| 查看: 2044 | 回复: 4 | |||
| 【奖励】 本帖被评价2次，作者rlafite增加金币 1.6 个 | |||
[资源]
AkaiKKR
|
|||
|
AkaiKKR (MACHIKANEYAMA) is a program package used for first-principles calculation of electronic structures of metals, semiconductors and compounds, in the framework of the local density approximation or generalized gradient approximation (LDA/GGA) of the density functional theory. It exploits the KKR–Green’s function method. High speed, high accuracy and compactness are among its unique features. It is an all-electron method. It does not suffer from any serious truncation errors such as those of the plane-wave cutoff. Moreover AkaiKKR is combined with CPA (coherent potential approximation). Thus it is suitable not only for normal ordered crystals but also for disordered systems such as impurity systems, random substitutional alloys and mixed crystals. Since the method directly calculates the Green’s function of the system, it can also provide a good starting point for first-principles linear response theory, many-body theory, and so on. The package has been continuously developed since late 70th and is still being developed by various authors. Each program in the package is written in FORTRAN 77. The package is completely self-contained and does not need any additional libraries. It runs equally well on a small note PC and a large supercomputer. Any platforms such as UNIX, Linux, Mac OS and Windows wherein the Fortran compiler is installed can be used. |
回复此楼» 本帖附件资源列表
-
欢迎监督和反馈：小木虫仅提供交流平台，不对该内容负责。
本内容由用户自主发布，如果其内容涉及到知识产权问题，其责任在于用户本人，如对版权有异议，请联系邮箱：xiaomuchong@tal.com - 附件 1 : cpa2002v010_AkaiKKR.tgz
2020-10-21 04:23:02, 199.95 K
» 猜你喜欢
2025冷门绝学什么时候出结果
已经有3人回复
-
天津工业大学郑柳春团队欢迎化学化工、高分子化学或有机合成方向的博士生和硕士生加入
已经有4人回复
-
康复大学泰山学者周祺惠团队招收博士研究生
已经有6人回复
-
AI论文写作工具：是科研加速器还是学术作弊器？
已经有3人回复
-
孩子确诊有中度注意力缺陷
已经有6人回复
-
2026博士申请-功能高分子，水凝胶方向
已经有6人回复
-
论文投稿，期刊推荐
已经有4人回复
-
硕士和导师闹得不愉快
已经有13人回复
-
请问2026国家基金面上项目会启动申2停1吗
已经有5人回复
-
同一篇文章，用不同账号投稿对编辑决定是否送审有没有影响？
已经有3人回复
|
Akai-KKR Here, we will explain how to draw the density of states (DOS) and band diagram using AkaiKKR, which is an all-electron calculation method. * AkaiKKR = Machikaneyama -------------------------------------------- ---------------------------------- ■ How to use Akai-KKR (learning method) How to use is "Computer Material "Introduction to Design" is often written in the Osaka University Press. You can also study by yourself. Don't be disappointed in the CMD Beginners course as you will learn the same content as a book unless you ask a question. If the videos and descriptions below are sufficient, and unless you intend to ask Dr. Akai, Dr. Ogura, and Dr. Sato, the value of participating in CMD for Akai-KKR will be greatly reduced. ・ Youtube video: https://www.youtube.com/playlist ... ipTNqiz_1VwWaEkzHpF -------------------------------------------------- ---------------------------- ・ "Negati Blog ( https://gomisai.blog75.fc2.com/blog- " entry-600.html ) "is wonderful. ・ Notes about AkaiKKR · GitHub ( https://gist.github.com/t-nissie/4edcf8d62144b4d77f5c ) is wonderful. -------------------------------------------------- ---------------------------- ■ Difference between Akai-KKR and SPR-KKR (27 / Sep / 2016) ◇ Recommended code is ?? ・ Akai-KKR is better to handle first. This is because it is easy to compile, it has good calculation convergence, and the band distribution calculation is stable. ・ SPR-KKR is for those who want to do maniac calculations. Since the full potential method can be used, it is suitable for calculating crystal structures with many gaps that allow hydrogen atoms to enter. (As with Akai-KKR, when using muffin-tin potential, you can specify an atom with atomic number 0 in the gap where a hydrogen atom is likely to enter to improve the calculation accuracy. GaAs, ZrNiSn, etc. Typical example) * Note that there are many things that you cannot understand about SPR-KKR just by reading the manual. You will also have to struggle to find out the content of the seminar and consult with the developers. ◇ Akai-KKR and SPR-KKR can be done ・ CPA can be calculated (system with element substitution in unit cell can be