Rosy

[cloud rivet]

I need to figure out a dev loop, I think a test makes sense, write a test, run spirv-val

I'm not making a shading language, I'm going to look into generating spirv with my application code

[cloud rivet]

going to give imhex another try

[cloud rivet]

ezpz

[astral hinge]

"someone" should make a .hexpat for spirv

I could try if you want

[cloud rivet]

oh if you want to. I am kind of figuring it out still

[astral hinge]

oh someone already made a spirv hexpat

[cloud rivet]

oh nice

[astral hinge]

you can find it in the content store

[cloud rivet]

https://docs.werwolv.net/imhex/views/pattern-editor is what I have been reading

[brisk chasm]

that reminds me of the #vulkan channel topic

oops itnt scroll

[astral hinge]

that's fine man

[cloud rivet]

cool thanks!

[astral hinge]

this hexpat could be improved

let me try something

try replacing the definition of Instruction near the bottom of the hexpat with this, then click the green arrow to evaluate the file again

struct Instruction {
    Op op;
    u16 words;
    u32 operands[words-1];
} [[format("FormatInstruction")]];

fn FormatInstruction(ref Instruction i) {
    return std::format("{}", i.op);
};

I just added a formatter so it shows the op name instead of { ... } in the value column

a more advanced formatter would look at the op and then the operands

[cloud rivet]

ohhh

I think I might be missing a feature

oh it says I can install the standard library

[astral hinge]

oh I forgot to give you another line

[cloud rivet]

oh had to add #include <std/string.pat>

[astral hinge]

or import std.io;

[cloud rivet]

awsome thank you

btw if you think generating shader from application code is weird, that's what I did for my blockens voxel game lol, the first thing I ever made with opengl

https://github.com/btipling/blockens/blob/main/src/game/gfx/vertex_shader_gen.zig#L169

        fn gen_inline_mesh_transforms(r: *runner) !void {
            if (r.cfg.mesh_transforms == null) return;
            var m = zm.identity();
            if (r.cfg.mesh_transforms) |mts| {
                const mt = mts[0];
                if (mt.scale) |s| {
                    m = zm.mul(m, zm.scalingV(s));
                }
            }
            const mr = zm.matToArr(m);
            var line = try shader_helpers.vec4_to_buf("    vec4 sc0 = vec4({d}, {d}, {d}, {d});\n", mr[0], mr[1], mr[2], mr[3]);
            r.l(&line);
            line = try shader_helpers.vec4_to_buf("    vec4 sc1 = vec4({d}, {d}, {d}, {d});\n", mr[4], mr[5], mr[6], mr[7]);
            r.l(&line);
            line = try shader_helpers.vec4_to_buf("    vec4 sc2 = vec4({d}, {d}, {d}, {d});\n", mr[8], mr[9], mr[10], mr[11]);
            r.l(&line);
            line = try shader_helpers.vec4_to_buf("    vec4 sc3 = vec4({d}, {d}, {d}, {d});\n", mr[12], mr[13], mr[14], mr[15]);
            r.l(&line);
            r.a("    mat4 scam = mat4(sc0, sc1, sc2, sc3);\n");
        }

:D

[astral hinge]

once c++ lets you reflect expressions (not in c++26 sadly), then we could generate whole shaders from c++ code

[brisk chasm]

i really like your hibtug pfp, bjorn

mine looks like a grumpy old man lol

[astral hinge]

yeah idk who that guy is but he looks very wise and kind

[cloud rivet]

you guys are wise and kind

I'm hoping this spirv detour won't be too much of a rabbithole

[astral hinge]

what are you doing exactly?

[cloud rivet]

since I'm not making a general feature for others to use

my goal is to have the pixels the way I want

[brisk chasm]

pixel.goHere(); pixel.goThere();

[cloud rivet]

but without having to use a shader language

yes like that

only more like drawSkimmer() or something

[brisk chasm]

bjorn->switchToRole("turtle");

[cloud rivet]

why turtle

[brisk chasm]

turtle grafics 😄

[astral hinge]

[cloud rivet]

man tomorrow is a workday I better go to bed, thanks for all the help! 🐢

[brisk chasm]

[cloud rivet]

🇬🇳

[cloud rivet]

Once I have a p_spv_write_header function and a p_spv_write_ops function that just takes an arena, context and ops code and a size and a pointer to bits the lowest level SPIR-V stuff is done. Then I just write some functions for each ops I want to use that just writes to 32 bits aligned memory. I will need an ssa table of values. Build out some helper functions for doing common things

Then bob’s your uncle

Maybe I can use _Generic and typeof for the data

[brisk chasm]

ops... ssa...

why do c people write cryptic code 🙂

and what is p_ supposed to mean, procedure? ie return void?

or is it function pointer prototypes? hence p_?

[pseudo dock]

That was a short amount of time between last message before sleep and first message of the day. I know that feeling when you are focused on a problem and even when you wake up you're still thinking about it ☺️

[brisk chasm]

indeed, barely 5hrs

[slim oak]

what if Bjorn is actually still asleep and is just sleeptexting

[hushed creek]

hmm

[cloud rivet]

ssa is for single static assignment

It’s in the SPIRV spec

Ops are also SPIRV

[brisk chasm]

forbidden cheese knowledge 🙂

[cloud rivet]

OpsCode etc

p_ is my application’s prefix because there are no namespaces

All my global functions and names have the prefix to avoid collisions

With win32 global names

One letter is probably not safe enough

[pseudo dock]

And 0x07230203?

[cloud rivet]

That’s the SPIRV magic number

All binaries begin with it

[brisk chasm]

Pjorn 🙂

[cloud rivet]

Five hours is a long amount of sleep for me lol

I basically slept in

I am well rested

[pseudo dock]

Living the dream! I used to be able to do that... Years ago... 👴

[broken fog]

that's not good for you

i'd know because i do the same

[cloud rivet]

I turn 50 in a year fwiw

[brisk chasm]

7 more years for me

[bronze socket]

maybe you have that low sleep gene

[brisk chasm]

the shorter the telomeres get the less schleep you need 😄

[bronze socket]

I need the full 8h, I can go off of 6h if I pay it back

[broken fog]

i hope i do cause otherwise i'm gonna die

[cloud rivet]

I used to sleep more

Sleeping less as you get older is common I think?

idk

[true moon]

yeah to some extent I think

I can function more or less fine off of anything more than 3 hours of sleep but yeah I need to pay it back asap

more than one day in a row of <7h sleep wears me down rapidly

really any sleep is better than none

I've pulled exactly one all nighter in my life and I slept through every class the following day

and got like a 60% on the assignment I stayed up all night to do anyways

but one or two full sleep cycles makes the next day more or less normal

[astral hinge]

I feel like doodoo all day if I get less than 8 hours of sleep

[true moon]

It really matters where in the cycle you wake up

I feel better off of 3 hours of sleep than I do off of 5

sleep cycles are ~1.5hr

[astral hinge]

Yeah I got like 5 hours last night and I feel ok. I'll probably become very sleepy soon though

[cloud rivet]

A few years ago I was doing really well getting 8 hours of sleep every day, and during that time if I slept anything less than 8 hours I would feel terrible. Now I have just normalized feeling terrible and it just becomes the new normal

wfm

[astral hinge]

homeostasis w I guess

now when you get 8 hours you'll feel incredible

[cloud rivet]

I'm sure I would be better off if I slept 8 hours, but I would only be able to live for work if I did that and I can't live like that.

[bronze socket]

it's a lot more tolerable to feel awful all the time if you don't exercise imo

it also contributes heavily

[cloud rivet]

I exercise a lot

5 days a week + 10K steps a day

I get exercise like at 11pm to 1 am lol

I can't exercise immediately after work I'm too beat up and tired

I do some game dev, that recharges me, I exercise and go to sleep

I am reducing my strength training a bit to lose weight for the next quarter or so

[cloud rivet]

I used GLSL to make some place holder spirv

to build a development render for the spv I'll generate

back to trongle I go

next is set up a test for generating valid spirv, then I'll replace the fragment shader with my own spirv

I took this GLSL directly from LOGL

for the fragment shader

I have regressed

1 step forward 5 months of steps back, I always say

I just need to generate

; SPIR-V
; Version: 1.0
; Generator: Google Shaderc over Glslang; 11
; Bound: 14
; Schema: 0
               OpCapability Shader
          %1 = OpExtInstImport "GLSL.std.450"
               OpMemoryModel Logical GLSL450
               OpEntryPoint Fragment %main "main" %FragColor
               OpExecutionMode %main OriginUpperLeft
               OpSource GLSL 330
               OpSourceExtension "GL_GOOGLE_cpp_style_line_directive"
               OpSourceExtension "GL_GOOGLE_include_directive"
               OpName %main "main"
               OpName %FragColor "FragColor"
               OpDecorate %FragColor Location 0
       %void = OpTypeVoid
          %3 = OpTypeFunction %void
      %float = OpTypeFloat 32
    %v4float = OpTypeVector %float 4
%_ptr_Output_v4float = OpTypePointer Output %v4float
  %FragColor = OpVariable %_ptr_Output_v4float Output
    %float_1 = OpConstant %float 1
  %float_0_5 = OpConstant %float 0.5
%float_0_200000003 = OpConstant %float 0.200000003
         %13 = OpConstantComposite %v4float %float_1 %float_0_5 %float_0_200000003 %float_1
       %main = OpFunction %void None %3
          %5 = OpLabel
               OpStore %FragColor %13
               OpReturn
               OpFunctionEnd

if this ends up not feeling like something I want to continue to do I may just go to GLSL

or idk, do something else with my time

maybe just do software rasterization

or do something with cuda

oh hrm

[pseudo dock]

What about HLSL? I know you rejected slang but I can't remember your reasons for disliking HLSL

[cloud rivet]

I don't know anything about HLSL

it's very different

I've been using slang not HLSL

I don't think it does bindless the same

slang has HLSL like syntax

hrmmmmmmmmmmmmmmmmmmmmm

I'm approaching a precipice

those RT tlas VVLs man, they suddenly just started annoying me so much I couldn't take it anymore, and then trying to fix it felt bad all the way down

[pseudo dock]

What are you thinking? You could ignore RT (and other graphics stuff) for now and keep working on the gameplay and mechanics?

