| /external/XNNPACK/src/f16-vmulcaddc/gen/ |
| D | c8-minmax-fma3-2x.c | 95 __m128i vh0 = _mm256_cvtps_ph(vacc0, _MM_FROUND_NO_EXC); in xnn_f16_vmulcaddc_minmax_ukernel_c8__fma3_2x() local
99 _mm_storel_epi64((__m128i*) o0, vh0); in xnn_f16_vmulcaddc_minmax_ukernel_c8__fma3_2x()
102 vh0 = _mm_unpackhi_epi64(vh0, vh0); in xnn_f16_vmulcaddc_minmax_ukernel_c8__fma3_2x()
109 _mm_storeu_si32(o0, vh0); in xnn_f16_vmulcaddc_minmax_ukernel_c8__fma3_2x()
112 vh0 = _mm_srli_epi64(vh0, 32); in xnn_f16_vmulcaddc_minmax_ukernel_c8__fma3_2x()
119 *o0 = (uint16_t) _mm_extract_epi16(vh0, 0); in xnn_f16_vmulcaddc_minmax_ukernel_c8__fma3_2x()
|
| D | c16-minmax-fma3-2x.c | 132 __m128i vh0 = _mm256_cvtps_ph(vacc0, _MM_FROUND_NO_EXC); in xnn_f16_vmulcaddc_minmax_ukernel_c16__fma3_2x() local
136 _mm_storel_epi64((__m128i*) o0, vh0); in xnn_f16_vmulcaddc_minmax_ukernel_c16__fma3_2x()
139 vh0 = _mm_unpackhi_epi64(vh0, vh0); in xnn_f16_vmulcaddc_minmax_ukernel_c16__fma3_2x()
146 _mm_storeu_si32(o0, vh0); in xnn_f16_vmulcaddc_minmax_ukernel_c16__fma3_2x()
149 vh0 = _mm_srli_epi64(vh0, 32); in xnn_f16_vmulcaddc_minmax_ukernel_c16__fma3_2x()
156 *o0 = (uint16_t) _mm_extract_epi16(vh0, 0); in xnn_f16_vmulcaddc_minmax_ukernel_c16__fma3_2x()
|
| /external/tensorflow/tensorflow/compiler/mlir/tensorflow/tests/ |
| D | resource-alias-analysis-test.mlir | 42 %vh0 = "tf.VarHandleOp"() {container = "c", shared_name = "v0"} : () -> !tf_res
47 %read0 = "tf.ReadVariableOp"(%vh0) : (!tf_res) -> tensor<32xf32>
49 %read1 = "tf.ReadVariableOp"(%vh0) : (!tf_res) -> tensor<32xf32>
51 "tf.AssignVariableOp"(%vh0, %read2) : (!tf_res, tensor<32xf32>) -> ()
79 %vh0 = "tf.VarHandleOp"() {container = "c", shared_name = "v0"} : () -> !tf_res
83 %replicated = "tf.TPUReplicatedInput"(%vh0, %vh1) : (!tf_res, !tf_res) -> (!tf_res)
84 "tf.AssignVariableOp"(%vh0, %arg0) : (!tf_res, tensor<32xf32>) -> ()
103 %vh0 = "tf.VarHandleOp"() {container = "c", shared_name = "v0"} : () -> !tf_res
104 %read0 = "tf.ReadVariableOp"(%vh0) : (!tf_res) -> tensor<32xf32>
108 %if:3 = "tf.If"(%read0, %arg1, %vh0) {
[all …]
|
| /external/XNNPACK/src/f32-f16-vcvt/gen/ |
| D | vcvt-neon-x8.c | 75 uint16x8_t vh0 = vaddq_u16(vmanth0, vexph0); in xnn_f32_f16_vcvt_ukernel__neon_x8() local
77 vh0 = vbslq_u16(vnanmaskh0, vnanh, vh0); in xnn_f32_f16_vcvt_ukernel__neon_x8()
79 vh0 = vorrq_u16(vh0, vsignh0); in xnn_f32_f16_vcvt_ukernel__neon_x8()
81 vst1q_u16(o, vh0); o += 8; in xnn_f32_f16_vcvt_ukernel__neon_x8()
|
| D | vcvt-scalar-fabsf-x2.c | 80 uint16_t vh0 = vexph0 + vmanth0; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x2() local
83 vh0 = vnanh; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x2()
88 vh0 |= vsignh0; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x2()
91 o[0] = vh0; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x2()
|
| D | vcvt-scalar-bitcast-x2.c | 78 uint16_t vh0 = vexph0 + vmanth0; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x2() local
81 vh0 = vnanh; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x2()
86 vh0 |= vsignh0; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x2()
89 o[0] = vh0; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x2()
