[GitHub] [tvm-rfcs] areusch commented on a change in pull request #22: [RFC][TIR] TIR Non-scalar Constants
areusch commented on a change in pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#discussion_r721602667 ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR: https://github.com/apache/tvm-rfcs/pull/22 +- GitHub Issue: TBD + +# 1. Summary + +This RFC proposes how non-scalar constants could be represented in TIR and used by passes in the lowering process. + +# 2. Motivation + +Currently, the non-scalar constants could be represented in Relay (relay.Constant) to be used by relay passes but not in TIR. Therefore, when performing lowering using TIR passes, we have to maintain a side-channel of tir::Var to constant non-scalar data mapping to perform transformations that could use the knowledge where some of the data are constants. + +Few example scenarios as further motivation : + +## Weight compression + +When lowering for accelerators (E.g. : [Arm(R) Ethos(TM)-U NPU](https://github.com/apache/tvm-rfcs/pull/11)), certain operations will need to get tiled to co-optimize performance and memory utilization. Such tiling patterns create slices of weights that need compressing that will end up with varying sizes. Therefore, the knowledge of some tir::Vars refer to constants are critical in the level of TIR to perform this. + +## Memory Planning + +The TIR program has the ability to express both inter and intra operator memory requirement, post-scheduling as explained further by [Unified Static Memory Planning RFC](https://github.com/apache/tvm-rfcs/pull/9). It would be better if the constants could be embedded to the TIR PrimFunc. Moreover, this allows various [target-dependent lowerings](https://github.com/apache/tvm-rfcs/pull/10), to produce TIR PrimFuncs with constants in it. + +## Winograd Constants + +The Winograd transformation (used for fast GEMMs) involves multiplication by a hard-coded constant tensor. This is currently accomplished in TE using a complicated TE compute expression with many nested selects. Being able to directly express a constant tensor here would significantly simplify this code. + + +# 3. Guide-level explanation + +This is not particularly a user-facing feature and this will allow constants to be 'linked' to TIR. Initially, we are planning to use this with gated on '-link-params' argument for relay.build and TVMC. + +# 4. Reference-level explanation Review comment: me neither anymore :/ -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
[GitHub] [tvm-rfcs] areusch commented on a change in pull request #22: [RFC][TIR] TIR Non-scalar Constants
areusch commented on a change in pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#discussion_r717813423 ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR: https://github.com/apache/tvm-rfcs/pull/22 +- GitHub Issue: TBD + +# 1. Summary + +This RFC proposes how non-scalar constants could be represented in TIR and used by passes in the lowering process. + +# 2. Motivation + +Currently, the non-scalar constants could be represented in Relay (relay.Constant) to be used by relay passes but not in TIR. Therefore, when performing lowering using TIR passes, we have to maintain a side-channel of tir::Var to constant non-scalar data mapping to perform transformations that could use the knowledge where some of the data are constants. + +Few example scenarios as further motivation : + +## Weight compression + +When lowering for accelerators (E.g. : [Arm(R) Ethos(TM)-U NPU](https://github.com/apache/tvm-rfcs/pull/11)), certain operations will need to get tiled to co-optimize performance and memory utilization. Such tiling patterns create slices of weights that need compressing that will end up with varying sizes. Therefore, the knowledge of some tir::Vars refer to constants are critical in the level of TIR to perform this. + +## Memory Planning + +The TIR program has the ability to express both inter and intra operator memory requirement, post-scheduling as explained further by [Unified Static Memory Planning RFC](https://github.com/apache/tvm-rfcs/pull/9). It would be better if the constants could be embedded to the TIR PrimFunc. Moreover, this allows various [target-dependent lowerings](https://github.com/apache/tvm-rfcs/pull/10), to produce TIR PrimFuncs with constants in it. Review comment: ok so i think here we are printing the NDArray contents to TVMScript but not to repr()--does that match your understanding @manupa-arm ? i'm generally ok with this PR as long as we don't cause all TIR printing to be consumed by constant data. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
[GitHub] [tvm-rfcs] areusch commented on a change in pull request #22: [RFC][TIR] TIR Non-scalar Constants