[solid grove]

work on gameplay and mechanics? in this server??

[silk pier]

its broken c++ with even more broken templates

imagine Slang, but:

1. templates, not generics
2. buggy and broken with spirv
3. fixing bugs takes a long time

https://www.abolishcrlf.org//horror/

[cloud rivet]

I think there's a path to get shader language free real time rendering on the GPU

without having any spirv involved

A Vulkan memory object exported using VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_BIT can be imported into CUDA using the NT handle associated with that object as shown below. Note that CUDA does not assume ownership of the NT handle and it is the application’s responsibility to close the handle when it is not required anymore. The NT handle holds a reference to the resource, so it must be explicitly freed before the underlying memory can be freed.

and

I would get step debugging on the gpu in visual studio

and, I'd get RT pipeline with OptiX

and it would render to memory owned by VK so I can just present it

and I don't have to write any shader code in anything but C

I'd have to give up clang for nvcc

[solid grove]

I think that extension is intended for sharing textures across process, like how video players do for security

[cloud rivet]

it's just memory

[solid grove]

eg the video decode and player service runs out-of-process and then the app process displays the rendered output

[pseudo dock]

(oh good, you're back, I was afraid you had fallen over the p_recipice!)

[cloud rivet]

across process?

An application may wish to reference device memory in multiple Vulkan logical devices or instances, in multiple processes, and/or in multiple APIs. This extension enables an application to export Windows handles from Vulkan memory objects and to import Vulkan memory objects from Windows handles exported from other Vulkan memory objects or from similar resources in other APIs.

the or

in and/or

leads me to think it's not about multiple processes

https://github.com/nvpro-samples/gl_vk_simple_interop/blob/6a306b198f4ca4835e4380cd01ebeec09edd64cd/compute.hpp#L251

gl and vk using it in the same process

https://www.gpultra.com/blog/vulkan-cuda-memory-interoperability/

HANDLE handle;

PFN_vkGetMemoryWin32HandleKHR vkGetMemoryWin32HandleKHR = 
    (PFN_vkGetMemoryWin32HandleKHR)instance.getProcAddress("vkGetMemoryWin32HandleKHR");
if (vkGetMemoryWin32HandleKHR == nullptr) {
    // handle missing function
}

VkMemoryGetWin32HandleInfoKHR info =
    { VK_STRUCTURE_TYPE_MEMORY_GET_WIN32_HANDLE_INFO_KHR };
info.memory = vkDeviceMemory;
info.handleType = VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_BIT;

if (vkGetMemoryWin32HandleKHR(vkDevice,&info,&handle) != VK_SUCCESS) {
    // handle error
}

CUexternalMemory externalMemory;
CUDA_EXTERNAL_MEMORY_HANDLE_DESC desc;

desc.type = CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32;
desc.handle.win32.handle = handle;
desc.size = size;

if (cuImportExternalMemory(&externalMemory,&desc) != CUDA_SUCCESS) {
    // handle error
}

cuda has a cudaWaitExternalSemaphoresAsync

https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__EXTRES__INTEROP.html#group__CUDART__EXTRES__INTEROP_1g52ecbdf57e75f0fde9fa44a2f9532c2d

https://github.com/NVIDIA/cuda-samples/blob/master/Samples/5_Domain_Specific/vulkanImageCUDA/README.md

This sample demonstrates Vulkan Image - CUDA Interop. CUDA imports the Vulkan image buffer, performs box filtering over it, and synchronizes with Vulkan through vulkan semaphores imported by CUDA. This sample depends on Vulkan SDK, GLFW3 libraries, for building this sample please refer to "Build_instructions.txt" provided in this sample's directory

https://github.com/NVIDIA/cuda-samples/blob/c94ff366aed18c797b8a85dfaac7817b0228b420/Samples/5_Domain_Specific/simpleVulkan/main.cpp#L483

        cudaExternalSemaphoreWaitParams waitParams = {};
        waitParams.flags                           = 0;
        waitParams.params.fence.value              = waitValue;

        cudaExternalSemaphoreSignalParams signalParams = {};
        signalParams.flags                             = 0;
        signalParams.params.fence.value                = signalValue;
        // Wait for vulkan to complete it's work
        checkCudaErrors(cudaWaitExternalSemaphoresAsync(&m_cudaTimelineSemaphore, &waitParams, 1, m_stream));
        // Now step the simulation
        m_sim.stepSimulation(time, m_stream);
        // Signal vulkan to continue with the updated buffers
        checkCudaErrors(cudaSignalExternalSemaphoresAsync(&m_cudaTimelineSemaphore, &signalParams, 1, m_stream));

this is the holy grail as far as I'm concerned

[silk pier]

you can compile with clang and nvcc

nvcc compiles your cuda stuff, you call that from your clang cpu stuff

[cloud rivet]

and then link dynamically?

[silk pier]

uhh, I don't know the specifics too well.
I've only ever used cuda via the magic tarp visual studio throws over it all so that it works seamlessly

[cloud rivet]

nice

[silk pier]

@ sam0oneau should know, I guess you'd ask in #gpu-compute

[cloud rivet]

thank you!

I want the ability to step through on the gpu in visual studio

that's really cool

I don't really need to hang on to clang, the issue would be, I think I'd be starting a new project I'm so deep into clang right now

which is fine

[solid grove]

i don't think the vs step debugger is a live debug, I'm pretty sure it's the same as renderdoc where it's a simulated replay

also the vs step debugger has been deprecated, PIX is the replacement

[cloud rivet]

nvidia ships a visual studio thing

[silk pier]

it can't be a 'live' debugger (per thread), that wouldn't make any sense given current hardware limitations (not without side effects that is)

[cloud rivet]

I think you're talking about something else

[solid grove]

NSight has a real live debug but it requires 2 gpus, one to run the game and one to run the debugger

[cloud rivet]

renderdoc debugger steps through a shading language

and it's bad at it

[silk pier]

I believe this is mostly the fault of the compiler (glslang, slang, etc...)

compilers are supposed to link each spirv op to a line in the non-compiled code by putting DebugOp's in the spirv (if they want complete and accurate stepping): https://registry.khronos.org/SPIR-V/specs/unified1/DebugInfo.html

the problem here is that this becomes a very messy ordeal

[cloud rivet]

hrm I see

[silk pier]

I don't really know too much about the debug ops though, haven't had any time to play with them-- someone else probably knows more on this than me.

[cloud rivet]

I'm gonna start a new project and see where it goes I guess.

there's no hope for me

[solid grove]

no more skimmer racing? 😢

[cloud rivet]

no I think I can copy a lot of that code over

idk

I think this is a cool project idea

I don't think many people are doing a thing like this, I could be wrong

[pseudo dock]

The precipice!

I feel similarly to Ravbug, though, I will miss the skimmer racing. But it's your own free time, you should definitely spend it doing what you want 🙂

[cloud rivet]

I should have never tried vcc tbh

:\ProgramData\NVIDIA Corporation\OptiX SDK 9.0.0\SDK>dir
 Volume in drive C is OS
 Volume Serial Number is 5C1A-ED43

 Directory of C:\ProgramData\NVIDIA Corporation\OptiX SDK 9.0.0\SDK

11/26/2025  10:43 PM    <DIR>          .
11/26/2025  10:43 PM    <DIR>          ..
11/26/2025  10:43 PM    <DIR>          CMake
01/29/2025  04:29 PM            25,337 CMakeLists.txt
11/26/2025  10:43 PM    <DIR>          cuda
11/26/2025  10:43 PM    <DIR>          data
01/29/2025  04:06 PM             1,132 INSTALL-LINUX.txt
01/29/2025  04:06 PM             2,540 INSTALL-WIN.txt
11/26/2025  10:43 PM    <DIR>          lib
11/26/2025  10:43 PM    <DIR>          optixBoundValues
11/26/2025  10:43 PM    <DIR>          optixCallablePrograms
11/26/2025  10:43 PM    <DIR>          optixClusterStructuredMesh
11/26/2025  10:43 PM    <DIR>          optixClusterUnstructuredMesh
11/26/2025  10:43 PM    <DIR>          optixCompileWithTasks
11/26/2025  10:43 PM    <DIR>          optixConsole
11/26/2025  10:43 PM    <DIR>          optixCurves
11/26/2025  10:43 PM    <DIR>          optixCustomPrimitive
11/26/2025  10:43 PM    <DIR>          optixCutouts
11/26/2025  10:43 PM    <DIR>          optixDenoiser
11/26/2025  10:43 PM    <DIR>          optixDynamicGeometry
11/26/2025  10:43 PM    <DIR>          optixDynamicMaterials
11/26/2025  10:43 PM    <DIR>          optixHair
11/26/2025  10:43 PM    <DIR>          optixHello
11/26/2025  10:43 PM    <DIR>          optixMeshViewer
11/26/2025  10:43 PM    <DIR>          optixMixSDKs
11/26/2025  10:43 PM    <DIR>          optixModuleCreateAbort
11/26/2025  10:43 PM    <DIR>          optixMotionGeometry
11/26/2025  10:43 PM    <DIR>          optixMultiGPU
11/26/2025  10:43 PM    <DIR>          optixNeuralTexture
11/26/2025  10:43 PM    <DIR>          optixNVLink
11/26/2025  10:43 PM    <DIR>          optixOpacityMicromap
11/26/2025  10:43 PM    <DIR>          optixOpticalFlow
11/26/2025  10:43 PM    <DIR>          optixPathTracer
11/26/2025  10:43 PM    <DIR>          optixRaycasting
11/26/2025  10:43 PM    <DIR>          optixRibbons
11/26/2025  10:43 PM    <DIR>          optixSimpleMotionBlur
11/26/2025  10:43 PM    <DIR>          optixSphere
11/26/2025  10:43 PM    <DIR>          optixTriangle
11/26/2025  10:43 PM    <DIR>          optixVolumeViewer
11/26/2025  10:43 PM    <DIR>          optixWhitted
01/29/2025  04:06 PM             2,512 sampleConfig.h.in
11/26/2025  10:43 PM    <DIR>          support
11/26/2025  10:43 PM    <DIR>          sutil
               4 File(s)         31,521 bytes

[silk pier]

what did vcc do to you 😭

[cloud rivet]

spoiled me

[silk pier]

why did you stop using vcc?

nvm, found the thread: #vulkan message

hopefully one day these languages will be fully implemented and made stable

one day...

