/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/x64/ |
H A D | jit_brgemm_post_ops.hpp | 72 const reg64_t reg_bias = r14; member 181 auto addr = ptr[reg_bias + bia_typesize_ * n_ * brg_.ld_block]; in loop_by_N() 190 auto addr = ptr[reg_bias + bia_typesize_ * n_ * brg_.ld_block]; in loop_by_N() 236 mov(reg_bias, ptr[param1 + GET_OFF(ptr_diff_bias)]); in generate() 244 add(reg_bias, bia_typesize_ * n_max_regs_ * brg_.ld_block); in generate() 352 const reg64_t reg_bias = r11; member 549 if (jcp.with_bias) mov(aux_reg_bias, reg_bias); in loop_by_N() 632 if (jcp.with_bias) mov(reg_bias, ptr[param1 + GET_OFF(ptr_bias)]); in generate()
|
H A D | jit_avx512_core_amx_1x1_conv_kernel.hpp | 77 const Xbyak::Reg64 ®_bias = r11; member 90 const Xbyak::Reg64 ®_scratch = reg_bias;
|
H A D | jit_gemm_inner_product_utils.cpp | 144 const Xbyak::Reg64 ®_bias = rbx; member 222 Xbyak::Address bias_ptr(const size_t offt) { return ptr[reg_bias + offt]; } in bias_ptr() 246 case arg_t::bias: return reg_bias; in get_reg_address() 813 if (this->do_bias()) add(reg_bias, offset * this->bias_data_type_size_); in compute_oc_channel_blk() 824 lea(reg_bias, ptr[reg_bias + offset * this->bias_data_type_size_]); in compute_oc_channel_blk() 841 lea(reg_bias, ptr[reg_bias + reg_oc * this->bias_data_type_size_]); in compute_oc_channel_blk() 1025 uni_vpbroadcastd(vmm_bias, ptr[reg_bias]); in compute_mb_blk() 1028 uni_vbroadcastss(vmm_bias, ptr[reg_bias]); in compute_mb_blk() 1112 mov(reg_bias, ptr[reg_param + PARAM_OFF(bias)]); in generate()
|
H A D | jit_sse41_conv_kernel_f32.cpp | 225 xword[reg_bias + sizeof(float) * ii * oc_blk]); in width_blk_step() 235 xword[reg_bias + sizeof(float) * ii * oc_blk]); in width_blk_step() 295 add(reg_bias, sizeof(float) * 4); in width_blk_step() 301 sub(reg_bias, sizeof(float) * 8); in width_blk_step() 359 if (jcp.with_bias) mov(reg_bias, ptr[this->param1 + GET_OFF(bias)]); in generate()
|
H A D | jit_uni_x8s8s32x_conv_kernel.hpp | 78 const Xbyak::Reg64 reg_bias = rdx; member 85 const Xbyak::Reg64 reg_icb = reg_bias;
|
H A D | jit_avx512_core_x8s8s32x_conv_kernel.hpp | 87 const Xbyak::Reg64 reg_bias = rdx; member 94 const Xbyak::Reg64 reg_icb = reg_bias;
|
H A D | jit_avx512_core_amx_1x1_conv_kernel.cpp | 292 mov(reg_bias, ptr[param1 + GET_OFF(bias)]); in store_output_vectors_int8() 294 auto bias_addr = EVEX_compress_addr(reg_bias, bias_offset); in store_output_vectors_int8() 391 mov(reg_bias, ptr[param1 + GET_OFF(bias)]); in store_output_vector_int8() 397 auto bias_addr = EVEX_compress_addr(reg_bias, bias_offset); in store_output_vector_int8() 443 mov(reg_bias, ptr[param1 + GET_OFF(bias)]); in store_output_vectors_bf16() 445 const auto bias_addr = EVEX_compress_addr(reg_bias, bias_offset); in store_output_vectors_bf16() 498 mov(reg_bias, ptr[param1 + GET_OFF(bias)]); in store_output_vector_bf16() 513 auto bias_addr = EVEX_compress_addr(reg_bias, bias_offset); in store_output_vector_bf16()
