simd post

2026-08-25 16:01:44 -04:00
1 parent 0dd4a41807
commit 11c9b1c15e
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain The answer exists."
+type: archive
+source: twitter
+source_url: "https://x.com/AgileJebrim/status/1944947900104237129"
+author: "Jebrim"
+handle: AgileJebrim
+post_id: "1944947900104237129"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain The answer exists."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/AgileJebrim/status/1944947900104237129
+- Author: Jebrim (@AgileJebrim)
+- Posted: 2025-07-15 02:31:29
+
+## Branch
+
+**1/** @AgileJebrim
+
+@nicbarkeragain
+
+The answer exists. It’s called ISPC.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Reduced instruction sets."
+type: archive
+source: twitter
+source_url: "https://x.com/BigWhoopGames/status/1945048842342219995"
+author: "Big Whoop Games"
+handle: BigWhoopGames
+post_id: "1945048842342219995"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Reduced instruction sets."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/BigWhoopGames/status/1945048842342219995
+- Author: Big Whoop Games (@BigWhoopGames)
+- Posted: 2025-07-15 09:12:35
+
+## Branch
+
+**1/** @BigWhoopGames
+
+@nicbarkeragain
+
+Reduced instruction sets. If we have op A and op B but B can be expressed as a series of As, just get rid of it. If all the parallel cores need to be running the same op at the same time, just optimize for the least instructions possible but more of them in a row.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain ISPC"
+type: archive
+source: twitter
+source_url: "https://x.com/DanielJCollier/status/1945037474742591968"
+author: "Daniel"
+handle: DanielJCollier
+post_id: "1945037474742591968"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain ISPC"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/DanielJCollier/status/1945037474742591968
+- Author: Daniel (@DanielJCollier)
+- Posted: 2025-07-15 08:27:25
+
+## Branch
+
+**1/** @DanielJCollier
+
+@nicbarkeragain
+
+ISPC
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Mojo lang  has SIMD aware data types and makes it a first class citizen."
+type: archive
+source: twitter
+source_url: "https://x.com/DeepknowledgeU/status/1944996456974217322"
+author: "Govind"
+handle: DeepknowledgeU
+post_id: "1944996456974217322"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Mojo lang  has SIMD aware data types and makes it a first class citizen."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/DeepknowledgeU/status/1944996456974217322
+- Author: Govind (@DeepknowledgeU)
+- Posted: 2025-07-15 05:44:26
+
+## Branch
+
+**1/** @DeepknowledgeU
+
+@nicbarkeragain
+
+Mojo lang  has SIMD aware data types and makes it a first class citizen. Pretty interesting model
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,79 @@
+---
+title: "My gamedev head just wants a sort of game mechanic lego for conveying a simd operation frame."
+type: archive
+source: twitter
+source_url: "https://x.com/Ed__dev/status/1944930386984734925"
+author: "Ed_"
+handle: Ed__dev
+post_id: "1944930386984734925"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "My gamedev head just wants a sort of game mechanic lego for conveying a simd operation frame."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/Ed__dev/status/1944930386984734925
+- Author: Ed_ (@Ed__dev)
+- Posted: 2025-07-15 01:21:53
+
+## Branch
+
+**1/** @Ed__dev
+
+My gamedev head just wants a sort of game mechanic lego for conveying a simd operation frame. We have carts to give the operation rail and we'll have it vend us the result in the storage bin. All these bins are slices. Everything inside the rail could behave like a shader's frame where you're only allowed to-do simd ops...
+So far that I can tell if its in an iterator, the iterator is almost always going to be succeeded after this op occurs so it can be optionally incremented implicitly or manually.  
+This may have been done already I haven't done that much studies on this.
+
+![](https://pbs.twimg.com/media/Gv3H3-IWkAAmiVO?format=png&name=orig)
+
+**2/** @Ed__dev
+
+You could do more composition by having calls to procs, but it would be the async/await situation where now that proc must be a rail proc for the specific width. 
+All operations within the procs must end up being simd, but the slices from arbitrary pointers can come from any struct in memory.
+
+**3/** @Ed__dev
+
+@nicbarkeragain
+
+![](https://pbs.twimg.com/media/Gv3L4JqXAAA22Q4?format=png&name=orig)
+
+**4/** @Ed__dev
+
+@nicbarkeragain
+
+You could probably do mixin of different simd widths within one rail block but I don't know how that would start to get messy so maybe provide some sort of layout struct like the shaders to specify what the rail expects width wise for a set of bound vars.
+
+**5/** @Ed__dev
+
+The nuance is that I know I want to feed these things a very specific width for a range of ops during hot-paths in a pipeline of an program. Its annoying to deal with this width and striding it. Shaders deal with this concept for you to a great degree. These array data layouts for operating on this array data should all be specificable at comp-time so you just need a way to triage that to compositional definitions. Specifying them constantly for every single var's access to their elements is annoying and add complexity to parsing what's going on visually.
+
+Specifically, since at many times any user is doing some weird strides/offsets on data to process, you want to specify on simd/rail blocks runtime offset vars that will be utilized with so they may be in a proc's args but then the rail knows that offset int is for a set of bound slices offset math in a layout definition for the rail/simd block.