[cloud rivet]

long thread a few months ago when I actually tried to contribute to vcc #1323084490997895198 message

I spent like 2 months on vcc

https://github.com/shady-gang/shady/pull/55

[silk pier]

this does connect some dots for me~ I recall you talking about contributing to Gob, had no idea you made a pr to shady at some point!

[cloud rivet]

vcc is amazing

I'm about to get vcc at home though

samples build

cool

uses #include <cuda_gl_interop.h>

D:\projects\cuda-samples>dir /s /b D:\projects\cuda-samples\Samples\*vulkan*
D:\projects\cuda-samples\Samples\5_Domain_Specific\simpleVulkan
D:\projects\cuda-samples\Samples\5_Domain_Specific\simpleVulkanMMAP
D:\projects\cuda-samples\Samples\5_Domain_Specific\vulkanImageCUDA
D:\projects\cuda-samples\Samples\5_Domain_Specific\simpleVulkan\VulkanBaseApp.cpp
D:\projects\cuda-samples\Samples\5_Domain_Specific\simpleVulkan\VulkanBaseApp.h
D:\projects\cuda-samples\Samples\5_Domain_Specific\simpleVulkanMMAP\VulkanBaseApp.cpp
D:\projects\cuda-samples\Samples\5_Domain_Specific\simpleVulkanMMAP\VulkanBaseApp.h
D:\projects\cuda-samples\Samples\5_Domain_Specific\simpleVulkanMMAP\VulkanCudaInterop.h
D:\projects\cuda-samples\Samples\5_Domain_Specific\vulkanImageCUDA\vulkanImageCUDA.cu

heyyyyy

[cloud rivet]

cuda vulkan interop

I got working OptiX examples and working vulkan memory sharing with cuda examples, this should be doable

tomorrow I start my 5th vulkan init

gonna use SDL and C++ though, no more NIH

so I can use vkbootstrap and go fast

this time I'm for sure going to get WSI right, for sure

imgui the whole deal

I'm gonna go exercise and go to bed, this is fun

[hushed creek]

You're not doing the cuberoach anymore?

[cloud rivet]

no, but this is the one, I'm sticking with this project, for sure

[cloud rivet]

Limited to ‘C++20’

That’s ok

nvcc sends host compilation steps to msvc

Also ok

Can’t actually write C as nvcc is not a C compiler

That’s fine

You can specify a different host compiler

So probably clang is possible as long as it’s C++20

you can specify older versions of the C++ std

nvcc creates statically linked binaries that are incompatible across major versions of NVIDIA GPUs

It just has to work on my machine

[hushed creek]

what's the new project about tho

[cloud rivet]

Same as all my other projects, just to learn and have fun

I am personally not interested in distributing my own software to others

[hushed creek]

yes but what will it do

[cloud rivet]

I am going to make a ray traced game

It has to be compiled from source for a specific major NVIDIA gpu version using MVDIAs CUDA compiler and the OptiX SDK and the Vulkan SDK

[cloud rivet]

I am over the handmade thing

Going to continue to use neovim, but going back to premake and using visual studio as my debugger

I think renderdoc is now useless to me

[hushed creek]

how so

ah because nvidia

you use their things

[broken fog]

sad

i liked the nih bjorn arc

[cloud rivet]

I gotta narrow my goals and focus

[broken fog]

yeah, that's fair enough, writing your own everything is fun but not a great way to finish, well, anything really

gotta find a middle ground

[cloud rivet]

it's already bad at RT, and yeah, it doesn't support the APIs I am going to use

TIL Meta has a Renderdoc fork https://developers.meta.com/horizon/documentation/unity/ts-renderdoc-for-oculus/

Meta now maintains its own fork of RenderDoc. This fork provides access to low-level GPU profiling data from Meta Quest’s Snapdragon 835, Meta Quest 2’s Snapdragon XR2, and Meta Quest Pro’s Snapdragon XR2+ chips, specifically information from its tile renderer.

https://github.com/baldurk/renderdoc/issues/1148#issuecomment-476600406

baldurk:

As a general rule I don't add support for vendor API extensions unless they're mostly trivial and 'free' to implement, which is definitely not the case for nvidia's RT extensions.

and that's just about the API extensions

that discussion is frustrating to read

baldurk making a very understandable case to someone who wants the RT support and isn't reading the very clear points baldurk is making

I could never maintain an open source project

the aftermath sdk can help explain lost device crashes cool

[astral hinge]

to make things worse, the guy is complaining because he can't use RenderDoc to profile his app

as if RenderDoc is even capable of accurate profiling

[cloud rivet]

right, and I was surprised profiling is the use case for the meta renderdoc fork

reading that I was thinking, but why renderdoc

does qualcom not have their profiling tool

https://www.qualcomm.com/developer/software/snapdragon-profiler

[astral hinge]

yeah idk

weird

[cloud rivet]

looks just like nsight

ok I guess I should write some code or something

I'll start with a blank premake lua file

and see what happens

[cloud rivet]

hrm latest premake that supports vs2026 doesn't build

that's fine, I'll just use the vs2022 option that builds in vs 2026

oh

--cc=VALUE     Choose a C/C++ compiler set; one of:
     clang       Clang (clang)
     gcc         GNU GCC (gcc/g++)
     mingw       MinGW GCC (gcc/g++)
     msc-v100    Microsoft compiler (Visual Studio 2010)
     msc-v110    Microsoft compiler (Visual Studio 2012)
     msc-v120    Microsoft compiler (Visual Studio 2013)
     msc-v140    Microsoft compiler (Visual Studio 2015)
     msc-v141    Microsoft compiler (Visual Studio 2017)
     msc-v142    Microsoft compiler (Visual Studio 2019)
     msc-v143    Microsoft compiler (Visual Studio 2022)
     msc-v80     Microsoft compiler (Visual Studio 2005)
     msc-v90     Microsoft compiler (Visual Studio 2008)

maybe I can't use it

fine cmake

[cloud rivet]

https://cmake.org/cmake/help/latest/prop_tgt/CUDA_STANDARD.html

https://cmake.org/cmake/help/latest/variable/CMAKE_CUDA_ARCHITECTURES.html

oh https://cmake.org/cmake/help/latest/module/FindVulkan.html

find_package(Vulkan COMPONENTS volk)

interesting

[cloud rivet]

cmake_minimum_required(VERSION 4.0)
project(PixelStorm LANGUAGES CXX CUDA)

set(VULKAN_SDK $ENV{VULKAN_SDK})
set(CUDAToolkit_ROOT "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0")
set(OPTIX_ROOT "C:/ProgramData/NVIDIA Corporation/OptiX SDK 9.0.0")
find_package(CUDAToolkit REQUIRED)
find_package(Vulkan REQUIRED COMPONENTS volk)

set(CMAKE_CXX_STANDARD 20)
set(CMAKE_CXX_STANDARD_REQUIRED 20)
set(CMAKE_CUDA_STANDARD 20)
set(CMAKE_CUDA_STANDARD_REQUIRED 20)
set(CMAKE_GENERATOR_INSTANCE "Visual Studio 17 2022")
set(CMAKE_CUDA_ARCHITECTURE 86)
set(CMAKE_EXPORT_COMPILE_COMMANDS ON)

add_library(SDL3 SHARED IMPORTED)
set_target_properties(SDL3 PROPERTIES
    IMPORTED_LOCATION "${VULKAN_SDK}/Bin/SDL3.dll"
    IMPORTED_IMPLIB "${VULKAN_SDK}/Lib/SDL3.lib"
    INTERFACE_INCLUDE_DIRECTORIES "${VULKAN_SDK}/Include"
)

add_subdirectory("external/vk-bootstrap")

add_executable(pixelStorm WIN32 src/main.cpp)

target_include_directories(pixelStorm PRIVATE
    ${OPTIX_ROOT}/include
    ${VULKAN_SDK}/Include
)

target_link_libraries(pixelStorm PRIVATE
    SDL3
    Vulkan::Vulkan
    Vulkan::volk
    CUDA::cuda_driver
    CUDA::cudart
    vk-bootstrap::vk-bootstrap
)

add_custom_command(TARGET pixelStorm POST_BUILD
    COMMAND ${CMAKE_COMMAND} -E copy_if_different
    "${VULKAN_SDK}/Bin/SDL3.dll"
    $<TARGET_FILE_DIR:pixelStorm>
)

