more lottes, rfluery, and nic.

This commit is contained in:
ed committed 2026-08-28 09:23:41 -04:00
1 parent b9769b7885
commit 8cffa8e23e
29 files changed
+1064 -84

No files matched your search

@@ -1,79 +0,0 @@
---
title: "@NOTimothyLottes These instrs are anything but fast and if you want both \"fast\" and compact you run a separate decoding pass where you expand custom bytecode into target instructions that suck less."
type: archive
source: twitter
source_url: "https://x.com/noop_dev/status/2078763409403683061"
author: "Boris Chuprin"
handle: noop_dev
post_id: "2078763409403683061"
date: 2026-07-19
archived: 2026-08-25
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "@NOTimothyLottes These instrs are anything but fast and if you want both \"fast\" and compact you run a separate decoding pass where you expand custom bytecode into target instructions that suck less."
in_reply_to: ""
parent_post_id: "2078729662021111958"
---
## Source
- URL: https://x.com/noop_dev/status/2078763409403683061
- Author: Boris Chuprin (@noop_dev)
- Posted: 2026-07-19 08:46:33
## Branch
**1/** **@noop_dev** ^2078763409403683061
**@NOTimothyLottes**
These instrs are anything but fast and if you want both "fast" and compact you run a separate decoding pass where you expand custom bytecode into target instructions that suck less.
**2/** **@NOTimothyLottes** ^2078851948543910353
**@noop_dev**
Basic truth: if the program is tiny and dependency free (ie intrinsically not just glue for libraries) it’s probably already fast even if the machine isn’t. And for the things that need perf there is always assembly
**3/** **@noop_dev** ^2078853226967781591
**@NOTimothyLottes**
Anyway, I am trying you to sell the idea of bytecode-driven macroassembler that expands "macros" on load. Like I did with 4K atari 2600 emu ~20 years ago.. some demosceners I knew also adopted the idea..
**4/** **@NOTimothyLottes** ^2078856306811691388
**@noop_dev**
Many of my other systems had been such that interpreted source generates raw code then executes the code (all at once after code is fully processed). Effectively forth as a macro language and instruction generator, runtime assembler.
**5/** **@NOTimothyLottes** ^2078858446409978111
**@noop_dev**
But for cold cache stuff it’s easy to burn more time in binary generation than it would cost to do simple interpreter.
**6/** **@noop_dev** ^2078860015323038149
**@NOTimothyLottes**
Not sure I can understand how cold cache matters for a linear transformation of a relatively small # of bytes.
**7/** **@NOTimothyLottes** ^2078861504284074431
**@noop_dev**
If it’s a byte that indexes into say a fixed size physical instruction, yeah it’s just a decompression of source code, sure easy to do fast. Im talking more the multi-branch miss per instruction stuff.
**8/** **@NOTimothyLottes** ^2078862606517895324
**@noop_dev**
Meaning “mov edi,[rax-0x32]” could be generated from 4 symbol lookups (2 for regs, one for offset, one for instruction). Also my intent here is for GPU code generation for AMD GPUs where youd have sometimes 8+ bitfields in an opcode. So the simple stuff won’t work there …
## Related
- Spine: [[archive/threads/NOTimothyLottes/2026-07-19-probably-shouldnt-consider-this-but-apparently]]
@@ -7,7 +7,7 @@ author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2078728600912630031"
date: 2026-07-19
archived: 2026-08-25
archived: 2026-08-28
draft: false
tags:
- archive
@@ -39,4 +39,50 @@ It's interesting because it cuts interpreted source size in half. Ie a stream of
The aim of course, keep the stuff that doesn't need to go fast optimized instead for low complexity and low size (aka interpreted forth), and keep the stuff that needs to go fast, at peak, assembly. Hits 2 extremes well.
Branches: [[archive/threads/NOTimothyLottes/2026-07-19-probably-shouldnt-consider-this-but-apparently/2026-07-19-noop_dev-these-instrs-are-anything-but-fast-and-if-you]]
**4/** **@noop_dev** ^2078763409403683061
**@NOTimothyLottes**
These instrs are anything but fast and if you want both "fast" and compact you run a separate decoding pass where you expand custom bytecode into target instructions that suck less.
**5/** **@NOTimothyLottes** ^2078851948543910353
**@noop_dev**
Basic truth: if the program is tiny and dependency free (ie intrinsically not just glue for libraries) it’s probably already fast even if the machine isn’t. And for the things that need perf there is always assembly
**6/** **@noop_dev** ^2078853226967781591
**@NOTimothyLottes**
Anyway, I am trying you to sell the idea of bytecode-driven macroassembler that expands "macros" on load. Like I did with 4K atari 2600 emu ~20 years ago.. some demosceners I knew also adopted the idea..
**7/** **@NOTimothyLottes** ^2078856306811691388
**@noop_dev**
Many of my other systems had been such that interpreted source generates raw code then executes the code (all at once after code is fully processed). Effectively forth as a macro language and instruction generator, runtime assembler.
**8/** **@NOTimothyLottes** ^2078858446409978111
**@noop_dev**
But for cold cache stuff it’s easy to burn more time in binary generation than it would cost to do simple interpreter.
**9/** **@noop_dev** ^2078860015323038149
**@NOTimothyLottes**
Not sure I can understand how cold cache matters for a linear transformation of a relatively small # of bytes.
**10/** **@NOTimothyLottes** ^2078861504284074431
**@noop_dev**
If it’s a byte that indexes into say a fixed size physical instruction, yeah it’s just a decompression of source code, sure easy to do fast. Im talking more the multi-branch miss per instruction stuff.
**11/** **@NOTimothyLottes** ^2078862606517895324
**@noop_dev**
Meaning “mov edi,[rax-0x32]” could be generated from 4 symbol lookups (2 for regs, one for offset, one for instruction). Also my intent here is for GPU code generation for AMD GPUs where youd have sometimes 8+ bitfields in an opcode. So the simple stuff won’t work there …
@@ -0,0 +1,66 @@
---
title: "Perhaps the most important aspect of AI is the legal framework it establishes to bypass patents and IP law."
