From: jgoerzen@complete.org 
  
 On Sat, Aug 16 2025, Simon McVittie wrote: 
  
 > On Sat, 16 Aug 2025 at 07:28:24 -0500, John Goerzen wrote: 
 >>Without getting into a debate over whether i386 should or should not be 
 >>dropped, as someone that runs other 32-bit archs, I wonder why armhf and 
 >>armel weren't similarly targeted? 
 > 
 > Each of these 32-bit ports has different characteristics and different 
 > tradeoffs. 
  
 Thank you, Simon.  Your message (and your other one about compilers 
 being unable to compile software within 4GB of VRAM) was very helpful 
 and clarified a lot of points for me. 
  
 John 
  
 > 
 > i386-without-SSE2 on real 32-bit CPUs (as opposed to i386 as a 
 compatibility 
 > architecture on newer CPUs) is a problematic porting target because it has 
 > non-IEEE floating-point behaviour that frequently causes trouble: test 
 suites 
 > often expect to see the same exact values, simulations (e.g. networked 
 games) 
 > want to see deterministic results from running the same simulation on 
 different 
 > CPUs, and the thing that finally prompted raising the baseline was that it 
 > turned out that it even harms soundness (ability to generate correct 
 machine 
 > code) in LLVM/rustc. i386-with-SSE2 is a lot easier, because it has the 
 same 
 > IEEE floating point as every other release architecture, but there were 
 only 
 a 
 > few CPU generations that have SSE2 but not 64-bit support (I think the 
 Pentium 4 
 > around 20 years ago was the most common example). 
 > 
 > armel is also on its way out, because it has a *different* characteristic 
 that 
 > makes it a problematic porting target: unlike every other release 
 architecture, 
 > it doesn't have atomic operations in the baseline instruction set, and they 
 have 
 > to be emulated. For example this is why mozjs/gjs had to be removed from 
 armel 
 > in trixie. I suspect we would have lost armel already if it wasn't for the 
 fact 
 > that it's the newest/highest-functionality Debian port that will run 
 unmodified 
 > on first-generation Raspberry Pi hardware, because ARMv6 is slightly too 
 old 
 for 
 > official Debian's armhf baseline; Raspberry Pi OS, a Debian derivative, 
 works 
 > around this by recompiling armhf with a lower baseline. 
 > 
 > armhf has had CPUs that can run it but not its successor manufactured more 
 > recently than i386 or armel, has normal IEEE floating-point unlike i386, 
 and 
 has 
 > normal atomic operations unlike armel, so it is both less painful to 
 support 
 and 
 > more likely to be useful as a "bare metal" OS when compared with either 
 i386 
 or 
 > armel. 
 > 
 > (I do think that armhf's useful lifetime is limited, too, though.) 
 > 
 >     smcv 
  
 --- SoupGate-Win32 v1.05 
  * Origin: you cannot sedate... all the things you hate (1:229/2)