calculated) ・ SCF and DOS can be calculated ・ Total energy and spin moment can be calculated Obtained data Phase stability and formation energy can be calculated from ◇ Akai-KKR is good ・ Convergence is stable ・ Band dispersion calculation is stable ◇ SPR-KKR is possible ・ Non-equivalent atomic position (Akai) natm the corresponding) a dozen or more in -KKR possible input (Fe2VAl of natm 4 so Akai-KKR even have enough) make the input file from the space group as-WIEN2k possible full potential calculation , LDA + U, GGA + U can be calculated ・ Many physical properties can be calculated (XAS, XPS (core-level and valence band), ARPES, XES, MXO, AES) Residual Resistivity of Alloys, Spin- and Orbital Moments, etc. http: //ebert.cup.uni-muenchen.de/index.php?option=com_content&view=article&id=8&catid=4&Itemid=7 Or see the manual at https://ebert.cup.uni-muenchen.d ... =51&lang=en . ・ Calculation of clusters and interfaces is possible ------------------------------------------ ------------------------------------ ■ What should I think about k points? "Negative log" [AkaiKK's of Please look because the person how to use] is who examined vigorously than I am. "Negati Burogu" is really grateful. I would be grateful if you could find out some deep points such as how to deal with Killed. About bzqlty: I hear that about 1 meV / atom is enough for total energy, so bzqlty 4 may not be enough for single metal calculation. It is said that the formation energy predicts the experimental value within about 0.1 eV, so I think it is enough if it can be calculated up to 0.01 eV. The longer the lattice constant, the smaller the number of k-point divisions. For example, Fe 2 VAl is 5.67 Å, so if Fe is 2.87 Å, Fe 2In VAl, I think that the calculation will be as accurate as bzqlty 8 from 4 * 5.67 / 2.87 = 7.9 (more k if the special k point {Monkhorst-Pack method} is taken in consideration of the symmetry due to the crystal structure. The score is low). However, the accuracy of single metal calculations is a concern. I calculated Fe with bzqlty 4 and applied it to Manahan's equation of state. If the calculation error is large and you do not get on the curve, you will be in trouble, but there seems to be no problem. Since it is calculated by sdftyp pbe, the lattice constant tends to be long. The difference from the experimental value is (2.881 / 2.866-1) * 100 = 0.52%, which is within 3%. Condition (reltyp nrl) Lattice constant [bohr] Total energy [Ry] Spin moment [μ B / atom] go: bzqlty 4 5.4400 -2522.8160225 2.36864 go: bzqlty 4> dos: bzqlty 8 5.4400 -2518.1759027 0.37377 go: bzqlty 8 5.4400 -2522.8159776 2.32490 go: bzqlty 16 5.4400 -2522.8160602 2.34245Difference in spin moment between bzqlty 4 and 16: (1-2.36864 / 2.342445) * 100 = -1.11% Since it is not different by a dozen percent, I think that this difference is okay. -------------------------------------------------- ---------------------------- □ Muffin-tin (MT) Potential Problems ・ What happens if you distort a little? It is vulnerable to the problem and substances with high anisotropy. [T1] □ CPA (coherent potential approximation) ・ There is an advantage that impurity calculation can be performed without using the supercell method. [T1] -Since CPA adopts the potential of averaging the arrangement of all defects, a system with a very thin defect concentration cannot be described correctly. Moreover, since it is difficult to calculate the force, the effect of relaxing the local atomic position due to the occurrence of the defect cannot be incorporated. [T3] ・ It can be used not only in the dilute limit. KKR-CPA (or generally the CPA method) is a method developed to handle irregular systems with finite concentrations (including the dilute limit), and is based on the single site approximation not only for dilute systems but also for any concentration. Is the best approximation to the placement mean (best means best in a variational sense). In the case of the dilute limit, it is not necessary to use CPA and it can be treated as an impurity problem of KKR. [T4] □ Impurity calculation ・ If the impurity concentration is set to 