# RFC **URL:** https://discuss.tvm.apache.org/c/development/rfc/8.md?page=2 [Latest](https://discuss.tvm.apache.org/latest.md) · [Categories](https://discuss.tvm.apache.org/categories.md) **Page:** 3 --- ## [Integration of muRISCV-NN kernel library in TVM](https://discuss.tvm.apache.org/t/integration-of-muriscv-nn-kernel-library-in-tvm/13124)
**Author:** [@PhilippvK](https://discuss.tvm.apache.org/u/PhilippvK)\ **Replies:** 3\ **Last updated:** [July 13, 2022, 10:01am UTC](https://discuss.tvm.apache.org/t/integration-of-muriscv-nn-kernel-library-in-tvm/13124 "2022-07-13T10:01:08Z")
Our team at the TU Munich (@r.stahl @fabian) has recently open-sourced our work on porting the ARM CMSIS-NN library to RISC-V targets: https://github.com/tum-ei-eda/muriscv-nn Let me shortly summarize the features: Su… --- ## [\[RFC\]\[BYOC\] An extended graph partitioning flow](https://discuss.tvm.apache.org/t/rfc-byoc-an-extended-graph-partitioning-flow/6028)
**Author:** [@mbaret](https://discuss.tvm.apache.org/u/mbaret)\ **Replies:** 4\ **Last updated:** [July 7, 2022, 10:13am UTC](https://discuss.tvm.apache.org/t/rfc-byoc-an-extended-graph-partitioning-flow/6028 "2022-07-07T10:13:30Z")
This RFC is built upon on the following prior RFCs : A1) https://discuss.tvm.ai/t/relay-improved-graph-partitioning-algorithm/5830 A2) https://discuss.tvm.ai/t/discuss-annotation-defined-subgraphs/5934/8 Myself and @… --- ## [Modularizing LLVM codegen/JIT](https://discuss.tvm.apache.org/t/modularizing-llvm-codegen-jit/12764)
**Author:** [@kparzysz](https://discuss.tvm.apache.org/u/kparzysz)\ **Replies:** 40\ **Last updated:** [June 28, 2022, 6:39pm UTC](https://discuss.tvm.apache.org/t/modularizing-llvm-codegen-jit/12764 "2022-06-28T18:39:59Z")
First of all, I’m not sure what path this discussion should take, i.e. whether RFC needs to follow or not. This post is to present the idea and get some initial feedback. Problem LLVM maintains global state, and that … --- ## [Asynchrnous semantics for TIR, and async pipeline](https://discuss.tvm.apache.org/t/asynchrnous-semantics-for-tir-and-async-pipeline/13011)
**Author:** [@masahi](https://discuss.tvm.apache.org/u/masahi)\ **Replies:** 0\ **Last updated:** [June 24, 2022, 2:35am UTC](https://discuss.tvm.apache.org/t/asynchrnous-semantics-for-tir-and-async-pipeline/13011 "2022-06-24T02:35:19Z")
Hi, I posted an RFC “Asynchronous stage in software pipeline” below: Currently, the scope of asynchrony is limited to the TIR software pipeline, but the discussion has been moving toward coming up with more “general a… --- ## [\[RFC\]\[BYOC\]NVIDIA CUTLASS Integration](https://discuss.tvm.apache.org/t/rfc-byoc-nvidia-cutlass-integration/9147)
**Author:** [@Laurawly](https://discuss.tvm.apache.org/u/Laurawly)\ **Replies:** 30\ **Last updated:** [March 25, 2022, 6:31am UTC](https://discuss.tvm.apache.org/t/rfc-byoc-nvidia-cutlass-integration/9147 "2022-03-25T06:31:43Z")
Motivation: Currently, the GEMM schedules searched by TVM auto scheduler on NVIDIA GPUs have some big performance gaps compared with NVIDIA CUTLASS library (benchmark table shown below). For each new shape, TVM needs to … --- ## [\[pre-RFC\] Polybench-TVM](https://discuss.tvm.apache.org/t/pre-rfc-polybench-tvm/12995)
**Author:** [@Wheest](https://discuss.tvm.apache.org/u/Wheest)\ **Replies:** 0\ **Last updated:** [June 21, 2022, 2:59pm UTC](https://discuss.tvm.apache.org/t/pre-rfc-polybench-tvm/12995 "2022-06-21T14:59:48Z")
My idea looks at adding official TVM tensor expression implementations of kernels from the polybench benchmarking system. It’s possible that some kernels do not make sense within the TVM programming model, however I sti… --- ## [\[RFC\] TVMScript Metaprogramming](https://discuss.tvm.apache.org/t/rfc-tvmscript-metaprogramming/12969)