areusch commented on a change in pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#discussion_r717813423 ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR: https://github.com/apache/tvm-rfcs/pull/22 +- GitHub Issue: TBD + +# 1. Summary + +This RFC proposes how non-scalar constants could be represented in TIR and used by passes in the lowering process. + +# 2. Motivation + +Currently, the non-scalar constants could be represented in Relay (relay.Constant) to be used by relay passes but not in TIR. Therefore, when performing lowering using TIR passes, we have to maintain a side-channel of tir::Var to constant non-scalar data mapping to perform transformations that could use the knowledge where some of the data are constants. + +Few example scenarios as further motivation : + +## Weight compression + +When lowering for accelerators (E.g. : [Arm(R) Ethos(TM)-U NPU](https://github.com/apache/tvm-rfcs/pull/11)), certain operations will need to get tiled to co-optimize performance and memory utilization. Such tiling patterns create slices of weights that need compressing that will end up with varying sizes. Therefore, the knowledge of some tir::Vars refer to constants are critical in the level of TIR to perform this. + +## Memory Planning + +The TIR program has the ability to express both inter and intra operator memory requirement, post-scheduling as explained further by [Unified Static Memory Planning RFC](https://github.com/apache/tvm-rfcs/pull/9). It would be better if the constants could be embedded to the TIR PrimFunc. Moreover, this allows various [target-dependent lowerings](https://github.com/apache/tvm-rfcs/pull/10), to produce TIR PrimFuncs with constants in it. Review comment: ok so i think here we are printing the NDArray contents to TVMScript but not to repr()--does that match your understanding @manupa-arm ? i'm generally ok with this PR as long as we don't cause all TIR printing to be consumed by constant data. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
[GitHub] [tvm-rfcs] areusch commented on a change in pull request #22: [RFC][TIR] TIR Non-scalar Constants
areusch commented on a change in pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#discussion_r703945618 ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR: https://github.com/apache/tvm-rfcs/pull/22 +- GitHub Issue: TBD + +# 1. Summary + +This RFC proposes how non-scalar constants could be represented in TIR and used by passes in the lowering process. + +# 2. Motivation + +Currently, the non-scalar constants could be represented in Relay (relay.Constant) to be used by relay passes but not in TIR. Therefore, when performing lowering using TIR passes, we have to maintain a side-channel of tir::Var to constant non-scalar data mapping to perform transformations that could use the knowledge where some of the data are constants. + +Few example scenarios as further motivation : + +## Weight compression + +When lowering for accelerators (E.g. : [Arm(R) Ethos(TM)-U NPU](https://github.com/apache/tvm-rfcs/pull/11)), certain operations will need to get tiled to co-optimize performance and memory utilization. Such tiling patterns create slices of weights that need compressing that will end up with varying sizes. Therefore, the knowledge of some tir::Vars refer to constants are critical in the level of TIR to perform this. + +## Memory Planning + +The TIR program has the ability to express both inter and intra operator memory requirement, post-scheduling as explained further by [Unified Static Memory Planning RFC](https://github.com/apache/tvm-rfcs/pull/9). It would be better if the constants could be embedded to the TIR PrimFunc. Moreover, this allows various [target-dependent lowerings](https://github.com/apache/tvm-rfcs/pull/10), to produce TIR PrimFuncs with constants in it. + +## Winograd Constants + +The Winograd transformation (used for fast GEMMs) involves multiplication by a hard-coded constant tensor. This is currently accomplished in TE using a complicated TE compute expression with many nested selects. Being able to directly express a constant tensor here would significantly simplify this code. + + +# 3. Guide-level explanation + +This is not particularly a user-facing feature and this will allow constants to be 'linked' to TIR. Initially, we are planning to use this with gated on '-link-params' argument for relay.build and TVMC. + +# 4. Reference-level explanation + +The proposal is quite simple and it could be explained as follows : + +``` +@tvm.script.tir +def myfunc(): + param = tir.allocate_const([1, 1, 1, 1, 1, 1, 1, 1, 1, 1], "int32", [10]) +``` + +This follows closely the semantics of tir.allocate and the difference being it represent a buffer filled with constants. Review comment: @junrushao1994 does this answer your questions? it does seem like we could specify placement of the parameters in terms of memory pool/scope in `tir.allocate_const` node if we wanted to, though. the main use case i could see for this is marking weights to be loaded into e.g. accelerator memory. i do think that sort of thing is related to this proposal, but i don't think anything here stops this feature from being implemented later on. so i think we can continue with the proposal as written -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