+This would make the block scope's math data arrays look like shader math, because all the slice specification for the SIMDs was dealt with via specifying some type forum on the simd slices. 
+
+This is equivalent to what structs did for C on doing data offset math on single data instructions.
+
+**6/** @Ed__dev
+
+@nicbarkeragain
+
+Figuring it out for simd will probably lead to systems langs being able to triage shader compilation and loading of those SPOs to the gpu in in the same lang syntax and you'd just have specific rules for compute, vertex, and frag shader railing blocks.
+
+**7/** @Ed__dev
+
+@nicbarkeragain
+
+Last dumb example I'll muse about. Def would be fun to try prototyping something like this...
+
+![](https://pbs.twimg.com/media/Gv4EmPLWAAADJDi?format=jpg&name=orig)
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain I have hope that this is true."
+type: archive
+source: twitter
+source_url: "https://x.com/JakeLegendXIII/status/1944967041368129746"
+author: "Jake Kirkbride"
+handle: JakeLegendXIII
+post_id: "1944967041368129746"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain I have hope that this is true."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/JakeLegendXIII/status/1944967041368129746
+- Author: Jake Kirkbride (@JakeLegendXIII)
+- Posted: 2025-07-15 03:47:33
+
+## Branch
+
+**1/** @JakeLegendXIII
+
+@nicbarkeragain
+
+I have hope that this is true. But damn have we tried and nobody has had a good answer so far. Hoping some of the AI distraction doesn’t take too much away from people still doing other computer science to make progress on actual programming.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain FFmpeg thinks otherwise."
+type: archive
+source: twitter
+source_url: "https://x.com/JebsSteve0x1/status/1944937394127470633"
+author: "Stove Jebs"
+handle: JebsSteve0x1
+post_id: "1944937394127470633"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain FFmpeg thinks otherwise."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/JebsSteve0x1/status/1944937394127470633
+- Author: Stove Jebs (@JebsSteve0x1)
+- Posted: 2025-07-15 01:49:44
+
+## Branch
+
+**1/** @JebsSteve0x1
+
+@nicbarkeragain
+
+FFmpeg thinks otherwise.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,41 @@
+---
+title: "@nicbarkeragain Have you ever studied APL?"
+type: archive
+source: twitter
+source_url: "https://x.com/SumErgoMonstro/status/1945143539224002588"
+author: "The Monster"
+handle: SumErgoMonstro
+post_id: "1945143539224002588"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Have you ever studied APL?"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/SumErgoMonstro/status/1945143539224002588
+- Author: The Monster (@SumErgoMonstro)
+- Posted: 2025-07-15 15:28:53
+
+## Branch
+
+**1/** @SumErgoMonstro
+
+@nicbarkeragain
+
+Have you ever studied APL?
+
+It has the best array-processing I've ever seen.  You can invert a matrix with a single operator.
+
+It's going to be something that treats arrays of data as objects to be operated upon.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain 128-bit seems to be pretty widespread, so it might be a good baseline? even in armv7 cores, not sure about risc-v and powerpc though."
+type: archive
+source: twitter
+source_url: "https://x.com/TGodfather1984/status/1944994288305098754"
+author: "TheGodfather1984"
+handle: TGodfather1984
+post_id: "1944994288305098754"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain 128-bit seems to be pretty widespread, so it might be a good baseline? even in armv7 cores, not sure about risc-v and powerpc though."
+in_reply_to: ""
+parent_post_id: "1944937375001666018"
+---
+
+## Source
+
+- URL: https://x.com/TGodfather1984/status/1944994288305098754
+- Author: TheGodfather1984 (@TGodfather1984)
+- Posted: 2025-07-15 05:35:49
+
+## Branch
+
+**1/** @TGodfather1984
+
+@nicbarkeragain
+
+128-bit seems to be pretty widespread, so it might be a good baseline? even in armv7 cores, not sure about risc-v and powerpc though. and then the compiler could just divide the number of iterations by 2 or 4, or am i missing something?
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain C# makes it fairly straightforward."
+type: archive
+source: twitter
+source_url: "https://x.com/TheBuzzSaw/status/1944927781512733078"
+author: "TheBuzzSaw"
+handle: TheBuzzSaw
+post_id: "1944927781512733078"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain C# makes it fairly straightforward."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/TheBuzzSaw/status/1944927781512733078
+- Author: TheBuzzSaw (@TheBuzzSaw)
+- Posted: 2025-07-15 01:11:32
+
+## Branch
+
+**1/** @TheBuzzSaw
+
+@nicbarkeragain
+
+C# makes it fairly straightforward.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Intel's ISPC"
+type: archive
+source: twitter
+source_url: "https://x.com/VPCOMPRESSB/status/1944933853707063570"
+author: "/i:'mɪər/"
+handle: VPCOMPRESSB
+post_id: "1944933853707063570"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Intel's ISPC"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/VPCOMPRESSB/status/1944933853707063570
+- Author: /i:'mɪər/ (@VPCOMPRESSB)
+- Posted: 2025-07-15 01:35:40
+
+## Branch
+
+**1/** @VPCOMPRESSB
+
+@nicbarkeragain
+
+Intel's ISPC
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Llm baked into gcc who’s only job is making things simd where possible."