**Author:** [@yelite](https://discuss.tvm.apache.org/u/yelite)\ **Replies:** 4\ **Last updated:** [June 18, 2022, 1:43pm UTC](https://discuss.tvm.apache.org/t/rfc-tvmscript-metaprogramming/12969 "2022-06-18T13:43:06Z")
Hi all, we sent out an RFC for TVMScript Metaprogramming at https://github.com/apache/tvm-rfcs/pull/79. Below is a section quoted from the RFC for better visibility, describing what metaprogramming features we want to su… --- ## [\[pre-RFC\] Name mangling in IRModules](https://discuss.tvm.apache.org/t/pre-rfc-name-mangling-in-irmodules/12944)
**Author:** [@areusch](https://discuss.tvm.apache.org/u/areusch)\ **Replies:** 14\ **Last updated:** [June 15, 2022, 2:40pm UTC](https://discuss.tvm.apache.org/t/pre-rfc-name-mangling-in-irmodules/12944 "2022-06-15T14:40:36Z")
Hi all, In \[mini-RFC\] Name mangling in AOT, we discussed ways to accommodate the runtime::Module tree in an embedded environment, where we prefer to call functions directly in the generated C using their symbol name rat… --- ## [\[RFC\]\[Relay\]\[Topi\] Hashtable support](https://discuss.tvm.apache.org/t/rfc-relay-topi-hashtable-support/5842)
**Author:** [@lfengad](https://discuss.tvm.apache.org/u/lfengad)\ **Replies:** 10\ **Last updated:** [June 8, 2022, 1:26am UTC](https://discuss.tvm.apache.org/t/rfc-relay-topi-hashtable-support/5842 "2022-06-08T01:26:58Z")
As in this PR, we propose to add HashTable support into TVM. We observe a great amount of TensorFlow NLP models used in our production environment, and we would like to improve model serving speed using TVM. Among these… --- ## [\[RFC\] TUNIP: TVMScript Unified Printer](https://discuss.tvm.apache.org/t/rfc-tunip-tvmscript-unified-printer/12841)
**Author:** [@junrushao](https://discuss.tvm.apache.org/u/junrushao)\ **Replies:** 0\ **Last updated:** [May 25, 2022, 10:56pm UTC](https://discuss.tvm.apache.org/t/rfc-tunip-tvmscript-unified-printer/12841 "2022-05-25T22:56:27Z")
--- ## [Export TIR to json](https://discuss.tvm.apache.org/t/export-tir-to-json/12329)
**Author:** [@SebastianBoblestETAS](https://discuss.tvm.apache.org/u/SebastianBoblestETAS)\ **Replies:** 14\ **Last updated:** [May 16, 2022, 8:56am UTC](https://discuss.tvm.apache.org/t/export-tir-to-json/12329 "2022-05-16T08:56:38Z")
We propose to implement an API to TVM that allows users to export the TIR representation of the generated kernels as json. Motivation Json is a standard file format and can easily be processed in python, for example. A… --- ## [\[RFC\] Python Dependencies in TVM CI Containers](https://discuss.tvm.apache.org/t/rfc-python-dependencies-in-tvm-ci-containers/9011)
**Author:** [@areusch](https://discuss.tvm.apache.org/u/areusch)\ **Replies:** 15\ **Last updated:** [May 13, 2022, 3:48pm UTC](https://discuss.tvm.apache.org/t/rfc-python-dependencies-in-tvm-ci-containers/9011 "2022-05-13T15:48:48Z")
Background This RFC is a follow-on to \[RFC\] Consolidating Python Dependencies, implemented in PR 7289. It will be followed with a third RFC describing how we might leverage changes introduced this RFC to produce version-… --- ## [\[RFC\] Target Pre-processing](https://discuss.tvm.apache.org/t/rfc-target-pre-processing/12739)