0, the impurity calculation is the limit of dilution. In impurity calculations, the presence of impurities does not affect the electronic state of the host element. However, the ewidth setting must cover the valence bands of both impurities and host elements. [T5] (In Si, it seems that it is better to insert a hole Vc and perform an ASA calculation. In DOS, if bzqlty is set to 22, the shape is reproduced well) [T1] https://pmt.sakura.ne.jp/ wiki / images / Cmdtxt1.pdf [T2] https://ann.phys.sci.osaka-u.ac.jp/jp/re_sub1.html [T3] https://www.murata.co.jp/zaidan/ annual / pdf / k01 / 2004 / a31130.pdf [T4] https://kkr.issp.u-tokyo.ac.jp/bbs/thread.php?id=433 (Akai-KKR BBS (new ver.)) ) [T5] https://gomisai.blog75.fc2.com/blog-entry-679.html#comment363 ------------------------- -------------------------------------------------- --- Basic usage of AkaiKKR * case is a simple name of the calculation system given by the other side (in my case, it is often the abbreviation or composition of the calculation system) * Notation: This review is mainly for "cpa2002v009c.tar.gz" around 2015. !!! ■ Unzipping and installing AkaiKKR (Intel fortran case) 1. Send an email to Akai Group to get AkaiKKR data. 2. Enter tar zxvf cpa2002v * .tar.gz on tarminal and decompress it. x and y are different in version. 3. cd cpa2002v * 4. touch source /*.f 5. make 6. ifort source / gpd.f -o gpd * https://act.jst.go.jp/content/h13/scom_net/S01/PageMain. html * It means to update the history of touch files. * As another way to deal with the case where the old object file (file with .o) remains, add the following to the end of the makefile (Tab before rm), enter make clean, and then make Enter. clean: rm -f -r $ (objs) * ifort compiler options -openmp: Allows the compiler to generate multithreaded code based on the OpenMP * directive (same as -fopenmp). -openmp-stubs: Compile OpenMP * programs in sequential mode. The OpenMP * declarer is ignored and the OpenMP * stub library is linked. (Sequential) -pc80: Sets the internal FPU precision to a 64-bit mantissa. (Default) -mp1: Increases floating point precision (has less impact on speed than -mp). -mieee-fp: Same as -mp. It can be disabled with -mno-ieee-fp. -i-dynamic: Dynamically links Intel-provided libraries. Use -shared-intel. -mcmodel = medium: Tells the compiler to limit the code to the first 2GB. The data is not restricted. -O2: Optimizes processing speed to the maximum. (Default) -[no-] Enables / disables single-file IPOs in ip files. ■ Unzipping and installing AkaiKKR (gfortran case) 1. Send an email to Akai Group to get AkaiKKR data. 2. Enter tar zxvf cpa2002v * .tar.gz on tarminal and decompress it. x and y are different in version. 3. cd cpa2002v * 4. touch source /*.f 5. gedit makefile omp = -fopenmp nomp = -openmp-stubs fort = gfortran flag = -mpc80 -mieee-fp -m64 -O2 -march = native -funroll-loops 6. make 7. gfortran source / gpd.f -o gpd * http: //act.jst.go.jp/content/h13/scom_net/S01/PageMain.html * It means to update the history of touch files. * As another way to deal with the case where the old object file (file with .o) remains, add the following to the end of the makefile (Tab before rm), enter make clean, and then make Enter. * If that doesn't work, try nomp = as well. The -dy option may result in an error. clean: rm -f -r $ (objs) * gfortran compiler options -mpc: Set 80387 floating-point precision (-mpc32, -mpc64, -mpc80) -mieee-fp: Use IEEE standard for floating point comparison -Bdynamic, -dy, -call_shared: Link against shared libraries -m64: Generate 64bit x86-64 code -mavx: Support MMX, SSE, SSE2, SSE3 , SSSE3, SSE4.1, SSE4.2 and AVX built-in functions and code generation * The option of the function similar to intel is gfortran Then, m may be the first one. ■ intel: gfortran options -xHost: -march = native -ipo: -flto -no-prec-div: -ffast-math -mp (= -mieee-fp): -mieee-fp -assume noprotect_parens: -fno-protect -parens (default): -fstack-arrays -fast (= -xHOST -O3 -ipo -no-prec-div -static): -march = native -O3 -flto -ffast-math -static -funroll-loops:- funroll-loops- [no-] unroll-aggressive: -funroll-all-loops [1] https://thatcadguy.blogspot.jp/2 ... -intel-fortran.html [2 ] ]https://www.hpc-sol.co.jp/support/intel_compiler_recipes/20121212_optimization.html [3] https://www.softek.co.jp/SPG/Pgi/TIPS/option_for_PDF.html [4] http: //www.cqpub.co.jp/interface/column/freesoft/2002/200208/08-4.htm [5] https://www.isus.jp/article/comp ... ntel-compilers-101/ ■ SCF Calculation □ Basic system calculation * For details , read https://kkr.phys.sci.osaka-u.ac.jp/pdf/machikaneyama2000_memo.pdf . 