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2079292875762717000"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2079292875762717000"
date: 2026-07-20
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "Perhaps the most important aspect of AI is the legal framework it establishes to bypass patents and IP law."
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2079292875762717000
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-07-20 19:50:27
## Thread
**1/** **@NOTimothyLottes** ^2079292875762717000
Perhaps the most important aspect of AI is the legal framework it establishes to bypass patents and IP law. Slap that “slop” logo on it (aka generated by prompt-grammars with no original human authoring) and its open season on stealing anything digital via derivatation
**2/** **@AgileJebrim** ^2079293021657403695
**@NOTimothyLottes**
This is why trade secrets are going to be the way to go.
**3/** **@NOTimothyLottes** ^2079293597719933301
**@AgileJebrim**
Secrets as in no applied patents and no disclosure of any kind, because certainly patent filings is one of the more important training data sources
**4/** **@NOTimothyLottes** ^2079294620857422248
**@AgileJebrim**
It used to be that graphics industry lived in this uncomfortable truce where the major devs would NOT apply for software patents and would still talk details at conferences. And major IHVs/OEMs had always been applying heavy pressure to patent
**5/** **@NOTimothyLottes** ^2079295407381762464
**@AgileJebrim**
NV broke the pattern and went trade secret for DLSS2+. No devs will ever talk about how it works, and likely there is no patents that describe enough of how it works either. So we’ve already seen the effects of the new tech trends on IP law/behavior
**6/** **@NOTimothyLottes** ^2079296288944800019
**@AgileJebrim**
The culture of competition between devs openly 1-uping each other with public conf talks stopped for one huge area of gfx tech, and unfortunately this trend will continue as more of dev gets black boxes by big tech
**7/** **@AgileJebrim** ^2079298228680294542
**@NOTimothyLottes**
E&S/Collins was always very secretive when I was there.
@@ -0,0 +1,46 @@
---
title: "Entertaining aspect of AI bubble is watching humans try to explain how profit & productivity is amplified by AI -> yes, most certainly if I was to go in your fridge and steal your hamburger meat, and just modify it with a different sauce, my burger stand looks more profitable too"
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2079950100986544273"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2079950100986544273"
date: 2026-07-22
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "Entertaining aspect of AI bubble is watching humans try to explain how profit & productivity is amplified by AI -> yes, most certainly if I was to go in your fridge and steal your hamburger meat, and just modify it with a different sauce, my burger stand looks more profitable too"
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2079950100986544273
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-07-22 15:22:02
## Thread
**1/** **@NOTimothyLottes** ^2079950100986544273
Entertaining aspect of AI bubble is watching humans try to explain how profit & productivity is amplified by AI -> yes, most certainly if I was to go in your fridge and steal your hamburger meat, and just modify it with a different sauce, my burger stand looks more profitable too
**2/** **@NOTimothyLottes** ^2079952730362179605
The sad part is how the desired outcome of legalized IP theft has to be bundled with all this expensive and power inefficient hardware -> ie packaged into ip theft subscription services
**3/** **@NOTimothyLottes** ^2079953589624045943
Remember according to corporate tech bros, actual intelligence isn't being the eagle that spends the day hunting and catching a fish, rather it is being the eagle that in a split second steals the fish from the other eagle and gets away with it
**4/** **@NOTimothyLottes** ^2079954856501018942
Problem with the tech bro "science" is that it's core it won't raise up humanity, because it is a system designed to subjugate those who actually contribute something novel while institutionalizing stealing by those who otherwise cannot. The reward system is inverted.
**5/** **@NOTimothyLottes** ^2079955534782865509
When the reward system is inverted in that way, the system at it's limit converges into a dystopia.
@@ -0,0 +1,34 @@
---
title: "Seriously fucking awesome! In a world of total enshittification, ModRetro turns the tables and shows us that doing it right is still possible."
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2082487980690706865"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2082487980690706865"
date: 2026-07-29
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "Seriously fucking awesome! In a world of total enshittification, ModRetro turns the tables and shows us that doing it right is still possible."
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2082487980690706865
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-07-29 15:26:40
## Thread
**1/** **@NOTimothyLottes** ^2082487980690706865
Seriously fucking awesome! In a world of total enshittification, ModRetro turns the tables and shows us that doing it right is still possible. Also it sounded like you'd be able to side load FGPA cores?, if so, this might not be only a N64 emulator . . .
**2/** **@NOTimothyLottes** ^2082492490255315100
Curious what size Artix UltraScale+ FPGA is in the M64 and the amount and bandwidth of the PSRAM ...
@@ -0,0 +1,43 @@
---
title: "@NOTimothyLottes Can you run your own code (without hacking it) though ?"
type: archive
source: twitter
source_url: "https://x.com/aussetg/status/2082521381674049872"
author: "Guillaume Ausset - @ausset.me"
handle: aussetg
post_id: "2082521381674049872"
date: 2026-07-29
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "@NOTimothyLottes Can you run your own code (without hacking it) though ?"
in_reply_to: ""
parent_post_id: "2082496468720521510"
---
## Source
- URL: https://x.com/aussetg/status/2082521381674049872
- Author: Guillaume Ausset - @ausset.me (@aussetg)
- Posted: 2026-07-29 17:39:23
## Branch
**1/** **@aussetg** ^2082521381674049872
**@NOTimothyLottes**
Can you run your own code (without hacking it) though ?
**2/** **@NOTimothyLottes** ^2082523022313845042
**@aussetg**
Sounds like they are serious about allowing side loading of the FPGA -> then it could in theory function similar to a MiSTer. Imagine side loading FPGA configuration for your own fantasy HW, then having a real physical game cartridge.
## Related
- Spine: [[archive/threads/NOTimothyLottes/2026-07-29-think-about-it-the-m64-is-probably-going-to-be]]
@@ -0,0 +1,71 @@
---
title: "Think about it - the M64 is probably going to be one of the best FPGA device platforms for custom fantasy HW platform development."
