Home
last modified time | relevance | path

Searched refs:vrndmask (Results 1 – 25 of 67) sorted by relevance

123

/external/XNNPACK/src/f32-vrnd/
Dvrndz-neon.c.in36 uint32x4_t vrndmask${ABC[N:N+4]} = vcaltq_f32(vx${ABC[N:N+4]}, vintegral_threshold);
42 vrndmask${ABC[N:N+4]} = vbicq_u32(vrndmask${ABC[N:N+4]}, vmovq_n_u32(UINT32_C(0x80000000)));
45 …const float32x4_t vy${ABC[N:N+4]} = vbslq_f32(vrndmask${ABC[N:N+4]}, vrndx${ABC[N:N+4]}, vx${ABC[N…
54 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); variable
56 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000)));
57 const float32x4_t vy = vbslq_f32(vrndmask, vrndx, vx);
63 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); variable
65 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000)));
66 const float32x4_t vy = vbslq_f32(vrndmask, vrndx, vx);
Dvrndne-neon.c.in34 uint32x4_t vrndmask${ABC[N:N+4]} = vcaltq_f32(vmagic_number, vx${ABC[N:N+4]});
40 vrndmask${ABC[N:N+4]} = vorrq_u32(vrndmask${ABC[N:N+4]}, vmovq_n_u32(UINT32_C(0x80000000)));
46 …const float32x4_t vy${ABC[N:N+4]} = vbslq_f32(vrndmask${ABC[N:N+4]}, vx${ABC[N:N+4]}, vrndabsx${AB…
55 uint32x4_t vrndmask = vcaltq_f32(vmagic_number, vx); variable
57 vrndmask = vorrq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000)));
59 const float32x4_t vy = vbslq_f32(vrndmask, vx, vrndabsx);
65 uint32x4_t vrndmask = vcaltq_f32(vmagic_number, vx); variable
67 vrndmask = vorrq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000)));
69 const float32x4_t vy = vbslq_f32(vrndmask, vx, vrndabsx);
Dvrndd-neon.c.in37 uint32x4_t vrndmask${ABC[N:N+4]} = vcaltq_f32(vx${ABC[N:N+4]}, vintegral_threshold);
43 vrndmask${ABC[N:N+4]} = vbicq_u32(vrndmask${ABC[N:N+4]}, vmovq_n_u32(UINT32_C(0x80000000)));
46 …const float32x4_t vrndx${ABC[N:N+4]} = vbslq_f32(vrndmask${ABC[N:N+4]}, vprerndx${ABC[N:N+4]}, vx$…
64 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); variable
66 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000)));
67 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx);
76 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); variable
78 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000)));
79 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx);
Dvrndu-neon.c.in37 uint32x4_t vrndmask${ABC[N:N+4]} = vcaltq_f32(vx${ABC[N:N+4]}, vintegral_threshold);
43 vrndmask${ABC[N:N+4]} = vbicq_u32(vrndmask${ABC[N:N+4]}, vmovq_n_u32(UINT32_C(0x80000000)));
46 …const float32x4_t vrndx${ABC[N:N+4]} = vbslq_f32(vrndmask${ABC[N:N+4]}, vprerndx${ABC[N:N+4]}, vx$…
67 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); variable
69 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000)));
70 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx);
81 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); variable
82 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000)));
83 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx);
Dvrndne-sse2.c.in38 …const __m128 vrndmask${ABC[N:N+4]} = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx${…
44 …{ABC[N:N+4]} = _mm_or_ps(_mm_and_ps(vx${ABC[N:N+4]}, vrndmask${ABC[N:N+4]}), _mm_andnot_ps(vrndmas…
57 … const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); variable
59 const __m128 vy = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vrndx));
67 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); variable
69 __m128 vy = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vrndx));
Dvrndz-sse2.c.in38 …const __m128 vrndmask${ABC[N:N+4]} = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx${…
44 …{ABC[N:N+4]} = _mm_or_ps(_mm_and_ps(vx${ABC[N:N+4]}, vrndmask${ABC[N:N+4]}), _mm_andnot_ps(vrndmas…
57 … const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); variable
59 const __m128 vy = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vrndx));
67 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); variable
69 __m128 vy = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vrndx));
Dvrndd-sse2.c.in39 …const __m128 vrndmask${ABC[N:N+4]} = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx${…
45 …{ABC[N:N+4]} = _mm_or_ps(_mm_and_ps(vx${ABC[N:N+4]}, vrndmask${ABC[N:N+4]}), _mm_andnot_ps(vrndmas…
61 … const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); variable
63 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx));