+type: archive
+source: twitter
+source_url: "https://x.com/WorldGhostC2C/status/1944921404933120353"
+author: "Catallus"
+handle: WorldGhostC2C
+post_id: "1944921404933120353"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Llm baked into gcc who’s only job is making things simd where possible."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/WorldGhostC2C/status/1944921404933120353
+- Author: Catallus (@WorldGhostC2C)
+- Posted: 2025-07-15 00:46:12
+
+## Branch
+
+**1/** @WorldGhostC2C
+
+@nicbarkeragain
+
+Llm baked into gcc who’s only job is making things simd where possible. Sinple
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Perhaps something like ISPC / how we author shaders in GPU could help ?"
+type: archive
+source: twitter
+source_url: "https://x.com/Zino2201_/status/1945169506805952589"
+author: "Zino"
+handle: Zino2201_
+post_id: "1945169506805952589"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Perhaps something like ISPC / how we author shaders in GPU could help ?"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/Zino2201_/status/1945169506805952589
+- Author: Zino (@Zino2201_)
+- Posted: 2025-07-15 17:12:04
+
+## Branch
+
+**1/** @Zino2201_
+
+@nicbarkeragain
+
+Perhaps something like ISPC / how we author shaders in GPU could help ?
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Relational model + columnar data."
+type: archive
+source: twitter
+source_url: "https://x.com/_Felipe/status/1945156422716535136"
+author: "Felipe O. Carvalho"
+handle: _Felipe
+post_id: "1945156422716535136"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Relational model + columnar data."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/_Felipe/status/1945156422716535136
+- Author: Felipe O. Carvalho (@_Felipe)
+- Posted: 2025-07-15 16:20:05
+
+## Branch
+
+**1/** @_Felipe
+
+@nicbarkeragain
+
+Relational model + columnar data. SQL lands itself very well to vectorization and scatter+gather algorithms. We need a general purpose language that can express SQL semantics in some places.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain languages that have comptime also have wonderful support for SIMD: mojo and zig"
+type: archive
+source: twitter
+source_url: "https://x.com/__morse/status/1945235507954295160"
+author: "Tommy D. Rossi"
+handle: __morse
+post_id: "1945235507954295160"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain languages that have comptime also have wonderful support for SIMD: mojo and zig"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/__morse/status/1945235507954295160
+- Author: Tommy D. Rossi (@__morse)
+- Posted: 2025-07-15 21:34:20
+
+## Branch
+
+**1/** @__morse
+
+@nicbarkeragain
+
+languages that have comptime also have wonderful support for SIMD: mojo and zig
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Its called array programming."
+type: archive
+source: twitter
+source_url: "https://x.com/agmethod75/status/1944929654340772147"
+author: "Rich Henry"
+handle: agmethod75
+post_id: "1944929654340772147"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Its called array programming."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/agmethod75/status/1944929654340772147
+- Author: Rich Henry (@agmethod75)
+- Posted: 2025-07-15 01:18:59
+
+## Branch
+
+**1/** @agmethod75
+
+@nicbarkeragain
+
+Its called array programming.  Fortran has been doing it since fortran90.  Odin and D have it.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Maybe Futhark, or array programming."
+type: archive
+source: twitter
+source_url: "https://x.com/andrew_v10209/status/1944934680404389997"
+author: "Andrew"
+handle: andrew_v10209
+post_id: "1944934680404389997"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Maybe Futhark, or array programming."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/andrew_v10209/status/1944934680404389997
+- Author: Andrew (@andrew_v10209)
+- Posted: 2025-07-15 01:38:57
+
+## Branch
+
+**1/** @andrew_v10209
+
+@nicbarkeragain
+
+Maybe Futhark, or array programming. There's an entire compiler that runs on the GPU written in Futhark (i.e. it does parsing, lexing, register allocation, etc on the GPU). It's called Pareas
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,49 @@
+---
+title: "@nicbarkeragain https://github.com/HigherOrderCO/Bend"
+type: archive
+source: twitter
+source_url: "https://x.com/another_sentry/status/1944914198397624515"
+author: "another sentry"
+handle: another_sentry
+post_id: "1944914198397624515"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain https://github.com/HigherOrderCO/Bend"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/another_sentry/status/1944914198397624515
+- Author: another sentry (@another_sentry)
+- Posted: 2025-07-15 00:17:34
+
+## Branch
+
+**1/** @another_sentry
+
+@nicbarkeragain
+
+https://github.com/HigherOrderCO/Bend
+
+**2/** @nicbarkeragain
+
+@another_sentry
+
+I've been keeping an eye on this & the HVM work, there is some really interesting stuff happening there. My issue however is this doesn't feel like a general purpose language, more something that you would reach for if you already knew you had a massively parallel problem.
