Riding the wind of rebirth

"This is a very insightful point of view." Zhou Zhi nodded to express his understanding: "But when upgrading from bits to blocks, a problem will inevitably occur."

"The memory grid on each integrated circuit has an upper limit on the number of times it can be erased and written. In the past, it was erased bit by bit. If one was damaged, it would jump to another one to continue. In this way, for the storage space of the chip, If the bad one skips the other one, there’s no big problem.”

"Now that bit reading and writing has become block reading and writing, the problem becomes that if a grid is damaged, a data block cannot be used. For the life quality requirements of flash memory, this kind of waste is too big."

"There is indeed this problem." Fujio Shigaoka frowned: "But we have also proposed a solution, which is to use the blocks as evenly as possible and use a random algorithm to access the next data block that can be read and written. Avoiding excessive reuse of the same data block solves a large part of the problem.”

"Indeed, from our Clover analysis of Toshiba NAND flash memory, we also found that the algorithm invented by Mr. Shuoka is quite clever and compromise." Zhou Zhi smiled: "But Mr. Shuoka has to admit that this It’s just a solution that only treats the symptoms but not the root cause. It will still happen that a point is broken and a block cannot be used. It is not the ultimate solution.”

"I'm sorry Zhou Sang, I've caused you trouble!" The problems pointed out by Zhou Zhi are indeed problems that the Shuogang team has not been able to solve in the first generation of NAND flash memory: "The problems Zhou Sang mentioned do exist, and we were originally planning to improve them, but now... …”

"But now even the NAND team is gone, right?" Matsui said.

"This... there is still a basic guarantee..." Fujio Sugaoka's face showed some embarrassment: "As far as I know, Toshiba is still planning to use the benefits of DRAM to repay NAND and maintain the ability to supply chips to Clover. …”

However, these words revealed that he was quite unconfident.

"If you have funds and a team, how will Mr. Shuogang solve this problem?" Zhou Zhi asked: "I'm asking about ideas."

"There are several ideas." Shuoka said: "The first is to increase the area of ​​the grid and reduce the flipping and drift of charges. This will not only make data storage more accurate, but also reduce the number of invalid erase and write times and increase the life of flash memory."

"This will reduce the most important volumetric performance of the chip. This is contrary to Mr. Shuokang's idea of ​​'turning bits into blocks' to increase volumetric performance." Zhou Zhi shook his head and said: "In this way, the performance of the chip has not been improved too much. There are more, but the previous advantage is gone.”

"The other thing is to modify the block storage algorithm and make the data blocks smaller."

"The data block becomes half the size and the addressing area doubles?" Zhou Zhi asked.

Fujio Shigaoka also shook his head slightly. This method will also lose a huge amount of storage space to store the addressing area. The idea is not much better than the previous method.

"The rest... then we can only use our brains on the addressing algorithm." Fujio Shioka said: "But such research and development requires a lot of investment. Now..."

"Mr. Shuoka, please look at this." Zhou Zhi took out an A4 paper with a code printed on it from his bag and handed it to Fujio Shuoka.

"This... this... this is the floating address temporary allocation algorithm! This is to achieve balanced erasing and data block reorganization through dynamic addresses!" Fujio Shigaoka took it and took a look. He was suddenly very excited, and then turned over the paper, disappointed. Suddenly, I discovered that there was a blank space at the back of the A4 paper: "Where is the rest of the code?!"

Zhou Zhi took the paper back in time and put it back in the bag: "The remaining code is in the company. If Mr. Shuoka is willing to become the technical director of our company's flash memory R&D department, he can have access to our improvement plans at will. "Now I can give Mr. Shuogang a general outline of our thinking." Zhou Zhi knows that for people who are addicted to alcohol, it is best to let them smell the smell of alcohol but not drink it, or even see it. Annoying: "The idea of ​​our algorithm is that only those BLOCKs that actually write data will get a temporarily allocated data address, and for the remaining spaces, the system will treat them as 'to be allocated' status, and It’s not like today’s flash memory, which is already divided.”

"When the data enters the cache, the algorithm will calculate the required storage space based on the size of the data, then temporarily calculate the addresses for these spaces, issue the addresses to these BLOCKs, and then write the data in the cache into it."

"Similar to the reverse principle, when data is erased, these blocks will be released, and the addresses they originally occupied will be reclaimed by the system, and the released space will return to the space pool to be allocated, waiting for the next combination and allocation. "

"If there are bad points, then these points will be discarded the next time the data block is reorganized. This will achieve the maximum utilization and achieve the same performance as EEPROM, while continuing to maintain the greatest advantage of flash memory."

"But this algorithm also needs to be stored separately, and it is always in the calculation logic." Fujio Shigaoka fell into thinking: "So how to solve the problem of the algorithm chip?"

"With ROM, the algorithm is solidified in the ROM, which is very cheap." Zhou Zhi said: "When turning on the computer, NAND will self-check the available space, and divide an addressing algorithm module in the space to transfer the addressing algorithm in the ROM. Boot into NAND and let NAND do the rest."

"Wonderful!" Fujio Shioka slapped his thigh: "Add a boot module to FLASH! But this also requires an operating instruction set... Hey! This is the strength of your Clover company!"

"This set of algorithms is still on the MiniCos system, and we have just passed the feasibility verification." Zhou Zhi said sincerely: "It is necessary to solidify it into ROM. This is Mr. Shuoka's strength, so our company wants to We have hired Mr. Wang as the technical director of our flash memory division to be responsible for this matter."

"That's it...but my invention and intellectual property rights belong to Toshiba." Saigaoka said: "If I continue to research, it will inevitably conflict with Toshiba's IP core, which may lead to huge intellectual property litigation..."

"Please rest assured, sir." Zhou Zhi said with a smile: "We have entrusted the Intellectual Property Division of AXA Fund to negotiate with Toshiba about the IP licensing of Mr.'s important inventions in the company. So far, Mr. Toshiba has a total of more than 500 inventions. AXA Fund has bought all the invention patents."

"As long as Mr. comes over, we can continue the original research and development. From hardware to software, I will provide him with everything."

"You...have you even thought of this?" Fujio Shioka asked with some anxiety: "I have a question I would like to ask. Of course, if Zhou Sang finds it difficult to answer, forget it."

"Mr. Shuoka, just ask."

"I would like to know how much it cost AXA Fund to obtain these authorizations from Toshiba?" asked Fujio Sugaoka.

"This amount may disappoint Mr. Shuogang." Zhou Zhi said: "AXA only spent less than 100,000 US dollars in total."

Sumoka's chest rose and fell several times, then he grabbed the small wine glass Ikedaoka put on the table and drank it in one gulp, with red eyes: "I have decided! Join Clover and continue to develop flash memory! I want to vent this bad breath!"

"Ah! Why is this wine so spicy!"