The news that matters, summarized in Telegram .

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA on Apple Silicon

• Open-source compiler, runtime, and compatibility libraries that run CUDA code on Apple's Metal GPU.

• No NVIDIA hardware needed: CUDA C++ or PTX goes in, Metal runs on the GPU.

• Experimental: unsupported paths fail explicitly rather than silently computing wrong results.

⚙️How it works

• Three entry paths: source recompilation (primary), PTX compatibility, and an opt-in libcuda.dylib binary shim.

• Fixed SIMD width of 32, mapping CUDA warp semantics onto Metal SIMD-group operations.

A registered test is not a passing test. A skip is not evidence of compatibility.

🔗 github.com⭐ 157 · last push 4h ago

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA on Apple Silicon

• Open-source compiler, runtime, and compatibility libraries that run CUDA code on Apple's Metal GPU.

• No NVIDIA hardware needed: CUDA C++ or PTX goes in, Metal runs on the GPU.

• Experimental: unsupported paths fail explicitly rather than silently computing wrong results.

⚙️How it works

• Three entry paths: source recompilation (primary), PTX compatibility, and an opt-in libcuda.dylib binary shim.

• Fixed SIMD width of 32, mapping CUDA warp semantics onto Metal SIMD-group operations.

A registered test is not a passing test. A skip is not evidence of compatibility.

🔗 github.com⭐ 157 · last push 4h ago

in your language

Add es, Spanish or Español.Add a language to your reply. These three all ask for Spanish:

Language code

Adriangithub.com/Lulzx/cuda-metal

English name

Adriangithub.com/Lulzx/cuda-metal

@xtldrbot spanish

Spanish name

Adriangithub.com/Lulzx/cuda-metal

@xtldrbot español

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA en Apple Silicon

• Compilador, runtime y librerías de compatibilidad que ejecutan código CUDA en la GPU Metal de Apple.

• Sin hardware NVIDIA: entra CUDA C++ o PTX y Metal lo ejecuta en la GPU.

• Experimental: los caminos no soportados fallan explícitamente en vez de dar resultados erróneos.

⚙️Cómo funciona

• Tres vías de entrada: recompilación de fuente (principal), compatibilidad PTX y un alias opcional libcuda.dylib.

• Anchura SIMD fija en 32, mapeando la semántica de warps CUDA a los SIMD-groups de Metal.

Un test registrado no es un test que pasa; un skip no es evidencia de compatibilidad.

🔗 github.com⭐ 157 · last push 4h ago

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA en Apple Silicon

• Compilador, runtime y librerías de compatibilidad que ejecutan código CUDA en la GPU Metal de Apple.

• Sin hardware NVIDIA: entra CUDA C++ o PTX y Metal lo ejecuta en la GPU.

• Experimental: los caminos no soportados fallan explícitamente en vez de dar resultados erróneos.

⚙️Cómo funciona

• Tres vías de entrada: recompilación de fuente (principal), compatibilidad PTX y un alias opcional libcuda.dylib.

• Anchura SIMD fija en 32, mapeando la semántica de warps CUDA a los SIMD-groups de Metal.

Un test registrado no es un test que pasa; un skip no es evidencia de compatibilidad.

🔗 github.com⭐ 157 · last push 4h ago

72languages

English
Español
Français
Deutsch
Italiano
Português
Nederlands
Русский
Polski
Українська
Čeština
Slovenčina
Română
Magyar
Ελληνικά
Svenska
Norsk
Dansk
Suomi
Íslenska
Türkçe
العربية
עברית
فارسی
हिन्दी
اردو
বাংলা
ਪੰਜਾਬੀ
தமிழ்
తెలుగు
मराठी
ગુજરાતી
ಕನ್ನಡ
മലയാളം
中文
粵語
日本語
한국어
Tiếng Việt
ไทย
Indonesia
Melayu
Filipino
Kiswahili
Afrikaans
Català
Euskara
Galego
Hrvatski
Српски
Slovenščina
Български
Lietuvių
Latviešu
Eesti
Shqip
Македонски
Azərbaycan
ქართული
Հայերեն
Қазақ тілі
O‘zbek
Монгол
नेपाली
සිංහල
ខ្មែរ
ລາວ
မြန်မာ
አማርኛ
Cymraeg
Gaeilge
Latina

summon it anywhere

get the context without the context switching

Reply to a link with @xtldrbot.

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA on Apple Silicon

• Open-source compiler, runtime, and compatibility libraries that run CUDA code on Apple's Metal GPU.

• No NVIDIA hardware needed: CUDA C++ or PTX goes in, Metal runs on the GPU.

• Experimental: unsupported paths fail explicitly rather than silently computing wrong results.

⚙️How it works

• Three entry paths: source recompilation (primary), PTX compatibility, and an opt-in libcuda.dylib binary shim.

• Fixed SIMD width of 32, mapping CUDA warp semantics onto Metal SIMD-group operations.

