# RFC **URL:** https://discuss.tvm.apache.org/c/development/rfc/8.md?page=1 [Latest](https://discuss.tvm.apache.org/latest.md) · [Categories](https://discuss.tvm.apache.org/categories.md) **Page:** 2 --- ## [\[RFC\]: Improve quantized convolution through mmla instruction](https://discuss.tvm.apache.org/t/rfc-improve-quantized-convolution-through-mmla-instruction/8336)
**Author:** [@giuseros](https://discuss.tvm.apache.org/u/giuseros)\ **Replies:** 2\ **Last updated:** [June 26, 2023, 7:58am UTC](https://discuss.tvm.apache.org/t/rfc-improve-quantized-convolution-through-mmla-instruction/8336 "2023-06-26T07:58:38Z")
Introduction and motivation This RFC is the third set of optimizations to enhance quantized convolution on Arm architectures. To give a brief summary: Basic Armv8-A convolution implementation (through gemm): \[RFC\] Impr… --- ## [Graph partitioning and Heterogeneous Execution](https://discuss.tvm.apache.org/t/graph-partitioning-and-heterogeneous-execution/504)
**Author:** [@zhiics](https://discuss.tvm.apache.org/u/zhiics)\ **Replies:** 35\ **Last updated:** [June 25, 2023, 10:46am UTC](https://discuss.tvm.apache.org/t/graph-partitioning-and-heterogeneous-execution/504 "2023-06-25T10:46:20Z")
Graph partitioning can be divided into two parts: partitioning, it partitions the graph into subgraphs that are suitable for different devices. There are two options for graph level optimizations after we obtain the p… --- ## [\[RFC\] MLIR Frontend](https://discuss.tvm.apache.org/t/rfc-mlir-frontend/6473)
**Author:** [@gmagogsfm](https://discuss.tvm.apache.org/u/gmagogsfm)\ **Replies:** 12\ **Last updated:** [May 31, 2023, 7:17am UTC](https://discuss.tvm.apache.org/t/rfc-mlir-frontend/6473 "2023-05-31T07:17:09Z")
RFC for Relay MLIR Frontend Authors: Yanan Cao, Yong Wu Contributors: Yida Wang, Haichen Shen, Yao Wang Summary We propose a solution that can give TVM/Relay top-notch model/op coverage for TensorFlow with affordable ef… --- ## [\[RFC\] SparseTIR as a new dialect in TVM](https://discuss.tvm.apache.org/t/rfc-sparsetir-as-a-new-dialect-in-tvm/14645)
**Author:** [@yzh119](https://discuss.tvm.apache.org/u/yzh119)\ **Replies:** 9\ **Last updated:** [April 19, 2023, 12:44am UTC](https://discuss.tvm.apache.org/t/rfc-sparsetir-as-a-new-dialect-in-tvm/14645 "2023-04-19T00:44:19Z")
Introduction Many Deep Learning workloads involve sparse/variable components, e.g. Mixture of Experts, Network Pruning, GNNs, and Sparse Conv. Currently, if users want to write these operators in TVM, they need to compos… --- ## [\[RFC\] Search-based Automated Quantization](https://discuss.tvm.apache.org/t/rfc-search-based-automated-quantization/5483)
**Author:** [@ziheng](https://discuss.tvm.apache.org/u/ziheng)\ **Replies:** 21\ **Last updated:** [April 14, 2023, 11:10am UTC](https://discuss.tvm.apache.org/t/rfc-search-based-automated-quantization/5483 "2023-04-14T11:10:05Z")
Search-based Automated Quantization Background One year before, I have implemented a quantization workflow in tvm: issue, pull. Brought the idea from some existing quantization frameworks, I choose to adopt the annotatio… --- ## [\[RFC\] UMA: Universal Modular Accelerator Interface](https://discuss.tvm.apache.org/t/rfc-uma-universal-modular-accelerator-interface/12039)
**Author:** [@MJKlaiber](https://discuss.tvm.apache.org/u/MJKlaiber)\ **Replies:** 39\ **Last updated:** [April 10, 2023, 3:50pm UTC](https://discuss.tvm.apache.org/t/rfc-uma-universal-modular-accelerator-interface/12039 "2023-04-10T15:50:30Z")
UMA: Universal Modular Accelerator Interface Feature Name: Universal Modular Accelerator Interface (UMA) Start Date: 2022 February Authors: Paul Palomero Bernardo @paulpb, Christoph Gerum @cgerum - University of Tübin… --- ## [TVMC - UMA Integration](https://discuss.tvm.apache.org/t/tvmc-uma-integration/14309)