**Author:** [@Mousius](https://discuss.tvm.apache.org/u/Mousius)\ **Replies:** 1\ **Last updated:** [May 11, 2022, 12:12pm UTC](https://discuss.tvm.apache.org/t/rfc-target-pre-processing/12739 "2022-05-11T12:12:48Z")
Posting here to raise awareness of a new RFC proposed against tvm-rfcs: --- ## [\[pre-RFC\]\[CRT\]\[microTVM\] Integration of platform-specific counters (or any data)](https://discuss.tvm.apache.org/t/pre-rfc-crt-microtvm-integration-of-platform-specific-counters-or-any-data/12724)
**Author:** [@fPecc](https://discuss.tvm.apache.org/u/fPecc)\ **Replies:** 2\ **Last updated:** [May 10, 2022, 10:54am UTC](https://discuss.tvm.apache.org/t/pre-rfc-crt-microtvm-integration-of-platform-specific-counters-or-any-data/12724 "2022-05-10T10:54:23Z")
Introduction I have stumbled across a use case which could be an interesting addition to the CRT runtime. Consider the following scenario: A host is running AutoTVM, executing the generated code in another platform th… --- ## [\[pre-RFC\] Compilation Configuration Representation](https://discuss.tvm.apache.org/t/pre-rfc-compilation-configuration-representation/11372)
**Author:** [@Mousius](https://discuss.tvm.apache.org/u/Mousius)\ **Replies:** 68\ **Last updated:** [April 28, 2022, 2:52pm UTC](https://discuss.tvm.apache.org/t/pre-rfc-compilation-configuration-representation/11372 "2022-04-28T14:52:09Z")
Inspired by the work of @mbs-octoml, I give you a new RFC for CompilationConfig! Summary This RFC supersedes Migrating IRModules to Attributes by replacing the various attributes associated with the Runtime and Executor … --- ## [\[Pre-RFC\] More meaningful names in rpc module in Python](https://discuss.tvm.apache.org/t/pre-rfc-more-meaningful-names-in-rpc-module-in-python/12618)
**Author:** [@pfk123](https://discuss.tvm.apache.org/u/pfk123)\ **Replies:** 3\ **Last updated:** [April 26, 2022, 10:24am UTC](https://discuss.tvm.apache.org/t/pre-rfc-more-meaningful-names-in-rpc-module-in-python/12618 "2022-04-26T10:24:10Z")
Hello, I’m reading code of RPC module in Python, and many elements are hard to understand for me. I’m not a big expert, but I think some of them should be more meaningfull, so I suggest to change their names. We can dis… --- ## [Relax: Co-Designing High-Level Abstraction Towards TVM Unity](https://discuss.tvm.apache.org/t/relax-co-designing-high-level-abstraction-towards-tvm-unity/12496)
**Author:** [@ziheng](https://discuss.tvm.apache.org/u/ziheng)\ **Replies:** 3\ **Last updated:** [April 20, 2022, 4:28pm UTC](https://discuss.tvm.apache.org/t/relax-co-designing-high-level-abstraction-towards-tvm-unity/12496 "2022-04-20T16:28:10Z")
Co-Authors: Ziheng Jiang (@ZihengJiang), Yuchen Jin (@YuchenJin), Denise Kutnick (@denise-k) We wrote this post on behalf of many(names/ids in this post are listed alphabetically) who are participating in this effort. … --- ## [\[RFC\] Generic Logger names in Python](https://discuss.tvm.apache.org/t/rfc-generic-logger-names-in-python/12611)
**Author:** [@pfk123](https://discuss.tvm.apache.org/u/pfk123)\ **Replies:** 9\ **Last updated:** [April 21, 2022, 8:15pm UTC](https://discuss.tvm.apache.org/t/rfc-generic-logger-names-in-python/12611 "2022-04-21T20:15:30Z")
Hello, my first RFC, in life, so be patient. When I run command find python -type f -name "\*.py" | xargs grep getLogger I can see, that names of Loggers are manually provided, e.g. Ethos-U, RPCTracker or auto\_scheduler.… --- ## [\[pre-RFC\]\[ BYOC\] RISC-V CSI-NN2 Compute Library integration](https://discuss.tvm.apache.org/t/pre-rfc-byoc-risc-v-csi-nn2-compute-library-integration/12610)
