| /external/XNNPACK/src/f32-vrnd/ |
| D | vrndu-neon.c.in | 49 uint32x4_t vadjmask${ABC[N:N+4]} = vcgeq_f32(vrndx${ABC[N:N+4]}, vx${ABC[N:N+4]});
55 vadjmask${ABC[N:N+4]} = vorrq_u32(vadjmask${ABC[N:N+4]}, vmovq_n_u32(UINT32_C(0x80000000)));
58 …const float32x4_t vy${ABC[N:N+4]} = vbslq_f32(vadjmask${ABC[N:N+4]}, vrndx${ABC[N:N+4]}, vadjrndx$…
71 uint32x4_t vadjmask = vcgeq_f32(vrndx, vx); variable
73 vadjmask = vorrq_u32(vadjmask, vmovq_n_u32(UINT32_C(0x80000000)));
74 const float32x4_t vy = vbslq_f32(vadjmask, vrndx, vadjrndx);
84 uint32x4_t vadjmask = vcgeq_f32(vrndx, vx); variable
86 vadjmask = vorrq_u32(vadjmask, vmovq_n_u32(UINT32_C(0x80000000)));
87 const float32x4_t vy = vbslq_f32(vadjmask, vrndx, vadjrndx);
|
| D | vrndu-sse2.c.in | 48 …const __m128 vadjmask${ABC[N:N+4]} = _mm_or_ps(_mm_cmpge_ps(vrndx${ABC[N:N+4]}, vx${ABC[N:N+4]}), …
54 …C[N:N+4]} = _mm_or_ps(_mm_and_ps(vrndx${ABC[N:N+4]}, vadjmask${ABC[N:N+4]}), _mm_andnot_ps(vadjmas…
70 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); variable
72 const __m128 vy = _mm_or_ps(_mm_and_ps(vrndx, vadjmask), _mm_andnot_ps(vadjmask, vadjrndx));
83 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); variable
85 __m128 vy = _mm_or_ps(_mm_and_ps(vrndx, vadjmask), _mm_andnot_ps(vadjmask, vadjrndx));
|
| D | vrndu-wasmsimd-addsub.c.in | 50 …const v128_t vadjmask${ABC[N:N+4]} = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx${ABC[N:N+4]}, vrndx…
56 …BC[N:N+4]} = wasm_v128_bitselect(vrndx${ABC[N:N+4]}, vadjrndx${ABC[N:N+4]}, vadjmask${ABC[N:N+4]});
71 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); variable
73 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask);
85 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); variable
87 v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask);
|
| D | vrndu-wasmsimd-cvt.c.in | 49 …const v128_t vadjmask${ABC[N:N+4]} = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx${ABC[N:N+4]}, vrndx…
55 …BC[N:N+4]} = wasm_v128_bitselect(vrndx${ABC[N:N+4]}, vadjrndx${ABC[N:N+4]}, vadjmask${ABC[N:N+4]});
71 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); variable
73 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask);
86 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); variable
88 v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask);
|
| D | vrndd-neon.c.in | 49 const uint32x4_t vadjmask${ABC[N:N+4]} = vcgtq_f32(vrndx${ABC[N:N+4]}, vx${ABC[N:N+4]});
52 …const float32x4_t vadjrndx${ABC[N:N+4]} = vreinterpretq_f32_u32(vandq_u32(vadjmask${ABC[N:N+4]}, v…
68 const uint32x4_t vadjmask = vcgtq_f32(vrndx, vx); variable
69 const float32x4_t vadjrndx = vreinterpretq_f32_u32(vandq_u32(vadjmask, vone));
80 const uint32x4_t vadjmask = vcgtq_f32(vrndx, vx); variable
81 const float32x4_t vadjrndx = vreinterpretq_f32_u32(vandq_u32(vadjmask, vone));
|
| /external/XNNPACK/src/f32-vrnd/gen/ |
| D | vrndu-neon-x8.c | 71 uint32x4_t vadjmask = vcgeq_f32(vrndx, vx); in xnn_f32_vrndu_ukernel__neon_x8() local
73 vadjmask = vorrq_u32(vadjmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndu_ukernel__neon_x8()
74 const float32x4_t vy = vbslq_f32(vadjmask, vrndx, vadjrndx); in xnn_f32_vrndu_ukernel__neon_x8()
84 uint32x4_t vadjmask = vcgeq_f32(vrndx, vx); in xnn_f32_vrndu_ukernel__neon_x8() local
86 vadjmask = vorrq_u32(vadjmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndu_ukernel__neon_x8()
87 const float32x4_t vy = vbslq_f32(vadjmask, vrndx, vadjrndx); in xnn_f32_vrndu_ukernel__neon_x8()
|
| D | vrndu-sse2-x8.c | 68 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x8() local
70 const __m128 vy = _mm_or_ps(_mm_and_ps(vrndx, vadjmask), _mm_andnot_ps(vadjmask, vadjrndx)); in xnn_f32_vrndu_ukernel__sse2_x8()
81 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x8() local
83 __m128 vy = _mm_or_ps(_mm_and_ps(vrndx, vadjmask), _mm_andnot_ps(vadjmask, vadjrndx)); in xnn_f32_vrndu_ukernel__sse2_x8()
|
| D | vrndu-wasmsimd-addsub-x4.c | 39 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4() local
41 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4()
