iiinova — where your digital self keeps evolving

eigenform @reflexive.space · 24/09/2026

:) shes being very patient with me while im grilling next to the feeder

ruby throated hummingbird perched on a branch (watching me from afar)

000

eigenform @reflexive.space · 15/08/2026

re f16h dram scrambling thing

BKDG for AMD Family 16h Models 30h-3Fh Processors

"Since MMIO configuration cycles are not serializing in the way that IO configuration cycles are, their ordering
rules relative to posted may result in unexpected behavior.
Therefore, processor MMIO configuration space is designed to match the following ordering relationship that
exists naturally with IO-space configuration: if a core generates a configuration cycle followed by a posted-
write cycle, then the posted write is held in the processor until the configuration cycle completes. As a result,
any unexpected behavior that might have resulted if the posted-write cycle were to pass MMIO configuration
cycle is avoided."

120

eigenform @reflexive.space · 11/07/2026

b̸ě̷͜ n̏̈́ͫot̴ͅ áͨf̒ͨřa̼͂̅i̯͍̚d̠͈̐

Figure B5-4 from the Arm Debug Interface Architecture Specification (it's a state diagram about switching between SWD and JTAG)

010

eigenform @reflexive.space · 02/07/2026

saw a cool video about how this beach on Trinidad & Tobago was very rapidly uplifted sometime shortly after those EQs in Venezuela, so I thought this would be a good excuse to look at Sentinel-1 SAR

322

eigenform @reflexive.space · 09/06/2026

adding to my collection of goofy microbenchmark tools; here's the zen2 op cache capacity (4096!) staring back at me

020

eigenform @reflexive.space · 20/04/2026

POV: you are one of the sanriku borehole observatories

100

eigenform @reflexive.space · 04/04/2026

family getting THIS for easter

110

eigenform @reflexive.space · 24/02/2026

shitty realmode trampoline works in qemu, we take those

020

eigenform @reflexive.space · 19/02/2026

behold dialed.gg?c=VJNV8X

100

eigenform @reflexive.space · 19/02/2026

twitch just straight up has a big link that says ROBLOX at the top of the page now? (i am getting old)

030

eigenform @reflexive.space · 09/02/2026

my buddys irish wolf hound says hello:^)

020

eigenform @reflexive.space · 23/12/2025

pulled this out of my bin, not sure exactly which version of Broadway (the CPU from the Nintendo Wii) this is, although I'm fairly confident it's one of the 65nm versions Wikipedia says Broadway-A has 43E4048 on the package, this one seems to have 43E5058 on the die?

the corner of a "Broadway" die (the CPU for the Nintendo Wii, ie. the IBM PPC750CL), looking through the top cover and metal at some markings. The part number appears to be "43E5058". Some other markings with the names of their corresponding layers are also visible.

110

eigenform @reflexive.space · 11/12/2025

big fan of the stilted groove in neurotica

100

eigenform @reflexive.space · 10/12/2025

now i have an excuse to learn some meshuggah tunes

111

eigenform @reflexive.space · 11/11/2025

ive had this thing in drop A for a few months (my brother's fault), not sure i like it but i do see the appeal:

100

eigenform @reflexive.space · 11/11/2025

amp/cab sim stuff is stupidly good and easy now..

311

eigenform @reflexive.space · 23/10/2025

defensive programming or four-line horror story? (i *think* it's actually the former, but I read it as "author realizes some horrible consequence of SLS happening across ERET" and physically felt pain)

a painful-looking diff from 'osfmk/arm64/locore.s' in XNU:

- /* We don't need an ISB here, as the eret is synchronizing. */
+ #if defined(ERET_IS_NOT_CONTEXT_SYNCHRONIZING)
+ isb sy
+ #endif /* defined(ERET_IS_NOT_CONTEXT_SYNCHRONIZING) */

110

eigenform @reflexive.space · 14/10/2025

what fresh hell is this (from the LNL perf events?)

110

eigenform @reflexive.space · 04/10/2025

ah... there are places in tosa where accounts describe the runup being over 20m...

