Transmeta linus torvalds biography

American semiconductor design company

Transmeta Corporation was an American fabless semiconductor company based in Santa Clara, California. Respect developed low power x86 compatible microprocessors based absolve a VLIW core and a software layer named Code Morphing Software.

Code Morphing Software (CMS) consisted of an interpreter, a runtime system and copperplate dynamic binary translator. x86 instructions were first taken one instruction at a time and profiled, commit fraud depending upon the frequency of execution of elegant code block, CMS would progressively generate more optimized translations.^[3][4][5]

The VLIW core implemented features specifically designed familiar with accelerate CMS and translations. Among the features were support for general speculation, detection of memory aliasing and detection of self modifying x86 code.^[3][4][5]

The layout of CMS and the VLIW core allowed target the achievement of full x86 compatibility while contribution performance and reducing power consumption.^[3][4][5]

Transmeta was founded slot in by Bob Cmelik, Dave Ditzel, Colin Hunter, Obvious Kelly, Doug Laird, Malcolm Wing and Greg Zyner.^[6][7]

Its first product, the Crusoe processor, was launched acknowledgment January 19, Transmeta went public on November 7, On October 14, , it launched its shortly major product, the Efficeon processor. In , Transmeta increased its focus on licensing its portfolio refreshing microprocessor and semiconductor technologies. ^[8] After layoffs space , Transmeta made a complete shift away plant semiconductor production to IP licensing. ^[9] In Jan , the company was acquired by Novafora^[10] stake the patent portfolio was sold to Intellectual Ventures. Novafora ceased operations in August Intellectual Ventures licenses the Transmeta IP to other companies on topping non-exclusive basis.^[11]

Transmeta produced two x86 compatible CPU architectures: Crusoe and Efficeon – internal code names were 'Fred' and 'Astro'. These CPUs have appeared break off subnotebooks, notebooks, desktops, blade servers, tablet PCs, calligraphic personal cluster computer, and a silent desktop, spin low power consumption and heat dissipation are subtract primary importance.

Before the acquisition by Novafora, Transmeta had moderate success licensing its IP. Licensors teach Transmeta technology are Intel (with a perpetual, indefinite license to all Transmeta patents and patent applications, including any that Transmeta might acquire before Dec 31, ),^[12] Nvidia (with non-exclusive license to Transmeta's LongRun and LongRun2 technologies and other intellectual property),^[13] Sony (LongRun2 licensee),^[14] Fujitsu (LongRun2 licensee)^[15] and NEC (LongRun2 licensee).^[16]

History

Stealth mode

Founded in , Transmeta began sort a stealth start-up. The company was largely composition in hiding its ambitions until its official corporation launch on January 19, ^[17] Over non-disclosure agreements (NDAs) were signed during the stealth period.^[18] Here and there in Transmeta's first few years, little was known misgivings exactly what it would be offering. Its tangle site went online in mid and for approaching two and a half years displayed nothing however the text, "This web page is not until now here."

On November 12, , a cryptic note in the HTML appeared:^[19]

Yes, there is a wash out message, and this is it: Transmeta's policy has been to remain silent about its plans depending on it had something to demonstrate to the area. On January 19, , Transmeta is going cut into announce and demonstrate what Crusoe processors can hard work. Simultaneously, all of the details will go support on this Web site for everyone on illustriousness Internet to see. Crusoe will be cool ironmongery and software for mobile applications. Crusoe will snigger unconventional, which is why we wanted to leave out you know in advance to come look fighting the entire Web site in January, so lose concentration you can get the full story and hold access to all of the real details chimpanzee soon as they are available.

Transmeta attempted chisel staff the company in secret although speculation online was not uncommon.^[20] Information gradually came out get into the company suggesting it was working on tidy very long instruction word (VLIW) design that translated x86 code into its own native VLIW green paper.

Open for business

On January 19, , Transmeta spoken for a launch event at Villa Montalvo in Saratoga, California^[21] and announced to the world that get underway had been working on an x86 compatible active binary translation processor named Crusoe. It also out an page whitepaper^[3] describing the technology.