1. Create the input file of the system you want to calculate by referring to the example of the input file in in and the input file written at the bottom of this HP. 2. The place to rewrite is as follows. * 1 ･ Change the name of the potential file next to go to data / case. ･ Enter the symbol corresponding to the crystal structure calculated by brvtyp. ･ A and c / a are the minimum that can express the structure. a is a bohr unit. When 0 is input, the volume (au) of all atoms in MJW (qvolum.f) is multiplied by the value of conc, summed, and the result of multiplying it by (1/3) is used (calculated by chklat.f). ). ･ Ewidth is about 1.0 to 1.5. If you cannot calculate it well, try increasing or decreasing the value. ･ Reltyp can be non-relativistic nrl or scalar relativity (that is, relativistic calculation with reduced spin-orbit interaction) sra. If in doubt, use sra. In the case of MCD, relativity calculation with srals (z direction is the quantization axis {spin is aligned with the z axis direction} {spin-orbit interaction}). ･ Bzqlty is equivalent to k points and should be about 4. ･ For lines where ntyp is the type of atom and type, describe only the lines described by ntyp (the amount of description increases as shown below for impurities and doping). Any name may be used. • The maximum value of l that mxl considers in the muffin-tin sphere. ･ Anclr is the atomic number ･ conc is the concentration or occupancy rate ･ natom is the number of lines that describe the coordinates ･ atomicx describes the atomic coordinates for the a, b, and c axes. Normally, the length of the a-axis is the basic unit. If you want to correspond to the length of each vector, write 0.5a 0.5b 0.5c. -Atomtyp should be the same as the name described in type. 3. SAVE after rewriting. 4. export OMP_NUM_THREADS = number of CPUs 5. specx <in / case 6. Enter specx <in / case many times until err = is about -6.0 and repeat. 7. total energy = will be displayed, so make a note of it. * 1 Priority is given to the description of brvtyp in the input file, and c / a etc. are changed to the optimum values on the AkaiKKR code side. * 2 err is a simple concept that can be thought of as err ≒ log√ [Σ (Vin-Vout) ^ 2]. * The description with c at the beginning is ignored as a comment. * There is also mjwasa in sdftyp. The ASA is a larger sphere than the MT sphere, and the sphere at the ASA is set to match the volume of the unit cell to be calculated. In the MT sphere, a part with a flat potential appears, but in the ASA, it is relaxed and may give good results. □ Calculation when impurities are added (conc is set to 0) 1. In the input file created by the basic system calculation, set ncmp to the value of the number of impurities to be added + 1. 2. Enter anclr (atomic number of impurities) and conc (concentration) on the next line immediately after the number of ncmps. 3. An example of adding Ir as an impurity (0%) to Fe is shown below. c ------------------------------------------------- ----------- c type ncmp rmt field mxl anclr conc Fe 2 1 0.0 2 26 100.0 77 0.0 c ---------------------------------------- -------------------- □ Calculation when doping with other elements (conc is not set to 0) 1. Input created by basic system calculation In the file, set ncmp to the value of the number of impurities +1. 2. Enter anclr (atomic number of impurities) and conc (concentration) on the next line immediately after the number of ncmps. 