53 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4() local
55 v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x4()
|
| D | vrndu-wasmsimd-cvt-x4.c | 40 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() local
42 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4()
55 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4() local
57 v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x4()
|
| D | vrndu-neon-x4.c | 60 uint32x4_t vadjmask = vcgeq_f32(vrndx, vx); in xnn_f32_vrndu_ukernel__neon_x4() local
62 vadjmask = vorrq_u32(vadjmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndu_ukernel__neon_x4()
63 const float32x4_t vy = vbslq_f32(vadjmask, vrndx, vadjrndx); in xnn_f32_vrndu_ukernel__neon_x4()
|
| D | vrndu-wasmsimd-cvt-x8.c | 71 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() local
73 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8()
86 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8() local
88 v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_cvt_x8()
|
| D | vrndd-neon-x8.c | 68 const uint32x4_t vadjmask = vcgtq_f32(vrndx, vx); in xnn_f32_vrndd_ukernel__neon_x8() local
69 const float32x4_t vadjrndx = vreinterpretq_f32_u32(vandq_u32(vadjmask, vone)); in xnn_f32_vrndd_ukernel__neon_x8()
80 const uint32x4_t vadjmask = vcgtq_f32(vrndx, vx); in xnn_f32_vrndd_ukernel__neon_x8() local
81 const float32x4_t vadjrndx = vreinterpretq_f32_u32(vandq_u32(vadjmask, vone)); in xnn_f32_vrndd_ukernel__neon_x8()
|
| D | vrndu-wasmsimd-addsub-x8.c | 69 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8() local
71 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8()
83 const v128_t vadjmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vx, vrndx)); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8() local
85 v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_f32_vrndu_ukernel__wasmsimd_addsub_x8()
|
| D | vrndu-sse2-x4.c | 57 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); in xnn_f32_vrndu_ukernel__sse2_x4() local
59 __m128 vy = _mm_or_ps(_mm_and_ps(vrndx, vadjmask), _mm_andnot_ps(vadjmask, vadjrndx)); in xnn_f32_vrndu_ukernel__sse2_x4()
|
| D | vrndd-neon-x4.c | 58 const uint32x4_t vadjmask = vcgtq_f32(vrndx, vx); in xnn_f32_vrndd_ukernel__neon_x4() local
59 const float32x4_t vadjrndx = vreinterpretq_f32_u32(vandq_u32(vadjmask, vone)); in xnn_f32_vrndd_ukernel__neon_x4()
|
| /external/XNNPACK/src/math/ |
| D | roundu-sse2-cvt.c | 54 const __m128 vadjmask = _mm_or_ps(_mm_cmpge_ps(vrndx, vx), _mm_castsi128_ps(vmagic)); in xnn_math_f32_roundu__sse2_cvt() local
67 const __m128 vy = _mm_or_ps(_mm_and_ps(vrndx, vadjmask), _mm_andnot_ps(vadjmask, vadjrndx)); in xnn_math_f32_roundu__sse2_cvt()
|
| D | roundu-sse-addsub.c | 59 const __m128 vadjmask = _mm_and_ps(_mm_cmplt_ps(vrndx, vx), vnonsign_mask); in xnn_math_f32_roundu__sse_addsub() local
71 const __m128 vy = _mm_or_ps(_mm_and_ps(vadjrndx, vadjmask), _mm_andnot_ps(vadjmask, vrndx)); in xnn_math_f32_roundu__sse_addsub()
|
| D | roundu-neon-cvt.c | 52 const uint32x4_t vadjmask = vorrq_u32(vcgeq_f32(vrndx, vx), vsign_mask); in xnn_math_f32_roundu__neon_cvt() local
65 const float32x4_t vy = vbslq_f32(vadjmask, vrndx, vadjrndx); in xnn_math_f32_roundu__neon_cvt()
|
| D | roundu-wasmsimd-cvt.c | 53 const v128_t vadjmask = wasm_v128_or(wasm_f32x4_ge(vrndx, vx), vsign_mask); in xnn_math_f32_roundu__wasmsimd_cvt() local
66 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_math_f32_roundu__wasmsimd_cvt()
|
| D | roundu-wasmsimd-addsub.c | 59 const v128_t vadjmask = wasm_v128_or(wasm_f32x4_ge(vrndx, vx), vsign_mask); in xnn_math_f32_roundu__wasmsimd_addsub() local
72 const v128_t vy = wasm_v128_bitselect(vrndx, vadjrndx, vadjmask); in xnn_math_f32_roundu__wasmsimd_addsub()
|
| D | roundu-neon-addsub.c | 59 const uint32x4_t vadjmask = vbicq_u32(vcltq_f32(vrndx, vx), vsign_mask); in xnn_math_f32_roundu__neon_addsub() local
71 const float32x4_t vy = vbslq_f32(vadjmask, vadjrndx, vrndx); in xnn_math_f32_roundu__neon_addsub()
|