Update convention for op handle.
This commit is contained in:
1 parent
ca16dae100
commit
0a281c571c
14 files changed
+841
-46
No files matched your search
-32
@@ -1,32 +0,0 @@
|
||||
---
|
||||
title: "@rianflo Explicit packed 16-bit code."
|
||||
type: archive
|
||||
source: twitter
|
||||
source_url: "https://x.com/NOTimothyLottes/status/1597798161665253376"
|
||||
author: "NOTimothyLottes"
|
||||
handle: NOTimothyLottes
|
||||
post_id: "1597798161665253376"
|
||||
date: 2022-11-30
|
||||
archived: 2026-08-23
|
||||
draft: false
|
||||
tags:
|
||||
- archive
|
||||
- twitter
|
||||
- NOTimothyLottes
|
||||
description: "@rianflo Explicit packed 16-bit code."
|
||||
in_reply_to: "1597720454000541696"
|
||||
---
|
||||
|
||||
## Source
|
||||
|
||||
- URL: https://x.com/NOTimothyLottes/status/1597798161665253376
|
||||
- Author: NOTimothyLottes (@NOTimothyLottes)
|
||||
- Posted: 2022-11-30 03:42:22
|
||||
|
||||
> @NOTimothyLottes Oh wait, you're saying you wrote the fp16 math manually?
|
||||
|
||||
## Thread
|
||||
|
||||
**1/**
|
||||
|
||||
@rianflo Explicit packed 16-bit code. FSR1 example: https://github.com/GPUOpen-Effects/FidelityFX-FSR/blob/master/ffx-fsr/ffx_fsr1.h - There are different 'F' (32-bit) and 'H' and 'Hx2' (packed 16-bit) functions.
|
||||
@@ -14,7 +14,6 @@ tags:
|
||||
- [[archive/threads/NOTimothyLottes/2022-11-08-example-of-bind-everything-once-layout-aliasing]]
|
||||
- [[archive/threads/NOTimothyLottes/2022-11-09-if-only-pc-vendors-with-open-isas-amd-intel-would]]
|
||||
- [[archive/threads/NOTimothyLottes/2022-11-13-my-prefered-method-of-managing-application-data]]
|
||||
- [[archive/threads/NOTimothyLottes/2022-11-30-explicit-packed-16-bit-code-fsr1-example-https]]
|
||||
- [[archive/threads/NOTimothyLottes/2023-04-26-can-make-a-custom-forth-to-binary-in-a-few-k]]
|
||||
- [[archive/threads/NOTimothyLottes/2023-06-30-nothing-oriented-programming-nop-where-the-goal]]
|
||||
- [[archive/threads/NOTimothyLottes/2023-08-03-spc-using-8-deep-swap-all-aliasing-same-bo-image]]
|
||||
|
||||
@@ -14,3 +14,4 @@ How to read a folder: [[notes/How threads are organized]].
|
||||
Example of the note shape (single spine, no branches): [[archive/threads/How to archive a thread]]
|
||||
|
||||
- [[archive/threads/NOTimothyLottes]]
|
||||
- [[archive/threads/rianflo]]
|
||||
+59
@@ -0,0 +1,59 @@
|
||||
---
|
||||
title: "\"The road to 16-bit floats GPU is paved with our blood\""
|
||||
type: archive
|
||||
source: twitter
|
||||
source_url: "https://x.com/rianflo/status/1597661267845865474"
|
||||
author: "Florian Hoenig"
|
||||
handle: rianflo
|
||||
post_id: "1597661267845865474"
|
||||
date: 2022-11-29
|
||||
archived: 2026-08-24
|
||||
draft: false
|
||||
tags:
|
||||
- archive
|
||||
- twitter
|
||||
- rianflo
|
||||
description: "\"The road to 16-bit floats GPU is paved with our blood\""
|
||||
in_reply_to: ""
|
||||
---
|
||||
|
||||
## Source
|
||||
|
||||
- URL: https://x.com/rianflo/status/1597661267845865474
|
||||
- Author: Florian Hoenig (@rianflo)
|
||||
- Posted: 2022-11-29 18:38:24
|
||||
|
||||
## Thread
|
||||
|
||||
**1/**
|
||||
|
||||
"The road to 16-bit floats GPU is paved with our blood"
|
||||
:-/
|
||||
|
||||
https://www.yosoygames.com.ar/wp/2022/01/the-road-to-16-bit-floats-gpu-is-paved-with-our-blood/
|
||||
|
||||
**2/**
|
||||
|
||||
@rianflo Explicit packed 16-bit works on AMD Vulkan Vega and up. I typically get up to 30% improvement on ALU bound stuff. Lots of occupancy wins. I don't use {HLSL, RenderDoc, Reflection, RADV, or VS/PS}. All constants are packed and aliased as UINT, so no coversion overheads.
|
||||
|
||||
**3/**
|
||||
|
||||
@NOTimothyLottes Oh I know the benefits. Just no simple clear way to write it in GLSL for vulkan.
|
||||
|
||||
**4/**
|
||||
|
||||
@rianflo Sure there is. CAS/FSR1/etc all shipped with fantastic GLSL versions using 16-bit packed math (I wrote those), all which at the time got fantastic code generation using AMD's drivers.
|
||||
|
||||
**5/**
|
||||
|
||||
@NOTimothyLottes What GLSL extension did you use?
|
||||
|
||||
**6/**
|
||||
|
||||
@NOTimothyLottes Oh wait, you're saying you wrote the fp16 math manually?
|
||||
|
||||
**7/**
|
||||
|
||||
@rianflo Explicit packed 16-bit code. FSR1 example: https://github.com/GPUOpen-Effects/FidelityFX-FSR/blob/master/ffx-fsr/ffx_fsr1.h - There are different 'F' (32-bit) and 'H' and 'Hx2' (packed 16-bit) functions.
|
||||
|
||||
|
||||
@@ -0,0 +1,12 @@
|
||||
---
|
||||
title: rianflo
|
||||
type: note
|
||||
draft: false
|
||||
description: Archived threads by rianflo.
|
||||
tags:
|
||||
- archive
|
||||
- twitter
|
||||
- rianflo
|
||||
---
|
||||
|
||||
- [[archive/threads/rianflo/2022-11-29-the-road-to-16-bit-floats-gpu-is-paved-with-our]]
|
||||
Reference in new issue
Block a user