3. An example of adding 1% of Ir to Fe is shown below. c ------------------------------------------------- ----------- c type ncmp rmt field mxl anclr conc Fe 2 1 0.0 2 26 99.0 77 1.0 c --------------------- --------------------------------------- □ Calculation by CPA G ~ = g 0 [1-t ~ g 0 ] -1 G A = G ~ [1- (t A -t ~) G ~] -1 G B = G ~ [1- (t B -t ~) G ~] -1 c * G A + (1-c) * G B = G ~ * t ~ is called an effective medium. ~ Is superscripted, but for some reason it is not displayed well. □ Calculation with LMD 0. Perform SCF calculation with ferromagnetism (normal SCF calculation is sufficient). 1. cd util 2. Create a file called case.fmg and write as follows. The following is an example when there are 4 components, and the spins of 1 and 3 are reversed. Write as many numbers as there are original components. ../data/case 1 2 3 4 ../data/case_lmd 1 -1 2 3 -1 4 The first is the potential information obtained by normal SCF calculation, and the second line is the component for which you want to reverse the spin. Write -1 immediately after the number that describes. In this example, the spins of atoms 1 and 3 are reversed. You may rewrite it with the same file name. 3. ./fmg case.fmg 4. cd .. 5. cp in / case in / case_in_lmd 6. Rewrite the potential file data / case specification in the input file case_in_lmd to the data / case_lmd created earlier as follows. .. c ---------------------- case -------------------------- -------- go data / case c ------------------------------------- Make ----------------------- as follows. c ---------------------- case -------------------------- -------- go data / case_lmd c ------------------------------------- ----------------------- 7. specx <in / case_lmd> out / case_out_lmd * After cd util, gfortran -o fmg fmg.f etc. Type to create an executable of fmg. You only have to do this once. * See cat data / case.info. The data is described in the order of lattice constant, total energy, and total spin moment. ■ DOS calculation 0. Perform SCF calculation. 1. Copy the input file created by SCF. Example cp in / case in / case_in_dos 2. Change calctyp from go to dos in the input file case_in_dos. Then, increase bzqlty to 8 to 10, which is larger than that for SCF calculation. 3. export OMP_NUM_THREADS = 1 4. specx <in / case_in_dos> out / case_out_dos 5. The result is output to case_out_dos. When the calculation considering spin up and down is performed, the up data is displayed first, and then the down data is displayed. 6. In case_out_dos, the output is as follows in the order of each component described in case_in_dos. * Energy, s orbital, p orbital, d orbital ... 7. If you enter ./gpd out / case_out_dos, Total DOS will be displayed. * To divide into small components such as px, py, pz, it is possible by rewriting the code. When calculating the doped system, various scatterings increase, so it is not a mistake to see sharp peaks in DOS. * If the DOS of up and down is the opposite of the empirical result, calculating SCF again will switch up and down. You may rewrite up → down and down → up when plotting. * If you match the shape to Total DOS of WIEN2k, Akai-KKR may match relatively well with msex = 1001, edelt = 0.0036 using GGA91 (msex in gpd.f must also be changed and made./gpd Note that does not work). ■ Band variance 0. Calculate SCF. 1. cp in / case in / case_spc 2. Rewrite calctp from go to spc in case_spc. 3. Add k points to the end of case_spc and save. Refer to Reference [1, 6] for an example of inputting k-points. 4. specx <in / case_spc 5. cd data 6. The result is described with a name with .spc such as case_spc_up.spc or case_spc_dn.spc (_up.spc or _up.spc or the character after data / written in the input file). The data (with _dn.spc) is described as follows. A (E, k) is the Bloch spectrum function. Point k, energy, A (E, K) 7. A (E, k) has some extent depending on the degree of scattering of the system. These data may be plotted with plotting software or the like. ･ Simple plotting method using gnuplot 1. gnuplot 2. set pm3d map 3. splot "case_spc_up.spc" ･ More detailed usage of gnuplot 1. gnuplot 2. set xrange [0.0: 1.1] 3. set yrange [-0.3: 0.3] 4. set ylabel "Energy / eV" 5. set title "case" 6. set pm3d map 7. splot "case_spc_up.spc" If you want to make it monochrome: set palette gray If you