{
    "version": 10,
    "cmakeMinimumRequired": {
        "major": 4,
        "minor": 0,
        "patch": 0
    },
    "configurePresets": [
        {
            "name": "dev",
            "displayName": "Release (Ninja)",
            "description": "development build using ninja",
            "generator": "Ninja",
            "binaryDir": "${sourceDir}/build-dev",
            "cacheVariables": {
                "CMAKE_EXPORT_COMPILE_COMMANDS": "ON",
                "CMAKE_BUILD_TYPE": "Debug"
            }
        },
        {
            "name": "release",
            "displayName": "Release (Ninja)",
            "description": "Relase build using ninja",
            "generator": "Ninja",
            "binaryDir": "${sourceDir}/build-release",
            "cacheVariables": {
                "CMAKE_BUILD_TYPE": "Release",
                "CMAKE_CXX_FLAGS_RELEASE": "/O2 /Ob3 /DNDEBUG",
                "CMAKE_CUDA_FLAGS_RELEASE": "-O3"
            }
        },
        {
            "name": "vs",
            "displayName": "Visual Studio",
            "description": "build using Visual Studio",
            "generator": "Visual Studio 17 2022",
            "binaryDir": "${sourceDir}/build"
        }
    ]
}

C:\Users\Bjorn\projects\code\pixel_storm>cmake --preset dev
-- The CXX compiler identification is MSVC 19.44.35221.0
-- The CUDA compiler identification is NVIDIA 13.0.88 with host compiler MSVC 19.44.35221.0
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.44.35207/bin/Hostx64/x64/cl.exe - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Detecting CUDA compiler ABI info
-- Detecting CUDA compiler ABI info - done
-- Check for working CUDA compiler: C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/bin/nvcc.exe - skipped
-- Detecting CUDA compile features
-- Detecting CUDA compile features - done
-- Found CUDAToolkit: C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include/cccl (found version "13.0.88")
-- Found Vulkan: C:/VulkanSDK/1.4.328.1/Lib/vulkan-1.lib (found version "1.4.328") found components: volk glslc glslangValidator
-- Configuring done (8.3s)
-- Generating done (0.0s)
-- Build files have been written to: C:/Users/Bjorn/projects/code/pixel_storm/build-dev

C:\Users\Bjorn\projects\code\pixel_storm>cmake --build build-dev
[6/6] Linking CXX executable pixelStorm.exe

C:\Users\Bjorn\projects\code\pixel_storm>start build-dev\pixelStorm.exe

gives me the hello SDL window right now

ready for vulkan init now

all I need from vulkan is some memory and a present

I'll blit the memory to the swapchain, submit and present and that's all vulkan will do

cuda and optix will write pixels to the memory

I will actually also add imgui, so I guess I do need a dynamic render pass

I'll do that after I have an OptiX generated triangle presented by vulkan on the screen

once I have a triangle I'll set up an acceleration structure

once I have that I'll add my track and skimmer back in

and we'll be back in business

ah all my controls have to get migrated to sdl3 stuff

[cloud rivet]

I have so many tabs open right now

too many

[cloud rivet]

man vk boostrap is pretty cool, fastest device init ever

took like 30 minutes

writing C++ again is so weird

C and C++ are not the same language

I hate that meme

I really don't like it, it is false

[astral hinge]

C/C++ is just ++

[cloud rivet]

1/++

[astral hinge]

forgive me deccer for I have sinned by using a regional indicator instead of a flag

[cloud rivet]

I think how fast I get back to having a track and a skimmer all depends on how much work it is to get through the cuda and optix stuff

I'm gonna be optimistic and say before the year is over

[astral hinge]

the hardest part will probably just be figuring out the new stuff (optix, cuda, cuda-vk interop)

cuda is actually pretty simple to use so probably just the other two things

[cloud rivet]

well I assume the cuda vk interop is solved based on the examples I have, it's a vk timeline semaphore and an win32 external memory handle

I think I can just copy paste the code

idk

[astral hinge]

yeah that's probably a single day problem at most

but it's still relatively unknown territory

[cloud rivet]

yes

I read through the optix programming guide and it reminds me a lot of all the NV RT extensions

for vulkan

which are slightly different from the KHR RT pipeline features

[bronze socket]

mine usually takes around 300ms and it feels slow to me :^)

[cloud rivet]

support for Visual Studio 2026 is still pretty far behind

cmake, premake, sdks, etc

have had trouble with pretty much everything, but just going with the Visual Studio 2022 generator in cmake solved all my problems and VS 2026 can handle it

I pasted my config above, I'm just ninja and rad debugger for the most part, it's just easier and faster, but if I need to debug in visual studio I have that option

today I have been setting up my laptop since I'll be traveling a bunch

haven't opened it in nearly a year had an hour's worth of windows updates to go through and am setting up my build system, making sure all the cuda optix stuff also works on this laptop

it's has a discrete RTX 3070 gpu for laptops, which is the same generation as my PC so hopefully works

[solid grove]

cmake has worked fine with vs 2026 for me

[cloud rivet]

my issue is with visual studio missing extensions

you're likely do not need these extensions

the extensions work with using the vs 2022 generator and then building inside vs 2026

the extensions are NVIDIAs

I don't really know where the fault is

I think there's problems all around in detecting the extensions, presenting the error correctly and building

[solid grove]

i think if you use cmake to create a vs2022 sln and open it in vs2026 it will still use the vs2022 compiler to build jt

[cloud rivet]

that's fine

I'm limited to C++20

that's the newest version supported by nvcc

I wonder if clangd is gonna have trouble with .cu files

I don't think it will

[broken fog]

are you using the vs generator just for debubbing?

isn't it easier to just use gdb at that point

[tight torrent]

cmake does support it

just requires cmake 4.1 i think?

or at least cmake 4

[cloud rivet]

Yeh

Hey we all have green names now

It’s not easy being green

[echo crystal]

[cloud rivet]

Staying in the SF W Hotel for fan expo, got a real nice view

that picture makes it look much brighter than it is

[silk pier]

stay safe and enjoy

make sure to stay away from the Tenderloin!

[cloud rivet]

nice today I got my laptop working with my new project, tomorrow I can make more progress on vulkan init

[brisk chasm]

nvidia's vs extensions always have been meh

the standalone tool worked

[cloud rivet]

https://www.youtube.com/watch?v=tiXyKDQ9Idk

interesting I guess

3 year old tech though

idk

not really a Gamers Nexus benchmark video

https://opendata.blender.org/benchmarks/query/?compute_type=OPTIX&compute_type=CUDA&compute_type=HIP&compute_type=METAL&compute_type=ONEAPI&group_by=device_name&blender_version=4.5.0

https://opendata.blender.org/benchmarks/query/?device_name=NVIDIA GeForce RTX 3060 Ti&compute_type=OPTIX&compute_type=CUDA&compute_type=HIP&compute_type=METAL&compute_type=ONEAPI&blender_version=4.5.0

hrm, the RT blender uses is optix

[solid grove]

it has a variety of backends

it used to be only CUDA, then nvidia added an optix backend, and after a while apple added a metalRT backend, and AMD and Intel added backends for their respective technologies

I don't think blender uses DXR or VulkanRT for anything

[cloud rivet]

https://www.youtube.com/watch?v=Hnn7scqXKno

this is also via blender

optix in this video is about twice as fast, but this isn't really rigorous science

[solid grove]

twice as fast compared to what?

[cloud rivet]

using CUDA

the blender CUDA backend

[solid grove]

afaik the CUDA backend in blender is purely a software trace and doesn't use any dedicated silicon

[cloud rivet]

right

idk RT hardware is faster at RT, not surprising I guess

[solid grove]

in my testing an M4 Max is about 10% faster than a 2070 super in hardware ray tracing. idk how much those numbers mean to u though

[cloud rivet]

the M4 is about 3 or 4 years newer than the 2070, and the 2070 is first gen RT?

not sure

I was just trying to understand what the future has in store for OptiX, they built it before Vulkan RT was a thing

I think?

like I'm getting invested in this proprietary API so just was looking for recent news, and all these Blender benchmarks popped up

which is not really what I'm interested in

but was kind of interesting

the latest release is in march https://github.com/NVIDIA/optix-dev/releases

which added clusters

https://developer.nvidia.com/blog/neural-rendering-in-nvidia-optix-using-cooperative-vectors/

idk

[silk pier]

optix won't be magically faster in any crazy way

there may be slight benefits from a few new knobs to fiddle with

but its the same hardware with largely the same features at the end of the day

[cloud rivet]

yeah

[broken fog]

that's nuts

[hushed creek]

wow that's a tiny amount

[broken fog]

iirc my m3 pro is only slightly slower than my 2060s too

[hushed creek]

I imagine 2070 super is nothing nowadays

[broken fog]

m3 onwards will make a big difference because they actually have hwrt

[broken fog]

it's still a desktop class card

m4 max has like a fifth of the power budget

for the entire soc

[hushed creek]

the price my man

m4 max macbook pros are well within the range of the highest end gaming laptops

for the desktop variants it's even worse

that's why I'm saying the amount is tiny

you can get a razer blade with RTX 5090 for the same price as the m4 max mbp

and I reckon the 5090 is more than 10% better than the 2070 super at hw rt

anyway I was just surprised

because I was under the impression that apple was much better than this at rt

[broken fog]

yea but it's a razer

i had a razer blade trust me you don't want one

but yea gpu is not apple's strong suit

[hushed creek]

then again none of their latest attempts at gaming really impressed anyone

except for death stranding on their phones, it's pretty crazy to play DS on a phone