+
+**3/** @AgileJebrim
+
+@nicbarkeragain @another_sentry
+
+Not even. Their performance is atrocious and their recursive programming model doesn’t map well to GPUs.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain It hasn't been stabilized yet but Rust's portable SIMD was pretty easy to work with https://doc.rust-lang.org/std/simd/index.html"
+type: archive
+source: twitter
+source_url: "https://x.com/asdhuqiwuhdais/status/1945222545709453526"
+author: "asdhuqiwuhdais 🇳🇱🇺🇦🇪🇺🏳️‍🌈"
+handle: asdhuqiwuhdais
+post_id: "1945222545709453526"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain It hasn't been stabilized yet but Rust's portable SIMD was pretty easy to work with https://doc.rust-lang.org/std/simd/index.html"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/asdhuqiwuhdais/status/1945222545709453526
+- Author: asdhuqiwuhdais 🇳🇱🇺🇦🇪🇺🏳️‍🌈 (@asdhuqiwuhdais)
+- Posted: 2025-07-15 20:42:50
+
+## Branch
+
+**1/** @asdhuqiwuhdais
+
+@nicbarkeragain
+
+It hasn't been stabilized yet but Rust's portable SIMD was pretty easy to work with https://doc.rust-lang.org/std/simd/index.html
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain more arrays"
+type: archive
+source: twitter
+source_url: "https://x.com/dezren39/status/1944982626667802713"
+author: "`drew (Parody)"
+handle: dezren39
+post_id: "1944982626667802713"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain more arrays"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/dezren39/status/1944982626667802713
+- Author: `drew (Parody) (@dezren39)
+- Posted: 2025-07-15 04:49:28
+
+## Branch
+
+**1/** @dezren39
+
+@nicbarkeragain
+
+more arrays
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain openmp does good enough job with compiler flags like march=avx2, i think the focus should be making compilers smarter, for the cases where parallelism is the primary concern one should use cuda"
+type: archive
+source: twitter
+source_url: "https://x.com/dodobyte/status/1945103550113054841"
+author: "Dodo"
+handle: dodobyte
+post_id: "1945103550113054841"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain openmp does good enough job with compiler flags like march=avx2, i think the focus should be making compilers smarter, for the cases where parallelism is the primary concern one should use cuda"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/dodobyte/status/1945103550113054841
+- Author: Dodo (@dodobyte)
+- Posted: 2025-07-15 12:49:59
+
+## Branch
+
+**1/** @dodobyte
+
+@nicbarkeragain
+
+openmp does good enough job with compiler flags like march=avx2, i think the focus should be making compilers smarter, for the cases where parallelism is the primary concern one should use cuda
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain I think ISPC more or less solved this and in a C-compatible way."
+type: archive
+source: twitter
+source_url: "https://x.com/etscrivner/status/1944916196429848910"
+author: "Eric Scrivner"
+handle: etscrivner
+post_id: "1944916196429848910"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain I think ISPC more or less solved this and in a C-compatible way."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/etscrivner/status/1944916196429848910
+- Author: Eric Scrivner (@etscrivner)
+- Posted: 2025-07-15 00:25:30
+
+## Branch
+
+**1/** @etscrivner
+
+@nicbarkeragain
+
+I think ISPC more or less solved this and in a C-compatible way.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Might be an oversimplification but I think the focus might need to be aimed more at grouping data efficiently."
+type: archive
+source: twitter
+source_url: "https://x.com/etzikun/status/1944931271077249360"
+author: "Etzikun"
+handle: etzikun
+post_id: "1944931271077249360"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Might be an oversimplification but I think the focus might need to be aimed more at grouping data efficiently."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/etzikun/status/1944931271077249360
+- Author: Etzikun (@etzikun)
+- Posted: 2025-07-15 01:25:24
+
+## Branch
+
+**1/** @etzikun
+
+@nicbarkeragain
+
+Might be an oversimplification but I think the focus might need to be aimed more at grouping data efficiently. Thinking more with association rather than just individual variables. There’s only so much optimisation a compiler can force. It’s (partially) a memory management issue.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Possible."
+type: archive
+source: twitter
+source_url: "https://x.com/examachine/status/1944938026037338610"
+author: "Eray Özkural, PhD - OG AI ⏭️🌟"
+handle: examachine
+post_id: "1944938026037338610"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Possible."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/examachine/status/1944938026037338610
+- Author: Eray Özkural, PhD - OG AI ⏭️🌟 (@examachine)
+- Posted: 2025-07-15 01:52:15
+
+## Branch
+
+**1/** @examachine
+
+@nicbarkeragain
+
+Possible. :)
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain So, FORTRAN?"
+type: archive
+source: twitter
+source_url: "https://x.com/groby/status/1944979658778239333"
+author: "Rachel Blum"
+handle: groby
+post_id: "1944979658778239333"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain So, FORTRAN?"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/groby/status/1944979658778239333
+- Author: Rachel Blum (@groby)
+- Posted: 2025-07-15 04:37:41
+
+## Branch
+
+**1/** @groby
+
+@nicbarkeragain
+
+So, FORTRAN?
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain the industry is trying to do that with vector-length-agnostic architectures, like SVE, RVV, and others."
+type: archive
+source: twitter
+source_url: "https://x.com/jon_roelofs/status/1944944456056365486"
+author: "jon"
+handle: jon_roelofs
+post_id: "1944944456056365486"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain the industry is trying to do that with vector-length-agnostic architectures, like SVE, RVV, and others."