**Author:** [@alter-xp](https://discuss.tvm.apache.org/u/alter-xp)\ **Replies:** 0\ **Last updated:** [April 20, 2022, 7:27am UTC](https://discuss.tvm.apache.org/t/pre-rfc-byoc-risc-v-csi-nn2-compute-library-integration/12610 "2022-04-20T07:27:54Z")
Summary Introduce CSI-NN2 Compute Library into TVM to accelerate the inference performance of RISC-V CPU with Vector Extension. Motivation Recently, in the latest Tiny v0.7 list released by AI benchmark MLPerf. Alibab… --- ## [\[pre-RFC\] New CI container: ci\_cpu\_asserts](https://discuss.tvm.apache.org/t/pre-rfc-new-ci-container-ci-cpu-asserts/12536)
**Author:** [@areusch](https://discuss.tvm.apache.org/u/areusch)\ **Replies:** 10\ **Last updated:** [April 13, 2022, 8:17pm UTC](https://discuss.tvm.apache.org/t/pre-rfc-new-ci-container-ci-cpu-asserts/12536 "2022-04-13T20:17:07Z")
Hi all, I wanted to discuss the idea of adding a new CI container: ci\_cpu\_asserts. This container would behave just like ci\_cpu, except that both TVM and various libraries within the container would be compiled with add… --- ## [\[RFC\] A proposed update to the Docker images ci\_\* tag pattern](https://discuss.tvm.apache.org/t/rfc-a-proposed-update-to-the-docker-images-ci-tag-pattern/12034)
**Author:** [@leandron](https://discuss.tvm.apache.org/u/leandron)\ **Replies:** 6\ **Last updated:** [April 8, 2022, 2:15pm UTC](https://discuss.tvm.apache.org/t/rfc-a-proposed-update-to-the-docker-images-ci-tag-pattern/12034 "2022-04-08T14:15:21Z")
For years we’ve been using this patern on our Docker images, which is composed by v0.INCREMENTAL\_NUMBER, e.g. tlcpack/ci\_cpu:v0.80, and when we update any image we bump the minor number, just as a way to release a new ve… --- ## [\[RFC\]\[Relay\] FP32 -\> FP16 Model Support](https://discuss.tvm.apache.org/t/rfc-relay-fp32-fp16-model-support/9994)
**Author:** [@AndrewZhaoLuo](https://discuss.tvm.apache.org/u/AndrewZhaoLuo)\ **Replies:** 17\ **Last updated:** [April 7, 2022, 6:23am UTC](https://discuss.tvm.apache.org/t/rfc-relay-fp32-fp16-model-support/9994 "2022-04-07T06:23:12Z")
This RFC discusses post-training quantization of FP32 models to an FP16 form. We do not focus on specific codegen but rather the relay representation of the models. Doing this transformation is useful for reducing model… --- ## [Collage RFC ready for comments](https://discuss.tvm.apache.org/t/collage-rfc-ready-for-comments/12308)
**Author:** [@mbs-octoml](https://discuss.tvm.apache.org/u/mbs-octoml)\ **Replies:** 5\ **Last updated:** [April 6, 2022, 4:12pm UTC](https://discuss.tvm.apache.org/t/collage-rfc-ready-for-comments/12308 "2022-04-06T16:12:22Z")
Hi folks, I’ve just put up a PR describing our plans for ‘Collage’ here at OctoML: This work is derived from the preprint: Collage: Automated Integration of Deep Learning Backends Byungsoo Jeon, Sunghyun Park, Pei… --- ## [\[RFC\] A general task extraction mechanism for auto\_scheduler](https://discuss.tvm.apache.org/t/rfc-a-general-task-extraction-mechanism-for-auto-scheduler/8444)
**Author:** [@comaniac](https://discuss.tvm.apache.org/u/comaniac)\ **Replies:** 18\ **Last updated:** [April 1, 2022, 4:41pm UTC](https://discuss.tvm.apache.org/t/rfc-a-general-task-extraction-mechanism-for-auto-scheduler/8444 "2022-04-01T16:41:18Z")
The current approach used by auto\_scheduler to extract tuning tasks leverages Relay op strategy. In short, auto\_scheduler registers an implementation in Relay op strategy as AutoTVM, but instead of using the TOPI schedul… --- ## [\[RFC\] Allow merging via PR comments](https://discuss.tvm.apache.org/t/rfc-allow-merging-via-pr-comments/12220)
