Update blog

Author: cmdr2

content/blog/2025-11-05-1762336053.md

@@ -0,0 +1,18 @@
+---
+title: "Post from Nov 05, 2025"
+date: 2025-11-05T09:47:33
+slug: "1762336053"
+tags:
+  - easydiffusion
+  - sdkit
+---
+
+Following up to the [deep-dive on ML compilers](https://cmdr2.github.io/notes/2025/11/1762335811/):
+
+sdkit v3 won't use general-purpose ML compilers. They aren't yet ready for sdkit's target platforms, and need a lot of work (well beyond sdkit v3's scope). But I'm quite certain that sdkit v4 will use them, and sdkit v3 will start making steps in that direction.
+
+For sdkit v3, I see two possible paths:
+1. Use an array of vendor-specific compilers (like TensorRT-RTX, MiGraphX, OpenVINO etc), one for each target platform.
+2. Auto-generate ggml code from onnx (or pytorch), and beat it on the head until it meets sdkit v3's [performance goals](https://cmdr2.github.io/notes/2025/10/1760085894/). Hand-tune kernels, contribute to ggml, and take advantage of ggml's multi-backend kernels.
+
+Both approaches provide a big step-up from sdkit v2 in terms of install size and performance. So it makes sense to tap into these first, and leave ML compilers for v4 (as another leap forward).