2090977998593827103

This commit is contained in:
ed committed 2026-08-27 19:08:31 -04:00
1 parent 0087cd5b5d
commit a98c86a675
5 files changed
+203

No files matched your search

@@ -0,0 +1,48 @@
---
title: "@AgileJebrim Why?"
type: archive
source: twitter
source_url: "https://x.com/_s_u_r_a_/status/2091080756806521239"
author: "asdb"
handle: _s_u_r_a_
post_id: "2091080756806521239"
date: 2026-08-22
archived: 2026-08-27
draft: false
tags:
- archive
- twitter
- AgileJebrim
description: "@AgileJebrim Why?"
in_reply_to: ""
parent_post_id: "2090977998593827103"
---
## Source
- URL: https://x.com/_s_u_r_a_/status/2091080756806521239
- Author: asdb (@_s_u_r_a_)
- Posted: 2026-08-22 08:31:17
## Branch
**1/** **@_s_u_r_a_** ^2091080756806521239
**@AgileJebrim**
Why?
**2/** **@AgileJebrim** ^2091110649526972527
Because they’re designed to be data parallel and work with GPUs from the start.
Because they don’t require integrating with an existing bloated library.
SPIR-V is just an incredibly simple and straightforward binary format. We write to it directly as an array of uint32s. You can even read the bytecode in a hex editor directly easily enough if you wanted. Think of it like Java bytecode; it’s portable across a wide range of hardware and vendors.
PTX unfortunately requires generating ASCII output to work with it, so a little more complicated but still nothing like linking to LLVM. It is good for accessing more low level control over NVIDIA GPUs.
We’re also creating our own IR that I suppose one could hypothetically target via code too. We’re doing it internally anyways until our visual editor is ready.
## Related
- Spine: [[archive/threads/AgileJebrim/2026-08-22-start-with-spir-v-or-ptx-instead]]
@@ -0,0 +1,85 @@
---
title: "@AgileJebrim ptx, my ass: https://redplait.blogspot.com/2026/08/parser-of-ptx-instructions.html"
type: archive
source: twitter
source_url: "https://x.com/real_redp/status/2091228915142472141"
author: "red plait"
handle: real_redp
post_id: "2091228915142472141"
date: 2026-08-22
archived: 2026-08-27
draft: false
tags:
- archive
- twitter
- AgileJebrim
description: "@AgileJebrim ptx, my ass: https://redplait.blogspot.com/2026/08/parser-of-ptx-instructions.html"
in_reply_to: ""
parent_post_id: "2090977998593827103"
---
## Source
- URL: https://x.com/real_redp/status/2091228915142472141
- Author: red plait (@real_redp)
- Posted: 2026-08-22 18:20:01
## Branch
**1/** **@real_redp** ^2091228915142472141
**@AgileJebrim**
ptx, my ass: https://redplait.blogspot.com/2026/08/parser-of-ptx-instructions.html
**2/** **@AgileJebrim** ^2091230269747397000
**@real_redp**
FWIW I target SPIR-V in part because it’s a binary format (ASCII asm sucks) and in part because it’s more portable than PTX.
**3/** **@AgileJebrim** ^2091230640842612865
**@real_redp**
The main advantage of PTX is that it gets you direct access to a lower level API into NVIDIA hardware than anything else available unless we manage to reverse engineer PSO binary caches (something I’m still debating attempting) to write directly in SASS.
**4/** **@real_redp** ^2091231795350044950
**@AgileJebrim**
he-he
I made tool to patch sass
**5/** **@AgileJebrim** ^2091233823228940511
**@real_redp**
Which generation? Is it available?
**6/** **@AgileJebrim** ^2091234694201331796
**@real_redp**
I see you have a disassembler. It looks to be CUDA-specific? I’d love to see you do one for Vulkan PSOs too, especially for Ampere.
https://github.com/redplait/denvdis/tree/master
**7/** **@AgileJebrim** ^2091235158682730697
**@real_redp**
Actually, I think this path might work as is combined with the Cubin extension in Vulkan.
“sed-like tool for inline patching of sass instructions within cubin files.”
Lots of possibilities! 🤔
How mature is this?
**8/** **@real_redp** ^2091235642898325872
**@AgileJebrim**
enough to do crazy things like https://redplait.blogspot.com/2026/07/optimization-of-sass-stall-counts.html
## Related
- Spine: [[archive/threads/AgileJebrim/2026-08-22-start-with-spir-v-or-ptx-instead]]
@@ -0,0 +1,37 @@
---
title: "@AgileJebrim Are people seriously starting from ptx 😭 I thought you went there after you got a good grip on cuda"
type: archive
source: twitter
source_url: "https://x.com/thisispiyushK/status/2091255763121721523"
author: "lsm_"
handle: thisispiyushK
post_id: "2091255763121721523"
date: 2026-08-22
archived: 2026-08-27
draft: false
tags:
- archive
- twitter
- AgileJebrim
description: "@AgileJebrim Are people seriously starting from ptx 😭 I thought you went there after you got a good grip on cuda"
in_reply_to: ""
parent_post_id: "2090977998593827103"
---
## Source
- URL: https://x.com/thisispiyushK/status/2091255763121721523
- Author: lsm_ (@thisispiyushK)
- Posted: 2026-08-22 20:06:42
## Branch
**1/** **@thisispiyushK** ^2091255763121721523
**@AgileJebrim**
Are people seriously starting from ptx 😭 I thought you went there after you got a good grip on cuda
## Related
- Spine: [[archive/threads/AgileJebrim/2026-08-22-start-with-spir-v-or-ptx-instead]]
@@ -0,0 +1,32 @@
---
title: "Start with SPIR-V or PTX instead."
type: archive
source: twitter
source_url: "https://x.com/AgileJebrim/status/2090977998593827103"
author: "Jebrim"
handle: AgileJebrim
post_id: "2090977998593827103"
date: 2026-08-22
archived: 2026-08-27
draft: false
tags:
- archive
- twitter
- AgileJebrim
description: "Start with SPIR-V or PTX instead."
in_reply_to: ""
---
## Source
- URL: https://x.com/AgileJebrim/status/2090977998593827103
- Author: Jebrim (@AgileJebrim)
- Posted: 2026-08-22 01:42:58
## Thread
**1/** **@AgileJebrim** ^2090977998593827103
Start with SPIR-V or PTX instead.
Branches: [[archive/threads/AgileJebrim/2026-08-22-start-with-spir-v-or-ptx-instead/2026-08-22-_s_u_r_a_-why]], [[archive/threads/AgileJebrim/2026-08-22-start-with-spir-v-or-ptx-instead/2026-08-22-real_redp-ptx-my-ass-https-redplait-blogspot-com-2026-08]], [[archive/threads/AgileJebrim/2026-08-22-start-with-spir-v-or-ptx-instead/2026-08-22-thisispiyushK-are-people-seriously-starting-from-ptx-i-thought]]
+1
View File
@@ -13,3 +13,4 @@ tags:
- [[archive/threads/AgileJebrim/2025-02-09-check-your-compiler]]
- [[archive/threads/AgileJebrim/2026-05-12-we-have-code-reviews-at-isochron-and-i-require-my]]
- [[archive/threads/AgileJebrim/2026-05-30-i-dont-think-hardware-vendors-are-going-to]]
- [[archive/threads/AgileJebrim/2026-08-22-start-with-spir-v-or-ptx-instead]]