[broken fog]

the rt is pretty decent for the tiny power budget

but yeah it's not even close to current nvidia tech

[broken fog]

gpu is drawing 15-16w while rendering with cycles so at full sustained load

tdp for the smolest blackwell gpu is three times that lol, no wonder it's a lot faster

[cloud rivet]

I had one too, in like 2015, it was very bad

razer's are overpriced so that actually supports FIrefly's argument

nobody is buying a mac for RT cores though

[broken fog]

mine was... 2018 iirc? some 6-core i7 and a mobile gtx 1060, perf was actually decent, the issue was with build quality

[cloud rivet]

yeah build quality

and the software was also bad

[broken fog]

razer cheaped out on the battery and stuck some no name chinesium in there rated for 40ºC

the cpu right next to it hit 95

[cloud rivet]

I think they deserve a class action lawsuit tbh

[broken fog]

so ofc battery got fucked after a year and it started inflating like a yummy lithium balloon

[cloud rivet]

also they shipped a crypto thing

oh wait mine was like in 2018 too? not 2015

[broken fog]

man i love my $2k laptop turning into a fire hazard overnight

oh also the gpu fan went after two years and started making a horrible grinding noise, so i had to disable it and then disable the gpu to stop it overheating

[cloud rivet]

they probably deserved a class action lawsuit for how horrible they are and the quality they ship given the price

[broken fog]

ended up selling the thing for less than half price because of all the issues it had

[cloud rivet]

mine just stopped turning on and I threw it away

[broken fog]

a year after that i got an m1 macbook for work and never looked back lmao

[cloud rivet]

there are no good windows laptops

I have a dell, it is "ok"

but it is more expensive than a mac

[broken fog]

oh btw i only got the razer blade because the dell xps i bought also had qc issues (awful fan noises day one) and it was the last one in stock

[cloud rivet]

I have an exps yeah

[broken fog]

so yea those xps look nice but i don't trust dell's qc either

[cloud rivet]

I haven't had any quality issues with this dell

I went through a lot of troubled windows laptops

I had an asus gaming laptop, the razer, two surfaces

before I finally got this dell and I have had this for like 3 or 4 years now with no issues

[broken fog]

tbh i think they're among the better non fruity laptops, still, only one i would actually get is a t or x series thinkpad

[cloud rivet]

the asus, razer, and surfaces were all very bad

I don't buy lenovo

[broken fog]

but those are also prices very close or more expensive than macs so,,, why

[broken fog]

any reason? the quality gap between the good thinkpads and everything else lenovo makes is massive

like they might as well be two different companies

[cloud rivet]

because I don't trust the manufactor to not put RoC stuff on there

[brisk chasm]

my business laptops are all thinkpads ever since, the cool (sic) thing about them is that you can flip them over during lunchtime to cook your meal on it

[broken fog]

i have two thinkpads both >12 years old

[cloud rivet]

the old thinkpads are fine before lenvo bought them

[broken fog]

awful screens, but they're excellent blunt weapons

i keep them around in case of home invasion

[brisk chasm]

yeah the old thinkpads had shitty lcd screens

[broken fog]

the ibm model m is my sword, the t420 is my shield

[brisk chasm]

hehe

literally

[broken fog]

these are lenovo but like 2nd-3rd gen after the buyout, those were still great

[brisk chasm]

i can use my p15 as a counterweight for the liebherr ltm 11900.1

[broken fog]

newer ones are fine but not as tanky

[brisk chasm]

trust me they are tanky, still

[broken fog]

though i hate the stupid camera bump on the newer thinkpads, it looks so dumb

why did they do that

[brisk chasm]

to appease the mac peasants

[broken fog]

? macs don't have a bump, they have a notch thing like the iphones

it doesn't physically stick out

[brisk chasm]

same thing 😄

[broken fog]

i mean this dumb thing

[brisk chasm]

i always went with the performance series of them thinkpads though, not sure about the lightweight ones

[broken fog]

it's ugly

[cloud rivet]

https://www.ftc.gov/news-events/news/press-releases/2018/01/ftc-gives-final-approval-lenovo-settlement

[brisk chasm]

the p15 doesnt have that one yet, not sure if that would bother me personally

[broken fog]

p series?

[brisk chasm]

ye

[broken fog]

ye those are awesome

[silk pier]

this is the wrong comparison

a lenovo 'gaming' laptop with 64gb of ddr5 ram and a 5060 is ~1300usd last time I checked

[broken fog]

the newer ones do have the bump apparently

[silk pier]

and a 5060 without doubt beats a m4 max pro gigachad whatever gpu

[brisk chasm]

i was eying a p16 gen2, not sure if they had a bump mcbumpington

[broken fog]

[cloud rivet]

what about the m2 ultra that ships in a mac pro?

[brisk chasm]

ah gen 3 has the bump

[brisk chasm]

keeps me on my toes feet 😛

[broken fog]

unsurprisingly with 3x the power budget yes

gpu perf is not what you buy a macbook for

let alone gaming lmao

[brisk chasm]

indeed

who games on a mac... natively

[broken fog]

i play rimworld on it sometimes

[cloud rivet]

I think the mac pro is where the target user would care about RT performance

[broken fog]

mac pro/studio have a higher power budget iirc, though if it's still stuck on m2 that has no hwrt whatsoever

[silk pier]

yes, Macs have good longevity on low powered workloads

but once you need high powered loads, macs falter hard

[cloud rivet]

M2 Ultra with 24-core CPU, 76-core GPU, and 32-core Neural Engine

hrm

[brisk chasm]

macbooks do be looking neat, at least the aluminium casing, speaking of aluminium, googles new os is called Aluminium too, not Aluminum xD

[silk pier]

I'll send something in a minute

[broken fog]

build quality is just light years ahead of everyone else tbh

like it just feels nice to hold the thing

i can't say that for literally any other laptop, except maybe the fancier dell xps

[brisk chasm]

i have not touched a single macbook, last one was a powerbook 150 or something

[silk pier]

@cloud rivet
https://www.notebookcheck.net/GeForce-RTX-5060-Laptop-vs-M2-Max-38-Core-GPU-vs-M2-Max-30-Core-GPU_12924_11574_11572.247598.0.html

In synthetics it looks close

In gaming (although not many cases that are shared, it's not even close)

[broken fog]

depends on the load, at cpu workloads they're pretty dang good for a laptop, gpu workloads yeah no that's not what you buy them for lol

[cloud rivet]

oh the mac studio ms an m3 ultra

[broken fog]

yeah i think the mac pro is ded

mac studio replaced it

[brisk chasm]

if i could put the intestines of my lenovo into a macbook shell, and use that beautiful display too, i would do it 😄

[broken fog]

not surprising, who would want the huge box when a tiny mac studio fits the exact same hw

[broken fog]

the mac cpu is way faster though, the chips are the best part

gpus are meh but the cpu cores are insanely good

[brisk chasm]

but the macos ecosystem sucks and asahi is probably not worth the hassle to use fulltime yet

[broken fog]

ymmv i'm drinking the hell out of this koolaid and it's tasty

[silk pier]

It's an okay build, I don't like the thin flat design personally.

The nonsense keyboard layout is the real annoyance for me

although-- I like the cpu efficiency 'portion' of the chip a lot

[cloud rivet]

it doesn't at all suck

not for developers

[brisk chasm]

looking forward to pollute my systray with 40 icons and apps of allah because you need to install all of them to make macos usable

[broken fog]

i've read stories about company gives all their employees apple silicon macs, one employee tries building the project on their new mac, oops it builds 2x faster than it does on the dedicated build server

i guess we support macos now

[brisk chasm]

hehe

[broken fog]

apple silicon cpus are amazing at compile workloads

i can build gcc in <5 mins and barely make a dent on the battery

and this is the cheapest m3 pro

[brisk chasm]

im just jealous hence the talking out of my ass

[cloud rivet]

your options as an engineer for a work computer at my company are a macbook or a macbook

[broken fog]

ye i do wish i could dualboot linux on it for some stuff (playing around with vk mostly heh)

[silk pier]

I recognize that.

My advice for ppl and laptops is a really simple pathway mapping of:

1. want macOS? get a mac
2. don't want macOS? get a non mac

if you hate macOS, don't get a mac, if you like it, get a mac

[broken fog]

what nonsense layout? it's pretty standard, other than well having the same modifier key layout macs have had since forever

[broken fog]

basically yes

[silk pier]

xcode sucks

[broken fog]

comparing between macs and anything else is kinda pointless because you don't decide based on berf or anything like that

[silk pier]

agreed

[broken fog]

shrimply don't use it

developing on macos is great, developing for macos is a pain

[brisk chasm]

use your favorite terminal.ai 😄

[broken fog]

but as long as you're not developing for ios (or macos and using swift and all the fancy shtuff ig) you can just not use xcode

[cloud rivet]

I use wezterm on my mac, ghostty is also ok

[broken fog]

kitty ftw

i went back to iterm2 tho for the global panel terminal thingy

[cloud rivet]

I like wezterm because it also runs on my windows machine

[broken fog]

wish kitty had that

[brisk chasm]

wish kitty had an option to center the fucking window when it spawns, but muhinder is reluctant because of some wayland excuse

[cloud rivet]

so I have the same terminal on mac and windows, obviously the shell is different, but I can share the dotfile config

[silk pier]

command and option is a horrible mess.

when I use programs I love my cmd menus. These are pretty standard across apps on windows.