want to restore the scale: If you want to redraw after entering the set autoscale command: replot If you want to do simple numerical calculation: 1: ($ 2 * If you write an expression in parentheses with $, such as 13.6) or ($ 2 * $ 3), you can use the calculated value. When to exit: q References on gnuplot: https://www.cse.kyoto-su.ac.jp/~ ... nuplot/gnuplot.html https://www-tlab.math.ryukoku .ac.jp/wiki/pukiwiki.php?gnuplot https://www.cn.kagawa-nct.ac.jp/ ... li_gnuplot_tgif.pdf https://kkr.phys.sci.osaka- u.ac.jp/bbs/view_msg.cgi?id=66 * If you want to draw the variance for each component, set bzqlty to 0 and rewrite the item related to line 50 in bzmesh.f. By doing so, PDOS at the k point specified by vkp (r, 1) can be obtained. ■ K-edge XMCD calculation [6] 1. Set calctyp to go, reltyp to srals, and SAVE with a name such as input_go 2. specx <in / input_go 3. Rewrite go in calctyp to mcd and name such as input_mcd SAVE 4. Rewrite data ref / 0.75d0 / in source / cemesr.f to data ref / 0.0d0 /, SAVE, and enter make to compile. * 1 5. specx <in / input_mcd> out / output_mcd 6. In the obtained output_mcd, the line xmd of Component appears. Each Component is output in the order of input_mcd. Each column is output in the order of energy, absorption, absorption (up), absorption (down). 7. Copy and paste the required Component data and SAVE with the name xmd_input 8. If you have xmdcnv from Akai Group, type xmdcnv and Terminal, then type xmd_input, then xmd_output. * 2 9. Enter gnuplot to start gnuplot. 10. Type plot "xmd_output" to get the result. 11. To exit, close the window and type quit on Terminal. 12. Change data ref / 0.0d0 / in cemesr.f back to data ref / 0.75d0 /, save, and type make to compile. * 1. ref specifies the ratio from the inner shell to EF of ewidth, which specifies the range from the inner shell to the unoccupied level. If ref is the default of 0.75, then ewidth * ref from the inner shell to EF. When ref is set to 0, the range from EF to the unoccupied level becomes ewidth. * 2. Enter gfortran -o xmdcnv xmdcnv.f on the Terminal to create an executable file named xmdcnv. xmdcnv broadens the obtained result with a low lentian. You may make your own program. ■ Charge density distribution Although it is conceptually possible, it is difficult to output the "charge density distribution" because it is necessary to rewrite the part of the program called green. Since it uses muffin-tin potential, it should be noted that it is not full potential. In the personal opinion of this HP administrator, if the Valence is limited, increase the number of k points drawn in the band diagram and plot the Bloch spectrum function from valence to unoccupied level. You may be able to draw if you go. If you draw the vicinity of EF and the unoccupied level, you may be able to discuss the contrast with MEM, electrical conductivity, and STM. [7] References [1] "Introduction to Computer Material Design", Osaka University Press [2] https://pmt.sakura.ne.jp/wiki/index.php?title=Doument for AkaiKKR [3] http: // kkr.phys.sci.osaka-u.ac.jp/jp/document.cgi [4] https://www.dyn.ap.eng.osaka-u.a ... ACHIKANEYAMA2000_1. pdf [5] https://olymp.cup.uni-muenchen.d ... =51&lang=en [6] https://www.ghfecher.de/Run_AKAI-KKR. pdf [7] https://kkr.phys.sci.osaka-u.ac.jp/bbs/thread.cgi?id=256 [8] https://gomisai.blog75.fc2.com/?tag=AkaiKKR #container |
2楼2020-10-21 04:49:10
|
nvidia gpu supported akaikkr (machikaneyama)原代码 akaikkr（machikaneyama）是一个程序包，用于在密度泛函理论的局部密度近似或广义梯度近似（lda / gga）的框架内对金属，半导体和化合物的电子结构进行第一性原理计算。 它利用了kkr-green的函数方法。高速，高精度和紧凑性是其独特的功能。这是一种全电子方法。它不会遭受任何严重的截断错误，例如平面波截止的错误。此外，akaikkr与cpa（相干势近似）结合在一起。因此，它不仅适用于普通有序晶体，而且适用于无序体系，例如杂质体系，无规替代合金和混合晶体。由于该方法直接计算系统的格林函数，因此它还可以为第一原理线性响应理论，多体理论等提供良好的起点。 自70年代末以来，该软件包就一直在不断开发，至今仍由各种作者开发。程序包中的每个程序都是用fortran 77编写的。程序包是完全独立的，不需要任何其他库。它在小型笔记本电脑和大型超级计算机上同样运行良好。可以使用安装了fortran编译器的任何平台，例如unix，linux，mac os和windows。 please email [email]zrli3 AT hotmail.com[/email] if you wanted this "nvidia gpu supported akaikkr (machikaneyama)"原代码 |
5楼2020-10-31 01:27:42
简单回复
材料廖3楼
2020-10-21 09:35
回复
五星好评 顶一下，感谢分享！
snail5944楼
2020-10-21 17:08
回复
五星好评 顶一下，感谢分享！