A registered test is not a passing test. A skip is not evidence of compatibility.

🔗 github.com⭐ 157 · last push 4h ago

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA on Apple Silicon

• Open-source compiler, runtime, and compatibility libraries that run CUDA code on Apple's Metal GPU.

• No NVIDIA hardware needed: CUDA C++ or PTX goes in, Metal runs on the GPU.

• Experimental: unsupported paths fail explicitly rather than silently computing wrong results.

⚙️How it works

• Three entry paths: source recompilation (primary), PTX compatibility, and an opt-in libcuda.dylib binary shim.

• Fixed SIMD width of 32, mapping CUDA warp semantics onto Metal SIMD-group operations.

A registered test is not a passing test. A skip is not evidence of compatibility.

🔗 github.com⭐ 157 · last push 4h ago

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA on Apple Silicon

• Open-source compiler, runtime, and compatibility libraries that run CUDA code on Apple's Metal GPU.

• No NVIDIA hardware needed: CUDA C++ or PTX goes in, Metal runs on the GPU.

• Experimental: unsupported paths fail explicitly rather than silently computing wrong results.

⚙️How it works

• Three entry paths: source recompilation (primary), PTX compatibility, and an opt-in libcuda.dylib binary shim.

• Fixed SIMD width of 32, mapping CUDA warp semantics onto Metal SIMD-group operations.

A registered test is not a passing test. A skip is not evidence of compatibility.

🔗 github.com⭐ 157 · last push 4h ago

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA on Apple Silicon

• Open-source compiler, runtime, and compatibility libraries that run CUDA code on Apple's Metal GPU.

• No NVIDIA hardware needed: CUDA C++ or PTX goes in, Metal runs on the GPU.

• Experimental: unsupported paths fail explicitly rather than silently computing wrong results.

⚙️How it works

• Three entry paths: source recompilation (primary), PTX compatibility, and an opt-in libcuda.dylib binary shim.

• Fixed SIMD width of 32, mapping CUDA warp semantics onto Metal SIMD-group operations.

A registered test is not a passing test. A skip is not evidence of compatibility.

🔗 github.com⭐ 157 · last push 4h ago

in your language

Add es, Spanish or Español.Add a language to your reply. These three all ask for Spanish:

Language code

Adriangithub.com/Lulzx/cuda-metal

English name

Adriangithub.com/Lulzx/cuda-metal

@xtldrbot spanish

Spanish name

Adriangithub.com/Lulzx/cuda-metal

@xtldrbot español

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA en Apple Silicon

• Compilador, runtime y librerías de compatibilidad que ejecutan código CUDA en la GPU Metal de Apple.

• Sin hardware NVIDIA: entra CUDA C++ o PTX y Metal lo ejecuta en la GPU.

• Experimental: los caminos no soportados fallan explícitamente en vez de dar resultados erróneos.

⚙️Cómo funciona

• Tres vías de entrada: recompilación de fuente (principal), compatibilidad PTX y un alias opcional libcuda.dylib.

• Anchura SIMD fija en 32, mapeando la semántica de warps CUDA a los SIMD-groups de Metal.

Un test registrado no es un test que pasa; un skip no es evidencia de compatibilidad.

🔗 github.com⭐ 157 · last push 4h ago

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA en Apple Silicon

• Compilador, runtime y librerías de compatibilidad que ejecutan código CUDA en la GPU Metal de Apple.

• Sin hardware NVIDIA: entra CUDA C++ o PTX y Metal lo ejecuta en la GPU.

• Experimental: los caminos no soportados fallan explícitamente en vez de dar resultados erróneos.

⚙️Cómo funciona

• Tres vías de entrada: recompilación de fuente (principal), compatibilidad PTX y un alias opcional libcuda.dylib.

• Anchura SIMD fija en 32, mapeando la semántica de warps CUDA a los SIMD-groups de Metal.

Un test registrado no es un test que pasa; un skip no es evidencia de compatibilidad.

🔗 github.com⭐ 157 · last push 4h ago

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal

via @xtldrbot

🍎CuMetal: CUDA en Apple Silicon

• Compilador, runtime y librerías de compatibilidad que ejecutan código CUDA en la GPU Metal de Apple.

• Sin hardware NVIDIA: entra CUDA C++ o PTX y Metal lo ejecuta en la GPU.

• Experimental: los caminos no soportados fallan explícitamente en vez de dar resultados erróneos.

⚙️Cómo funciona

• Tres vías de entrada: recompilación de fuente (principal), compatibilidad PTX y un alias opcional libcuda.dylib.

• Anchura SIMD fija en 32, mapeando la semántica de warps CUDA a los SIMD-groups de Metal.

Un test registrado no es un test que pasa; un skip no es evidencia de compatibilidad.

🔗 github.com⭐ 157 · last push 4h ago

you're going to like this one

Adriangithub.com/Lulzx/cuda-metal