+in_reply_to: ""
+parent_post_id: "1944937375001666018"
+---
+
+## Source
+
+- URL: https://x.com/jon_roelofs/status/1944944456056365486
+- Author: jon (@jon_roelofs)
+- Posted: 2025-07-15 02:17:48
+
+## Branch
+
+**1/** @jon_roelofs
+
+@nicbarkeragain
+
+the industry is trying to do that with vector-length-agnostic architectures, like SVE, RVV, and others.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain I agree"
+type: archive
+source: twitter
+source_url: "https://x.com/jshurmer/status/1944928429071053201"
+author: "Jordan Shurmer"
+handle: jshurmer
+post_id: "1944928429071053201"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain I agree"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/jshurmer/status/1944928429071053201
+- Author: Jordan Shurmer (@jshurmer)
+- Posted: 2025-07-15 01:14:07
+
+## Branch
+
+**1/** @jshurmer
+
+@nicbarkeragain
+
+I agree
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain The \"average programmer\" you mentioned uses OOP and other things that make their programs so slower and fatter that SIMD would make literally zero impact within that scale of inefficiency through bad programming practices."
+type: archive
+source: twitter
+source_url: "https://x.com/meglio/status/1944957913279422515"
+author: "meglio"
+handle: meglio
+post_id: "1944957913279422515"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain The \"average programmer\" you mentioned uses OOP and other things that make their programs so slower and fatter that SIMD would make literally zero impact within that scale of inefficiency through bad programming practices."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/meglio/status/1944957913279422515
+- Author: meglio (@meglio)
+- Posted: 2025-07-15 03:11:16
+
+## Branch
+
+**1/** @meglio
+
+@nicbarkeragain
+
+The "average programmer" you mentioned uses OOP and other things that make their programs so slower and fatter that SIMD would make literally zero impact within that scale of inefficiency through bad programming practices. You are attacking the wrong problem.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain @Modular Mojo has a first class SIMD support AFAIK but I've never used it https://docs.modular.com/mojo/stdlib/builtin/simd/SIMD#implemented-traits"
+type: archive
+source: twitter
+source_url: "https://x.com/meoyawn/status/1945231466549363169"
+author: "Adel"
+handle: meoyawn
+post_id: "1945231466549363169"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain @Modular Mojo has a first class SIMD support AFAIK but I've never used it https://docs.modular.com/mojo/stdlib/builtin/simd/SIMD#implemented-traits"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/meoyawn/status/1945231466549363169
+- Author: Adel (@meoyawn)
+- Posted: 2025-07-15 21:18:16
+
+## Branch
+
+**1/** @meoyawn
+
+@nicbarkeragain @Modular
+
+Mojo has a first class SIMD support AFAIK but I've never used it https://docs.modular.com/mojo/stdlib/builtin/simd/SIMD#implemented-traits
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain I agree, there is something weird with SIMD accessibility today, looks like there is room for innovation"
+type: archive
+source: twitter
+source_url: "https://x.com/quentinoeoeoe/status/1944919638719987940"
+author: "Quentin"
+handle: quentinoeoeoe
+post_id: "1944919638719987940"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain I agree, there is something weird with SIMD accessibility today, looks like there is room for innovation"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/quentinoeoeoe/status/1944919638719987940
+- Author: Quentin (@quentinoeoeoe)
+- Posted: 2025-07-15 00:39:11
+
+## Branch
+
+**1/** @quentinoeoeoe
+
+@nicbarkeragain
+
+I agree, there is something weird with SIMD accessibility today, looks like there is room for innovation
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,39 @@
+---
+title: "@nicbarkeragain I think we need a system visualizer where devs can view their implementation running through the processor/gpus/memory/network."
+type: archive
+source: twitter
+source_url: "https://x.com/salonen/status/1945116028540784694"
+author: "Austin Salonen"
+handle: salonen
+post_id: "1945116028540784694"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain I think we need a system visualizer where devs can view their implementation running through the processor/gpus/memory/network."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/salonen/status/1945116028540784694
+- Author: Austin Salonen (@salonen)
+- Posted: 2025-07-15 13:39:34
+
+## Branch
+
+**1/** @salonen
+
+@nicbarkeragain
+
+I think we need a system visualizer where devs can view their implementation running through the processor/gpus/memory/network.
+
+Then examples could highlight good and bad designs for standard hardware layouts.  Exercises for the reader (AI?) could be to "optimize for yours"
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,73 @@
+---
+title: "@nicbarkeragain have you spent time writing SIMD algorithms for general-purpose tasks? it does not seem possible to make this stuff accessible, because the set of available doodads and their costs influences your algorithm design"
+type: archive
+source: twitter
+source_url: "https://x.com/sasuke___420/status/1944916085217878130"
+author: "sasuke⚡420"
+handle: sasuke___420
+post_id: "1944916085217878130"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain have you spent time writing SIMD algorithms for general-purpose tasks? it does not seem possible to make this stuff accessible, because the set of available doodads and their costs influences your algorithm design"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/sasuke___420/status/1944916085217878130
+- Author: sasuke⚡420 (@sasuke___420)
+- Posted: 2025-07-15 00:25:04
+
+## Branch
+
+**1/** @sasuke___420
+
+@nicbarkeragain
+
+have you spent time writing SIMD algorithms for general-purpose tasks? it does not seem possible to make this stuff accessible, because the set of available doodads and their costs influences your algorithm design
+
+**2/** @nicbarkeragain
+
+@sasuke___420
+
+Yeah I have and I agree, it's clear that the way I've learned programming hasn't given me the correct "mental tools" to think about solving problems with a SIMD mindset. Maybe it's possible or maybe it isn't, but it feels like we don't have quite the right framing at the moment.
