/external/XNNPACK/src/math/ |
D | expminus-neonfma-rr2-lut2048-p1.c | 66 …float32x2_t vl01 = vld1_dup_f32((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_2048 + (uin… in xnn_math_f32_expminus__neonfma_rr2_lut2048_p1() local 68 …vl01 = vld1_lane_f32((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_2048 + (uint32_t) (vid… in xnn_math_f32_expminus__neonfma_rr2_lut2048_p1() 70 const float32x4_t vl = vcombine_f32(vl01, vl23); in xnn_math_f32_expminus__neonfma_rr2_lut2048_p1()
|
D | expminus-neonfma-rr2-lut64-p2.c | 66 …float32x2_t vl01 = vld1_dup_f32((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint3… in xnn_math_f32_expminus__neonfma_rr2_lut64_p2() local 68 …vl01 = vld1_lane_f32((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx0… in xnn_math_f32_expminus__neonfma_rr2_lut64_p2() 70 const float32x4_t vl = vcombine_f32(vl01, vl23); in xnn_math_f32_expminus__neonfma_rr2_lut64_p2()
|
D | exp-neonfma-rr2-lut64-p2.c | 72 …float32x2_t vl01 = vld1_dup_f32((const float*) ((uintptr_t) xnn_table_exp2_k_over_64 + (uint32_t) … in xnn_math_f32_exp__neonfma_rr2_lut64_p2() local 74 …vl01 = vld1_lane_f32((const float*) ((uintptr_t) xnn_table_exp2_k_over_64 + (uint32_t) (vidx01 >> … in xnn_math_f32_exp__neonfma_rr2_lut64_p2() 76 float32x4_t vl = vcombine_f32(vl01, vl23); in xnn_math_f32_exp__neonfma_rr2_lut64_p2()
|
/external/XNNPACK/src/f32-raddstoreexpminusmax/gen/ |
D | neon-lut64-p2-x8.c | 84 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx01]); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x8() local 89 vl01 = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x8() 91 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x8()
|
D | neonfma-lut64-p2-x8-acc2.c | 84 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx01]); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x8_acc2() local 89 vl01 = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x8_acc2() 91 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x8_acc2()
|
D | neon-lut64-p2-x8-acc2.c | 85 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx01]); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x8_acc2() local 90 vl01 = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x8_acc2() 92 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x8_acc2()
|
D | neonfma-lut64-p2-x8.c | 83 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx01]); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x8() local 88 vl01 = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x8() 90 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x8()
|
D | neonfma-lut64-p2-x12-acc2.c | 91 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx01]); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x12_acc2() local 98 vl01 = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x12_acc2() 100 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x12_acc2()
|
D | neon-lut64-p2-x12.c | 91 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx01]); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x12() local 98 vl01 = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x12() 100 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x12()
|
D | neonfma-lut64-p2-x12.c | 90 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx01]); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x12() local 97 vl01 = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x12() 99 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x12()
|
D | neon-lut64-p2-x12-acc3.c | 93 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx01]); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x12_acc3() local 100 vl01 = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x12_acc3() 102 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x12_acc3()
|
D | neonfma-lut64-p2-x12-acc3.c | 92 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx01]); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x12_acc3() local 99 vl01 = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x12_acc3() 101 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x12_acc3()
|
/external/XNNPACK/src/f32-sigmoid/gen/ |
D | neonfma-rr1-lut64-p2-div-x8.c | 55 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_div_x8() local 62 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_div_x8() 64 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_div_x8()
|
D | neonfma-rr1-lut2048-p1-div-x8.c | 54 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x8() local 61 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x8() 63 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x8()
|
D | neonfma-rr1-lut64-p2-nr2recps-x8.c | 55 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2recps_x8() local 62 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2recps_x8() 64 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2recps_x8()
|
D | neon-rr2-lut64-p2-nr2recps-x8.c | 56 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neon_rr2_lut64_p2_nr2recps_x8() local 63 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neon_rr2_lut64_p2_nr2recps_x8() 65 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neon_rr2_lut64_p2_nr2recps_x8()
|
D | neon-rr2-lut2048-p1-nr2recps-x8.c | 55 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neon_rr2_lut2048_p1_nr2recps_x8() local 62 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neon_rr2_lut2048_p1_nr2recps_x8() 64 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neon_rr2_lut2048_p1_nr2recps_x8()
|
D | neonfma-rr1-lut64-p2-nr1recps1fma-x8.c | 55 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr1recps1fma_x8() local 62 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr1recps1fma_x8() 64 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr1recps1fma_x8()
|
D | neonfma-rr1-lut2048-p1-nr2recps-x8.c | 54 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2recps_x8() local 61 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2recps_x8() 63 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2recps_x8()
|
D | neonfma-rr1-lut2048-p1-nr2fma-x8.c | 54 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2fma_x8() local 61 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2fma_x8() 63 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2fma_x8()
|
D | neonfma-rr1-lut2048-p1-nr1recps1fma-x8.c | 54 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr1recps1fma_x8() local 61 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr1recps1fma_x8() 63 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr1recps1fma_x8()
|
D | neonfma-rr1-lut64-p2-nr2fma-x8.c | 55 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2fma_x8() local 62 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2fma_x8() 64 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2fma_x8()
|
D | neonfma-rr1-lut2048-p1-div-x12.c | 59 float32x2_t vl01 = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx01]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x12() local 70 vl01 = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx01 >> 32)], vl01, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x12() 72 const float32x4_t vl0123 = vcombine_f32(vl01, vl23); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x12()
|
/external/XNNPACK/src/f32-velu/gen/ |
D | velu-neonfma-rr1-lut16-p3-x8.c | 61 …int32x2_t vl01 = vld1_dup_s32((const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint3… in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x8() local 63 …vl01 = vld1_lane_s32((const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vid… in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x8() 65 const int32x4_t vl0123 = vcombine_s32(vl01, vl23); in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x8()
|
D | velu-neon-rr2-lut16-p3-x8.c | 62 …int32x2_t vl01 = vld1_dup_s32((const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint3… in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x8() local 64 …vl01 = vld1_lane_s32((const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vid… in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x8() 66 const int32x4_t vl0123 = vcombine_s32(vl01, vl23); in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x8()
|