// 
// sub_1025D4 from 0x1025d4 to 0x1026f0 (284 bytes)
// 
// 1 xrefs:
// > <???> @ 0x11a110
// 
// 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
// 

float32x4_t *__fastcall sub_1025D4(float32x4_t *result, __int64 a2, __int64 a3, __int64 a4, __int64 a5)
{
  __int64 i; // x30
  int v6; // w8
  int8x16_t v7; // q0
  int64x2_t v8; // q1
  int8x16_t v9; // q2
  float32x4_t v10; // q3
  __int64 v11; // x5
  unsigned __int64 v12; // d4
  unsigned __int64 v13; // d5
  __int64 v14; // x9
  int8x16_t v15; // q4
  int8x16_t v16; // q6
  __int64 v17; // x9
  int8x16_t v18; // q5
  float32x4_t v19; // q18
  int32x4_t v20; // q7
  float32x4_t v21; // q6
  float v22; // w9
  float v23; // s0
  __int64 v24; // x1
  int v25; // w8

  if ( (int)a2 >> 2 )
  {
    result = (float32x4_t *)sub_102CE0(result, a2, a3, a4, a5, i);
    for ( i = v11; v6; result += 2 )
    {
      v12 = vshl_n_s32(vzip2_s32((int32x2_t)v7.n128_u64[0], (int32x2_t)vextq_s8(v7, v7, 8u).n128_u64[0]), 1u).n64_u64[0];
      v13 = vshl_n_s32(vzip2_s32((int32x2_t)v9.n128_u64[0], (int32x2_t)vextq_s8(v9, v9, 8u).n128_u64[0]), 1u).n64_u64[0];
      v14 = HIDWORD(v12);
      v15 = *(int8x16_t *)(a3 + 4LL * (unsigned int)v12);
      v16 = *(int8x16_t *)(a3 + 4 * v14);
      v17 = 4LL * HIDWORD(v13);
      v18 = *(int8x16_t *)(a3 + 4LL * (unsigned int)v13);
      v19 = vextq_s8(v15, vextq_s8(v16, v15, 8u), 8u);
      v15.n128_u64[1] = v16.n128_u64[0];
      v20 = vmulq_f32(vcvtq_f32_u32(vshrq_n_u32(vuzp1q_s32(v7, v9), 1u)), v10);
      v21 = vextq_s8(v18, vextq_s8(*(int8x16_t *)(a3 + v17), v18, 8u), 8u);
      v18.n128_u64[1] = *(_QWORD *)(a3 + v17);
      --v6;
      v7 = vaddq_s64(v7, v8);
      v9 = vaddq_s64(v9, v8);
      *result = vmlaq_f32(v15, vzip1q_s32(v20, v20), vsubq_f32(v19, v15));
      result[1] = vmlaq_f32(v18, vzip2q_s32(v20, v20), vsubq_f32(v21, v18));
    }
    *(_QWORD *)a4 = v7.n128_u64[0];
  }
  v22 = 2.3283e-10;
  v23 = 1.0;
  if ( (a2 & 3) != 0 )
  {
    do
    {
      result->n128_u64[0] = vmla_n_f32(
                              vmul_n_f32(
                                *(float32x2_t *)(a3 + 4LL * (unsigned int)(2 * *(_DWORD *)(a4 + 4) + 2)),
                                (float)*(unsigned int *)a4 * v22),
                              *(float32x2_t *)(a3 + 4LL * (unsigned int)(2 * *(_DWORD *)(a4 + 4))),
                              v23 - (float)((float)*(unsigned int *)a4 * v22)).n64_u64[0];
      result = (float32x4_t *)sub_102E10(&result->n128_i8[8], i);
      i = v24;
    }
    while ( v25 );
  }
  return result;
}