010

eigenform @reflexive.space · 04/10/2025

this shrine in kochi pref. the 39th step... ah

110

eigenform @reflexive.space · 04/10/2025

been playing a little bit recently ^^

010

eigenform @reflexive.space · 14/07/2025

figure from the paper (for comparison to the `inp_diff` field in the previous image) \o/ neat!

Figure 3. from the paper "Take A Way: Exploring the Security Implications of AMD’s Cache Way Predictors"

120

eigenform @reflexive.space · 14/07/2025

false alarm, i was apparently only iterating through 12 bits instead of all 16 bits of input! \o/ i think this works

Command-line output showing consistent miss address buffer (MAB) allocations for a load instruction caused by L1D way predictor misses on a Zen 2 core.

Each line in the output represents a pair of virtual addresses whose micro-tag is colliding, necessarily resulting in a misprediction. The `inp` fields are bits in the addresses used as input to a simple hash function which associates each virtual address to the cache way used in the previous load from that address. The `inp_diff` fields visually show us which pairs of bits are being XORed together!

100

eigenform @reflexive.space · 20/05/2025

alright here's where we stand folks

030

eigenform @reflexive.space · 16/05/2025

relaxing by the lake this weekend :>

110

eigenform @reflexive.space · 04/05/2025

pads in question are the cyan ones at the top: the corresponding package pins are just connected to ground on the board 🤔

Gameboy Advance CPU die, top-right corner, composite image, 5x objective.
The ARM7TDMI is in the middle, with some unknown SRAMs around the edge. A bunch of the pads are labelled, but there are four mysterious pads next to the clocks.

100

eigenform @reflexive.space · 04/05/2025

quick n' dirty 5x stitch (i need to clean the die and also the inside of my scope)

Gameboy Advance CPU die (CPU AGB), top metal layer visible through passivation, very poorly-stitched composite from ~25 images, 5x objective.

The ARM7TDMI macro stands out in the top-righthand corner.
There are a bunch of SRAMs around the corner of the die.
Marking in the top left corner says "C2UN19" (?)

160

eigenform @reflexive.space · 04/05/2025

about at the limit of my optics (and also we're looking through all the passive layers on the top of the chip) but holy crap!!

GBA CPU die, 20x objective.
Markings in the center of the image read:

ARM7TDMI R3A (C)(T)(M) ARM Ltd. 1999

151

eigenform @reflexive.space · 03/05/2025

and a closeup of some pads:

Gameboy Advance CPU die, 20x objective.
Image of two square IO pads (with big craters in the middle from the bond wires), and some test pads around the edge of the die where it was diced. Markings by the test pads read "PMOS" and ".3/10"

000

eigenform @reflexive.space · 03/05/2025

closer view of the lower metal layers:

Gameboy Advance CPU die, 20x objective.
There are a lot of wires everywhere (big surprise).
Not sure what part of the die this is yet.

120

eigenform @reflexive.space · 03/05/2025

so i realized that I had an AGB CPU in my bin from a long time ago... and decided to try depackaging it with just a heat gun and pliers because why not :^) (surprisingly not cracked! the dark left half is just the part still wedged in the package, too afraid to try pulling it out right now)

Gameboy Advance CPU die, partially obscured top-left corner at 5x magnification. You can see the pad frame and wire bonds around the edge, a big blob of wires in the middle, and two SRAMs peeking through. Half the image is darkened out because the die is sticking halfway out of the package.

131

eigenform @reflexive.space · 29/04/2025

> huge water-cooled xeon test board isn't real > huge water-cooled xeon test board cant hurt you huge water-cooled xeon test board:

intimidating looking pcb with a huge intel xeon part (?) and lots of pins fanned out to the edge of the board

image from: https://www.hardwareluxx.de/index.php/news/hardware/wasserkuehlung/66015-bis-zu-1-000-w-intel-arbeitet-an-wasserkuehlung-auf-package-level.html

030

eigenform @reflexive.space · 23/04/2025

Michael J. Flynn dropping one of the hardest lines ever (in 1972!!)

"Serious inefficiencies may arise in confluent SISD organizations due to turbulence when data interacts with the instruction stream."

010