| /external/XNNPACK/src/f32-vrnd/gen/ |
| D | vrndz-wasmsimd-addsub-x4.c | 38 const v128_t vadjustment = wasm_v128_and(wasm_f32x4_lt(vabsx, vrndabsx), vone); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x4() local
39 const v128_t vflrabsx = wasm_f32x4_sub(vrndabsx, vadjustment); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x4()
51 const v128_t vadjustment = wasm_v128_and(wasm_f32x4_lt(vabsx, vrndabsx), vone); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x4() local
52 const v128_t vflrabsx = wasm_f32x4_sub(vrndabsx, vadjustment); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x4()
|
| D | vrndz-wasmsimd-addsub-x8.c | 65 const v128_t vadjustment = wasm_v128_and(wasm_f32x4_lt(vabsx, vrndabsx), vone); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x8() local
66 const v128_t vflrabsx = wasm_f32x4_sub(vrndabsx, vadjustment); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x8()
78 const v128_t vadjustment = wasm_v128_and(wasm_f32x4_lt(vabsx, vrndabsx), vone); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x8() local
79 const v128_t vflrabsx = wasm_f32x4_sub(vrndabsx, vadjustment); in xnn_f32_vrndz_ukernel__wasmsimd_addsub_x8()
|
| /external/XNNPACK/src/f32-vsqrt/gen/ |
| D | avx512f-nr1fma1adj-x16.c | 39 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x16() local
40 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x16()
59 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x16() local
60 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x16()
|
| D | fma3-nr1fma1adj-x8.c | 40 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x8() local
41 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x8()
59 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x8() local
60 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x8()
|
| D | avx512f-nr1fma1adj-x32.c | 70 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x32() local
71 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x32()
90 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x32() local
91 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x32()
|
| D | fma3-nr1fma1adj-x16.c | 71 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x16() local
72 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x16()
90 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x16() local
91 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x16()
|
| D | avx512f-nr1fma1adj-x48.c | 80 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x48() local
81 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x48()
100 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x48() local
101 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x48()
|
| D | fma3-nr1fma1adj-x24.c | 81 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x24() local
82 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x24()
100 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x24() local
101 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x24()
|
| D | avx512f-nr1fma1adj-x64.c | 90 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x64() local
91 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x64()
110 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x64() local
111 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x64()
|
| D | fma3-nr1fma1adj-x32.c | 91 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x32() local
92 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x32()
110 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x32() local
111 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x32()
|
| D | avx512f-nr1fma1adj-x80.c | 100 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x80() local
101 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x80()
120 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x80() local
121 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x80()
|
| D | fma3-nr1fma1adj-x40.c | 101 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x40() local
102 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x40()
120 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x40() local
121 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x40()
|
| D | avx512f-nr1fma1adj-x96.c | 110 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x96() local
111 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x96()