**Author:** [@r.stahl](https://discuss.tvm.apache.org/u/r.stahl)\ **Replies:** 18\ **Last updated:** [March 1, 2023, 3:17pm UTC](https://discuss.tvm.apache.org/t/tvmc-uma-integration/14309 "2023-03-01T15:17:27Z")
Let’s discuss how UMA could be integrated into TVMC. I did a first brute force attempt to figure out the challenges. CLI A simple approach would be to just give TVMC an additional argument, e.g. --uma-backends that adds… --- ## [\[RFC\]\[Quantization\] A new quantization framework in TVM: initial RFC (1/4)](https://discuss.tvm.apache.org/t/rfc-quantization-a-new-quantization-framework-in-tvm-initial-rfc-1-4/9775)
**Author:** [@electriclilies](https://discuss.tvm.apache.org/u/electriclilies)\ **Replies:** 36\ **Last updated:** [January 13, 2023, 7:29am UTC](https://discuss.tvm.apache.org/t/rfc-quantization-a-new-quantization-framework-in-tvm-initial-rfc-1-4/9775 "2023-01-13T07:29:31Z")
A new quantization framework in TVM: Initial RFC In this and subsequent RFCs, we will present a new framework for doing static, data-aware quantization on relay graphs. Previously, I put up an RFC to discuss the new qua… --- ## [\[RFC\] Further Unify Packed and Object in TVM Runtime](https://discuss.tvm.apache.org/t/rfc-further-unify-packed-and-object-in-tvm-runtime/14169)
**Author:** [@tqchen](https://discuss.tvm.apache.org/u/tqchen)\ **Replies:** 2\ **Last updated:** [January 9, 2023, 1:46am UTC](https://discuss.tvm.apache.org/t/rfc-further-unify-packed-and-object-in-tvm-runtime/14169 "2023-01-09T01:46:30Z")
Authors: @cloud-mxd, @junrushao, @tqchen Feature Name: Further Unify Packed and Object in TVM Runtime Start Date: 2023-01-08 RFC PR: apache/tvm-rfcs#0097 GitHub Issue: apache/tvm#0000 Summary This RFC proposes to fur… --- ## [\[RFC\] Relay Containers Array/Map/String](https://discuss.tvm.apache.org/t/rfc-relay-containers-array-map-string/7560)
**Author:** [@xiandi](https://discuss.tvm.apache.org/u/xiandi)\ **Replies:** 21\ **Last updated:** [December 28, 2022, 8:28am UTC](https://discuss.tvm.apache.org/t/rfc-relay-containers-array-map-string/7560 "2022-12-28T08:28:56Z")
I am a mlsys engineer who mainly provides server support for the nlp-related model for the company. In the nlp-related business scenario, we found that tvm can provide superior tensor computing performance, so we intend … --- ## [\[RFC\] Meta Schedule (AutoTensorIR)](https://discuss.tvm.apache.org/t/rfc-meta-schedule-autotensorir/10120)
**Author:** [@junrushao](https://discuss.tvm.apache.org/u/junrushao)\ **Replies:** 21\ **Last updated:** [December 22, 2022, 7:41am UTC](https://discuss.tvm.apache.org/t/rfc-meta-schedule-autotensorir/10120 "2022-12-22T07:41:33Z")
\[RFC\] Meta Schedule (AutoTensorIR) Feature Name: Meta Schedule (AutoTensorIR) Start Date: 2021-05-28 RFC PR: TBD (apache/tvm-rfcs#0000) GitHub Issue: TBD (apache/tvm-rfcs#0000) 1. Summary This proposal introduces Meta … --- ## [\[ RFC \] \[Tensorize\] AMX-TMUL integration (u8s8s32)](https://discuss.tvm.apache.org/t/rfc-tensorize-amx-tmul-integration-u8s8s32/14119)
**Author:** [@error\_Customshop](https://discuss.tvm.apache.org/u/error_Customshop)\ **Replies:** 2\ **Last updated:** [December 19, 2022, 9:15pm UTC](https://discuss.tvm.apache.org/t/rfc-tensorize-amx-tmul-integration-u8s8s32/14119 "2022-12-19T21:15:23Z")
What is AMX? AMX(Advanced Matrix Extensions) is a new 64-bit programming paradigm consisting of two components: A set of 2-dimensional registers (tiles) representing sub-arrays from a larger 2-dimensional memory imag… --- ## [\[RFC\] Type-Directed Relay Fuzzing Library](https://discuss.tvm.apache.org/t/rfc-type-directed-relay-fuzzing-library/12234)