|
| D | vcvt-neon-x16.c | 100 uint16x8_t vh0 = vaddq_u16(vmanth0, vexph0); in xnn_f32_f16_vcvt_ukernel__neon_x16() local
103 vh0 = vbslq_u16(vnanmaskh0, vnanh, vh0); in xnn_f32_f16_vcvt_ukernel__neon_x16()
106 vh0 = vorrq_u16(vh0, vsignh0); in xnn_f32_f16_vcvt_ukernel__neon_x16()
109 vst1q_u16(o, vh0); o += 8; in xnn_f32_f16_vcvt_ukernel__neon_x16()
|
| D | vcvt-scalar-bitcast-x3.c | 92 uint16_t vh0 = vexph0 + vmanth0; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x3() local
96 vh0 = vnanh; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x3()
104 vh0 |= vsignh0; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x3()
108 o[0] = vh0; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x3()
|
| D | vcvt-scalar-fabsf-x3.c | 95 uint16_t vh0 = vexph0 + vmanth0; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x3() local
99 vh0 = vnanh; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x3()
107 vh0 |= vsignh0; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x3()
111 o[0] = vh0; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x3()
|
| D | vcvt-scalar-fabsf-x4.c | 110 uint16_t vh0 = vexph0 + vmanth0; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x4() local
115 vh0 = vnanh; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x4()
126 vh0 |= vsignh0; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x4()
131 o[0] = vh0; in xnn_f32_f16_vcvt_ukernel__scalar_fabsf_x4()
|
| D | vcvt-scalar-bitcast-x4.c | 106 uint16_t vh0 = vexph0 + vmanth0; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x4() local
111 vh0 = vnanh; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x4()
122 vh0 |= vsignh0; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x4()
127 o[0] = vh0; in xnn_f32_f16_vcvt_ukernel__scalar_bitcast_x4()
|
| D | vcvt-neon-x24.c | 125 uint16x8_t vh0 = vaddq_u16(vmanth0, vexph0); in xnn_f32_f16_vcvt_ukernel__neon_x24() local
129 vh0 = vbslq_u16(vnanmaskh0, vnanh, vh0); in xnn_f32_f16_vcvt_ukernel__neon_x24()
133 vh0 = vorrq_u16(vh0, vsignh0); in xnn_f32_f16_vcvt_ukernel__neon_x24()
137 vst1q_u16(o, vh0); o += 8; in xnn_f32_f16_vcvt_ukernel__neon_x24()
|
| D | vcvt-neonfp16-x8.c | 34 … const uint16x8_t vh0 = vreinterpretq_u16_f16(vcombine_f16(vcvt_f16_f32(vf0), vcvt_f16_f32(vf1))); in xnn_f32_f16_vcvt_ukernel__neonfp16_x8() local
36 vst1q_u16(o, vh0); o += 8; in xnn_f32_f16_vcvt_ukernel__neonfp16_x8()
|
| D | vcvt-sse2-x16.c | 94 __m128i vh0 = _mm_and_si128(vnanh, vnanmaskh0); in xnn_f32_f16_vcvt_ukernel__sse2_x16() local
102 vh0 = _mm_or_si128(vh0, vsignh0); in xnn_f32_f16_vcvt_ukernel__sse2_x16()
128 vh0 = _mm_or_si128(vh0, _mm_andnot_si128(vnanmaskh0, vnonsignh0)); in xnn_f32_f16_vcvt_ukernel__sse2_x16()
131 _mm_storeu_si128((__m128i*) o, vh0); in xnn_f32_f16_vcvt_ukernel__sse2_x16()
|
| D | vcvt-neon-x32.c | 150 uint16x8_t vh0 = vaddq_u16(vmanth0, vexph0); in xnn_f32_f16_vcvt_ukernel__neon_x32() local
155 vh0 = vbslq_u16(vnanmaskh0, vnanh, vh0); in xnn_f32_f16_vcvt_ukernel__neon_x32()
160 vh0 = vorrq_u16(vh0, vsignh0); in xnn_f32_f16_vcvt_ukernel__neon_x32()
165 vst1q_u16(o, vh0); o += 8; in xnn_f32_f16_vcvt_ukernel__neon_x32()
|
