45     const float32x4_t x1v = vsubq_f32(a01v, a23v);  in cft1st_128_neon()
47     const float32x4_t x3v = vsubq_f32(a45v, a67v);  in cft1st_128_neon()
51     x0v = vsubq_f32(x0v, x2v);  in cft1st_128_neon()
88     const float32x4_t x1r0_1i0_1r1_x1i1 = vsubq_f32(a_00_32, a_08_40);  in cftmdl_128_neon()
96     const float32x4_t x3r0_3i0_3r1_x3i1 = vsubq_f32(a_16_48, a_24_56);  in cftmdl_128_neon()
98     const float32x4_t xx1 = vsubq_f32(x0r0_0i0_0r1_x0i1, x2r0_2i0_2r1_x2i1);  in cftmdl_128_neon()
146       const float32x4_t x1r0_1i0_1r1_x1i1 = vsubq_f32(a_00_32, a_08_40);  in cftmdl_128_neon()
154       const float32x4_t x3r0_3i0_3r1_x3i1 = vsubq_f32(a_16_48, a_24_56);  in cftmdl_128_neon()
156       const float32x4_t xx1 = vsubq_f32(x0r0_0i0_0r1_x0i1, x2r0_2i0_2r1_x2i1);  in cftmdl_128_neon()
199     const float32x4_t wkrt = vsubq_f32(mm_half, c_k1);   // 28, 29, 30, 31,  in rftfsub_128_neon()
212     const float32x4_t xr_ = vsubq_f32(a_j2_p.val[0], a_k2_p0);  in rftfsub_128_neon()
223     const float32x4_t yr_ = vsubq_f32(a_, b_);  // 2-126, 4-124, 6-122, 8-120,  in rftfsub_128_neon()
233     const float32x4_t a_k2_p1n = vsubq_f32(a_k2_p1, yi_);  in rftfsub_128_neon()
240     a_j2_p.val[0] = vsubq_f32(a_j2_p.val[0], yr_);  in rftfsub_128_neon()
242     a_j2_p.val[1] = vsubq_f32(a_j2_p.val[1], yi_);  in rftfsub_128_neon()
279     const float32x4_t wkrt = vsubq_f32(mm_half, c_k1);  // 28, 29, 30, 31,  in rftbsub_128_neon()
292     const float32x4_t xr_ = vsubq_f32(a_j2_p.val[0], a_k2_p0);  in rftbsub_128_neon()
304     const float32x4_t yi_ = vsubq_f32(c_, d_);  // 3-127, 5-125, 7-123, 9-121,  in rftbsub_128_neon()
313     const float32x4_t a_k2_p1n = vsubq_f32(yi_, a_k2_p1);  in rftbsub_128_neon()
321     a_j2_p.val[0] = vsubq_f32(a_j2_p.val[0], yr_);  in rftbsub_128_neon()
323     a_j2_p.val[1] = vsubq_f32(yi_, a_j2_p.val[1]);  in rftbsub_128_neon()