| /external/XNNPACK/src/f32-vrnd/gen/ |
| D | vrndu-neon-x4.c | 41 const float32x4_t vrndx0123 = vbslq_f32(vrndmask0123, vprerndx0123, vx0123); in xnn_f32_vrndu_ukernel__neon_x4() local
43 uint32x4_t vadjmask0123 = vcgeq_f32(vrndx0123, vx0123); in xnn_f32_vrndu_ukernel__neon_x4()
45 const float32x4_t vadjrndx0123 = vaddq_f32(vrndx0123, vone); in xnn_f32_vrndu_ukernel__neon_x4()
49 const float32x4_t vy0123 = vbslq_f32(vadjmask0123, vrndx0123, vadjrndx0123); in xnn_f32_vrndu_ukernel__neon_x4()
|
| D | vrndu-sse2-x4.c | 40 …const __m128 vrndx0123 = _mm_or_ps(_mm_and_ps(vx0123, vrndmask0123), _mm_andnot_ps(vrndmask0123, v… in xnn_f32_vrndu_ukernel__sse2_x4() local
42 … const __m128 vadjmask0123 = _mm_or_ps(_mm_cmpge_ps(vrndx0123, vx0123), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x4()
44 const __m128 vadjrndx0123 = _mm_add_ps(vrndx0123, vone); in xnn_f32_vrndu_ukernel__sse2_x4()
46 …const __m128 vy0123 = _mm_or_ps(_mm_and_ps(vrndx0123, vadjmask0123), _mm_andnot_ps(vadjmask0123, v… in xnn_f32_vrndu_ukernel__sse2_x4()
|
| D | vrndu-wasmsimd-cvt-x8.c | 46 const v128_t vrndx0123 = wasm_v128_bitselect(vprerndx0123, vx0123, vrndmask0123); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() local
49 const v128_t vadjmask0123 = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx0123, vrndx0123)); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8()
52 const v128_t vadjrndx0123 = wasm_f32x4_add(vrndx0123, vone); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8()
55 const v128_t vy0123 = wasm_v128_bitselect(vrndx0123, vadjrndx0123, vadjmask0123); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8()
|
| D | vrndu-wasmsimd-addsub-x8.c | 45 const v128_t vrndx0123 = wasm_v128_bitselect(vx0123, vrndabsx0123, vrndmask0123); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8() local
48 const v128_t vadjmask0123 = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx0123, vrndx0123)); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8()
51 const v128_t vadjrndx0123 = wasm_f32x4_add(vrndx0123, vone); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8()
54 const v128_t vy0123 = wasm_v128_bitselect(vrndx0123, vadjrndx0123, vadjmask0123); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8()
|
| D | vrndd-sse2-x4.c | 40 …const __m128 vrndx0123 = _mm_or_ps(_mm_and_ps(vx0123, vrndmask0123), _mm_andnot_ps(vrndmask0123, v… in xnn_f32_vrndd_ukernel__sse2_x4() local
42 const __m128 vy0123 = _mm_sub_ps(vrndx0123, _mm_and_ps(_mm_cmpgt_ps(vrndx0123, vx0123), vone)); in xnn_f32_vrndd_ukernel__sse2_x4()
|
| D | vrndd-neon-x4.c | 41 const float32x4_t vrndx0123 = vbslq_f32(vrndmask0123, vprerndx0123, vx0123); in xnn_f32_vrndd_ukernel__neon_x4() local
43 const uint32x4_t vadjmask0123 = vcgtq_f32(vrndx0123, vx0123); in xnn_f32_vrndd_ukernel__neon_x4()
47 const float32x4_t vy0123 = vsubq_f32(vrndx0123, vadjrndx0123); in xnn_f32_vrndd_ukernel__neon_x4()
|
| D | vrndu-neon-x8.c | 46 const float32x4_t vrndx0123 = vbslq_f32(vrndmask0123, vprerndx0123, vx0123); in xnn_f32_vrndu_ukernel__neon_x8() local
49 uint32x4_t vadjmask0123 = vcgeq_f32(vrndx0123, vx0123); in xnn_f32_vrndu_ukernel__neon_x8()
52 const float32x4_t vadjrndx0123 = vaddq_f32(vrndx0123, vone); in xnn_f32_vrndu_ukernel__neon_x8()
58 const float32x4_t vy0123 = vbslq_f32(vadjmask0123, vrndx0123, vadjrndx0123); in xnn_f32_vrndu_ukernel__neon_x8()
|
| D | vrndu-sse2-x8.c | 44 …const __m128 vrndx0123 = _mm_or_ps(_mm_and_ps(vx0123, vrndmask0123), _mm_andnot_ps(vrndmask0123, v… in xnn_f32_vrndu_ukernel__sse2_x8() local