130 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x96() local
131 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x96()
|
| D | fma3-nr1fma1adj-x48.c | 111 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x48() local
112 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x48()
130 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x48() local
131 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__fma3_nr1fma1adj_x48()
|
| D | avx512f-nr1fma1adj-x112.c | 120 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x112() local
121 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x112()
140 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x112() local
141 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_f32_vsqrt_ukernel__avx512f_nr1fma1adj_x112()
|
| /external/XNNPACK/src/f32-vrnd/ |
| D | vrndz-wasmsimd-addsub.c.in | 47 …const v128_t vadjustment${ABC[N:N+4]} = wasm_v128_and(wasm_f32x4_lt(vabsx${ABC[N:N+4]}, vrndabsx${…
50 …const v128_t vflrabsx${ABC[N:N+4]} = wasm_f32x4_sub(vrndabsx${ABC[N:N+4]}, vadjustment${ABC[N:N+4]…
67 const v128_t vadjustment = wasm_v128_and(wasm_f32x4_lt(vabsx, vrndabsx), vone); variable
68 const v128_t vflrabsx = wasm_f32x4_sub(vrndabsx, vadjustment);
80 const v128_t vadjustment = wasm_v128_and(wasm_f32x4_lt(vabsx, vrndabsx), vone); variable
81 const v128_t vflrabsx = wasm_f32x4_sub(vrndabsx, vadjustment);
|
| /external/XNNPACK/src/f32-vsqrt/ |
| D | avx512f-nr1fma1adj.c.in | 51 …const __m512 vadjustment${ABC[N]} = _mm512_fnmadd_ps(vsqrtx${ABC[N]}, vsqrtx${ABC[N]}, vx${ABC[N]}…
54 …const __m512 vy${ABC[N]} = _mm512_fmadd_ps(vhalfrsqrtx${ABC[N]}, vadjustment${ABC[N]}, vsqrtx${ABC…
71 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); variable
72 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx);
91 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); variable
92 const __m512 vy = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx);
|
| D | fma3-nr1fma1adj.c.in | 52 …const __m256 vadjustment${ABC[N]} = _mm256_fnmadd_ps(vsqrtx${ABC[N]}, vsqrtx${ABC[N]}, vx${ABC[N]}…
55 …const __m256 vy${ABC[N]} = _mm256_fmadd_ps(vhalfrsqrtx${ABC[N]}, vadjustment${ABC[N]}, vsqrtx${ABC…
72 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); variable
73 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx);
91 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); variable
92 const __m256 vy = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx);
|
| /external/XNNPACK/src/math/ |
| D | roundz-neon-addsub.c | 52 …const float32x4_t vadjustment = vreinterpretq_f32_u32(vandq_u32(vone, vcgtq_f32(vrndabsx, vabsx))); in xnn_math_f32_roundz__neon_addsub() local
55 const float32x4_t vflrabsx = vsubq_f32(vrndabsx, vadjustment); in xnn_math_f32_roundz__neon_addsub()
|
| D | sqrt-avx512f-nr1fma1adj.c | 43 const __m512 vadjustment = _mm512_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_math_f32_sqrt__avx512f_nr1fma1adj() local
44 vsqrtx = _mm512_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_math_f32_sqrt__avx512f_nr1fma1adj()
|
| D | sqrt-fma3-nr1fma1adj.c | 43 const __m256 vadjustment = _mm256_fnmadd_ps(vsqrtx, vsqrtx, vx); in xnn_math_f32_sqrt__fma3_nr1fma1adj() local
44 vsqrtx = _mm256_fmadd_ps(vhalfrsqrtx, vadjustment, vsqrtx); in xnn_math_f32_sqrt__fma3_nr1fma1adj()
|
| D | roundz-wasmsimd-addsub.c | 52 const v128_t vadjustment = wasm_v128_and(wasm_f32x4_gt(vrndabsx, vabsx), vone); in xnn_math_f32_roundz__wasmsimd_addsub() local
55 const v128_t vflrabsx = wasm_f32x4_sub(vrndabsx, vadjustment); in xnn_math_f32_roundz__wasmsimd_addsub()
|
| D | roundz-sse-addsub.c | 52 const __m128 vadjustment = _mm_and_ps(vone, _mm_cmpgt_ps(vrndabsx, vabsx)); in xnn_math_f32_roundz__sse_addsub() local
55 const __m128 vflrabsx = _mm_sub_ps(vrndabsx, vadjustment); in xnn_math_f32_roundz__sse_addsub()
|
| D | sqrt-neonfma-nr1rsqrts1fma1adj.c | 47 const float32x4_t vadjustment = vfmsq_f32(vx, vsqrtx, vsqrtx); in xnn_math_f32_sqrt__neonfma_nr1rsqrts1fma1adj() local
48 vsqrtx = vfmaq_f32(vsqrtx, vhalfrsqrtx, vadjustment); in xnn_math_f32_sqrt__neonfma_nr1rsqrts1fma1adj()
|
| D | sqrt-neonfma-nr2fma1adj.c | 46 const float32x4_t vadjustment = vfmsq_f32(vx, vsqrtx, vsqrtx); in xnn_math_f32_sqrt__neonfma_nr2fma1adj() local
47 vsqrtx = vfmaq_f32(vsqrtx, vhalfrsqrtx, vadjustment); in xnn_math_f32_sqrt__neonfma_nr2fma1adj()
|