|
H A D | jit_avx512_core_x8s8s32x_deconvolution.hpp | 96 const Xbyak::Reg64 reg_bias = rdx; member 97 const Xbyak::Reg64 reg_icb = reg_bias;
|
H A D | jit_sse41_conv_kernel_f32.hpp | 54 reg64_t reg_bias = rbx; member
|
H A D | jit_avx512_common_conv_kernel.hpp | 70 reg64_t reg_bias = rdx; member 447 reg64_t reg_bias = rbx; member
|
/dports/math/onednn/oneDNN-2.5.1/src/cpu/x64/ |
H A D | jit_brgemm_post_ops.hpp | 72 const reg64_t reg_bias = r14; member 181 auto addr = ptr[reg_bias + bia_typesize_ * n_ * brg_.ld_block]; in loop_by_N() 190 auto addr = ptr[reg_bias + bia_typesize_ * n_ * brg_.ld_block]; in loop_by_N() 236 mov(reg_bias, ptr[param1 + GET_OFF(ptr_diff_bias)]); in generate() 244 add(reg_bias, bia_typesize_ * n_max_regs_ * brg_.ld_block); in generate() 353 const reg64_t reg_bias = r11; member 560 if (jcp.with_bias) mov(aux_reg_bias, reg_bias); in loop_by_N() 643 if (jcp.with_bias) mov(reg_bias, ptr[param1 + GET_OFF(ptr_bias)]); in generate()
|
H A D | jit_gemm_inner_product_utils.cpp | 140 const Xbyak::Reg64 ®_bias = rbx; member 219 Xbyak::Address bias_ptr(const size_t offt) { return ptr[reg_bias + offt]; } in bias_ptr() 245 case arg_t::bias: return reg_bias; in get_reg_address() 793 if (this->do_bias()) add(reg_bias, offset * this->bias_data_type_size_); in compute_oc_channel_blk() 804 lea(reg_bias, ptr[reg_bias + offset * this->bias_data_type_size_]); in compute_oc_channel_blk() 827 lea(reg_bias, ptr[reg_bias + reg_oc * this->bias_data_type_size_]); in compute_oc_channel_blk() 1010 case s32: uni_vpbroadcastd(vmm_bias, ptr[reg_bias]); break; in compute_mb_blk() 1011 case f32: uni_vbroadcastss(vmm_bias, ptr[reg_bias]); break; in compute_mb_blk() 1093 mov(reg_bias, ptr[reg_param + PARAM_OFF(bias)]); in generate()
|
H A D | jit_avx512_core_amx_1x1_conv_kernel.hpp | 77 const Xbyak::Reg64 ®_bias = r11; member 91 const Xbyak::Reg64 ®_scratch = reg_bias;
|
H A D | jit_sse41_conv_kernel_f32.cpp | 218 xword[reg_bias + sizeof(float) * ii * oc_blk]); in width_blk_step() 228 xword[reg_bias + sizeof(float) * ii * oc_blk]); in width_blk_step() 288 add(reg_bias, sizeof(float) * 4); in width_blk_step() 294 sub(reg_bias, sizeof(float) * 8); in width_blk_step() 352 if (jcp.with_bias) mov(reg_bias, ptr[this->param1 + GET_OFF(bias)]); in generate()
|
H A D | jit_uni_x8s8s32x_conv_kernel.hpp | 79 const Xbyak::Reg64 reg_bias = rdx; member 86 const Xbyak::Reg64 reg_icb = reg_bias;
|
H A D | jit_avx512_core_x8s8s32x_conv_kernel.hpp | 88 const Xbyak::Reg64 reg_bias = rdx; member 95 const Xbyak::Reg64 reg_icb = reg_bias;
|