On MacOS? everyone has a different opinion on what OPTION+SOMETHING and CMD+SOMETHING is supposed to do (and which one should be the defacto CMD menu button)

as a result I need to fix the key binds of a bunch of programs manually, it annoys me a lot

[brisk chasm]

yeah that is reasonable

[broken fog]

use fish

[brisk chasm]

you have that ctl, ctl+alt, alt, ctrl+shift, alt+shift, ctrl+alt+shift bs on other osses too

[broken fog]

what kinda software? ime it's pretty consistent

[silk pier]

😭 I use a Mac when I have to do the for part

[broken fog]

the one exception is blender

but it's stupid on other os too

[brisk chasm]

i blame the developers of non os software, not adhering to os guidelines

[broken fog]

my condolences

xcode is truly ass

[cloud rivet]

you can't use fish on windows without wsl

[brisk chasm]

but it has a decent metal debugger 😄

[broken fog]

install linux

[silk pier]

normally on other OS's its ctrl for everything

on Mac it's cmd or option for the same thing that was once ctrl

why did they reinvent the key board layout... thats all I am asking

[brisk chasm]

fish is fucked on linux too, you cant just run any bash in it, because of fish is doing some things differently

[cloud rivet]

are you memeing

[silk pier]

nah, blender is great

just switch to 2.79 key binds

[silk pier]

its pretty bad in my experience, it just crashes... all the time...

Also, you can't mix debugging and validation

[broken fog]

doesn't it just work if you have the #!

[broken fog]

they didn't? they've had those keys since the 80s

back then it was kinda wild west and everyone did whatever

[broken fog]

not at following os guidelines for keyboard shortcuts lol

[broken fog]

yes because i know you're windowspilled

but also no

[brisk chasm]

i never bothered to look into it

[solid grove]

lol blender’s shortcuts feel like they were designed to slow people down
“hmm instead of clustering common actions next to each other so that one hand can operate all of them, lets scatter them randomly around the keyboard!”

slowly making my own key map

though I guess blender’s shortcuts are more equitable for left handed people. The industry keymap is designed for right handed people while blender’s shortcuts are equally terrible for everyone

[cloud rivet]

do left handed people have issues with wasd?

never considered that

[silk pier]

I like spacebar menu

[solid grove]

(for reference I’m not left handed)
games don’t have that many shortcuts so some people just use wasd with their right hand
But one time I sprained my wrist and Maya because unusable to me because they arranged they keys assuming your left hand is over the left side of the keyboard

[cloud rivet]

I use the marking menu in maya

I don't use keyboard shortcuts outside of switching between rotate, translate, scale and some other minor things

you don't need a keyboard at all besides hitting the space and cntrol and shift buttons

I feel like a wizard using the marking menu

[solid grove]

I also use set key frame (s), group (g), and the shading modes (1 2 3 4)
Probably some more since they are muscle memory now

X to toggle the snap mode

[cloud rivet]

yeah shading modes

I use those

other than that it's all marking menu, you got the full power of maya in it

I'll use the menu bar for snapping windows

[solid grove]

oh yeah D to change the pivot of an object

F to frame selected

[cloud rivet]

have you tried the marking menu

[solid grove]

that’s the space bar menu right?

[cloud rivet]

wow

that's one of them

[solid grove]

oh this thing?

lol image didn’t paste

I use this all the time

[cloud rivet]

I do everything with the marking menus

including creating cubes/geometry, manage UVs

I tool switch like from quadtool to select to whatever

[broken fog]

if you have #!/bin/bash in the first line that should tell the shell (os?) to run it with bash

[cloud rivet]

everything

[solid grove]

one of my major annoyances with blender is that the modes are global. Vs here you can have different objects in different modes at the same time and select things in both simultaneously

flipping between modes constantly in blender is a major source of slowdown

[brisk chasm]

yeah no idea, tbh, i never went back to fish since the first incident a decade ago or whenever itwas i tried fish

[solid grove]

and you also can’t go from any mode to any other mode in blender either. If you’re doing skin weights and you’re in mesh edit mode, you have to first go to object mode, deselect everything, select the armature and the object in the correct order, and then go to weight mode. 4 actions instead of 1

[brisk chasm]

yeah that mode nonsense is quite weird

i think they started to work on that, according to blendercon from january 2025?

[solid grove]

the modes are tied to what view you’re in too. If you change from layout to compositing and you were in edit mode it might take you to object mode and lose your selection. infuriating

[solid grove]

big if true

[brisk chasm]

time to rewrite a blender clone from scratch!

screw your minecraft clone and openworld mmo rpg

hehe we should suggest that to beginners actually, those who want to compete with ue5 nanite and lumen

make a blender clone instead 🙂

how hard can it be

replace python with c# on dotnet

or, alternatively. replace the existing gltf importer/exporter with something fast 😄

[cloud rivet]

freya holmer is doing that

https://half-edge.handmade.network/

she hasn't updated that page in a while but on her bluesky she does https://bsky.app/profile/freya.bsky.social/post/3m6ggmojkss2c

Freya Holmér (@freya.bsky.social)

I've *almost* got dual quaternion interpolation going!

DQs naturally interpolate using a screw motion, which works for any two rigid poses

Likes

202

▶ Play video

Bluesky · Nov 25, 2025, 3:47 AM

[brisk chasm]

nah shes just working on a subset of tools

and the source code is quite spaghetti

[cloud rivet]

Is she working on half edge still?

[tight torrent]

yes

[brisk chasm]

yep

[cloud rivet]

Nice

[brisk chasm]

she streamed yesterday while working on it

[tight torrent]

she was streaming it yesterday (or a day ago? idk)

[echo crystal]

https://lotusspring.substack.com/p/how-i-wrote-the-fastest-blender-exporter prety cool post i just found

[brisk chasm]

i think i shared the videos some time ago

that embedded yt looks familiar

could you check if this was i posted this in #graphics-resources and if not also link it there, i think this is some cool stuff

ah nevermind

that article is just few days old 🙂

i must have posted only the yt videos back then

[echo crystal]

just shared it

[brisk chasm]

i remember that blender ackchually builds quite fast

when jakov and i enhanced the existing gltf plugin lol

[echo crystal]

thank u for ur service

[cloud rivet]

I don't hate cmake

it's ok

I'm not doing a ton with it

I just kept avoiding it because everyone always complained about it

but just my first cmake project, idk, it's fine, it works

[brisk chasm]

yeah, its quite ok for small projects

once you put your butt down for half an hour to see what you need and how to use those bits, at least when it comes to lunix and binbows, no idea how fucked it is for android or mac

[solid grove]

it’s ok for big ones too. for some reason people love putting tons of extra crap in their cmake that makes it worse and I don’t know why

[solid grove]

works the same on mac as it does for Linux. On android it’s painful but that’s because doing anything with Android dev is painful

[broken fog]

mac is pretty much the same as loonix

[brisk chasm]

i also had no trouble to add foreign code which didnt come with cmake support, but cmakeified it so to speak

[brisk chasm]

i thought so

[solid grove]

On Android you have to do some backflips to get gradle to invoke cmake, and some gradle functions like “clean” don’t work correctly

[broken fog]

i could never get the find package stuff to work, probably cmake skill issue

[brisk chasm]

i also thought so about android 😄 i did touch it 100 years ago for a month or two

[broken fog]

i hate cmake though

for a very small project i'd rather just write a makefile

and for something bigger i uh

[brisk chasm]

i never rely on findpackage, i just compile shit myself, and sometimes you have to anyway

[broken fog]

use cmake and really wish i had a better build system

[brisk chasm]

hehe

[broken fog]

i don't like writing cipipi to begin with anyway

[solid grove]

me waiting for Google to expose everything to NativeActivity so we never have to deal with Java again (I will die of old age)

[cloud rivet]

makefile doesn't work on windows, but it's ok too

[brisk chasm]

just join tccpcpcpp server and checkout #showcase, every other week some 17 year old reinvented a c++ build/package manager system

makefile also works on windows, oyu need to have your PATH properly set for that, or have some, god forbid, cygwin/wsl shit setup

[cloud rivet]

no

[solid grove]

afaik on Android your app starts life by having the jvm load your main class via reflection. It’s cooked out the gate

[cloud rivet]

makefile on windows is a shitshow unless you're using wsl 2

[brisk chasm]

i remember using make on binbows, in 2006 or so to add native ntlm support to TortoiseSVN

[cloud rivet]

it doesn't work

[bronze socket]

in theory it should at least work with the nicer JVM langs, like kotlin

[cloud rivet]

like I can just build my work's anything on windows with any of our makefiles

[broken fog]

makefile on windows is a shitshow unless you're using wsl 2
ftfy

[broken fog]

if only kotlin was any better

[bronze socket]

don't you do fullstack stuff

[broken fog]

it isn't

[brisk chasm]

kotlin itself is 10000 times better than jaba

[bronze socket]

why does your work have makefiles

[cloud rivet]

because we use go

[solid grove]

I want to meet the person at Google who decided “let’s run 50 JVM instances on this memory-constrained device”

[cloud rivet]

and go projects use Makefiles

[brisk chasm]

go go make-o-mobile

[bronze socket]

isn't there an embedded version of java

[solid grove]

and yes, every app or process is its own JVM instance. It has to be, for security

[cloud rivet]

https://github.com/ConductorOne/baton-demo/blob/main/Makefile

like that

[broken fog]

i disagree, i mean, it's marginally better but also worse in some ways

the syntax decisions they made are utterly baffling

[brisk chasm]

its what java should have been 20 years ago already, at least

[broken fog]

it does at least get true generics ig

[cloud rivet]

that one actually works on windows

so bad example

[broken fog]

but the syntax is awful, ambiguity everywhere, kotlin code is just super hard to parse for a human

[brisk chasm]

https://github.com/ConductorOne/baton-demo/blob/main/Makefile#L17C15-L17C20 xD ./... looks funny

[broken fog]

does bongows even have make?

