You are viewing a plain text version of this content. The canonical link for it is here.
Posted to discuss-archive@tvm.apache.org by Pratik Fegade via Apache TVM Discuss <no...@discuss.tvm.ai> on 2020/09/07 23:36:49 UTC
[Apache TVM Discuss] [Application] LLVM target triple for Graviton2
CPUs
I am using TVM on the AWS ARM machines and am wondering the correct target triple is for them (I have been using -mcpu=cortex-a76 -mattr=+neon). Specifically the Graviton CPUs use the 64-bit ARM Neoverse N1 cores. Thanks!
---
[Visit Topic](https://discuss.tvm.apache.org/t/llvm-target-triple-for-graviton2-cpus/7829/1) to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/8829aa934eb56f0490d590b8849b91dd491ef7a153c9743a6b5746d3ac5be319).
[Apache TVM Discuss] [Application] LLVM target triple for Graviton2
CPUs
Posted by Giuseppe Rossini via Apache TVM Discuss <no...@discuss.tvm.ai>.
Hello @pratikfegade,
This is what I always use (which also enables the dot-product support. If you don't want it, just remove the `+dotprod`):
`"llvm -device=arm_cpu -mtriple=aarch64-linux-gnu -mattr=+neon,+v8.2a,+dotprod"`
A good source for me was [this tutorial](https://tvm.apache.org/docs/tutorials/get_started/cross_compilation_and_rpc.html#declare-and-cross-compile-kernel-on-local-machine).
Hope this helps
---
[Visit Topic](https://discuss.tvm.apache.org/t/llvm-target-triple-for-graviton2-cpus/7829/4) to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/1715900d72e3b1fa6d28231c2c8f0f3c1008e308dc846d71090655670e2ce5f0).
[Apache TVM Discuss] [Application] LLVM target triple for Graviton2
CPUs
Posted by "Cody H. Yu via Apache TVM Discuss" <no...@discuss.tvm.ai>.
This is the one we used on EC2 m6g:
`llvm -device=arm_cpu -mtriple=aarch64-linux-gnu -mattr=+v8.2a,+fullfp16,+fp-armv8,+dotprod,+crc,+crypto,+neon`
---
[Visit Topic](https://discuss.tvm.apache.org/t/llvm-target-triple-for-graviton2-cpus/7829/5) to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/878b3e26ac3f51f480494581608d5b32b47ce2f3f96e443b943cd946a68348c5).
[Apache TVM Discuss] [Application] LLVM target triple for Graviton2
CPUs
Posted by Junru Shao via Apache TVM Discuss <no...@discuss.tvm.ai>.
Hmmm it should just follow the LLVM’s convention
---
[Visit Topic](https://discuss.tvm.apache.org/t/llvm-target-triple-for-graviton2-cpus/7829/2) to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/83fcb00869e7da6dbfa51f444a5ff74247c779600cdd95292b3dc9a1ddaaca17).
[Apache TVM Discuss] [Application] LLVM target triple for Graviton2
CPUs
Posted by Ramana Radhakrishnan via Apache TVM Discuss <no...@discuss.tvm.ai>.
-mcpu=neoverse-n1 should be fine depending on the version of llvm . I can't remember if you need anything in the -mattr to specifically allow for the dot product or fp16 .
@giuseros has probably some experience here
Ramana
---
[Visit Topic](https://discuss.tvm.apache.org/t/llvm-target-triple-for-graviton2-cpus/7829/3) to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/903623437dbde6fdf83c3614fd32579f76e8c5c41ae2a49641a22937840c911b).
[Apache TVM Discuss] [Application] LLVM target triple for Graviton2
CPUs
Posted by Pratik Fegade via Apache TVM Discuss <no...@discuss.tvm.ai>.
Thanks all for the suggestions. I'll try these.
---
[Visit Topic](https://discuss.tvm.apache.org/t/llvm-target-triple-for-graviton2-cpus/7829/6) to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/c00387647737dd14f893b4d18b69d576effa685e3423b830ab7be53893f3bdaa).