| D | vcvt-neonfp16-x16.c | 36 … const uint16x8_t vh0 = vreinterpretq_u16_f16(vcombine_f16(vcvt_f16_f32(vf0), vcvt_f16_f32(vf1))); in xnn_f32_f16_vcvt_ukernel__neonfp16_x16() local
39 vst1q_u16(o, vh0); o += 8; in xnn_f32_f16_vcvt_ukernel__neonfp16_x16()
|
| D | vcvt-sse2-x24.c | 114 __m128i vh0 = _mm_and_si128(vnanh, vnanmaskh0); in xnn_f32_f16_vcvt_ukernel__sse2_x24() local
125 vh0 = _mm_or_si128(vh0, vsignh0); in xnn_f32_f16_vcvt_ukernel__sse2_x24()
161 vh0 = _mm_or_si128(vh0, _mm_andnot_si128(vnanmaskh0, vnonsignh0)); in xnn_f32_f16_vcvt_ukernel__sse2_x24()
165 _mm_storeu_si128((__m128i*) o, vh0); in xnn_f32_f16_vcvt_ukernel__sse2_x24()
|
| D | vcvt-sse2-x32.c | 134 __m128i vh0 = _mm_and_si128(vnanh, vnanmaskh0); in xnn_f32_f16_vcvt_ukernel__sse2_x32() local
148 vh0 = _mm_or_si128(vh0, vsignh0); in xnn_f32_f16_vcvt_ukernel__sse2_x32()
194 vh0 = _mm_or_si128(vh0, _mm_andnot_si128(vnanmaskh0, vnonsignh0)); in xnn_f32_f16_vcvt_ukernel__sse2_x32()
199 _mm_storeu_si128((__m128i*) o, vh0); in xnn_f32_f16_vcvt_ukernel__sse2_x32()
|
| /external/XNNPACK/src/f16-f32-vcvt/gen/ |
| D | vcvt-neonfp16-x16.c | 31 const float16x8_t vh0 = vreinterpretq_f16_u16(vld1q_u16(i)); i += 8; in xnn_f16_f32_vcvt_ukernel__neonfp16_x16() local
34 const float32x4_t vf0 = vcvt_f32_f16(vget_low_f16(vh0)); in xnn_f16_f32_vcvt_ukernel__neonfp16_x16()
35 const float32x4_t vf1 = vcvt_f32_f16(vget_high_f16(vh0)); in xnn_f16_f32_vcvt_ukernel__neonfp16_x16()
|
| D | vcvt-scalar-x2.c | 38 const uint16_t vh0 = i[0]; in xnn_f16_f32_vcvt_ukernel__scalar_x2() local
42 const uint32_t vw0 = (uint32_t) vh0 << 16; in xnn_f16_f32_vcvt_ukernel__scalar_x2()
|
| D | vcvt-wasmsimd-int16-x16.c | 38 const v128_t vh0 = wasm_v128_load(i); in xnn_f16_f32_vcvt_ukernel__wasmsimd_int16_x16() local
42 const v128_t vsign0 = wasm_v128_and(vh0, vsign_mask); in xnn_f16_f32_vcvt_ukernel__wasmsimd_int16_x16()
45 const v128_t vnonsign0 = wasm_v128_xor(vh0, vsign0); in xnn_f16_f32_vcvt_ukernel__wasmsimd_int16_x16()
|
| D | vcvt-scalar-x3.c | 38 const uint16_t vh0 = i[0]; in xnn_f16_f32_vcvt_ukernel__scalar_x3() local
43 const uint32_t vw0 = (uint32_t) vh0 << 16; in xnn_f16_f32_vcvt_ukernel__scalar_x3()
|
| D | vcvt-sse41-int16-x16.c | 38 const __m128i vh0 = _mm_loadu_si128((const __m128i*) i); in xnn_f16_f32_vcvt_ukernel__sse41_int16_x16() local
42 const __m128i vsign0 = _mm_and_si128(vh0, vsign_mask); in xnn_f16_f32_vcvt_ukernel__sse41_int16_x16()
45 const __m128i vnonsign0 = _mm_xor_si128(vh0, vsign0); in xnn_f16_f32_vcvt_ukernel__sse41_int16_x16()
|
| D | vcvt-neon-int32-x16.c | 37 const uint16x8_t vh0 = vld1q_u16(i); i += 8; in xnn_f16_f32_vcvt_ukernel__neon_int32_x16() local
40 const uint32x4_t vw0 = vshll_n_u16(vget_low_u16(vh0), 16); in xnn_f16_f32_vcvt_ukernel__neon_int32_x16()
41 const uint32x4_t vw1 = vshll_n_u16(vget_high_u16(vh0), 16); in xnn_f16_f32_vcvt_ukernel__neon_int32_x16()
|
| D | vcvt-avx-int16-x16.c | 38 const __m128i vh0 = _mm_loadu_si128((const __m128i*) i); in xnn_f16_f32_vcvt_ukernel__avx_int16_x16() local
42 const __m128i vsign0 = _mm_and_si128(vh0, vsign_mask); in xnn_f16_f32_vcvt_ukernel__avx_int16_x16()
45 const __m128i vnonsign0 = _mm_xor_si128(vh0, vsign0); in xnn_f16_f32_vcvt_ukernel__avx_int16_x16()
|