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2082496468720521510"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2082496468720521510"
date: 2026-07-29
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "Think about it - the M64 is probably going to be one of the best FPGA device platforms for custom fantasy HW platform development."
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2082496468720521510
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-07-29 16:00:23
## Thread
**1/** **@NOTimothyLottes** ^2082496468720521510
Think about it - the M64 is probably going to be one of the best FPGA device platforms for custom fantasy HW platform development.
![](https://pbs.twimg.com/media/HOaDf8xWEAEOfhU?format=jpg&name=orig)
Branches: [[archive/threads/NOTimothyLottes/2026-07-29-think-about-it-the-m64-is-probably-going-to-be/2026-07-29-aussetg-can-you-run-your-own-code-without-hacking-it]]
**2/** **@NOTimothyLottes** ^2082501415977566241
Anyone want to guess on what size FPGA they used?
![](https://pbs.twimg.com/media/HOaHoR7XQAAazym?format=png&name=orig)
**3/** **@NOTimothyLottes** ^2082502244348358790
Hint just pin count:
![](https://pbs.twimg.com/media/HOaIvwtWcAA_iqE?format=jpg&name=orig)
**4/** **@NOTimothyLottes** ^2082506788927054144
Looks like an FFVB676 package, which means it is at least an AU10P (cannot be the AU7P)
![](https://pbs.twimg.com/media/HOaMRaFXYAAyfnA?format=png&name=orig)
**5/** **@NOTimothyLottes** ^2082521843479543861
Looks like AP Memory 24-pin PSRAMs. If it was DDR it might be 16-bit per clock, so effectively a "64-bit bus" perhaps at 133 MHz - so very rough guess ballpark paper spec peak around 1 GiB/s.
![](https://pbs.twimg.com/media/HOaWvq7XwAAQBA0?format=jpg&name=orig)
![](https://pbs.twimg.com/media/HOaZ844XEAAcfem?format=jpg&name=orig)
**6/** **@NOTimothyLottes** ^2082527433958285351
Looks like the FPGA in the M64 is an AU15P
![](https://pbs.twimg.com/media/HOafQ28XEAADcYX?format=png&name=orig)
![](https://pbs.twimg.com/media/HOafpgWXQAAFbA_?format=png&name=orig)
**7/** **@NOTimothyLottes** ^2082531267094917151
So ballpark FPGA capacity is like
{32 CLBs + 2 DSPs, per 2KiB BRAM} x 288 BRAMs
@@ -0,0 +1,37 @@
---
title: "Ratchet and Clank is the only remaining callback to the good days of Crash Bandicoot and that kind of whimsy game design direction."
type: archive
source: twitter
source_url: "https://x.com/szspagna/status/2083341841999462807"
author: "szspagna"
handle: szspagna
post_id: "2083341841999462807"
date: 2026-07-31
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "Ratchet and Clank is the only remaining callback to the good days of Crash Bandicoot and that kind of whimsy game design direction."
in_reply_to: ""
parent_post_id: "2083259553995460709"
---
## Source
- URL: https://x.com/szspagna/status/2083341841999462807
- Author: szspagna (@szspagna)
- Posted: 2026-07-31 23:59:36
## Branch
**1/** **@szspagna** ^2083341841999462807
Ratchet and Clank is the only remaining callback to the good days of Crash Bandicoot and that kind of whimsy game design direction.
Unfortunately there is a popular and pernicious idea that in order for something to be for adult audiences, it must be deeply cynical, dysfunctional, or violent. Somehow the wild success of LoTR is lost on these people.
## Related
- Spine: [[archive/threads/NOTimothyLottes/2026-07-31-one-unfortunate-shift-in-1st-party-sony-content]]
@@ -0,0 +1,56 @@
---
title: "One unfortunate shift in 1st party Sony content has been from family friendly to a core focus on either extreme graphic violence or say non binary or etc."
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2083252794190901690"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2083252794190901690"
date: 2026-07-31
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "One unfortunate shift in 1st party Sony content has been from family friendly to a core focus on either extreme graphic violence or say non binary or etc."
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2083252794190901690
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-07-31 18:05:45
## Thread
**1/** **@NOTimothyLottes** ^2083252794190901690
One unfortunate shift in 1st party Sony content has been from family friendly to a core focus on either extreme graphic violence or say non binary or etc. Ie, once upon a time there was Jak, then Uncharted, then firmly planted into adult themes by Last of Us and Intergalactic …
**2/** **@NOTimothyLottes** ^2083254001743630544
At some point the escape desired is actually to return to the whimsical or kid friendly non serious stuff that was the primary of say the PS2 generation, but Sony doesn’t really fund that kind of content any more.
**3/** **@NOTimothyLottes** ^2083255152094359929
Pre CG film had found a good balance of horror, they would show only just enough to let the imagination fill in most of it. With CG the focus was to show all, and thus suspense fell into the uncanny valley along with the scare.
**4/** **@NOTimothyLottes** ^2083256389019771238
With 1st party games focusing on adult-topic “reality” they just faceplant into the uncanny valley. When the point to buy a platform is just a more detailed uncanny valley, sorry, but the platform already rotted in many respects.
**5/** **@NOTimothyLottes** ^2083257233236713977
The best the “follow NV+UE crowd” (which Sony is firmly planted in now) has come up with is more AI AA and post process reskinning? Just more uncanny stuff? What about using technology for more fun instead?