47 … const __m128 vadjmask0123 = _mm_or_ps(_mm_cmpge_ps(vrndx0123, vx0123), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x8()
50 const __m128 vadjrndx0123 = _mm_add_ps(vrndx0123, vone); in xnn_f32_vrndu_ukernel__sse2_x8()
53 …const __m128 vy0123 = _mm_or_ps(_mm_and_ps(vrndx0123, vadjmask0123), _mm_andnot_ps(vadjmask0123, v… in xnn_f32_vrndu_ukernel__sse2_x8()
|
| D | vrndd-sse2-x8.c | 44 …const __m128 vrndx0123 = _mm_or_ps(_mm_and_ps(vx0123, vrndmask0123), _mm_andnot_ps(vrndmask0123, v… in xnn_f32_vrndd_ukernel__sse2_x8() local
47 const __m128 vy0123 = _mm_sub_ps(vrndx0123, _mm_and_ps(_mm_cmpgt_ps(vrndx0123, vx0123), vone)); in xnn_f32_vrndd_ukernel__sse2_x8()
|
| D | vrndd-wasmsimd-addsub-x8.c | 45 const v128_t vrndx0123 = wasm_v128_bitselect(vx0123, vrndabsx0123, vrndmask0123); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x8() local
48 …const v128_t vy0123 = wasm_f32x4_sub(vrndx0123, wasm_v128_and(wasm_f32x4_lt(vx0123, vrndx0123), vo… in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x8()
|
| D | vrndd-wasmsimd-cvt-x8.c | 46 const v128_t vrndx0123 = wasm_v128_bitselect(vprerndx0123, vx0123, vrndmask0123); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8() local
49 const v128_t vadj0123 = wasm_v128_and(wasm_f32x4_lt(vx0123, vrndx0123), vone); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8()
52 const v128_t vy0123 = wasm_f32x4_sub(vrndx0123, vadj0123); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x8()
|
| D | vrndd-neon-x8.c | 46 const float32x4_t vrndx0123 = vbslq_f32(vrndmask0123, vprerndx0123, vx0123); in xnn_f32_vrndd_ukernel__neon_x8() local
49 const uint32x4_t vadjmask0123 = vcgtq_f32(vrndx0123, vx0123); in xnn_f32_vrndd_ukernel__neon_x8()
55 const float32x4_t vy0123 = vsubq_f32(vrndx0123, vadjrndx0123); in xnn_f32_vrndd_ukernel__neon_x8()
|
| D | vrndz-neon-x4.c | 36 const float32x4_t vrndx0123 = vcvtq_f32_s32(vintx0123); in xnn_f32_vrndz_ukernel__neon_x4() local
40 const float32x4_t vy0123 = vbslq_f32(vrndmask0123, vrndx0123, vx0123); in xnn_f32_vrndz_ukernel__neon_x4()
|
| D | vrndz-sse2-x4.c | 37 const __m128 vrndx0123 = _mm_cvtepi32_ps(vintx0123); in xnn_f32_vrndz_ukernel__sse2_x4() local
39 …_m128 vy0123 = _mm_or_ps(_mm_and_ps(vx0123, vrndmask0123), _mm_andnot_ps(vrndmask0123, vrndx0123)); in xnn_f32_vrndz_ukernel__sse2_x4()
|
| D | vrndne-sse2-x4.c | 37 const __m128 vrndx0123 = _mm_cvtepi32_ps(vintx0123); in xnn_f32_vrndne_ukernel__sse2_x4() local
39 …_m128 vy0123 = _mm_or_ps(_mm_and_ps(vx0123, vrndmask0123), _mm_andnot_ps(vrndmask0123, vrndx0123)); in xnn_f32_vrndne_ukernel__sse2_x4()
|
| D | vrndz-wasmsimd-cvt-x8.c | 40 const v128_t vrndx0123 = wasm_f32x4_convert_i32x4(vintx0123); in xnn_f32_vrndz_ukernel__wasmsimd_cvt_x8() local
45 const v128_t vy0123 = wasm_v128_bitselect(vrndx0123, vx0123, vrndmask0123); in xnn_f32_vrndz_ukernel__wasmsimd_cvt_x8()
|
| D | vrndz-neon-x8.c | 39 const float32x4_t vrndx0123 = vcvtq_f32_s32(vintx0123); in xnn_f32_vrndz_ukernel__neon_x8() local
45 const float32x4_t vy0123 = vbslq_f32(vrndmask0123, vrndx0123, vx0123); in xnn_f32_vrndz_ukernel__neon_x8()
|
| D | vrndz-sse2-x8.c | 40 const __m128 vrndx0123 = _mm_cvtepi32_ps(vintx0123); in xnn_f32_vrndz_ukernel__sse2_x8() local
43 …_m128 vy0123 = _mm_or_ps(_mm_and_ps(vx0123, vrndmask0123), _mm_andnot_ps(vrndmask0123, vrndx0123)); in xnn_f32_vrndz_ukernel__sse2_x8()
|
| D | vrndne-sse2-x8.c | 40 const __m128 vrndx0123 = _mm_cvtepi32_ps(vintx0123); in xnn_f32_vrndne_ukernel__sse2_x8() local
43 …_m128 vy0123 = _mm_or_ps(_mm_and_ps(vx0123, vrndmask0123), _mm_andnot_ps(vrndmask0123, vrndx0123)); in xnn_f32_vrndne_ukernel__sse2_x8()
|