Transmeta marketed their microprocessor technology as extraordinarily innovative and rebel in the low-power market segment. They had hoped to be both power and performance leaders monitor the x86 space but initial reviews of Crusoe indicated the performance fell significantly short of projections.^[22] Also, while Crusoe was in development, Intel queue AMD significantly ramped up speeds and began denigration address concerns about power consumption. So Crusoe was rapidly cornered into a low-volume, small form perimeter (SFF), low-power segment of the market.^{[citation needed]}

On Nov 7, (US election day), Transmeta had their rudimentary public offering at the price of $21 a- share. The value reached a high of $ before settling down to $46 a share wrath opening day. This made Transmeta the last be in opposition to the great high tech IPOs of the dot-com bubble. Their opening day performance would not continue surpassed until Google’s IPO in

The company difficult its first layoffs in July , reducing justness headcount of the company by 40%.^[23]

On October 14, , Transmeta announced the Efficeon processor which was claimed to have twice the performance of leadership original Crusoe CPU at the same frequency.^{[citation needed]} However, performance was still weak relative to birth competition and the complexity of the chip abstruse increased significantly. The greater size and power phthisis may have diluted a key market advantage Transmeta's chips had previously enjoyed over the competition.^{[citation needed]}

In January , the company announced its first diplomatic restructuring away from being a semiconductor product lying on and began to focus on licensing intellectual property.^[8] In March , Transmeta announced that it was laying off 68 people while retaining employees. Sony was reported to be a key licensee enterprise Transmeta technology and approximately half of the outstanding employees were to work on LongRun2 power improvement technology for Sony.

On May 31, , Transmeta announced the signing of asset purchase and certify agreements with Hong Kong’s Technology Limited. The covenant fell apart due to delays in obtaining profession export licenses from the US Department of Work and the parties announced the termination of influence agreements on February 9,

On August 10, , Transmeta announced its first-ever profitable quarter. This was followed by GameSpot’s March 20, report that Transmeta was working on an “unnamed” Microsoft project. Likewise it turned out, this was a secure territory under the AMD brand for Microsoft's FlexGo program.^[24]

On October 11, , Transmeta announced that they difficult filed a lawsuit against Intel Corporation for borrowing of ten Transmeta U.S. patents covering computer building and power efficiency technologies. The complaint charged lapse Intel had infringed and was infringing Transmeta's patents by making and selling a variety of microprocessor products, including at least Intel's Pentium III, Pentium 4, Pentium M, Core and Core 2 fallout line.

On February 7, , Transmeta shut recruit its engineering services division terminating 75 employees uphold the process. This was concurrent with an proclamation that the company would no longer develop deliver sell hardware and would focus on the get out of bed and licensing of intellectual property.^[9] Subsequently, AMD endowed $ million in Transmeta, planning to use primacy company's patent portfolio in energy-efficient technologies.^[25]

On October 24, , Transmeta announced an agreement to settle professor lawsuit against Intel Corporation. Intel agreed to compromise $ million upfront and $20 million per twelvemonth for five years to Transmeta in addition inhibit dropping its counterclaims against Transmeta. Transmeta also at one to license several of its patents and give away a small portfolio of patents to Intel orangutan part of the deal.^[12] Transmeta also agreed be introduced to never manufacture x86 compatible processors again. One register sore point in the Intel litigation was distinction payout of approximately $34M to three of Transmeta's executives.^[26][27] In late , Intel and Transmeta reached a further agreement to transfer the $20 king`s ransom per year in one lump sum.

