profile
viewpoint
Ashley expenses Berlin https://expenses.github.io Graphics programmer and occasional gamedev living in Berlin, DE

expenses/cheese 37

A RTS about mice on the moon made for the 2020 Github Game Off

expenses/face_recognition 26

💀 UNMAINTAINED 💀 Rust bindings to dlibs face recognition tools

expenses/art-downloader 5

A semi-automatic downloading script for Google Arts and Culture images.

expenses/bloom 4

A vulkan implementation of the bloom method presented in Next Generation Post Processing in Call of Duty: Advanced Warfare.

expenses/async-rl 3

Tensorflow + Keras + OpenAI Gym implementation of 1-step Q Learning from "Asynchronous Methods for Deep Reinforcement Learning"

expenses/expenses.github.io 1

https://expenses.github.io

expenses/169-pwa-trusted-web-activity 0

Episode 169 - Deploy a PWA to as an Android Package to the Google Play Store

expenses/Akumuli 0

Time-series database

push eventexpenses/ash-opinionated-abstractions

Ashley Ruglys

commit sha 3da2793f2f4661277f193272dfa1e98b29338432

Update gpu-allocator and ash

view details

push time in 2 days

push eventexpenses/rust-gpu-cli-builder

Ashley Ruglys

commit sha 17c7e06d273cd368595c9e0bc24cf174ac4f65d6

Update rust gpu

view details

push time in 7 days

push eventexpenses/transmission-renderer

Ashley Ruglys

commit sha 57c2ab9207c9cc347dc882fd4c2f8fb4bfab8846

Update rust-gpu

view details

push time in 7 days

push eventexpenses/kayteex

Ashley

commit sha 9af4e0943316e8f06a452bd895229f7fa9cdced1

Parse more things

view details

Ashley

commit sha 15b897d3f23b76884182c47c9712ed16c9187ab3

Added VkFormat, etc etc

view details

push time in 8 days

create barnchexpenses/kayteex

branch : master

created branch time in 11 days

created repositoryexpenses/kayteex

A KTX2 parser and decompressor

created time in 11 days

issue openedKhronosGroup/KTX-Specification

The example file in the spec is invalid with regards to `dfdByteLength` equalling `dfdTotalSize`.

[https://github.khronos.org/KTX-Specification/#_an_example_ktx_version_2_file](The file included in the spec) is not valid according to the same specification.

3.9.2. dfdByteLength states

dfdByteLength must equal dfdTotalSize.

However, in the example dfdByteLength (inconsistently referred to as dfdByteSize) is 0x0000005C (92) while dfdTotalSize is 0x3C (60).

created time in 12 days

PR opened EmbarkStudios/rust-gpu

Add a `OpAtomicIIncrement` function to arch

I thought I'd just upstream this function from my shader code. While implementing all the atomic instructions with a nice api would of course be nice, this is all you need for doing simple bump allocations into a storage buffer e.g. for frustum or light culling.

I've been using it with the Device scope in my code but that's not allowed in the compiletests without a new capability so the test just uses Workgroup for now:

error: error:0:0 - Use of device scope with VulkanKHR memory model requires the VulkanMemoryModelDeviceScopeKHR capability

+40 -0

0 comment

2 changed files

pr created time in 12 days

create barnchexpenses/rust-gpu

branch : atomic-i-increment

created branch time in 12 days

push eventexpenses/transmission-renderer

Ashley Ruglys

commit sha f1f32851552ced51b1c2096429b9cb539402c67f

Switch rust-gpu back to upstream

view details

push time in 12 days

push eventexpenses/rust-gpu-cli-builder

Ashley Ruglys

commit sha 3bf7797c91a4570e8e91cdc58d211730044a5f4d

Switch branch back to upstream

view details

push time in 12 days

push eventexpenses/transmission-renderer

Ashley Ruglys

commit sha 53d587925aaa652e38a4420f4983471c3a2daab9

Bunch of misc changes, tidying up etc

view details

Ashley Ruglys

commit sha e9252afbe943a9e7c3cd3772dbabc1dc8db37a0f

Cargo fmt

view details

Ashley Ruglys

commit sha 118f3f6ac8cefd1d120f42545613e28f51e62f1f

Remove mean remapping

view details

Ashley Ruglys

commit sha 19dfdfa88a63802fc4f4007a17aaf13e12edac93

Preserve bindings so that the code works without --ray-tracing

view details

Ashley Ruglys

commit sha b722452f3584d66fbee190c6040253a371cf37a4

Update ash

view details

Ashley Ruglys

commit sha bc440ac3e14fe461f5b8151fc94c5cec1d691b0e

Allow for setting a gltf variant to use

view details

push time in 12 days

Pull request review commentEmbarkStudios/kajiya

Allow scale to be set of an instance mesh

 impl WorldRenderer {             .instances             .iter()             .flat_map(|inst| {-                let inst_position = inst.position;-                let inst_rotation = inst.rotation;+                let (_scale, roation, translation) =
                let (_scale, rotation, translation) =
TuckerBMorgan

comment created time in 15 days

Pull request review commentEmbarkStudios/kajiya

Allow scale to be set of an instance mesh

 impl WorldRenderer {             .instances             .iter()             .flat_map(|inst| {-                let inst_position = inst.position;-                let inst_rotation = inst.rotation;+                let (_scale, roation, translation) =+                    inst.transformation.to_scale_rotation_translation();+                let inst_position = translation;+                let inst_rotation = roation;
                let inst_rotation = rotation;
TuckerBMorgan

comment created time in 15 days

PullRequestReviewEvent
PullRequestReviewEvent

startedterhechte/postsack

started time in 16 days

issue commentexpenses/ray-tracing-gallery

[Question] can the project run on a specific GPU or a general GPU?

