LONDON A clone of the ARM7 32-bit RISC processor core, previously available free for download from the Internet, has been taken down or hidden pending discussions between the core's designer and a Chinese representative of ARM Holdings plc (Cambridge, England).
Meanwhile, ARM has decided not to offer a lightweight version of the ARM processor for students and university researchers. The existence of an officially approved "ARM-lite" processor might have dissuaded some students from attempting to clone ARM instruction-set architectures but could have disrupted the commercial market, an ARM executive said.
Since July, an ARM clone in the form of a synthesizable Verilog language description, designed by graduate student Shengyu Shen, had been available at the OpenCores Web site. At press time, however, the page stated only that "the team members of nnARM are currently [in discussions] wi th a company to sort out some issues" and that the Web page would not be available until further notice. It is believed Shen has taken down or hidden the nnARM clone at the OpenCores Web site.
The "issues" to be sorted out with ARM likely include technology and potential patent infringement, but to date no details have emerged. Shen, a graduate student at the National University of Defense Technology at Changsha, in China's Hunan province, is listed as the maintainer of the project.
According to sources, a personal meeting or teleconference between Shen and ARM representatives is planned that will likely decide the fate of the core. It is believed that a Chinese representative of ARM is trying to set up a meeting between Shen and Matt Lee, university program manager at ARM.
"I have spoken to our Chinese representative, but I haven't yet been invited to [participate in a conference]," said Lee. Asked why ARM sought to confer with Shen, Lee said, "He's a very bright chap, so we want to talk to him ."
When documentation on the nnARM was publicly available, the core was described as "not an ARM clone" but was claimed to be a near clone of the ARM7.
OpenCores, of which Shen is a member, is a loosely knit global organization of hobbyists, students and professional engineers that aims to produce open-source synthesizable circuit descriptions. Its Web site shows more than 50 hardware cores as "projects" and lists 26 as complete. They include processors, peripherals and cryptographic cores.
OpenCores has an on-chip bus policy, based on the Wishbone interconnect standard. Wishbone was developed by Silicore Corp. (Minneapolis), which placed the specification into the public domain in January.
Clones are unlikely to be commercially significant, ARM's Lee said, because while they might be able to run some compiled code they cannot be guaranteed to run all compiled code.
"It's unlikely that clones will be complete. The easy bits can be done quickly, but completeness and full compatibility are hard," he said.
Earlier this year, when the nnARM first became available, Shen said the core could run all ALU, multiply, multiply-accumulate and program-status register instructions, as well as support all single-data transfer instructions, branches and certain forms of conditional execution. But the core did not support interrupt and exception handling, single-data swaps, block-data transfers or coprocessor instructions. Neither did the nnARM support the Thumb 16-bit version instruction set.
Shen reportedly had planned to improve the core and implement it in an FPGA.
Meanwhile, ARM executives have said that the company is supportive of academic research around the ARM architecture provided the research is carried out within an appropriate legal framework but that it does not take commercial cloning of the ARM architecture lightly.
"We are licensing ac ademics to allow them to pursue research, providing [them with] software simulators, licenses for limited manufacture, formal methods for modeling and phantom layouts with detail omitted," said Lee. "We had even batted around the idea of an 'ARM-lite' core for university research" a stripped-down ARM core that would have been suitable for student research projects but would have omitted many refinements needed in a robust commercial core.
But Lee said ARM has decided not to proceed with the ARM-lite core. Though such an offering might have guarded against students' being tempted to develop their own clones, it would have risked confusing the commercial market, the company concluded.
This year, three students at Malardalen University (Vasteras, Sweden) said they had developed a lightweight ARM-style processor core named BlackARM and had used the design for a reconfigurable-multiprocessing project. But the core description was not ma de available, and the trio has since moved on to the Royal Institute of Technology (Stockholm, Sweden) to study low-power computer architectures and operating systems.