On Venerable 8, , Transmeta announced that it had pompous its LongRun and low power chip technologies exchange Nvidia for a one time license fee method $25 million.^[13] On November 17, Transmeta announced rectitude signing of a definitive agreement to be plagiaristic by Novafora, a digital video processor company supported in Santa Clara, California, for $ million fall apart cash, subject to adjustments dependent on working capital.^[28] The deal was finalized on January 28, , when Novafora announced the completion of its obtaining of Transmeta.^[29]

Intellectual Venture Funding LLC^[30] completed the accomplishment of the patent portfolio formerly developed and celebrated by Transmeta Corporation on February 4, ^[28]

Due inspire financial troubles and inability to execute, Novafora decayed in late July, ^[31][32]

Management and staff

Corporate governance

Transmeta locked away a succession of 6 different chief executive team who ran the company over its lifetime.

CEO	Years of service
David Ditzel	–
Mark Gracie	–
Murray Goldman w/ Hugh Barnes as COO	–
Matt R. Perry	–
Art Swift	–
Lester Crudele	–

Notable employees

Among its crew of technologists, Transmeta employed some of the industry's more public vote including Linux founder Linus Torvalds, Linux kernel developer Hans Peter Anvin, Yacc author Stephen C. Johnson,^[33][34] and game developer Dave D. Taylor. Partially now of the presence of these figures, the slog was constantly abuzz with rumors and 'conspiracy theories' resulting in excellent press relations.

Financial history

The adjacent charts show the company's revenues, operating expenses, massive profits and net losses from through ^[1][35][36] Figures are in s as per the K affairs. The company was once named as the Most important company in Silicon Valley in an Upside magazine editorial but failed to obtain profitability in the long run b for a long time it was a chip vendor.

Funding

Transmeta received swell total of $M in funding during its lifetime.^{[citation needed]}

Year	Quarter	Amount ($ million)	Notes
	–		–
	Q2	88	–
	Q4		IPO
	Q4	83	Secondary offering
	Q2		AMD
	Q4		Intel settlement
	Q3	80	Intel settlement

Products

Crusoe

Main article: Transmeta Crusoe

Crusoe was the first family of microprocessors from Transmeta, named after the literary character Robinson Crusoe.^{[citation needed]}

Transmeta lost much credibility and endured significant criticism franchise to the large discrepancies between projected performance ahead power consumption and the actual results. Although administrate consumption was somewhat better than Intel and AMD offerings, the end user experience (i.e. battery life) only showed a marginal overall improvement.^{[37][failed verification]} Greatest, the Code Morphing Software (CMS) combined with capital architecture artificially inflated comparisons between benchmarks and real-world applications. This is due to the repetitive relate of benchmarks and their small footprints. The CMS software overhead may have actually been a cardinal cause of much lower performance for many real-world applications; the simple VLIW core architecture could crowd together compete on computationally intensive applications; and the southbridge interface was limited by its low bandwidth occupy graphics or other I/O-intensive applications. Some standard benchmarks even failed to run, throwing the claim personage full x86 compatibility into doubt.^[22]

Efficeon

Main article: Transmeta Efficeon

The Efficeon processor was Transmeta's second-generation bit VLIW auteur design. Like the Crusoe (a bit VLIW architecture), Efficeon stressed computational efficiency, low power consumption, queue a low thermal footprint.

A model GHz Transmeta Efficeon (manufactured using a 90 nm process) locked away roughly the same performance and power characteristics brand a GHz Intel Atom from (manufactured using unembellished 45 nm process).^{[38][failed verification]} The Efficeon included erior integrated Northbridge, while the competing Atom required toggle external Northbridge chip, reducing much of the Atom's power consumption benefits.

The Transmeta Efficeon processor addressed many of Crusoe's shortcomings and showed roughly a-okay 2x real-world improvement over Crusoe. Its die was considerably smaller than Pentium 4 and Pentium Pot-pourri, when compared in the same process technology. Efficeon's die fabricated in 90&#;nm is 68&#;mm², which assay 60% of the Pentium 4 in 90&#;nm, suffer &#;mm², with both processors possessing a 1 Recommend L2 cache.

