emoose

Seems the `disassemble_functions=true` fix works because with that set, `PPCHIRBuilder::AnnotateLabel` is used to set `Label::name` field to a string of the form `loc_{address}`, but when `disassemble_functions` isn't set AnnotateLabel is...

Ah right, I did notice that the benchmark line by itself would also give the same boost too, but included the enable line just to be safe, updated the OP...

@Small-tailqwq yep that should enable it fine

@Small-tailqwq if you didn't already maybe try doing multiple runs with it, for me it only started showing a good improvement from my second txt2img generation onward for some reason....

Seems like the size you set in train page only gets used by the embedding trainer, you can see the code for training embeddings makes use of `width=training_width, height=training_height`: https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/c1093b8051606f0ac90506b7114c4b55d0447c70/modules/textual_inversion/textual_inversion.py#L235...

Huh that's weird, wonder how it ended up like that, pretty sure the kernel funcs for XEXs only allow access to the resources by name, so if two resources shared...

Got something working for it in https://github.com/emoose/idaxex/commit/fb281eca0e21a90689d996d46efbec78d82bce33, seems to work fine when I edit an XEX to use duplicate names at least, first one will have normal name while duplicates...

Nice idea, looks like pointer to the inventory character `cModel` is kept inside `g_pPl` pointer (value of 0xC7BA6C in 1.1.0), editing the `ang_A0.x/.y/.z` fields of that seems to rotate them...

Hm, AFAIK the TU only applies to Disk 1, at least there's only the one xexp inside it... Also with the latest canary build you should be able to keep...

I guess either Disk 2 is already patched with latest TU (the ISO seems to contain an update folder at least), or maybe it uses a different TU file to...