[cloud rivet]

yes

[brisk chasm]

not by default

[bronze socket]

I have msys 1.0 mapped into my path so makefiles work better than otherwise

but you still have trouble from the more GNU-y builds

worth it though pretty much solely for grep

[solid grove]

Switching to kotlin doesn’t make android dev with native code any easier. You still have to deal with the dual build systems and JNI-whispering to do anything

[cloud rivet]

C:\Users\swart\projects\code\pixel_storm>winget search gnuwin32.make
Name           Id            Version Source
--------------------------------------------
GnuWin32: Make GnuWin32.Make 3.81    winget

[broken fog]

is there an actual build or are you using mingw or some memery like that where you pretend it's a real os

[bronze socket]

how is mingw not an actual build

mingw isn't solely contained in msys 2.0, they have exes for native windows everything, pretty well maintained and supported

I use GCC via cmd.exe, not directly but via cmake/ninja

[cloud rivet]

Í don't know why you need mingw when there's wsl 2

[bronze socket]

but you absolutely need msys 2.0 to build weird stuff that's designed to live in GNU-y environments (ffmpeg is my chief example) though pretty much entirely for the shell emulation

[broken fog]

wsl is just cope tho

like just install linux

[cloud rivet]

stop lol

[bronze socket]

I don't wanna build for linux, if I wanna make a linux build I boot into linux

[broken fog]

yea you're not gonna target binbows from wsl right?

[bronze socket]

msys lets me work around shitty linux-oriented build systems but still generate an entirely windows native result

again e.g. ffmpeg, I can build it with msys and I link it from my regular windows builds

especially since it's just a DLL with a C API I can use it from either MSVC or my gcc builds

[cloud rivet]

I don't wan't or care about linux, I don't use linux I don't want to build anything for linux, but sometimes some tools are only available/work well on linux, as in maybe ffmpeg (I don't know) in that case wsl 2 is going to be fine

[bronze socket]

ffmpeg is available for windows, could not live without it

building it from source while on windows is just a bit annoying

[cloud rivet]

right I don't have an example of a linux only thing

I think the openai cli tool is an example

codex

[bronze socket]

the only one I can think of is valgrind, but it relies on being in the linux env

[cloud rivet]

that's only on windows via wsl 2 right now

[brisk chasm]

windows sdk also provides make for windows, but lets not talk about that

[bronze socket]

but I just have my laptop that I pretty much only boot into linux and I just build linux native stuff there

[cloud rivet]

cool, I don't have a need to do that personally

[broken fog]

i will respect your decisions because this is your own project thread and you're a cool frog

but inshallah we will wipe windows off the face of this earth 🙏

[bronze socket]

I am so used to the windows crust and have so many ways of coping that it doesn't bother me at all anymore

that being said it's crazy how much linux literally is by and for C/C++ devs

(I mean basically all OSes are by C/C++ devs when you get down to it, but you know what I mean)

[wraith urchin]

I was thinking of stepping back into Linux but the friction is still so high

[bronze socket]

I just use kubuntu, everything is like ubuntu under the hood but it uses KDE and doesn't have regular ubuntu's sponsored bloat

could not use it as my daily driver because there's a lot of programs I like on windows

[wraith urchin]

I have a Manjaro box, I was thinking of installing it on my main machine for a bit

[bronze socket]

but it's my "it just works" options that I open and do dev stuff and don't care

[wraith urchin]

I guess I'm gonna find out what the windows things I rely on are

[bronze socket]

yeah I mean if you're using high friction distros that's that much more stopping you

[wraith urchin]

I mean already #1 is probably Visual Studio

[bronze socket]

using an ubuntu-based distro any issue I have is like the first result when I search

[wraith urchin]

Is manjaro high friction? it was the smoothest linux experience I had in a while

that was on AMD though

Nvidia gpu is probably gonna make it worse

[bronze socket]

does it not have similar DIY setup friction as arch

[wraith urchin]

nah

its just an installer

Its like arch for dummies

which is what I want

[brisk chasm]

ubuntu based distros usually have "old"er packages installed, especially compilers

if thats no concern for you then you are good

[wraith urchin]

Yeah I hate that shit

[brisk chasm]

archinstall's installer is also very straight forward these days

nvidia gpus are no issue anymore, since 560.xx or 540.xx even

[solid grove]

when I set up a fedora install I had to run that stupid .run file to install the Nvidia drivers

so it’s still a pain there

[wraith urchin]

This is my definition of no issue, immediately after installing I can go watch a YouTube video and there will be no screen tearing

[tight torrent]

theyve had some fuck ups in the past (ex. their gui package manager had bugs that caused the arch aur to effectively get ddos'ed twice) personally id reccomend one of the other "arch with a gui installer" distros like cachy or endeavor

[wraith urchin]

You have no idea how many distros I've tried that had full screen tearing right out of the box

[brisk chasm]

cachy is one of those distros made by 18 year olds, god knows when they burn out and then this goes to shit

[bronze socket]

yeah I actually just checked and I only have gcc 13.3.0 on my laptop, though I have 14.6.0 on windows and I haven't updated mingw in forever

[brisk chasm]

hehe isnt it at 17 already

ah no, 15.2

fedora is also ok otherwise

[bronze socket]

I'm also using LTS so that might contribute to how far behind it is

[brisk chasm]

yeah

[solid grove]

also I find modern gnome very annoying to use. and you have to install 3rd party plugins to fix those behaviors. sucks though because gnome actually handles DPI scaling correctly

[brisk chasm]

26.04 is next LTS iirc

[bronze socket]

that's why I use kubuntu, you get KDE so you don't get gnomed

[brisk chasm]

yeah gnome sucks somehow, thats why i just go with XFCE, i like its uglyness and it has some neat even more ugly themes which i enjoy for nostalgic raisins

[bronze socket]

and no preinstalled amazon shop button or whatever else they do these days

[solid grove]

last time I used KDE it did not handle DPI scaling very well, though it’s been a few years

[brisk chasm]

kde has HDR support at least, which somehow works good enough

[tight torrent]

with nvidia that might happen until you install the proprietary drivers, distros probably cant ship them by default (and require something like a button in the installer or an install thingy after the install) for legal reasons or something

[brisk chasm]

steam also runs out of the box on arch at least, i could imagine same on fedora, not so sure about oeboentoe based ones because of outdated packages perhaps

[bronze socket]

yeah I forget how I installed official nv drivers, probably off their website though

[solid grove]

was helping debug something for someone and it turned out cachy broke something when it was building all of its various forks for its package manager

[brisk chasm]

sudo pacman -S nvidia-dkms or select nonfree dribers in archinstaller

cachy seems to break shit quite often, i was lurking on their discord for a week because i thought i give it a try

all i do now, is check archlinux.org once a week or two, to see if some manual intervention is required before updating the os, which was also very trivial in the past, and happened once a year or so

and i have to disable auto update every 3 months or so for discord, because discord claims thres an update, but the package has not been updated to latest version yet, which usually happens a day later then

unless you use it in ze browser only

[broken fog]

oops i accidentally unleashed a linuxshed it seems

sorry

[brisk chasm]

Rosy the new Linux Distro 😛

[broken fog]

would install

[brisk chasm]

but yeah, sorry bjorn

[wraith urchin]

I remember last time, not only did I have to install the official drivers, but I also had to edit some arcane config file

[brisk chasm]

this is a thing from the past

no more x11 config bs

[wraith urchin]

Oh thank god

[broken fog]

wayland is the way

[brisk chasm]

even xfce is almost wayland-support-complete

im still on x11 though, until 100% is done and tested for some time

[tight torrent]

i just use kde (100% shameless kde shill)

[brisk chasm]

Kshameless KDE Kshill, Kplah

[wraith urchin]

Ok I might try the arch installer, but if I run into issues I'm blaming you @brisk chasm

[tight torrent]

i have cachyos on my laptop because i like the kde cachyos nord theme, can always revert to base arch if need be

[tight torrent]

the built in arch installer can be accessed by just typing archinstall inside the archiso enviornment

[brisk chasm]

yeah

[wraith urchin]

wait it isn't a button?

[bronze socket]

of course not, it's arch

[brisk chasm]

its a little menu driven thing

[wraith urchin]

you aren't selling me on this

[brisk chasm]

it guides you through all the important bits

[wraith urchin]

Manjaro it is then

[tight torrent]

nope its a TUI (terminal user interface) installer

[bronze socket]

just wait another few years

[tight torrent]

oh dear god no

[brisk chasm]

elementaryOS i would allow, but manjaro is dead for me

elementary comes with a grafickel installer too

and its main color is some sort of purple somehow which i like

[bronze socket]

I've heard of artix before, I think it's just arch made easier to use

never bothered to investigate it though

[tight torrent]

theres also endeavoros which ive heard good things about

[broken fog]

guipilled take, become one with the terminal

[wraith urchin]

no

I'm just saying any distro that doesn't have a GUI installer is bound to have a whole bunch of other shit I don't like

[broken fog]

fair, arch is not for people who value their time

[wraith urchin]

I like the way manjaro looks out of the box

The complaints about manjaro seem

[brisk chasm]

artix is arch sans systemd, instead if uses openrc as the init system or whats it called, it also has its own repos, like manjaro

[tight torrent]

systemd is very good tho

[broken fog]

the thing with kinda obscure forks is can you trust them to stay maintained

i'd stick to well known and supported distros for that reason

[wraith urchin]

I wish there wasn't this whole repo nonsense with linux, I wish I could just download shit from the developer's website like you do on windows

instead you gotta trust someone is maintaining the packages

[broken fog]

you mean like package managers?