+
+**3/** @kellabyte
+
+@nicbarkeragain @sasuke___420
+
+Data flow oriented programming instead of “OOP” where objects play roles and interact. Think like data and memory and layer your business rules on top of data structures that are sympathetic to the hardware.
+
+**4/** @sasuke___420
+
+@kellabyte @nicbarkeragain
+
+yeah, a big thing in the proposed "do something like APL" solution is that they can have pretty complicated verbs that end up being implemented with operations that are normally exposed in libraries in normal languages
+
+**5/** @pelavarre
+
+@sasuke___420 @kellabyte @nicbarkeragain
+
+I think
+
+a ) SIMD Lexxing & Reg Ex Grep & so on runs lots faster than recursive-descent
+
+b ) Practically nobody codes it that way
+
+I wonder why not
+
+**6/** @sasuke___420
+
+@pelavarre @kellabyte @nicbarkeragain
+
+it's an esoteric hobby. and solving general-ish problems like for regex engines instead of hand-coding specific parsers is like a lifetime of work. the guy whose lifetime of work it is, geoff langdale, is on here if you want to ask him
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,43 @@
+---
+title: "@nicbarkeragain have you looked at Odin’s SIMD?"
+type: archive
+source: twitter
+source_url: "https://x.com/slendidev/status/1944952001516716329"
+author: "Slendi"
+handle: slendidev
+post_id: "1944952001516716329"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain have you looked at Odin’s SIMD?"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/slendidev/status/1944952001516716329
+- Author: Slendi (@slendidev)
+- Posted: 2025-07-15 02:47:47
+
+## Branch
+
+**1/** @slendidev
+
+@nicbarkeragain
+
+have you looked at Odin’s SIMD?
+
+**2/** @nicbarkeragain
+
+@xslendix
+
+Yeah I really like the simplicity of the array programming in Odin, and there are a number of other languages that provide similar things. It could just be because my mindset is narrow, but that type of approach feels a bit specific and niche, I use it mostly for 3d math.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Whats wrong with using inline ASM?"
+type: archive
+source: twitter
+source_url: "https://x.com/sorek_UK/status/1944943796006838648"
+author: "Maciej 'sorek' Łoboda"
+handle: sorek_UK
+post_id: "1944943796006838648"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Whats wrong with using inline ASM?"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/sorek_UK/status/1944943796006838648
+- Author: Maciej 'sorek' Łoboda (@sorek_UK)
+- Posted: 2025-07-15 02:15:10
+
+## Branch
+
+**1/** @sorek_UK
+
+@nicbarkeragain
+
+Whats wrong with using inline ASM?
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Just enable arithmetic operations on fixed sized arrays and optimise this to use SIMD where possible"
+type: archive
+source: twitter
+source_url: "https://x.com/splizard/status/1945085884254105812"
+author: "Quentin Quaadgras"
+handle: splizard
+post_id: "1945085884254105812"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Just enable arithmetic operations on fixed sized arrays and optimise this to use SIMD where possible"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/splizard/status/1945085884254105812
+- Author: Quentin Quaadgras (@splizard)
+- Posted: 2025-07-15 11:39:47
+
+## Branch
+
+**1/** @splizard
+
+@nicbarkeragain
+
+Just enable arithmetic operations on fixed sized arrays and optimise this to use SIMD where possible
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Have you met the average programmer?"
+type: archive
+source: twitter
+source_url: "https://x.com/thejoephase/status/1945065810763899008"
+author: "the joe phase"
+handle: thejoephase
+post_id: "1945065810763899008"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Have you met the average programmer?"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/thejoephase/status/1945065810763899008
+- Author: the joe phase (@thejoephase)
+- Posted: 2025-07-15 10:20:01
+
+## Branch
+
+**1/** @thejoephase
+
+@nicbarkeragain
+
+Have you met the average programmer?
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain"
+type: archive
+source: twitter
+source_url: "https://x.com/theogcb405/status/1945247807578759590"
+author: "kit 🇱🇮"
+handle: theogcb405
+post_id: "1945247807578759590"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/theogcb405/status/1945247807578759590
+- Author: kit 🇱🇮 (@theogcb405)
+- Posted: 2025-07-15 22:23:12
+
+## Branch
+
+**1/** @theogcb405
+
+@nicbarkeragain
+
+![](https://pbs.twimg.com/media/Gv7okrWWUAAPlGF?format=jpg&name=orig)
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain What popped into my head is languages like Uiua"
+type: archive
+source: twitter
+source_url: "https://x.com/2BitSalute/status/1945487552011722857"
+author: "2BitSalute"
+handle: 2BitSalute
+post_id: "1945487552011722857"
+date: 2025-07-16
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain What popped into my head is languages like Uiua"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/2BitSalute/status/1945487552011722857
+- Author: 2BitSalute (@2BitSalute)
+- Posted: 2025-07-16 14:15:52
+
+## Branch
+
+**1/** @2BitSalute
+
+@nicbarkeragain
+
+What popped into my head is languages like Uiua
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,39 @@
+---
+title: "@nicbarkeragain But is it really that hard to use though? Or what scenario do you mean? If we are to make use of SIMD we will be doing operations on collections of numbers."