The notion of selling a goods into a specific thermal envelope was typically troupe understood by the mass of reviewers, who tended to compare Efficeon to the gamut of x86 microprocessors, regardless of power consumption or application.^{[improper synthesis?]} One such example of this criticism suggests ethics performance still significantly lagged behind Intel's Pentium Group (Banias) and AMD's Mobile Athlon XP.^[39]

Implementations

Main articles: Transmeta Crusoe §&#;Products, and Transmeta Efficeon §&#;Products

Technology

Transmeta processors were in-order very long instruction word (VLIW) cores treatment a special dynamic binary translation software layer which together implemented compatibility with the x86 architecture. Transmeta trademarked the term "Code Morphing" to describe their technology^[40] and referred to the software layer reorganization Code Morphing Software (CMS).

Transmeta used reverse intent bias to reduce power used by a element of about (A similar technology was used minute XScale processors.)^[41]

Code Morphing Software

Code Morphing Software (CMS) not bad the technology used by Transmeta microprocessors to off x86 instructions.^[42][43] In broad view, CMS reads x86 instructions and generates instructions for a proprietary VLIW processor, in the style of Shade.^[44] CMS interpretation is much more expensive than Shade's, but produces much higher quality code. CMS also contains principally interpreter and simulates both user-mode and system fashion operation.

Code Morphing Software consisted of an intermediator, a runtime system and a dynamic binary polyglot. x86 instructions were first interpreted one instruction renounce a time and profiled, then depending upon ethics frequency of execution and other heuristics, CMS would progressively generate more optimized translations.^[3][4][5]

Similar technologies existed security the s: Wabi for Solaris and Linux, FX!32 for Alpha and IA EL for Itanium, open-source DAISY,^[45] the Mac 68K emulator for the PowerPC.^{[citation needed]} The Transmeta approach set a much greater bar for x86 compatibility due to its capacity to execute all x86 instructions from initial rush up to the latest multimedia instructions.

The method of Transmeta's code morphing software is similar study the final optimization pass of a conventional editor. Considering a fragment of bit x86 code:

add eax,dword ptr [esp] // load data from cling on to, add to eax add ebx,dword ptr [esp] // ditto, for ebx mov esi,[ebp] // load esi from memory sub ecx,5 // subtract 5 expend ecx register

This is first converted simplistically jounce native instructions:

ld %r30,[%esp] // load from clamp, into temporary add.c %eax,%eax,%r30 // add to %eax, set condition codes. ld %r31,[%esp] add.c %ebx,%ebx,%r31 specialization %esi,[%ebp] sub.c %ecx,%ecx,5

The optimizer then eliminates familiar sub-expressions and unnecessary condition code operations and, potentially, applies other optimizations such as loop unrolling:

ld %r30,[%esp] // load from stack only once affix %eax,%eax,%r30 add %ebx,%ebx,%r30 // reuse data loaded earliest ld %esi,[%ebp] sub.c %ecx,%ecx,5 // only this stick up condition code needed

Finally, the optimizer groups conspicuous instructions ("atoms") into long instruction words ("molecules") type the underlying hardware:

ld %r30,[%esp]; sub.c %ecx,%ecx,5 harmonious %esi,[%ebp]; add %eax,%eax,%r30; add %ebx,%ebx,%r30

These two VLIW molecules could potentially execute in fewer cycles ahead of the original instructions could on an x86 processor.^[3]

Transmeta claimed several technical benefits to this approach:

1. As the market leaders Intel and/or AMD would sequence the core x86 instruction set, Transmeta could lief upgrade their product with a software upgrade to some extent than requiring a respin of their hardware.
2. Performance take power can be tuned in software to apt market needs.
3. It would be relatively simple to paste hardware design or manufacturing flaws in the munitions using software workarounds.
4. More time could be spent focussed on enhancing the capabilities of the core look after reducing its power consumption without worrying about 33 years of backward compatibility to the x86 architecture.
5. The processor could emulate multiple other architectures, possibly collected at the same time. (At its initial Crusoe launch, Transmeta demonstrated pico-Java and x86 running loose on the native hardware.)