[wraith urchin]

Like on windows when I want a program I go to the developers website and I can just download an installer

but on linux theres all this extra bullshit

like oh, that version of the package isn't available on your distro

[broken fog]

on linux i just do yay thing and bam it's installed

that's the single reason arch is my favorite distro btw the official repos + aur has everything under the sun and it's all up to date

[bronze socket]

but who is behind making yay thing work

[tight torrent]

yay install openjdk-latest is something i had to do for my younger brother yesterday

[wraith urchin]

if it works

[tight torrent]

fedora

[broken fog]

i've never had it not work

[brisk chasm]

on windows a lot of shit can be installed via winget/chocolatey too, no manual msi bs where you download and run and click through 100 Next buttons of allah

[bronze socket]

maybe dev stuff

[wraith urchin]

I never use that chocolately stuff

[true moon]

I kind of like the 100 next buttons of allah

[wraith urchin]

I hate it

just give me an installer

[bronze socket]

I miss when they were called wizards

[broken fog]

idk i've always disliked installers tbh

[brisk chasm]

i always hated it

[bronze socket]

that was back when computing was cool

[wraith urchin]

Like I want a binary installer

[true moon]

My favorite is the classic installer wizard when there is a secondary window that takes up the whole screen with a fancy blue gradient background and then the regular installer wizard appears above that

[wraith urchin]

I miss those days

[brisk chasm]

i also have ptsd from building installers xD

[true moon]

I still have no idea what it means to even install a program lmao mine just run out of a folder

[broken fog]

i hate not knowing what the hell the install is doing, windows installers will create three folders in different places in your filesystem, mess with the registry and do god knows what else

[bronze socket]

you install it by cramming that folder somewhere in appdata or program files

or both

[true moon]

there must be more to it than that though

I figure there's some registry BS or something going on

[broken fog]

i like the macos approach where stuff is packaged and self contained and you literally just copy it to applications folder

[true moon]

associating file extensions or whatever else

yeah

[broken fog]

want to uninstall? delete it

done

[true moon]

I don't necessarily like or dislike that but it's definitely different

[wraith urchin]

It depends on the installer

[brisk chasm]

i also have fond memories of having all my fav tools (TeraCopy, Getright, IrfanView et al) on some lan drive, and a .reg file to "install"

[true moon]

windows feels like it's inserting tendrils of chaos into your computer and if you just delete the folder in program files it's like leaving the head of the tick in your skin

[bronze socket]

I like that windows is standard/monolithic enough that you can do nontrivial stuff like that, imagine trying to associate a browser link with a program on linux

e.g. a default magnet:// program or something

[brisk chasm]

yeah, i mean, if you never did that on binbows before its as hard/trivial

[bronze socket]

I mean programs can ship that'll do that with certainty, try shipping a linux binary that can do that

[tight torrent]

https://portland.freedesktop.org/doc/xdg-settings.html

xdg-settings set default-web-browser google-chrome.desktop

[bronze socket]

I remember when I was trying out matrix servers there was this rust one I was curious about because theoretically it would've been less resource intensive than synapse

[brisk chasm]

or graffical if you want

[bronze socket]

they claimed to have an ubuntu build but it was literally just a binary from their CI in one of their CI dump folders, and I had no fucking clue how to get it to work and they had no docs

but hey there was technically a static standalone binary

except it was server software so I would've needed to figure out how to configure the service for it from scratch, so fuck that

[brisk chasm]

i remember these standalone .exe files which were some standalone virtualized os thingy

it wasnt vmware and wasnt virtualbox, something enterprisey/commercial

[tight torrent]

hyper-v?

[brisk chasm]

a 500mb exe or so, you just run it, and inside oyu have 1 application running in another os

no

[bronze socket]

was it like proto-docker

[brisk chasm]

ja way before

mid/late 2000s

if not late 90s already

you could also "dockerize" whole applications with it, and stuff everything into that one .exe

[bronze socket]

neat, never heard of that

I remember like 10ish years ago it was cool and fancy tech to have seamless RDP for single applications, citrix type stuff

very commonplace now

[brisk chasm]

yeah xenapp/xendesktop

i hate it

luckily i dont have to use it much

[echo crystal]

freebsd merch is pretty funny hehe

[brisk chasm]

congratulations dodo 🙂

[silk pier]

you can have different views all at the same time?

oh wow, I forgot to scroll down for a while...

[cloud rivet]

An installer can basically do anything, including editing the registry and badly behaving programs can just leave a mess behind because the incentive to providing a good uninstaller is not is understood or not appreciated by some companies

[broken fog]

needing an uninstaller at all is fishy

ideally i should be able to just rm a directory and poof gone

[bronze socket]

how do you deal with the case of stuff you interact with centralized OS state for

like file extension association or registering services

[broken fog]

yeah, that's a fair point

at some point you're gonna have to register something with the os

[bronze socket]

I don't like that they're binaries though, if installers/uninstallers were transparently shell scripts you could at least deal with the majority of the crust

[broken fog]

ig at that point you'd want to make it minimal enough that the os can clean up after an uninstalled program

[solid grove]

also uninstallers have the opportunity to clear out caches and other junk that would be left behind if you just deleted the dir

though if you write your cache to the correct location the OS is supposed to clean it up for you

[broken fog]

like, if you have an associated extension to a program that's no longer there os just goes "well ig that's gone, let's just use the next one and get rid of this entry"

[solid grove]

though conversely if it was a shell script then the user having an incompatible shell would mean it wouldn't work

eg users who uninstall bash and then wonder why nothing works anymore

[broken fog]

that's what we call a skill issue

uninstalling bash is like deleting random files in system32

if you're that dumb you do not deserve a working os

[bronze socket]

if you have a custom shell that can't even run compat for bash you should figure out how to edit your own installer scripts

because that's clearly the message you're sending

[solid grove]

idk, on linux every component is supposed to be replacable so I don't think the desire to uninstall bash is that unreasonable

[bronze socket]

yeah but in practice that's exactly why the user experience of linux is dogshit

[solid grove]

having a manifest system for programs to tell the OS where they write all their config and temp data would mean that an uninstaller wouldn't need to itself be an executable program

because it's not like the temp files location changes for a given program

[brisk chasm]

im always kind of amazed when i take a binary i cooked on arch box1 and let it run on some oeboentoe test vm and it just works out of ze box, or other way around

[broken fog]

why do the programs need to write shit anywhere other than their binary dir or a global temp dir anyway

store the config next to the binary, shrimple

[solid grove]

i would be amazed too, because i have never had this type of cross-distro running work

[brisk chasm]

XDG_BASE_DIRECTORY you mean 🙂

[wraith urchin]

Sorry about the thread Bjorn, we went a lil off topic

[solid grove]

because the install location is not always writable

esp if you got installed via a package manager

[broken fog]

hrm fair

[solid grove]

there is no excuse to dump your cache in my documents folder though

[brisk chasm]

mozilla fixed a 21 year old request recently, which was around ~/.mozilla on linux not adhering to XDG_BASE_DIRECTORY

[broken fog]

the only two sane options imo are either store it next to the binary and ensure the program has access to its own directory, or global (or user-wide) config dir

[brisk chasm]

yeah, in a multiuser environment having it all in the app dir is weird, but, if i am the only user, yes, lets go

btw next time we derail ravbugs fred

[cloud rivet]

I don’t mind

I am not making a lot of progress this weekend just hanging out with my kid at the convention

[true moon]

what's the convention

[silk pier]

always important to spend time with the kids

code can be written tomorrow

kids won't always be there tomorrow

[solid grove]

bikeshedders anonymous

[true moon]

I guess something in SF from his picture yesterday

I was working the first aid station at SJMADE on Friday

[brisk chasm]

San Jancisco

[true moon]

lol

[solid grove]

This, based on the lanyard https://fanexpohq.com/fanexposanfrancisco/

[cloud rivet]

SF Fanexpo

[silk pier]

people do things in San Jose?
I thought the city was mostly empty...

interesting.

[solid grove]

a lot of it is your typical suburban sprawl

SJ is also home to the nations worst light rail system

[silk pier]

most expensive as well*

()

[broken fog]

at least it has a light rail system

that's better than most of the us

[solid grove]

I rode the VTA light rail once. should’ve walked instead

the plan originally was to build a circular system like in paris. but they only built a quarter of the circle

[silk pier]

In California you either drive or uber to places

every other option has severe downsides

[solid grove]

i've used SF Muni, caltrain, and bart and those have all been fine. but yeah everywhere else you're toast

[silk pier]

you still need to uber to the rail system

(and from)

[cloud rivet]

I take the AC Transit bus to work

I live near Berkeley

[silk pier]

USA buses are not very enjoyable to use (worse than Canada/europe at the very least)

[solid grove]

i would take the bus to work but I'd have to connect halfway and the busses arrive within 2 minutes of each other. so if the first one is late then i'm stuck at that street corner for 30 minutes

not worth it

esp since my entire commute by car is 30 minutes

[cloud rivet]

The transbay AC bus is a pretty fancy bus

Karl Urban is doing talk that’s cool

Cast from the Boyz

[brisk chasm]

https://tenor.com/view/dredd-burn-watching-apathy-karl-urban-gif-20285124

[true moon]

I didn't think so but there were 12,000 people there lol

I think me and my partner were the only white people in the entire building though

About the demographics of SJ you'd expect

Pretty neat event

Cool that people line up like crazy for convention center stuff I guess nature is healing