**6/** **@NOTimothyLottes** ^2083258031886807240
The platform disconnect isn’t just the loss of physical media. It is much deeper than this …
**7/** **@NOTimothyLottes** ^2083259553995460709
Western kids and adults would turn to Japanese anime legends like the Studio Ghibli films, for both timeless and ageless universal experiences that would enrich the life of all who experienced it. And one would hope a gaming platform would also strive for that kind of mastery
Branches: [[archive/threads/NOTimothyLottes/2026-07-31-one-unfortunate-shift-in-1st-party-sony-content/2026-07-31-szspagna-ratchet-and-clank-is-the-only-remaining-callback]]
@@ -0,0 +1,52 @@
---
title: "@NOTimothyLottes https://n64brew.dev/wiki/Parallel_Interface"
type: archive
source: twitter
source_url: "https://x.com/Ed__dev/status/2083930871740928371"
author: "Ed_"
handle: Ed__dev
post_id: "2083930871740928371"
date: 2026-08-02
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "@NOTimothyLottes https://n64brew.dev/wiki/Parallel_Interface"
in_reply_to: ""
parent_post_id: "2083930291505131687"
---
## Source
- URL: https://x.com/Ed__dev/status/2083930871740928371
- Author: Ed_ (@Ed__dev)
- Posted: 2026-08-02 15:00:12
## Branch
**1/** **@Ed__dev** ^2083930871740928371
**@NOTimothyLottes**
https://n64brew.dev/wiki/Parallel_Interface
https://digital.pcea.net/issues/september-2024/gonzales/
https://easyfpga.blog/posts/high-speed-parallel-interface-design/
**2/** **@Ed__dev** ^2083934218418004107
**@NOTimothyLottes**
Only constraint maybe is the sdcard they use on the carts... or if the cart is doing something complicated beyond data pushing?
https://summercart64.dev/features.html
**3/** **@NOTimothyLottes** ^2083936799571808332
**@Ed__dev**
Same conclusion here, but they didn't yet get into what technology is on the M64 carts, so it's still an open question. When they say "immune" to bit rot, then I'm thinking the official carts don't use flash. But certainly a modern flash solution could max out the bus.
## Related
- Spine: [[archive/threads/NOTimothyLottes/2026-08-02-im-really-curious-what-the-m64-cart-bus-can]]
@@ -0,0 +1,44 @@
---
title: "I'm really curious what the M64 cart bus can actually pull off in DMA bandwidth (and capacity)."
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2083789486740079096"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2083789486740079096"
date: 2026-08-02
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "I'm really curious what the M64 cart bus can actually pull off in DMA bandwidth (and capacity)."
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2083789486740079096
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-08-02 05:38:23
## Thread
**1/** **@NOTimothyLottes** ^2083789486740079096
I'm really curious what the M64 cart bus can actually pull off in DMA bandwidth (and capacity). I thought vintage N64 carts had been limited by the storage tech on cart itself, not the bus ... how did that change with the M64?
**2/** **@Ed__dev** ^2083930291505131687
**@NOTimothyLottes**
Tried to look into it but so long as nothing was constrained on the underlying way to FPGA is wired up to that board the theoretical throughput is around 110–150 MB/s?
Branches: [[archive/threads/NOTimothyLottes/2026-08-02-im-really-curious-what-the-m64-cart-bus-can/2026-08-02-Ed__dev-https-n64brew-dev-wiki-parallel-interface-https]]
**3/** **@NOTimothyLottes** ^2083936224138457409
**@Ed__dev**
This is what I was thinking, but then I read something like real N64 games had been as low as 5 MiB/s to as high as 50 MiB/s which is quite a ways off from the estimated spec peak
@@ -0,0 +1,61 @@
---
title: "@NOTimothyLottes Wasn't that pretty much what the Saturn did?"
type: archive
source: twitter
source_url: "https://x.com/mbur82/status/2083886133923000690"
author: "MB82"
handle: mbur82
post_id: "2083886133923000690"
date: 2026-08-02
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "@NOTimothyLottes Wasn't that pretty much what the Saturn did?"
in_reply_to: ""
parent_post_id: "2083777028449263884"
---
## Source
- URL: https://x.com/mbur82/status/2083886133923000690
- Author: MB82 (@mbur82)
- Posted: 2026-08-02 12:02:25
## Branch
**1/** **@mbur82** ^2083886133923000690
**@NOTimothyLottes**
Wasn't that pretty much what the Saturn did?
**2/** **@NOTimothyLottes** ^2083923204146401663
**@mbur82**
Maybe but I don’t think any super-scaler did anything intelligent to handle overdraw (which is probably the critical missing element).
**3/** **@mbur82** ^2083924126926860519
**@NOTimothyLottes**
Nope. They had INSANE memory bandwidth for the time to deal with all the overdraw. Outrun from 1986 had an 80-bit VRAM bus which ran at 25MHz. That’s more than the PCI bus.
**4/** **@NOTimothyLottes** ^2083935522624319968
**@mbur82**
Haha, actually that was my point exactly, they solved overdraw with brute force. But had to scale back for say the Saturn's 48-bit VDP1 at 28MHz if my interneting was accurate.
**5/** **@NOTimothyLottes** ^2083937509717803511
**@mbur82**
It would be nice to get SGSSAA (super-sampling on the MSAA pattern) with 2.5D depth sprites, and hex grid for bitmaps instead of regular grid. Which is probably well beyond the possibility of brute force ...
## Related
- Spine: [[archive/threads/NOTimothyLottes/2026-08-02-m64-the-physical-fpga-psram-hardware-offers-such]]
@@ -7,7 +7,7 @@ author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2083774527255740917"
date: 2026-08-02
archived: 2026-08-25
archived: 2026-08-28
draft: false
tags:
- archive
@@ -41,6 +41,8 @@ The obvious stuff, like on M64, you have >200x {36-bit x 512 entry} memories, an
Most of the 3D HW that sits firmly in the post 2D to 3D transition region that has low enough clocks to FPGA emulate, seems quite strangled. Meaning an evolved sprite super-scalar (perhaps depth sprites) might actually have been a better alternative timeline to go after ...
Branches: [[archive/threads/NOTimothyLottes/2026-08-02-m64-the-physical-fpga-psram-hardware-offers-such/2026-08-02-mbur82-wasnt-that-pretty-much-what-the-saturn-did]]
**5/** **@NOTimothyLottes** ^2083778960945562067
Most of the puzzle of targeting a machine with massive ALU capacity is a compression problem, how to have a small number of 'things' transform into a lot of 'stuff' without round-tripping through memory
@@ -0,0 +1,42 @@
---
title: "Read index-able shift register saves on not needing address lines for the store, just need a clock enable bit."