H A D | jit_avx512_core_amx_1x1_conv_kernel.cpp | 300 mov(reg_bias, ptr[param1 + GET_OFF(bias)]); in store_output_vectors_int8() 302 auto bias_addr = EVEX_compress_addr(reg_bias, bias_offset); in store_output_vectors_int8() 409 mov(reg_bias, ptr[param1 + GET_OFF(bias)]); in store_output_vector_int8() 415 auto bias_addr = EVEX_compress_addr(reg_bias, bias_offset); in store_output_vector_int8() 461 mov(reg_bias, ptr[param1 + GET_OFF(bias)]); in store_output_vectors_bf16() 463 const auto bias_addr = EVEX_compress_addr(reg_bias, bias_offset); in store_output_vectors_bf16() 516 mov(reg_bias, ptr[param1 + GET_OFF(bias)]); in store_output_vector_bf16() 531 auto bias_addr = EVEX_compress_addr(reg_bias, bias_offset); in store_output_vector_bf16()
|
H A D | jit_avx512_core_x8s8s32x_deconvolution.hpp | 99 const Xbyak::Reg64 reg_bias = rdx; member 100 const Xbyak::Reg64 reg_icb = reg_bias;
|
H A D | jit_sse41_conv_kernel_f32.hpp | 54 reg64_t reg_bias = rbx; member
|
H A D | jit_avx512_common_conv_kernel.hpp | 70 reg64_t reg_bias = rdx; member 451 reg64_t reg_bias = rbx; member
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/mkldnn/src/cpu/aarch64/ |
H A D | jit_sve_512_x8s8s32x_conv_kernel.hpp | 100 const XReg reg_bias = x2; member 107 const XReg reg_icb = reg_bias;
|
H A D | jit_uni_dw_conv_kernel_f32.cpp | 54 add_imm(reg_tmp_addr, reg_bias, b_off * sizeof(float), in load_src() 214 if (jcp.with_bias) mov(reg_bias_stack, reg_bias); in compute_loop() 228 add_imm(reg_bias, reg_bias, bias_stride, reg_tmp_imm); in compute_loop() 243 if (jcp.with_bias) mov(reg_bias, reg_bias_stack); in compute_loop() 335 if (jcp.with_bias) { ldr(reg_bias, ptr(abi_param1, GET_OFF(bias))); } in generate()
|
/dports/math/onednn/oneDNN-2.5.1/src/cpu/aarch64/ |
H A D | jit_sve_512_x8s8s32x_conv_kernel.hpp | 100 const XReg reg_bias = x2; member 107 const XReg reg_icb = reg_bias;
|
H A D | jit_uni_dw_conv_kernel_f32.cpp | 54 add_imm(reg_tmp_addr, reg_bias, b_off * sizeof(float), in load_src() 214 if (jcp.with_bias) mov(reg_bias_stack, reg_bias); in compute_loop() 228 add_imm(reg_bias, reg_bias, bias_stride, reg_tmp_imm); in compute_loop() 243 if (jcp.with_bias) mov(reg_bias, reg_bias_stack); in compute_loop() 335 if (jcp.with_bias) { ldr(reg_bias, ptr(abi_param1, GET_OFF(bias))); } in generate()
|
/dports/math/libxsmm/libxsmm-1.16.3/src/ |
H A D | generator_mateltwise_avx_avx512.c | 2965 unsigned int reg_shift = 31, reg_bias = 30, reg_scale = 29; in libxsmm_generator_scale_avx512_microkernel() local 3097 reg_bias, 0, 1, 0 ); in libxsmm_generator_scale_avx512_microkernel() 3135 reg_n, reg_bias, reg_n ); in libxsmm_generator_scale_avx512_microkernel() 3222 reg_bias, 0, 1, 0 ); in libxsmm_generator_scale_avx512_microkernel() 3260 reg_n, reg_bias, reg_n ); in libxsmm_generator_scale_avx512_microkernel() 3315 reg_bias, 0, 1, 0 ); in libxsmm_generator_scale_avx512_microkernel() 3352 reg_m, reg_bias, reg_m ); in libxsmm_generator_scale_avx512_microkernel()
|