It should be able to be run on any of the GPUs listed here: http://vulkan.gpuinfo.org/listdevicescoverage.php?extension=VK_KHR_ray_tracing_pipeline&platform=all

ccworld1000

comment created time in 17 days

issue commentEmbarkStudios/kajiya

run kajiya in OSX

Ah, you're trying to load the x86 dynamic library but you're on a M1 Mac. I'll try and build a universal library when I get home. Alternatively you could build it yourself.

@tlightsky okay, I've pushed a new commit (https://github.com/EmbarkStudios/kajiya/commit/6599d2ed70e5ae2925cd44f7e546e1c0aa57f3ac) and it should be using a universal binary now. Try running it again.

file lib/libdxcompiler.3.7.dylib:

libdxcompiler.dynlib: Mach-O universal binary with 2 architectures: [x86_64:Mach-O 64-bit dynamically linked shared library x86_64] [arm64]
libdxcompiler.dynlib (for architecture x86_64):	Mach-O 64-bit dynamically linked shared library x86_64
libdxcompiler.dynlib (for architecture arm64):	Mach-O 64-bit dynamically linked shared library arm64

For the record, this can be build like so:

git clone https://github.com/Microsoft/DirectXShaderCompiler --recursive
cd DirectXShaderCompiler
mkdir build
cd build
cmake ..  -GNinja -DCMAKE_BUILD_TYPE=Release '-DCMAKE_OSX_ARCHITECTURES=arm64;x86_64' -C ../cmake/caches/PredefinedParams.cmake
ninja
file lib/libdxcompiler.3.7.dylib
tlightsky

comment created time in 18 days

push eventexpenses/kajiya

Ashley

commit sha 6599d2ed70e5ae2925cd44f7e546e1c0aa57f3ac

Switch to universal libdxcompiler.dynlib

view details

push time in 18 days

issue commentEmbarkStudios/kajiya

run kajiya in OSX

Ah, you're trying to load the x86 dynamic library but you're on a M1 Mac. I'll try and build a universal library when I get home. Alternatively you could build it yourself.

tlightsky

comment created time in 18 days

issue commentMaikKlein/ash

Using builders with conditional blocks

You want

let mut create_info = ...;

and

create_info = create_info.push_next(debug_messenger_create_info);

Instead.

dawsonc623

comment created time in 19 days

PR opened gltf-rs/gltf

Support `KHR_materials_variants`
+260 -4

0 comment

11 changed files

pr created time in 19 days

issue commentEmbarkStudios/kajiya

run kajiya in OSX

I've been looking into porting the code to macOS. I've got a fork here with my progress: https://github.com/EmbarkStudios/kajiya/compare/main...expenses:rtx-off. At the moment I'm just getting a timeout:

[mvk-error] VK_ERROR_DEVICE_LOST: Command buffer 0x7fc47b8e8fe0 "vkQueueSubmit CommandBuffer on Queue 0-0" execution failed (code 2): Caused GPU Timeout Error (00000002:kIOAccelCommandBufferCallbackErrorTimeout)
thread 'main' panicked at 'Wait for fence failed.: ERROR_DEVICE_LOST', crates/lib/kajiya-backend/src/vulkan/device.rs:483:22
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

but perhaps it would work with a better GPU.

To run it you'll need ossp-uuid installed (brew install ossp-uuid) to avoid a note: ld: library not found for -luuid error. This might not work on the M1 macs though (https://github.com/Traverse-Research/hassle-rs/issues/23#issuecomment-1003597930).

Then run ./bake.sh and ./build_and_run.sh battle as per normal.

The first commit (https://github.com/EmbarkStudios/kajiya/commit/60f78a6c5ee83864d6fcafe8d6f88619f0567d81) also seemed to work okay on Windows on a non-RTX GPU.

tlightsky

comment created time in 20 days

push eventexpenses/kajiya

Ashley

commit sha dccf1a46c09a2581f62097a0a588ef9c957aaa61

Adjust script

view details

push time in 20 days

push eventexpenses/kajiya

Ashley

commit sha 5bc2a1814e03dc0707bc87635718e94e65b67eac

Getting a device timeout

view details

push time in 20 days

issue commentopenphotogrammetry/meshroomcl

'Reading configuration' in MultiviewStereoCL stalls at the start of SteroeFusion

Yeah, I upped it to 8GB or so and it started paging, but that's fine. I think the situations where you want the Data Cache Size option to be active are when allocating buffers for OpenCL, as paging seems to be a problem there.

expenses

comment created time in 20 days

issue commentTraverse-Research/hassle-rs

Linking fails on MacOSX (missing -luuid)

I am, yeah.

hrydgard

comment created time in 20 days

issue commentTraverse-Research/hassle-rs

Linking fails on MacOSX (missing -luuid)

Installing ossp-uuid (brew install ossp-uuid) fixed this for me.

hrydgard

comment created time in 20 days

create barnchexpenses/kajiya

branch : rtx-off

created branch time in 20 days

more