72 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); variable
74 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx));
Dvrndu-sse2.c.in39 …const __m128 vrndmask${ABC[N:N+4]} = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx${…
45 …{ABC[N:N+4]} = _mm_or_ps(_mm_and_ps(vx${ABC[N:N+4]}, vrndmask${ABC[N:N+4]}), _mm_andnot_ps(vrndmas…
67 … const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); variable
69 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx));
80 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); variable
82 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx));
Dvrndz-wasmsimd-cvt.c.in42 …const v128_t vrndmask${ABC[N:N+4]} = wasm_v128_andnot(wasm_f32x4_lt(vabsx${ABC[N:N+4]}, vmagic_num…
45 … vy${ABC[N:N+4]} = wasm_v128_bitselect(vrndx${ABC[N:N+4]}, vx${ABC[N:N+4]}, vrndmask${ABC[N:N+4]});
59 const v128_t vrndmask = wasm_v128_andnot(wasm_f32x4_lt(vabsx, vmagic_number), vsign_mask); variable
60 const v128_t vy = wasm_v128_bitselect(vrndx, vx, vrndmask);
71 const v128_t vrndmask = wasm_v128_andnot(wasm_f32x4_lt(vabsx, vmagic_number), vsign_mask); variable
72 v128_t vy = wasm_v128_bitselect(vrndx, vx, vrndmask);
/external/XNNPACK/src/f32-vrnd/gen/
Dvrndz-neon-x8.c54 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); in xnn_f32_vrndz_ukernel__neon_x8() local
56 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndz_ukernel__neon_x8()
57 const float32x4_t vy = vbslq_f32(vrndmask, vrndx, vx); in xnn_f32_vrndz_ukernel__neon_x8()
63 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); in xnn_f32_vrndz_ukernel__neon_x8() local
65 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndz_ukernel__neon_x8()
66 const float32x4_t vy = vbslq_f32(vrndmask, vrndx, vx); in xnn_f32_vrndz_ukernel__neon_x8()
Dvrndne-neon-x8.c56 uint32x4_t vrndmask = vcaltq_f32(vmagic_number, vx); in xnn_f32_vrndne_ukernel__neon_x8() local
58 vrndmask = vorrq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndne_ukernel__neon_x8()
60 const float32x4_t vy = vbslq_f32(vrndmask, vx, vrndabsx); in xnn_f32_vrndne_ukernel__neon_x8()
66 uint32x4_t vrndmask = vcaltq_f32(vmagic_number, vx); in xnn_f32_vrndne_ukernel__neon_x8() local
68 vrndmask = vorrq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndne_ukernel__neon_x8()
70 const float32x4_t vy = vbslq_f32(vrndmask, vx, vrndabsx); in xnn_f32_vrndne_ukernel__neon_x8()
Dvrndd-neon-x8.c64 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); in xnn_f32_vrndd_ukernel__neon_x8() local
66 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndd_ukernel__neon_x8()
67 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndd_ukernel__neon_x8()
76 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); in xnn_f32_vrndd_ukernel__neon_x8() local
78 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndd_ukernel__neon_x8()
79 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndd_ukernel__neon_x8()
Dvrndu-neon-x8.c67 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); in xnn_f32_vrndu_ukernel__neon_x8() local
69 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndu_ukernel__neon_x8()
70 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndu_ukernel__neon_x8()
81 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); in xnn_f32_vrndu_ukernel__neon_x8() local
82 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndu_ukernel__neon_x8()
83 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndu_ukernel__neon_x8()
Dvrndz-sse2-x8.c55 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndz_ukernel__sse2_x8() local
57 const __m128 vy = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vrndx)); in xnn_f32_vrndz_ukernel__sse2_x8()
65 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndz_ukernel__sse2_x8() local
67 __m128 vy = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vrndx)); in xnn_f32_vrndz_ukernel__sse2_x8()
Dvrndne-sse2-x8.c55 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndne_ukernel__sse2_x8() local
57 const __m128 vy = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vrndx)); in xnn_f32_vrndne_ukernel__sse2_x8()
65 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndne_ukernel__sse2_x8() local
67 __m128 vy = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vrndx)); in xnn_f32_vrndne_ukernel__sse2_x8()
Dvrndd-sse2-x8.c59 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndd_ukernel__sse2_x8() local
