;
        v8 = v31;
        *v9 = vmlaq_f32(vaddq_f32(*v30, v32), vcvtq_f32_u32(vandq_s8(v33, vshlq_s16(v28, i))), v34);
      }
      break;
    case 4:
    case 5:
    case 6:
    case 7:
      v11 = sub_104780();
      v13 = v12 >> 3;
      for ( j = sub_1047CC(v11); v13; v9 += 2 )
      {
        *(long double *)&v15 = sub_104768(j);
        v16 = vshlq_s16(v15, v43);
        v17 = vmlaq_f32(v49, vcvtq_f32_u32(vandq_s8(v50, vshlq_s16(v15, v45))), v48);
        v18 = vmlaq_f32(v49, vcvtq_f32_u32(vandq_s8(v15, v50)), v48);
        v19 = vmlaq_f32(v49, vcvtq_f32_u32(vandq_s8(v50, vshlq_s16(v15, v44))), v48);
        *v8 = vaddq_f32(vzip1q_s32(v18, v19), *v8);
        v20 = vmlaq_f32(v49, vcvtq_f32_u32(vandq_s8(v50, v16)), v48);
        *v9 = vaddq_f32(vzip1q_s32(v17, v20), *v9);
        --v13;
        v8[1] = vaddq_f32(vzip2q_s32(v18, v19), v8[1]);
        v8 += 2;
        v9[1] = vaddq_f32(v9[1], vzip2q_s32(v17, v20));
      }
      break;
    case 8:
      for ( k = a6 >> 3; k; v9 += 2 )
      {
        sub_1047B8(a1);
        sub_1043B8(a1, a5, v52, &v51);
        --k;
        v36 = vmlaq_f32(v49, vcvtq_f32_u32(vandq_s8(v52[1], v50)), v48);
        v37 = vmlaq_f32(v49, vcvtq_f32_u32(vandq_s8(v52[0], v50)), v48);
        v38 = vmlaq_f32(v49, vcvtq_f32_u32(vandq_s8(v52[2], v50)), v48);
        v39 = vmlaq_f32(v49, vcvtq_f32_u32(vandq_s8(v51, v50)), v48);
        v40 = vaddq_f32(vuzp1q_s32(v37, v39), v8[1]);
        *v8 = vaddq_f32(*v8, vuzp1q_s32(v38, v36));
        v8[1] = v40;
        v8 += 2;
        v41 = vaddq_f32(v9[1], vuzp2q_s32(v37, v39));
        *v9 = vaddq_f32(*v9, vuzp2q_s32(v38, v36));
        v9[1] = v41;
      }
      break;
    default:
      return 0;
  }
  return 0;
}