[Simh] VAX/VMS

[Thomas Merritt]

That is a pretty Intel centric view of history.  The original architecture was developed by AMD.  The architecture was initially called x86-64 by AMD and was renamed right before the public launch to AMD64.  SUSE was involved in the development of the gcc port and linux port, didn’t wan’t to take the risk of of changing all of the x86_64 references to amd64, so the x86_64 name persists.  The first Intel chips got a number of items wrong that were never issues on the released AMD chips.  The NX bit was not in all of AMD’s development spins, but was discussed publicly on the x86_64 mailing list prior to the release of the first Athlon64 chips.  When Intel later released their first x86_64 compatible chips they did not include the NX bit and would not run Linux correctly.  Microsoft, having early access to the Intel chips, did not have issues with the 64-bit version of Windows as I recall.  Both Intel and AMD have made enhancements over the years to the AMD64 (ne x86_64) architecture.  The AMD and Intel flavors of virtualization support are different, AMD’s Pacifica is well thought out, Intel took four iterations on Virtualization to get to what they have today.  Because of the quick response on the Intel side, they were able to stay competitive in the marketplace.  If they had continued to pursue the Itanium route for 64-bit, they would ultimately have lost the market to AMD.  Fortunately for Intel, they included enough AMD compatibility to remain competitive.  They had huge advantages on the semiconductors side of the house, that has allowed them to be the dominate player that they are today.

— TJ

> On Feb 16, 2016, at 2:44 PM, Clement T. Cole <clemc at ccc.com> wrote:
> 
> Depends on how you look at it,  AMD did developed an early 64 bit extensions on to the 32 bit ISA. But that was over 10 years ago and I was not there at the time.   IMO thankfully when Core/INTEL*64 was developed much of it made to be the same in the desire to keep user binaries to run.   Since that time Intel has taken and continues to extend the ISA. Simply put, INTEL*64 is different - there are whole sections of the ISA that are not implemented on all processors (even at Intel). For instance Intel's Phi brings in a number of new instructions.  INTEL*64 is the official name (trademark name) for the ISA (although some folks refuse to acknowledge that fact). 
> 
> Also in the case of privileged ISA features there are some significant difference which the OS's have to handle.  For instance the VT subsystems have some differences.
> 
> As Tim points out the real cost of compatibly is the architectural tests suites and effort to ensure that things just work across the board.  In the case of x86 it's even more difficult then just the instructions and BIOS because it means whole HW sections have to be made virtual also so that old code (like ones for DOS) do keep working.  
> 
> Similarly, Regardless of which Intel produced processor for that ISA is the output target,
> Intel's compilers generate code for the INTEL*64 ISA and perform optimization for same.  When user mode binaries are run on non-intel manufactured processors they should "just work" if the others manufactures have implemented equivalent functionality (There is no truth to the sometimes stated comment that Intel's compilers check for non-intel manufactured and do bad things).  That said the Intel development suite does not do specific optimizations for non intel manufactured processors but they do make an attempt to ensure things execute correctly. 
> 
> The neutral term is x86_64 which does not acknowledge either AMD or Intel.  But in print, I am fairly certain that the ISA's trademarked name is INTEL*64 when referring to that ISA.  
> 
> Sent from my iPad
> 
>> On Feb 16, 2016, at 5:02 PM, Rhialto <rhialto at falu.nl> wrote:
>> 
>>> On Tue 16 Feb 2016 at 11:25:37 -0500, Clem Cole wrote:
>>> Unless you are using a cell phone, I'm willing to bet that you are typing
>>> your messages on a INTEL*64 architecture system, even if the processor is
>>> not made by Intel.
>> 
>> Was the 64-bit mode not designed by AMD? I'm typing this on NetBSD/amd64
>> after all...
>> 
>> -Olaf.
>> -- 
>> ___ Olaf 'Rhialto' Seibert  -- The Doctor: No, 'eureka' is Greek for
>> \X/ rhialto/at/xs4all.nl    -- 'this bath is too hot.'
> _______________________________________________
> Simh mailing list
> Simh at trailing-edge.com
> http://mailman.trailing-edge.com/mailman/listinfo/simh