#Rasterization
Rasterization of an idea
November 21, 2026 at 2:21 AM
Rasterization cut the exported PDF size of each graph from 10mb down to 27kb. This is not a joke. I probably should've realized a single graph taking up more space than an entire book was weird but I didn't.
November 19, 2026 at 2:20 PM
This WebGL app uses the vertex processor to do GPU compute. Behind the scene, a plane with 9 points is used to create 2 textures of blended vectors. The magic happens when per-vertex rasterization interpolates across the surface of the plane to create potentially millions of blended vectors for free
WebGL orthographic plane with rotation textures
Shared via Claude, an AI assistant from Anthropic
claude.ai
November 16, 2026 at 2:36 PM
I don't blame anything other than programming careers and bootcamps not teaching data structures and compsci fundamentals, NVIDIA's push towards realtime-inviable rendering techniques, and Epic Games's engine design decisions. So much awesome rasterization tech is stuck in custom engine limbo.
November 16, 2026 at 7:02 PM
Why do vector graphics stay sharp when resized? Unlike pixel-based images, they use mathematical paths and Bézier curves to render shapes at different resolutions. Explore how SVG, rasterization and antialiasing work—and where scaling still has limits.

techamerica.ai/technology-a...
How Vector Graphics Rendering Works | TechAmerica
A logo can stay crisp from a tiny screen to a large sign. Explore vector paths, SVG, rendering limits, file formats and accessible design.
techamerica.ai
November 19, 2026 at 1:17 AM
Why do vector graphics stay sharp when resized? Unlike pixel-based images, they use mathematical paths and Bézier curves to render shapes at different resolutions. Explore how SVG, rasterization and antialiasing work—and where scaling still has limits.

techamerica.ai/technology-a...
How Vector Graphics Rendering Works | TechAmerica
A logo can stay crisp from a tiny screen to a large sign. Explore vector paths, SVG, rendering limits, file formats and accessible design.
techamerica.ai
November 19, 2026 at 1:04 AM
RTX 5070 or RX 9070 XT? The 9070 XT wins on raster performance and VRAM (16GB vs 12GB); the 5070 wins on ray tracing and DLSS 4 frame generation. Which one belongs in your build depends on which of those you'll actually use.
RTX 5070 vs RX 9070 XT: Which GPU Wins in 2026? | PlanMyPC Guides
RX 9070 XT wins rasterization, VRAM and price-per-frame; RTX 5070 wins ray tracing and DLSS 4 MFG. Benchmarks, VRAM analysis and which card belongs in your build.
planmypc.com
November 14, 2026 at 3:53 PM
🚨 EUVD-2026-121441
📊 8.8/10
🏢 Microsoft

📝 Access of resource using incompatible type ('type confusion') in Windows Advanced Rasterization Platform allows an authorized attacker to elevate priv...

🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-121441

#cybersecurity #infosec #cve #euvd
November 12, 2026 at 10:02 PM
🚨 EUVD-2026-121612
📊 7.8/10
🏢 Microsoft

📝 Heap-based buffer overflow in Windows Advanced Rasterization Platform allows an authorized attacker to elevate privileges locally.

🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-121612

#cybersecurity #infosec #cve #euvd
November 12, 2026 at 10:04 PM
Working on making #klartraum rendering fully differentiable by moving from #glsl to #slang. Plan is to distill outputs of a generative model into neural assets usable in klartraum. The example combines rasterization of a cube using a neural texture followed by a U-net decoder.
November 12, 2026 at 7:30 PM
LightOnOCR-3-0.8B hits 85.5 on olmOCR-Bench, yet old scans fall to 48.7 and bursty RAG pipelines still pay the rasterizer and idle-GPU bill. Full breakdown on when to self-host vs calling an async PDF-to-Markdown API.
Self-Hosted OCR Models vs. Hosted PDF APIs for Bursty Agent Workflows
LightOnOCR-3-0.8B hits 85.5 on olmOCR-Bench, yet old scans drop to 48.7 and bursty RAG pipelines still pay the rasterization and idle-GPU tax.
reidmarlow.com
November 12, 2026 at 2:11 PM
I think I read somewhere that it could be faster this way ? But I'm not sure, since I need to reupload the atlas, wheras Slug would allow to do the rasterization on the GPU texture directly
November 9, 2026 at 12:54 PM
This is turning out to be a complete rewrite of my rasterization core 😅
November 9, 2026 at 4:08 AM
Avoiding rasterization fallback for transparency and blending is huge for PDF portability. Once a graphics device flattens plots into bitmaps or skips native font embedding, text selection breaks and zoomed charts turn muddy across PDFium and Preview.
November 9, 2026 at 5:00 PM
barycentric coordinates are important enough in software rasterization to fund this effort, and will allow me to free up a ton of registers and stack management ceremony in my rasterization code. Groups of dedicated registers passed through functions is always a smell when it happens anyway.
November 7, 2026 at 11:30 PM
Ok so the issue I have right now in my rasterizer is that I have these artifacts that are the result of false negatives in barycentric coord tests, and I assume it is a precision issue. I need to invest more effort into my barycentric coordinate assembly code
November 7, 2026 at 11:30 PM
Checking DevTools Network is the cleanest way to verify on-device tools. Merging is mostly xref stitching and /Pages tree pointer updates—it doesn't require server-side rasterization. Running it locally via JS or WebAssembly handles multi-megabyte docs in memory without outbound requests.
November 7, 2026 at 3:04 PM
I think NVIDIA and AMD are already planning to get rid of legacy hardware like rasterization to free up more space for FP4 units.
November 6, 2026 at 4:09 PM
November 3, 2026 at 7:17 PM
Rasterization-based discontinuous integration on the GPU
https://arxiv.org/pdf/2611.05186
Manuel Saberi, Muqtasid Tashfin, Andreas Vogel.
https://arxiv.org/abs/2611.05186
arXiv abstract link
arxiv.org
November 4, 2026 at 9:32 AM
Manuel Saberi, Muqtasid Tashfin, Andreas Vogel: Rasterization-based discontinuous integration on the GPU https://arxiv.org/abs/2611.05186 https://arxiv.org/pdf/2611.05186 https://arxiv.org/html/2611.05186
November 4, 2026 at 6:54 AM
There's a name-worthy concept here;

I'd call it Infinite-Depth Rasterization.
November 2, 2026 at 3:05 PM
Chrome/Electron (Blink+Skia) keeps layout tighter because the engine is shared, but font rasterization still varies by platform backend, so weight, sharpness and occasional wrapping differ. LO's variance is larger for full office docs that target print fidelity.
November 3, 2026 at 3:32 AM
World kernel:
Canonical state, validated mutations, deterministic evolution

Representation:
Multiple independent renderers, shared snapshots

Computation:
Exact banded rasterization and pooled fitting

Persistence:
Save/restore, deterministic replay
October 30, 2026 at 11:45 PM