+type: archive
+source: twitter
+source_url: "https://x.com/farre_x/status/1945597104895778866"
+author: "Simon"
+handle: farre_x
+post_id: "1945597104895778866"
+date: 2025-07-16
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain But is it really that hard to use though? Or what scenario do you mean? If we are to make use of SIMD we will be doing operations on collections of numbers."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/farre_x/status/1945597104895778866
+- Author: Simon (@farre_x)
+- Posted: 2025-07-16 21:31:11
+
+## Branch
+
+**1/** @farre_x
+
+@nicbarkeragain
+
+But is it really that hard to use though? Or what scenario do you mean? If we are to make use of SIMD we will be doing operations on collections of numbers. That is fairly trivial with intrinsics
+
+Isn't the hard part figuring out if SIMD even makes sense when solving problem foo?
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,71 @@
+---
+title: "@nicbarkeragain Hmmn."
+type: archive
+source: twitter
+source_url: "https://x.com/geofflangdale/status/1945401263774162962"
+author: "Geoff Langdale"
+handle: geofflangdale
+post_id: "1945401263774162962"
+date: 2025-07-16
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Hmmn."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/geofflangdale/status/1945401263774162962
+- Author: Geoff Langdale (@geofflangdale)
+- Posted: 2025-07-16 08:32:59
+
+## Branch
+
+**1/** @geofflangdale
+
+@nicbarkeragain
+
+Hmmn. I feel most of the PL design community - who have many folks I have the greatest admiration for, to be clear - is really fundamentally not that interested in the reality of the actual instructions on the machine.
+
+If the choice is to have SIMD ...
+
+**2/** @geofflangdale
+
+@nicbarkeragain
+
+programming smothered in a bureaucratic pile of type theory or some intrusive, hard-to-understand "magical autovec" system, my response to their help is, like the proverbial French merchants, "Laissez-nous faire".
+
+What I'd really like to see - a small but modest step - is a ...
+
+**3/** @geofflangdale
+
+@nicbarkeragain
+
+run-time that pads out the start and end of every region (stack, globals, heap)  - *not* every data item - with 63 bytes of padding so that any SIMD access can be done without the weird risk that we're right next to a unmapped page or some device with weird semantics ("on ...
+
+**4/** @joseph_h_garvin
+
+@geofflangdale @nicbarkeragain
+
+I'm an aspiring PL+perf person that does care but I seem to be rare 😅 I think the situation is worse than this though...
+
+**5/** @joseph_h_garvin
+
+@geofflangdale @nicbarkeragain
+
+Even if your allocator/linker/etc guarantee there's always an extra 63 zero bytes the C(++)/Rust lang semantics still allow the optimizer to launch missiles when you go out of bounds. That's UB, so now it can do whatever it wants instead. So you need lang semantics to change too.
+
+**6/** @geofflangdale
+
+@joseph_h_garvin @nicbarkeragain
+
+Yes, I'd include this with the run-time changes. Personally, I think a good deal of UB (not all, of course) feels somewhat fatuous, but I don't really study the area closely enough to understand how much we could do without.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,40 @@
+---
+title: "@nicbarkeragain I’ve been wondering about this for over twenty years, but never found anything I felt was truly accessible."
+type: archive
+source: twitter
+source_url: "https://x.com/rtharston/status/1945604998022533294"
+author: "Reed Harston"
+handle: rtharston
+post_id: "1945604998022533294"
+date: 2025-07-16
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain I’ve been wondering about this for over twenty years, but never found anything I felt was truly accessible."
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/rtharston/status/1945604998022533294
+- Author: Reed Harston (@rtharston)
+- Posted: 2025-07-16 22:02:33
+
+## Branch
+
+**1/** @rtharston
+
+@nicbarkeragain
+
+I’ve been wondering about this for over twenty years, but never found anything I felt was truly accessible.
+If I used SIMD more I’d probably try to come up with something to help.
+But as long as it’s a pain to use I often just don’t.
+It’s almost a chicken and egg problem.