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2084817993213583512"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2084817993213583512"
date: 2026-08-05
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "Read index-able shift register saves on not needing address lines for the store, just need a clock enable bit."
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2084817993213583512
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-08-05 01:45:18
## Thread
**1/** **@NOTimothyLottes** ^2084817993213583512
Read index-able shift register saves on not needing address lines for the store, just need a clock enable bit. Makes it temping for a bank of a register file, and a SLICEM gets 32-bit/s LUT, simple-dual-port on ultrascale+ peaks at 7/8ths that.
**2/** **@NOTimothyLottes** ^2084818888043135455
Array of shift registers can function as both an operand and result cache simultaneously. It can also provide sub-word addressing, depending on how operand and result bits are programmibly distributed across shift registers (for SIMD, requiring complex instruction encoding).
**3/** **@NOTimothyLottes** ^2084819391850377635
Of course the side effect of using a set of sub-word addressable multi-bank shift registers as a flexible register file, the index of their contents changes on each clock-enable. Meaning a human needs help programming ...
**4/** **@NOTimothyLottes** ^2084821615636836684
For context, yes it would be easy to just use 3 BRAMs to produce a 3 bank configuration for operand fetch, but then the BRAM:DSP ratio would be 3:1 which ultimately wouldn't be good utilization of a machine that has a physical 1:2 BRAM:DSP ratio.
@@ -7,7 +7,7 @@ author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2085367363373850680"
date: 2026-08-06
archived: 2026-08-25
archived: 2026-08-28
draft: false
tags:
- archive
@@ -7,7 +7,7 @@ author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2085365714412556730"
date: 2026-08-06
archived: 2026-08-25
archived: 2026-08-28
draft: false
tags:
- archive
@@ -0,0 +1,36 @@
---
title: "Silly quick test of what super-scalar sprites might have looked like with 8xSGSSAA (so nearest and still no mips) - LEFT [1x] | RIGHT [8x] - https://www.shadertoy.com/view/7fK3RV - Actually quite nice."
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2086458680086528056"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2086458680086528056"
date: 2026-08-09
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "Silly quick test of what super-scalar sprites might have looked like with 8xSGSSAA (so nearest and still no mips) - LEFT [1x] | RIGHT [8x] - https://www.shadertoy.com/view/7fK3RV - Actually quite nice."
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2086458680086528056
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-08-09 14:24:48
## Thread
**1/** **@NOTimothyLottes** ^2086458680086528056
Silly quick test of what super-scalar sprites might have looked like with 8xSGSSAA (so nearest and still no mips) - LEFT [1x] | RIGHT [8x] - https://www.shadertoy.com/view/7fK3RV - Actually quite nice.
![](https://pbs.twimg.com/media/HPSWoUBWYAAqbJp?format=png&name=orig)
**2/** **@NOTimothyLottes** ^2086460017469730977
Just a personal preference: for HW - rather drop texture filtering and up capacity of non-filtered for SGSSAA than the other way around - actual super-sampling is way more important
@@ -0,0 +1,90 @@
---
title: "It takes a special kind of stupidity to believe that police are needed because of inherent human nature, aka some people will do bad things in certain situations, and simultaneously believe that same truth doesn't apply to those controlling 1984 style surveillance."
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2090987060886654984"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2090987060886654984"
date: 2026-08-22
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "It takes a special kind of stupidity to believe that police are needed because of inherent human nature, aka some people will do bad things in certain situations, and simultaneously believe that same truth doesn't apply to those controlling 1984 style surveillance."
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2090987060886654984
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-08-22 02:18:58
## Thread
**1/** **@NOTimothyLottes** ^2090987060886654984
It takes a special kind of stupidity to believe that police are needed because of inherent human nature, aka some people will do bad things in certain situations, and simultaneously believe that same truth doesn't apply to those controlling 1984 style surveillance.
**2/** **@NOTimothyLottes** ^2090990982682575005
Basic calculus, what a system converges to at the limit of it's timeline. Even lacking in foresight, the vintage human minds have well explored what is happening now in the form of paper fiction, do we really want that world for our children?
**3/** **@NOTimothyLottes** ^2090992642867802327
Do we want to build a system that allows for absolute coercion of any targeted person? Something only people in power control. The critical slope is quite low, before natural convergence is unavoidable after establishing that searchable composite of enough of anyone's life.
**4/** **@NOTimothyLottes** ^2090993971963048061
Soon automatic fines for all the times you rolled a stop sign, or happened to be speeding and not actually harming anyone. Insurance rates that ratchet up based on accumulated prior behavior, like braking fast, even if no actual accident happened.
**5/** **@NOTimothyLottes** ^2090995589060190587
Eventually you'd get tickets for things that happened long ago, but they didn't have the computation time or say city water budget to detect at the time.
**6/** **@NOTimothyLottes** ^2090997035386261770
Then the concept of "socially acceptable" pre-crime enforcement extends. Today it's just targeting the minority group of people who like to drive fast, in the future analysis of your in-camera reaction time and it will strip seniors of their right to drive early.
**7/** **@NOTimothyLottes** ^2090998271380439520
At some point with the mosaic of your life and connections, auto-detection of things becomes possible. Like the high correlation between married men in power and cheating on their wives, there will be a full searchable history of that behavior. Hidden, to be used when needed.
**8/** **@NOTimothyLottes** ^2090999112611090504
Eventually universal eye tracking history of anyone on camera, combined with audio tracking, they will auto detect how you feel based on your response to your external stimulus. Listen to a political broadcast, your body language will expose what you are thinking.
**9/** **@NOTimothyLottes** ^2091002152101597221
1984 is a gravity trap. It's easy to approach the event horizon, and trigger the convergence. Just build enough of the tech and human nature does the rest automatically across time. Starts by criminalizing those who don't do actual harm but just disrupt the monitoring system.
**10/** **@NOTimothyLottes** ^2091003682393395339
It will be brutal in the US because of the flip flop of political power. Those groups targeted by the monitoring system will change with the political cycle. You might be safe today, and find your recorded history criminalizes you in the future.