61 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndd_ukernel__sse2_x8()
70 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndd_ukernel__sse2_x8() local
72 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndd_ukernel__sse2_x8()
Dvrndu-sse2-x8.c65 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndu_ukernel__sse2_x8() local
67 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndu_ukernel__sse2_x8()
78 const __m128 vrndmask = _mm_castsi128_ps(_mm_or_si128(vmagic, _mm_cmpeq_epi32(vintx, vmagic))); in xnn_f32_vrndu_ukernel__sse2_x8() local
80 const __m128 vrndx = _mm_or_ps(_mm_and_ps(vx, vrndmask), _mm_andnot_ps(vrndmask, vprerndx)); in xnn_f32_vrndu_ukernel__sse2_x8()
Dvrndz-neon-x4.c47 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); in xnn_f32_vrndz_ukernel__neon_x4() local
49 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndz_ukernel__neon_x4()
50 const float32x4_t vy = vbslq_f32(vrndmask, vrndx, vx); in xnn_f32_vrndz_ukernel__neon_x4()
Dvrndne-wasmsimd-addsub-x4.c35 const v128_t vrndmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vmagic_number, vabsx)); in xnn_f32_vrndne_ukernel__wasmsimd_addsub_x4() local
37 const v128_t vy = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndne_ukernel__wasmsimd_addsub_x4()
46 const v128_t vrndmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vmagic_number, vabsx)); in xnn_f32_vrndne_ukernel__wasmsimd_addsub_x4() local
48 v128_t vy = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndne_ukernel__wasmsimd_addsub_x4()
Dvrndz-wasmsimd-cvt-x4.c37 const v128_t vrndmask = wasm_v128_andnot(wasm_f32x4_lt(vabsx, vmagic_number), vsign_mask); in xnn_f32_vrndz_ukernel__wasmsimd_cvt_x4() local
38 const v128_t vy = wasm_v128_bitselect(vrndx, vx, vrndmask); in xnn_f32_vrndz_ukernel__wasmsimd_cvt_x4()
49 const v128_t vrndmask = wasm_v128_andnot(wasm_f32x4_lt(vabsx, vmagic_number), vsign_mask); in xnn_f32_vrndz_ukernel__wasmsimd_cvt_x4() local
50 v128_t vy = wasm_v128_bitselect(vrndx, vx, vrndmask); in xnn_f32_vrndz_ukernel__wasmsimd_cvt_x4()
Dvrndne-neon-x4.c48 uint32x4_t vrndmask = vcaltq_f32(vmagic_number, vx); in xnn_f32_vrndne_ukernel__neon_x4() local
50 vrndmask = vorrq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndne_ukernel__neon_x4()
52 const float32x4_t vy = vbslq_f32(vrndmask, vx, vrndabsx); in xnn_f32_vrndne_ukernel__neon_x4()
Dvrndd-wasmsimd-addsub-x4.c36 const v128_t vrndmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vmagic_number, vabsx)); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x4() local
38 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x4()
48 const v128_t vrndmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vmagic_number, vabsx)); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x4() local
50 const v128_t vrndx = wasm_v128_bitselect(vx, vrndabsx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_addsub_x4()
Dvrndd-wasmsimd-cvt-x4.c38 const v128_t vrndmask = wasm_v128_andnot(wasm_f32x4_lt(vabsx, vmagic_number), vsign_mask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4() local
39 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4()
52 const v128_t vrndmask = wasm_v128_andnot(wasm_f32x4_lt(vabsx, vmagic_number), vsign_mask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4() local
53 const v128_t vrndx = wasm_v128_bitselect(vprerndx, vx, vrndmask); in xnn_f32_vrndd_ukernel__wasmsimd_cvt_x4()
Dvrndz-wasmsimd-addsub-x4.c36 const v128_t vrndmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vmagic_number, vabsx)); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x4() local
40 const v128_t vy = wasm_v128_bitselect(vx, vflrabsx, vrndmask); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x4()
49 const v128_t vrndmask = wasm_v128_or(vsign_mask, wasm_f32x4_le(vmagic_number, vabsx)); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x4() local
53 v128_t vy = wasm_v128_bitselect(vx, vflrabsx, vrndmask); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x4()
Dvrndd-neon-x4.c54 uint32x4_t vrndmask = vcaltq_f32(vx, vintegral_threshold); in xnn_f32_vrndd_ukernel__neon_x4() local
56 vrndmask = vbicq_u32(vrndmask, vmovq_n_u32(UINT32_C(0x80000000))); in xnn_f32_vrndd_ukernel__neon_x4()
57 const float32x4_t vrndx = vbslq_f32(vrndmask, vprerndx, vx); in xnn_f32_vrndd_ukernel__neon_x4()

123