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,37 @@
+---
+title: "@nicbarkeragain Honestly it will definitely be a random idea someone has at 3 am and everyone will just go \"how did we not think about that earlier?\" and then a revolution begins"
+type: archive
+source: twitter
+source_url: "https://x.com/themovercell/status/1945386815537459355"
+author: "The Mover Cell🇷🇺"
+handle: themovercell
+post_id: "1945386815537459355"
+date: 2025-07-16
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "@nicbarkeragain Honestly it will definitely be a random idea someone has at 3 am and everyone will just go \"how did we not think about that earlier?\" and then a revolution begins"
+in_reply_to: ""
+parent_post_id: "1944913662499516737"
+---
+
+## Source
+
+- URL: https://x.com/themovercell/status/1945386815537459355
+- Author: The Mover Cell🇷🇺 (@themovercell)
+- Posted: 2025-07-16 07:35:35
+
+## Branch
+
+**1/** @themovercell
+
+@nicbarkeragain
+
+Honestly it will definitely be a random idea someone has at 3 am and everyone will just go "how did we not think about that earlier?" and then a revolution begins
+
+## Related
+
+- Spine: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
@@ -0,0 +1,42 @@
+---
+title: "I can't help but have this feeling that we're missing something in programming language design that will make SIMD much easier for the average programmer to use."
+type: archive
+source: twitter
+source_url: "https://x.com/nicbarkeragain/status/1944913662499516737"
+author: "Nic Barker"
+handle: nicbarkeragain
+post_id: "1944913662499516737"
+date: 2025-07-15
+archived: 2026-08-25
+draft: false
+tags:
+  - archive
+  - twitter
+  - nicbarkeragain
+description: "I can't help but have this feeling that we're missing something in programming language design that will make SIMD much easier for the average programmer to use."
+in_reply_to: ""
+---
+
+## Source
+
+- URL: https://x.com/nicbarkeragain/status/1944913662499516737
+- Author: Nic Barker (@nicbarkeragain)
+- Posted: 2025-07-15 00:15:26
+
+## Thread
+
+**1/** @nicbarkeragain
+
+I can't help but have this feeling that we're missing something in programming language design that will make SIMD much easier for the average programmer to use. And that it will likely seem simple in hindsight.
+
+Branches: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-another_sentry-https-github-com-higherorderco-bend]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-sasuke___420-have-you-spent-time-writing-simd-algorithms-for]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-etscrivner-i-think-ispc-more-or-less-solved-this-and-in-a-c]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-quentinoeoeoe-i-agree-there-is-something-weird-with-simd]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-WorldGhostC2C-llm-baked-into-gcc-whos-only-job-is-making-things]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-TheBuzzSaw-c-makes-it-fairly-straightforward]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-jshurmer-i-agree]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-agmethod75-its-called-array-programming-fortran-has-been]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-Ed__dev-my-gamedev-head-just-wants-a-sort-of-game]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-etzikun-might-be-an-oversimplification-but-i-think-the]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-VPCOMPRESSB-intels-ispc]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-andrew_v10209-maybe-futhark-or-array-programming-theres-an]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-JebsSteve0x1-ffmpeg-thinks-otherwise]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-examachine-possible]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-sorek_UK-whats-wrong-with-using-inline-asm]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-AgileJebrim-the-answer-exists-its-called-ispc]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-slendidev-have-you-looked-at-odins-simd]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-meglio-the-average-programmer-you-mentioned-uses-oop-and]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-JakeLegendXIII-i-have-hope-that-this-is-true-but-damn-have-we]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-groby-so-fortran]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-dezren39-more-arrays]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-DeepknowledgeU-mojo-lang-has-simd-aware-data-types-and-makes-it]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-DanielJCollier-ispc]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-BigWhoopGames-reduced-instruction-sets-if-we-have-op-a-and-op-b]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-thejoephase-have-you-met-the-average-programmer]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-splizard-just-enable-arithmetic-operations-on-fixed-sized]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-dodobyte-openmp-does-good-enough-job-with-compiler-flags]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-salonen-i-think-we-need-a-system-visualizer-where-devs]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-SumErgoMonstro-have-you-ever-studied-apl-it-has-the-best-array]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-_Felipe-relational-model-columnar-data-sql-lands-itself]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-Zino2201_-perhaps-something-like-ispc-how-we-author-shaders]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-asdhuqiwuhdais-it-hasnt-been-stabilized-yet-but-rusts-portable]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-meoyawn-mojo-has-a-first-class-simd-support-afaik-but-ive]], [[archive/threLine truncated
+
+**2/** @nicbarkeragain
+
+It's easy to think of structured programming (i.e. text based, with conditionals like if / else, loops, local variables, etc) as a simple and obvious innovation over the top of assembly, but I suspect if you were programming in 1950 it might not have seemed so obvious.
+
+**3/** @nicbarkeragain
+
+Ironically it might take settling on a standard width among hardware vendors for us to be able to make reasonable headway there. I have no idea whether there are physical limitations to how wide we could go, or whether this would constitute a "pause" on hardware progress...
+
+Branches: [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-jon_roelofs-the-industry-is-trying-to-do-that-with-vector]], [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were/2025-07-15-TGodfather1984-128-bit-seems-to-be-pretty-widespread-so-it-might]]
@@ -9,5 +9,6 @@ tags:
  - nicbarkeragain
 ---

+- [[archive/threads/nicbarkeragain/2025-07-15-i-cant-help-but-have-this-feeling-that-were]]
 - [[archive/threads/nicbarkeragain/2026-08-03-imo-basically-every-object-creation-api-you-have]]
 - [[archive/threads/nicbarkeragain/2026-08-05-an-example-of-this-a-classic-way-to-construct-a]]