**11/** **@NOTimothyLottes** ^2091005515589443935
Think about the scale of coercion with the COVID vax, forcing 2/3 of the population to be in a medical experiment with huge death toll and post-experiment injury, with no one held accountable? And that was pre-surveillance state, image what happens next time?
**12/** **@NOTimothyLottes** ^2091007731171729479
Do we really want our economic success to be tied to how well we build out the technology for 1984? Or instead perhaps we should reaffirm our original constitutional rights and expand on it with protections for this digital era so this remains a country of the free.
**13/** **@NOTimothyLottes** ^2091010803981770916
And for those who don't get the convergence slope function, like in COVID, even good doctors had to choose between loosing their license or just going along with it. Most did the human thing, what is best for self interest, keep quite ...
**14/** **@NOTimothyLottes** ^2091011614807937056
... Eventually if you disagree it comes at the cost of your ability to make a living or even your life, like perhaps any Germans in WW2 that didn't go along with it. That's the slope, the start of when the system goes out of human control.
**15/** **@NOTimothyLottes** ^2091012087723987129
When every aspect of your life is searchable, anything you do, say, think, etc, starts cutting off your future if you are anything other than exactly what the system wants in it's convergence. That's the dangerous slope.
**16/** **@NOTimothyLottes** ^2091013713780208120
Privacy in many respects is what enables individuality, what enables people with radically different thinking or beliefs to actually co-exist peacefully.
@@ -0,0 +1,46 @@
---
title: "The sticky slope to tyrantry starts with people that fear being targeted for expressing an opinion -> self sensoring, etc -> and one can already see this with flock, people don't want to sign a petition because they know that will become an instant list of targeted individuals"
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2091676077533897054"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2091676077533897054"
date: 2026-08-23
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "The sticky slope to tyrantry starts with people that fear being targeted for expressing an opinion -> self sensoring, etc -> and one can already see this with flock, people don't want to sign a petition because they know that will become an instant list of targeted individuals"
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2091676077533897054
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-08-23 23:56:53
## Thread
**1/** **@NOTimothyLottes** ^2091676077533897054
The sticky slope to tyrantry starts with people that fear being targeted for expressing an opinion -> self sensoring, etc -> and one can already see this with flock, people don't want to sign a petition because they know that will become an instant list of targeted individuals
**2/** **@NOTimothyLottes** ^2091677136624955844
We already see flock cameras placed to record just people entering a shooting range, sure not a problem while under republican rule, but when the political flip flop happens, the dems already have their next target list
**3/** **@NOTimothyLottes** ^2091678107912528237
Hardcore dems and republicans never decide elections because they are fixed in their ways, the reality is those that don't map well to either end up swinging left or right in mass and deciding things for everyone else
**4/** **@NOTimothyLottes** ^2091680926052815147
Right now it seems like more dems are involved in cancelling flock than the republicans, who if they double down, might force the hand of libertarians/etc to switch sides
**5/** **@NOTimothyLottes** ^2091682632773857441
This sticky slope issue is core in the convergence to a dystopia, any system designed for absolute crushing of difference of opinion converges into the dark ages. Once humans have trade their lively hood to make the right choice, history shows a seriously horrible outcome
@@ -0,0 +1,42 @@
---
title: "Tech overlords have long since realized just like chips hit power and thermal walls, and benefit/cost of process shrinks is leveling off, that the amount of $ they can extract from eyeball ad time was going asymptotic, and could go negative with population decline ->"
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2092405327845109889"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2092405327845109889"
date: 2026-08-26
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "Tech overlords have long since realized just like chips hit power and thermal walls, and benefit/cost of process shrinks is leveling off, that the amount of $ they can extract from eyeball ad time was going asymptotic, and could go negative with population decline ->"
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2092405327845109889
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-08-26 00:14:39
## Thread
**1/** **@NOTimothyLottes** ^2092405327845109889
Tech overlords have long since realized just like chips hit power and thermal walls, and benefit/cost of process shrinks is leveling off, that the amount of $ they can extract from eyeball ad time was going asymptotic, and could go negative with population decline ->
**2/** **@NOTimothyLottes** ^2092406097055293647
So naturally tech overlords are all trying to finish their evolution of full government infiltration, because if they can become something that gets taxed, that their customers would finally be fully enslaved and there is little in the way of maximum $ extraction ->
**3/** **@NOTimothyLottes** ^2092407490776301797
In the US we've seen the infiltration of many industries into things like building codes - the irony of weaponizing the epa to force the periodic landfilling of all the AC units every decade as new refrigerants are forced, simultaneously outlawing geothermal cooling, etc -
**4/** **@NOTimothyLottes** ^2092409254883856830
This idea of a "free economy" or "capitalism" almost doesn't really exist any more in the US, it's more "governmentalism" at this stage. We have just enough eroding freedom to almost have a shot at the American Dream --IF-- one operates outside a goven-monopolized industry.
@@ -0,0 +1,48 @@
---
title: "Thinking about visible pipelining in the ISA."
type: archive
source: twitter
source_url: "https://x.com/NOTimothyLottes/status/2092844404004266058"
author: "NOTimothyLottes"
handle: NOTimothyLottes
post_id: "2092844404004266058"
date: 2026-08-27
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- NOTimothyLottes
description: "Thinking about visible pipelining in the ISA."
in_reply_to: ""
---
## Source
- URL: https://x.com/NOTimothyLottes/status/2092844404004266058
- Author: NOTimothyLottes (@NOTimothyLottes)
- Posted: 2026-08-27 05:19:23
## Thread
**1/** **@NOTimothyLottes** ^2092844404004266058
Thinking about visible pipelining in the ISA.
For example,
IMAD p=a*b+c
Use {A,B} reg file indexes from instruction[0]
Use {C} reg file indexes from instruction[1]
To save on flipflops (CLBs), one less pipeline stage for C
C has 1 less clock of input latency
**2/** **@NOTimothyLottes** ^2092845415980826677
My plan for conditionals is to have a {U,V} register which is pipelined in parallel with the DSP, and a CLB between DSP.p and dRAM stores. The CLB conditionally selects the store between {mux(u,v) based on P.MSB, and P}. With {U,V} update conditional on opcode bits (CE control).