[GitHub] [tvm-rfcs] areusch commented on a change in pull request #22: [RFC][TIR] TIR Non-scalar Constants
areusch commented on a change in pull request #22: URL: https://github.com/apache/tvm-rfcs/pull/22#discussion_r692378556 ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR: https://github.com/apache/tvm-rfcs/pull/22 +- GitHub Issue: TBD + +# 1. Summary + +This RFC proposes how non-scalar constants could be represented in TIR and used by passes in the lowering process. + +# 2. Motivation + +Currently, the non-scalar constants could be represented in Relay (relay.Constant) to be used by relay passes but not in TIR. Therefore, when performing lowering using TIR passes, we have to maintain a side-channel of tir::Var to constant non-scalar data mapping to perform transformations that could use the knowledge where some of the data are constants. Review comment: nit: say "non-scalar constants are represented in Relay" ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR: https://github.com/apache/tvm-rfcs/pull/22 +- GitHub Issue: TBD + +# 1. Summary + +This RFC proposes how non-scalar constants could be represented in TIR and used by passes in the lowering process. + +# 2. Motivation + +Currently, the non-scalar constants could be represented in Relay (relay.Constant) to be used by relay passes but not in TIR. Therefore, when performing lowering using TIR passes, we have to maintain a side-channel of tir::Var to constant non-scalar data mapping to perform transformations that could use the knowledge where some of the data are constants. Review comment: flip the second sentence around: "Therefore, when a compiler pass is operating on TIR, it needs additional input to determine whether a `tir::Var` is runtime-supplied data or a compile-time constant (and what that constant data is). Should that pass want to transform the constant data, the additional input also becomes an output. In order to provide this data, the compiler has to maintain a side-channel map, complicating the Pass interface." ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR: https://github.com/apache/tvm-rfcs/pull/22 +- GitHub Issue: TBD + +# 1. Summary + +This RFC proposes how non-scalar constants could be represented in TIR and used by passes in the lowering process. + +# 2. Motivation + +Currently, the non-scalar constants could be represented in Relay (relay.Constant) to be used by relay passes but not in TIR. Therefore, when performing lowering using TIR passes, we have to maintain a side-channel of tir::Var to constant non-scalar data mapping to perform transformations that could use the knowledge where some of the data are constants. + +Few example scenarios as further motivation : + +## Weight compression + +When lowering for accelerators (E.g. : [Arm(R) Ethos(TM)-U NPU](https://github.com/apache/tvm-rfcs/pull/11)), certain operations will need to get tiled to co-optimize performance and memory utilization. Such tiling patterns create slices of weights that need compressing that will end up with varying sizes. Therefore, the knowledge of some tir::Vars refer to constants are critical in the level of TIR to perform this. + +## Memory Planning + +The TIR program has the ability to express both inter and intra operator memory requirement, post-scheduling as explained further by [Unified Static Memory Planning RFC](https://github.com/apache/tvm-rfcs/pull/9). It would be better if the constants could be embedded to the TIR PrimFunc. Moreover, this allows various [target-dependent lowerings](https://github.com/apache/tvm-rfcs/pull/10), to produce TIR PrimFuncs with constants in it. + +## Winograd Constants + +The Winograd transformation (used for fast GEMMs) involves multiplication by a hard-coded constant tensor. This is currently accomplished in TE using a complicated TE compute expression with many nested selects. Being able to directly express a constant tensor here would significantly simplify this code. + + +# 3. Guide-level explanation + +This is not particularly a user-facing feature and this will allow constants to be 'linked' to TIR. Initially, we are planning to use this with gated on '-link-params' argument for relay.build and TVMC. + +# 4. Reference-level explanation Review comment: does the example from the other RFC still hold? if so, worth copying here? ## File path: rfcs/0022-tir-non-scalar-constants.md ## @@ -0,0 +1,107 @@ + +- Feature Name: tir_non_scalar_constants +- Start Date: 2021-06-01 +- RFC PR: https://github.com/apache/tvm-rfcs/pull/22 +- GitHub Issue: TBD + +# 1. Summary + +This RFC proposes how non-scalar constants could be represented in TIR and used by passes in the lowering