**Author:** [@slyubomirsky](https://discuss.tvm.apache.org/u/slyubomirsky)\ **Replies:** 29\ **Last updated:** [December 16, 2022, 3:58am UTC](https://discuss.tvm.apache.org/t/rfc-type-directed-relay-fuzzing-library/12234 "2022-12-16T03:58:59Z")
RFC: Type-Directed Relay Fuzzing Library Summary This RFC proposes to employ fuzzing (mass generation of random programs) for Relay in order to test the compiler. Fuzz testing for Relay can expose bugs and improve confid… --- ## [\[DISCUSS\]\[torch.fx\] Support pytorch's new frontend torch.fx](https://discuss.tvm.apache.org/t/discuss-torch-fx-support-pytorchs-new-frontend-torch-fx/12498)
**Author:** [@Lyken17](https://discuss.tvm.apache.org/u/Lyken17)\ **Replies:** 12\ **Last updated:** [November 23, 2022, 7:40pm UTC](https://discuss.tvm.apache.org/t/discuss-torch-fx-support-pytorchs-new-frontend-torch-fx/12498 "2022-11-23T19:40:39Z")
Currently loading a pytorch module to tvm (relay) follows the path torch.nn.Module -\> torch.jit.trace -\> torchsript -\> tvm.relay , which works work for most vision and nlp models. However, such conversion has its own lim… --- ## [Introducing Hexagon backend](https://discuss.tvm.apache.org/t/introducing-hexagon-backend/2421)
**Author:** [@kparzysz](https://discuss.tvm.apache.org/u/kparzysz)\ **Replies:** 36\ **Last updated:** [November 3, 2022, 10:37am UTC](https://discuss.tvm.apache.org/t/introducing-hexagon-backend/2421 "2022-11-03T10:37:09Z")
We have implemented basic support for generating and executing Hexagon code from TVM. We would like to upstream our work, and we are asking for community’s feedback about how to proceed. Below is some background inform… --- ## [Introducing TY-NNP backend with end2end TensorIR integration](https://discuss.tvm.apache.org/t/introducing-ty-nnp-backend-with-end2end-tensorir-integration/11807)
**Author:** [@wrongtest](https://discuss.tvm.apache.org/u/wrongtest)\ **Replies:** 12\ **Last updated:** [November 1, 2022, 3:48am UTC](https://discuss.tvm.apache.org/t/introducing-ty-nnp-backend-with-end2end-tensorir-integration/11807 "2022-11-01T03:48:17Z")
Hi, all~ This RFC is to upstream the support for our TY-NNP accelerator backend. We are from the AI accelerator toolchain team of Intellifusion, who has been focusing on developing vision processor that accelerates deep… --- ## [Embedded Rust Interface](https://discuss.tvm.apache.org/t/embedded-rust-interface/13759)
**Author:** [@Mousius](https://discuss.tvm.apache.org/u/Mousius)\ **Replies:** 0\ **Last updated:** [October 20, 2022, 4:08pm UTC](https://discuss.tvm.apache.org/t/embedded-rust-interface/13759 "2022-10-20T16:08:05Z")
Building upon the great work of the TVM community in making microTVM a success, the natural companion for the \[RFC\] \[uTVM\] Embedded C Runtime Interface is the Embedded Rust API. Allowing us to write Rust applications suc… --- ## [\[rfc\] Developer Survey 2022](https://discuss.tvm.apache.org/t/rfc-developer-survey-2022/13743)
**Author:** [@driazati](https://discuss.tvm.apache.org/u/driazati)\ **Replies:** 2\ **Last updated:** [October 19, 2022, 3:25pm UTC](https://discuss.tvm.apache.org/t/rfc-developer-survey-2022/13743 "2022-10-19T15:25:59Z")
To better engage with TVM’s developer and user community we’ve made two anonymous surveys, one for TVM users (link) and one for TVM developers (link). The user survey is intended to help us get some usage data on TVM to … --- ## [\[Process RFC\] Empowering New Scoped Module to the Project](https://discuss.tvm.apache.org/t/process-rfc-empowering-new-scoped-module-to-the-project/13617)