**3/** **@NOTimothyLottes** ^2092845964029521939
Meaning one can save a prior value into U or V, and then later do the test which sets a sign bit which controls which bank gets the U vs the V. So one can get a full compare and swap in one clock. Or store just one and discard the other (for predicated logic).
**4/** **@NOTimothyLottes** ^2092846391387193766
Meaning single clock store simultaneous min and max, etc. Stuff today's GPUs don't even have.
+12
View File
@@ -126,12 +126,24 @@ tags:
- [[archive/threads/NOTimothyLottes/2026-07-17-oh-f-it-the-urge-to-drop-the-c-baggage-was-too]]
- [[archive/threads/NOTimothyLottes/2026-07-18-all-0-6-argument-syscall-options-in-32-bytes]]
- [[archive/threads/NOTimothyLottes/2026-07-19-probably-shouldnt-consider-this-but-apparently]]
- [[archive/threads/NOTimothyLottes/2026-07-20-perhaps-the-most-important-aspect-of-ai-is-the]]
- [[archive/threads/NOTimothyLottes/2026-07-22-entertaining-aspect-of-ai-bubble-is-watching]]
- [[archive/threads/NOTimothyLottes/2026-07-22-working-through-rough-draft-of-assembly-for-a]]
- [[archive/threads/NOTimothyLottes/2026-07-29-seriously-fucking-awesome-in-a-world-of-total]]
- [[archive/threads/NOTimothyLottes/2026-07-29-think-about-it-the-m64-is-probably-going-to-be]]
- [[archive/threads/NOTimothyLottes/2026-07-31-one-unfortunate-shift-in-1st-party-sony-content]]
- [[archive/threads/NOTimothyLottes/2026-08-02-im-really-curious-what-the-m64-cart-bus-can]]
- [[archive/threads/NOTimothyLottes/2026-08-02-m64-the-physical-fpga-psram-hardware-offers-such]]
- [[archive/threads/NOTimothyLottes/2026-08-02-re-fpga-land-alus-im-firmly-distracted-by-an]]
- [[archive/threads/NOTimothyLottes/2026-08-05-read-index-able-shift-register-saves-on-not]]
- [[archive/threads/NOTimothyLottes/2026-08-06-seems-like-one-of-the-pain-points-with-xlinix]]
- [[archive/threads/NOTimothyLottes/2026-08-09-silly-quick-test-of-what-super-scalar-sprites]]
- [[archive/threads/NOTimothyLottes/2026-08-10-been-saying-this-for-a-decade-now-hu-has-a-really]]
- [[archive/threads/NOTimothyLottes/2026-08-18-cheers]]
- [[archive/threads/NOTimothyLottes/2026-08-22-it-takes-a-special-kind-of-stupidity-to-believe]]
- [[archive/threads/NOTimothyLottes/2026-08-23-operands-time-perf]]
- [[archive/threads/NOTimothyLottes/2026-08-23-the-sticky-slope-to-tyrantry-starts-with-people]]
- [[archive/threads/NOTimothyLottes/2026-08-25-one-downsides-to-being-a-left-justified-machine]]
- [[archive/threads/NOTimothyLottes/2026-08-26-for-an-integer-dsp-if-one-wants-2-negative]]
- [[archive/threads/NOTimothyLottes/2026-08-26-tech-overlords-have-long-since-realized-just-like]]
- [[archive/threads/NOTimothyLottes/2026-08-27-thinking-about-visible-pipelining-in-the-isa]]
+1
View File
@@ -57,6 +57,7 @@ Example of the note shape (single spine, no branches): [[archive/threads/How to
- [[archive/threads/rianflo]]
- [[archive/threads/rygorous]]
- [[archive/threads/SebAaltonen]]
- [[archive/threads/sh4forever]]
- [[archive/threads/SheriefFYI]]
- [[archive/threads/simplex_fx]]
- [[archive/threads/TheGingerBill]]
@@ -0,0 +1,74 @@
---
title: "I dev with the debugger attached 100% of the time, and unless I'm doing something visual like UI, I literally always put a breakpoint at the beginning of the new code I've just written and step through it before I ever let it \"run\"."
type: archive
source: twitter
source_url: "https://x.com/nicbarkeragain/status/2090577570916184127"
author: "Nic Barker"
handle: nicbarkeragain
post_id: "2090577570916184127"
date: 2026-08-20
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- nicbarkeragain
description: "I dev with the debugger attached 100% of the time, and unless I'm doing something visual like UI, I literally always put a breakpoint at the beginning of the new code I've just written and step through it before I ever let it \"run\"."
in_reply_to: ""
---
## Source
- URL: https://x.com/nicbarkeragain/status/2090577570916184127
- Author: Nic Barker (@nicbarkeragain)
- Posted: 2026-08-20 23:11:48
## Thread
**1/** **@nicbarkeragain** ^2090577570916184127
I dev with the debugger attached 100% of the time, and unless I'm doing something visual like UI, I literally always put a breakpoint at the beginning of the new code I've just written and step through it before I ever let it "run". About half the time I find a bug immediately.
**2/** **@MrJayLC** ^2090913197964341464
**@nicbarkeragain**
idk if you've seen @NOTimothyLottes video on C, but the "talk" function really *speaks* to me.
It's essentially what I do: Log statements everywhere and run the code (usually debugger attached). Either it all works, or I get some exception and iterate.