**Author:** [@driazati](https://discuss.tvm.apache.org/u/driazati)\ **Replies:** 8\ **Last updated:** [March 30, 2022, 6:33pm UTC](https://discuss.tvm.apache.org/t/rfc-allow-merging-via-pr-comments/12220 "2022-03-30T18:33:33Z")
Summary PRs authors must currently wait for changes to be manually merged by a committer clicking the “merge” button. A large amount of time can often elapse between CI going green, a PR being approved, and this merge bu… --- ## [\[RFC\] TensorIR: A schedulable IR for TVM](https://discuss.tvm.apache.org/t/rfc-tensorir-a-schedulable-ir-for-tvm/7872)
**Author:** [@Hzfengsy](https://discuss.tvm.apache.org/u/Hzfengsy)\ **Replies:** 98\ **Last updated:** [March 27, 2022, 1:41am UTC](https://discuss.tvm.apache.org/t/rfc-tensorir-a-schedulable-ir-for-tvm/7872 "2022-03-27T01:41:16Z")
Background and Motivation TVM is an end-to-end deep learning compiler with two levels of IR and optimization. TVM translates popular DL frameworks into Relay and optimizes the computation graph, after which it lowers eac… --- ## [BYOC: Add Verisilicon's NPU support](https://discuss.tvm.apache.org/t/byoc-add-verisilicons-npu-support/11097)
**Author:** [@sven](https://discuss.tvm.apache.org/u/sven)\ **Replies:** 9\ **Last updated:** [March 25, 2022, 1:33am UTC](https://discuss.tvm.apache.org/t/byoc-add-verisilicons-npu-support/11097 "2022-03-25T01:33:47Z")
Summary Support Versilicon’s NPU with BYOC framework. Bring the TVM ecosystem to my customers. Motivation Verislicon’s NPU applied on the edge broadly with many SoC vendors. Meanwhile, more and more customers want to bu… --- ## [\[RFC\] tlcpack: Thirdparty Binary Packages](https://discuss.tvm.apache.org/t/rfc-tlcpack-thirdparty-binary-packages/7903)
**Author:** [@haichen](https://discuss.tvm.apache.org/u/haichen)\ **Replies:** 24\ **Last updated:** [March 23, 2022, 4:19pm UTC](https://discuss.tvm.apache.org/t/rfc-tlcpack-thirdparty-binary-packages/7903 "2022-03-23T16:19:17Z")
So far we only released source code in past tvm releases, and we will continue to do so. As we continue to develop tvm, we also see some demand for convenient binary packages, such as wheel or docker binaries. One impor… --- ## [\[RFC\] Verification and (possibly) program synthesis of expression rewriting rules](https://discuss.tvm.apache.org/t/rfc-verification-and-possibly-program-synthesis-of-expression-rewriting-rules/12319)
**Author:** [@yzh119](https://discuss.tvm.apache.org/u/yzh119)\ **Replies:** 1\ **Last updated:** [March 16, 2022, 12:56am UTC](https://discuss.tvm.apache.org/t/rfc-verification-and-possibly-program-synthesis-of-expression-rewriting-rules/12319 "2022-03-16T00:56:47Z")
Currently, TVM uses the following way to define expression simplifying rules: This approach is error-prone and not scalable: The rewrite rules were added manually, the number of possible rewriting rules might be too… --- ## [\[RFC\] Rebuild Docker images per commit](https://discuss.tvm.apache.org/t/rfc-rebuild-docker-images-per-commit/12047)
**Author:** [@driazati](https://discuss.tvm.apache.org/u/driazati)\ **Replies:** 6\ **Last updated:** [March 11, 2022, 6:17pm UTC](https://discuss.tvm.apache.org/t/rfc-rebuild-docker-images-per-commit/12047 "2022-03-11T18:17:47Z")
Summary Rebuild Docker images per-build rather than use the pinned Docker Hub images in the Jenkinsfile. Guide Note: This is a spin off discussion from https://discuss.tvm.apache.org/t/rfc-a-proposed-update-to-the-docke… [Previous page](https://discuss.tvm.apache.org/c/development/rfc/8.md?page=1) [Next page](https://discuss.tvm.apache.org/c/development/rfc/8.md?page=3)