**Author:** [@Hzfengsy](https://discuss.tvm.apache.org/u/Hzfengsy)\ **Replies:** 46\ **Last updated:** [October 19, 2022, 5:52am UTC](https://discuss.tvm.apache.org/t/process-rfc-empowering-new-scoped-module-to-the-project/13617 "2022-10-19T05:52:15Z")
Background Machine Learning Compilation (MLC) is an emerging field in fast development. With the tremendous help from the whole community, it’s exciting to see that TVM delivers significant needs from and to developers, … --- ## [\[meta-RFC\] Vector length agnostic (VLA) vectorization](https://discuss.tvm.apache.org/t/meta-rfc-vector-length-agnostic-vla-vectorization/13596)
**Author:** [@elenkalda-arm](https://discuss.tvm.apache.org/u/elenkalda-arm)\ **Replies:** 0\ **Last updated:** [September 28, 2022, 11:26am UTC](https://discuss.tvm.apache.org/t/meta-rfc-vector-length-agnostic-vla-vectorization/13596 "2022-09-28T11:26:59Z")
This Meta-RFC proposes steps to enable VLA support in TVM by first implementing SVE support in codegen and then adding a generic VLA support into TIR. For an explanation and examples of SVE, see the introduction to SVE2 … --- ## [Issue Triage Workflow](https://discuss.tvm.apache.org/t/issue-triage-workflow/13451)
**Author:** [@hpanda](https://discuss.tvm.apache.org/u/hpanda)\ **Replies:** 1\ **Last updated:** [August 31, 2022, 4:04pm UTC](https://discuss.tvm.apache.org/t/issue-triage-workflow/13451 "2022-08-31T16:04:33Z")
Issue Triage Feature Name: Update Issue Triage Workflow Co-authored with @areusch, @denise Summary This RFC proposes expanding functionality to the existing “New Issue” templates in order to allow for more a precise lo… --- ## [Establish TVM Unity Connection — A Technical Strategy](https://discuss.tvm.apache.org/t/establish-tvm-unity-connection-a-technical-strategy/13344)
**Author:** [@tqchen](https://discuss.tvm.apache.org/u/tqchen)\ **Replies:** 13\ **Last updated:** [August 30, 2022, 2:45pm UTC](https://discuss.tvm.apache.org/t/establish-tvm-unity-connection-a-technical-strategy/13344 "2022-08-30T14:45:52Z")
Co-authored with @denise, @jwfromm Also thanks to helpful inputs from @jknight @jroesch @areusch @yuchenj @junrushao @psrivas2 @sunggg @slyubomirsky @yongwww @ziheng and many other members of the community. Summary At… --- ## [An enabling framework for int8 quantization](https://discuss.tvm.apache.org/t/an-enabling-framework-for-int8-quantization/13269)
**Author:** [@ulfhanebutte](https://discuss.tvm.apache.org/u/ulfhanebutte)\ **Replies:** 8\ **Last updated:** [August 26, 2022, 3:13pm UTC](https://discuss.tvm.apache.org/t/an-enabling-framework-for-int8-quantization/13269 "2022-08-26T15:13:48Z")
Dear TVM community, We, at Marvell, like to continue the discussion on Quantization within TVM. As starting point, we take Masahi’s response to the question by JosseVanDelm (Dec 2021): status on quantization in TVM http… --- ## [\[RFC\] Relax Upstreaming](https://discuss.tvm.apache.org/t/rfc-relax-upstreaming/13351)
**Author:** [@yuchenj](https://discuss.tvm.apache.org/u/yuchenj)\ **Replies:** 1\ **Last updated:** [August 18, 2022, 4:21pm UTC](https://discuss.tvm.apache.org/t/rfc-relax-upstreaming/13351 "2022-08-18T16:21:15Z")
Hi all, We sent out an RFC for upstreaming Relax (Relay Next): https://github.com/apache/tvm-rfcs/pull/89. We would love to hear your thoughts and feedback, feel free to review and leave comments on the RFC! After this… --- ## [\[RFC\] Relay to ONNX](https://discuss.tvm.apache.org/t/rfc-relay-to-onnx/6101)