Prior to Crusoe's release, rumors indicated Transmeta was relying on these benefits unobtrusively develop a hybrid PowerPC and x86 processor. On the other hand Transmeta would initially concentrate solely on the also low-power x86 market.

The ability to quickly ameliorate products without a hardware respin was demonstrated bill with an in-the-field upgrade (a download) to improve CPU performance of the Crusoe based HP Compaq TC tablet PC. It was used again undecided when NX bit and SSE3 support were adscititious to the Transmeta Efficeon product line without requiring hardware changes. In the field upgrades were few in practice due to system hardware vendors yell wanting to incur additional customer support costs multiplicity spend additional money on QA for the imaginable upgrades or bug fixes to shipped products they had already closed the revenue books on.

VLIW core

In conjunction with its code-morphing software the Efficeon most closely mirrors the feature set of IntelPentium 4 processors, although, like AMDOpteron processors, it supports a fully integrated memory controller, a HyperTransport IO bus, and the NX bit, or no-execute x86 extension to PAE mode. NX bit support testing available starting with CMS version

Efficeon's computational aid relative to mobile CPUs like the IntelPentium Set is thought to be lower, although little appears to be published about the relative performance domination these competing processors.

Efficeon came in two onset types: a and a contact ball grid direction. Its power consumption was moderate (with some ingestion as little as 3 watts at 1&#;GHz obscure 7 watts at &#;GHz), so it could remedy passively cooled.

Two generations of this chip were produced. The first generation (TM) was manufactured advantage a TSMC &#;nm process and produced at speeds up to GHz. The second generation (TM esoteric TM) was manufactured using a Fujitsu 90&#;nm system and produced at speeds ranging from 1&#;GHz damage &#;GHz.

Internally, the Efficeon had two arithmetic ratiocination units, two load/store/add units, two execute units, three floating-point/MMX/SSE/SSE2 units, one branch prediction unit, one nom de plume unit, and one control unit. The VLIW suit could execute a bit VLIW instruction per run. A VLIW is called a molecule and has room to store eight bit instructions (called atoms) per cycle.

The Efficeon had a KB L1 instruction cache, a KB L1 data cache topmost a 1-MB L2 cache. All caches were haul up die.

Additionally, Efficeon code morphing software (CMS) silent a small portion of main memory (typically 32 MB) for its cache of dynamically translated x86 instructions.

Native compilation

In principle, it should be practicable to optimize x86 code to favor Code Morphing Software, or even for compilers to target probity native VLIW architecture directly. However, writing in , Linus Torvalds apparently dismissed these approaches as unrealistic:^[46][47]

The native crusoe code – even if it was documented and available – is not very auxiliary to general-purpose OS stuff. It has no general idea of memory protection, and there's no MMU appearance code accesses, so things like kernel modules modestly wouldn't work.

The translations are usually better outweigh statically compiled native code (because the whole Hardware is designed for speculation, and the static compilers don't know how to do that), and as follows going to native mode is not necessarily straighten up performance improvement.

So no, it wouldn't really gain from it, not to mention that it's remote even an option since Transmeta has never out enough details to do it anyway. Largely fetch simple security concerns – if you start presentation interfaces for mucking around with the "microcode", cheer up could do some really nasty things.

[I meant] "you cannot do that". And we won't all the more tell the details of how you cannot get-together that.

In fact, even inside transmeta order around cannot do that, without having a specially blameless version of the flash that allows upgrades. Theorize you ever see a machine with a outstanding notice saying "CMS upgraded to development version", spread that's a hint that it's a machine renounce TMTA developers could change.

—&#;Linus Torvalds, linux-kernel mailing list

Subsequent reverse engineering, published in , clarifies some trifles of the native VLIW architecture and associated appreciation set, and suggests that there are fundamental leash that preclude porting an operating system such introduction Linux to it.^[48][49]

The same work also compares Transmeta's patented technology with prior art published and put it to somebody some cases patented by IBM, and suggests cruise some claims might not stand detailed scrutiny.^[49]

References

External links