**3/** **@NOTimothyLottes** ^2090964526325731761
**@MrJayLC** **@nicbarkeragain**
Oh the classic problem. New programmer hired into a mountain of technical dept with no documentation and code that obfusticates meaning (like say Unreal). So yeah you'd use that debugger as an archaeologist tool to try to understand enough of a subset of it to make changes ->
**4/** **@NOTimothyLottes** ^2090965834185211915
**@MrJayLC** **@nicbarkeragain**
Some lucky people get good code bases with good docs and transcend the need for classic debuggers, typically need a new subset of tooling because the issues being solved are parallel or timing related optimizations -> need to run on a release build, etc
**5/** **@NOTimothyLottes** ^2090966335710732378
**@MrJayLC** **@nicbarkeragain**
And then need to see the relationship of behavior between parallel things. Then you want effectively minimal cost reporting or manual injection. Not 'printf' because that is a boat anchor, and writing to console screws timing ->
**6/** **@NOTimothyLottes** ^2090966786950807966
**@MrJayLC** **@nicbarkeragain**
Instead as minimal runtime effect as possible -> one atomic to pick the next line, and a store to a fixed size cacheline in a mapped page that could say be mapped even across multiple processes (so relationship between host and client is order accurate)
**7/** **@NOTimothyLottes** ^2090967640550289917
**@MrJayLC** **@nicbarkeragain**
No kernel calls or even page misses (mlockall) on runtime messaging, and when your project recompiles at runtime without exit, or even restarts in less than a second, you can track down issues faster than it takes to load visual studio.
**8/** **@NOTimothyLottes** ^2090968558394061175
**@MrJayLC** **@nicbarkeragain**
Seriously pro devs build the tooling into the engine, so they always know the timing and runtime behavior and work with eyes open, a full understanding of what is happening, ability to reason about why something might fail without even loading a debugger
+1
View File
@@ -16,4 +16,5 @@ tags:
- [[archive/threads/nicbarkeragain/2026-05-10-psa-if-youve-never-done-graphics-programming]]
- [[archive/threads/nicbarkeragain/2026-08-03-imo-basically-every-object-creation-api-you-have]]
- [[archive/threads/nicbarkeragain/2026-08-05-an-example-of-this-a-classic-way-to-construct-a]]
- [[archive/threads/nicbarkeragain/2026-08-20-i-dev-with-the-debugger-attached-100-of-the-time]]
- [[archive/threads/nicbarkeragain/2026-08-25-i-think-one-of-the-biggest-level-ups-i-had-in]]
@@ -0,0 +1,62 @@
---
title: "I wasn’t joking!"
type: archive
source: twitter
source_url: "https://x.com/rfleury/status/2086955595907891673"
author: "Ryan Fleury"
handle: rfleury
post_id: "2086955595907891673"
date: 2026-08-10
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- rfleury
description: "I wasn’t joking!"
in_reply_to: ""
---
## Source
- URL: https://x.com/rfleury/status/2086955595907891673
- Author: Ryan Fleury (@rfleury)
- Posted: 2026-08-10 23:19:22
## Thread
**1/** **@rfleury** ^2086955595907891673
I wasn’t joking!
![](https://pbs.twimg.com/media/HPZbDOiWoAADQqT?format=jpg&name=orig)
**2/** **@NOTimothyLottes** ^2086974195846496384
**@rfleury**
Move up a level, don't limit to N64 on M64, build your own bitstream and see what can be done on the raw FPGA hardware
**3/** **@Karyuutensei** ^2087072033142157422
**@NOTimothyLottes** **@rfleury**
What do you mean? Can the m64 FPGA be reconfigured?
**4/** **@NOTimothyLottes** ^2087177541249356110
**@Karyuutensei** **@rfleury**
Modretro says yes, youd be able to load other bitstreams on it similar to a Mister. With open HW docs. It’s a really low price device with a quite powerful FPGA and 4 parallel PSRAMs. Perfect for fantasy console HW dev.
**5/** **@Karyuutensei** ^2087240941425877352
**@NOTimothyLottes** **@rfleury**
I’ve implemented the Pokémon mini on mister but implementing a CPU from a spec + other information sources is one thing, coming up with your own design is another. Where do you even begin?
**6/** **@NOTimothyLottes** ^2087254893761343529
**@Karyuutensei** **@rfleury**
I’d suggest working backwards from knowledge of what doesn’t map well to the FPGA and start removing stuff that is not needed, see how simple something can become and still be useful, then make that as parallel as possible …
+1
View File
@@ -51,3 +51,4 @@ tags:
- [[archive/threads/rfleury/2026-03-05-this-ignores-an-important-shift-in-the-gaming]]
- [[archive/threads/rfleury/2026-06-02-the-first-way-people-usually-learn-code]]
- [[archive/threads/rfleury/2026-08-07-a-good-way-to-evaluate-one-axis-of-quality-of-an]]
- [[archive/threads/rfleury/2026-08-10-i-wasnt-joking]]
@@ -0,0 +1,34 @@
---
title: "I've had a few comments in posts which were really thoughtful questions and ideas."
type: archive
source: twitter
source_url: "https://x.com/sh4forever/status/1725666232861696116"
author: "sh4"
handle: sh4forever
post_id: "1725666232861696116"
date: 2023-11-18
archived: 2026-08-28
draft: false
tags:
- archive
- twitter
- sh4forever
description: "I've had a few comments in posts which were really thoughtful questions and ideas."
in_reply_to: ""
---
## Source
- URL: https://x.com/sh4forever/status/1725666232861696116
- Author: sh4 (@sh4forever)
- Posted: 2023-11-18 00:04:25
## Thread
**1/** **@sh4forever** ^1725666232861696116
I've had a few comments in posts which were really thoughtful questions and ideas. For what it's worth, once we reach the point that texturing and reasonable performance have been achieved, one of my main goals for the project is to produce a few videos about the architecture.
**2/** **@sh4forever** ^1725666234140877077
I think this would be interesting for anyone curious about how GPUs can work, why immediate mode rendering wasn't an option, how we solved the follow-up problem of binning, rasterization techniques and optimizations we're doing, how the VPU shader cores work, etc.
+12
View File
@@ -0,0 +1,12 @@
---
title: sh4forever
type: note
draft: false
description: Archived threads by sh4forever.
tags:
- archive
- twitter
- sh4forever
---
- [[archive/threads/sh4forever/2023-11-18-ive-had-a-few-comments-in-posts-which-were-really]]