**Author:** [@maheshambule](https://discuss.tvm.apache.org/u/maheshambule)\ **Replies:** 28\ **Last updated:** [August 17, 2022, 2:47am UTC](https://discuss.tvm.apache.org/t/rfc-relay-to-onnx/6101 "2022-08-17T02:47:27Z")
Motivation: We want to port the DL models in Relay IR. For that, we want to serialize the Relay IR to disk. Once serialized third-party frameworks, compilers should be able to import those. We want the serialization form… --- ## [\[RFC\]\[v2\] Commit Message Guideline](https://discuss.tvm.apache.org/t/rfc-v2-commit-message-guideline/13249)
**Author:** [@gromero](https://discuss.tvm.apache.org/u/gromero)\ **Replies:** 8\ **Last updated:** [August 15, 2022, 4:03pm UTC](https://discuss.tvm.apache.org/t/rfc-v2-commit-message-guideline/13249 "2022-08-15T16:03:12Z")
This RFC replaces Commit Message Guideline This RFC proposes the addition of the following commit message guidelines to pull\_request.rst, under section Submit a Pull Request, below subsection Guidelines, as a subsection… --- ## [Establishing a home for TVM dev tools](https://discuss.tvm.apache.org/t/establishing-a-home-for-tvm-dev-tools/13212)
**Author:** [@areusch](https://discuss.tvm.apache.org/u/areusch)\ **Replies:** 15\ **Last updated:** [August 9, 2022, 3:47am UTC](https://discuss.tvm.apache.org/t/establishing-a-home-for-tvm-dev-tools/13212 "2022-08-09T03:47:24Z")
Hi all, I’d like to start a home in the TVM repo for TVM developer tools–that is, scripts that folks working on TVM might like to run locally. At present, TVM’s developer tools are quite scattered: A Makefile is in th… --- ## [Commit Message Guideline](https://discuss.tvm.apache.org/t/commit-message-guideline/12334)
**Author:** [@gromero](https://discuss.tvm.apache.org/u/gromero)\ **Replies:** 17\ **Last updated:** [August 1, 2022, 11:56pm UTC](https://discuss.tvm.apache.org/t/commit-message-guideline/12334 "2022-08-01T23:56:14Z")
The proposal below is meant to be an initial draft for the Commit Message Guideline and hopefully it will help us to reach a consensus about the rules and standards that the Apache TVM Community wants to attain to ensure… --- ## [\[RFC\]FamilySeer: A new search method for Auto-scheduler](https://discuss.tvm.apache.org/t/rfc-familyseer-a-new-search-method-for-auto-scheduler/11877)
**Author:** [@shanjunZ](https://discuss.tvm.apache.org/u/shanjunZ)\ **Replies:** 4\ **Last updated:** [July 27, 2022, 7:02am UTC](https://discuss.tvm.apache.org/t/rfc-familyseer-a-new-search-method-for-auto-scheduler/11877 "2022-07-27T07:02:09Z")
Motivation: Auto-scheduler (Ansor) uses code sketch and optimization rules to generate a large search space. The search space defined by Ansor has shown great opportunities and therefore the search quality and the search… --- ## [\[RFC\]\[Runtime\] Bring \`PackedFunc\` into TVM Object System](https://discuss.tvm.apache.org/t/rfc-runtime-bring-packedfunc-into-tvm-object-system/11816)
**Author:** [@cyx](https://discuss.tvm.apache.org/u/cyx)\ **Replies:** 11\ **Last updated:** [July 19, 2022, 2:36pm UTC](https://discuss.tvm.apache.org/t/rfc-runtime-bring-packedfunc-into-tvm-object-system/11816 "2022-07-19T14:36:10Z")
Summary This RFC allows developers to use PackedFunc as TVM objects, which completes the last missing step of TVM runtime object system. Motivation Historically, several fundamental data structures in TVM are not part o… [Previous page](https://discuss.tvm.apache.org/c/development/rfc/8.md) [Next page](https://discuss.tvm.apache.org/c/development/rfc/8.md?page=2)