Aart J.C. Bik received his PhD from the Leiden University in 1996. From 1998 to 2007, he worked at Intel, where he was the lead compiler architect of automatic vectorization in the Intel C++/Fortran compilers. In 2007, he moved to Google, where he has worked on compilers for ART and the Dart VM. He is now a member of the MLIR codegen team, focusing on compiler support for sparse tensor computations.
User Details
User Details
- User Since
- Jan 10 2020, 10:13 AM (168 w, 1 d)
Yesterday
Yesterday
Thu, Mar 30
Thu, Mar 30
Wed, Mar 29
Wed, Mar 29
aartbik added inline comments to D147192: [mlir][sparse] avoid creating temporary unordered COO buffer when reshape sparse tensor..
aartbik added inline comments to D147011: [mlir][sparse] Replace Element with ElementId for sorting..
aartbik added inline comments to D146684: [mlir][sparse] Preliminary code changes for ExprId, LatPointId, LatSetId newtypes.
aartbik added inline comments to D146684: [mlir][sparse] Preliminary code changes for ExprId, LatPointId, LatSetId newtypes.
Tue, Mar 28
Tue, Mar 28
aartbik added inline comments to D147074: [mlir][sparse] convert a sparse tensor slice to sparse tensor correctly..
Mon, Mar 27
Mon, Mar 27
Wed, Mar 22
Wed, Mar 22
aartbik added inline comments to D146083: [mlir][sparse] Updating the `Merger::{exp,lat,set}` methods to return const.
Tue, Mar 21
Tue, Mar 21
aartbik accepted D146563: [mlir][sparse] fix crash when using pure constant index in indexing mapping (fixes #61530).
can you add a before fail/after pass test?
aartbik committed rGa2c63d7f0b77: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs (authored by aartbik).
[mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs
aartbik updated the diff for D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs.
rebased with main
Mon, Mar 20
Mon, Mar 20
aartbik added inline comments to D146086: [mlir][sparse] Making `TensorExp::Kind` a nested enum-class.
aartbik added inline comments to D146086: [mlir][sparse] Making `TensorExp::Kind` a nested enum-class.
aartbik accepted D142928: [mlir][sparse] add merger/topo sort support for slice-based affine sparse index codegen.
Wait to see if Wren is also okay with how her comments were addressed.
aartbik added inline comments to D142928: [mlir][sparse] add merger/topo sort support for slice-based affine sparse index codegen.
Fri, Mar 17
Fri, Mar 17
aartbik updated the diff for D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs.
bit of code cleanup, also used smaller values to avoid fp16 issues
aartbik added inline comments to D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs.
aartbik retitled D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs from [WIP] [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs to [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs.
aartbik added a reviewer for D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs: K-Wu.
aartbik updated the summary of D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs.
aartbik updated the diff for D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs.
dotted all the i's (thanks to Chris); ready for review!
aartbik committed rGa3bb693420c5: [mlir][gpu][nvvm] refined sparsity selector test and verification of mma.sp (authored by aartbik).
[mlir][gpu][nvvm] refined sparsity selector test and verification of mma.sp
aartbik updated the diff for D146319: [mlir][gpu][nvvm] refined sparsity selector test and verification of mma.sp.
rebased with main (old main was broken)
aartbik updated the diff for D146319: [mlir][gpu][nvvm] refined sparsity selector test and verification of mma.sp.
rebased with main
aartbik committed rG4e4af1338da5: [mlir][gpu][nvvm] fixed bug with literal for inline asm for mma instruction (authored by aartbik).
[mlir][gpu][nvvm] fixed bug with literal for inline asm for mma instruction
aartbik added a comment to D146110: [mlir][gpu][nvvm] fixed bug with literal for inline asm for mma instruction.
Thu, Mar 16
Thu, Mar 16
aartbik updated the diff for D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs.
made code actually running, still needs a few i's dotted....
aartbik added inline comments to D142928: [mlir][sparse] add merger/topo sort support for slice-based affine sparse index codegen.
aartbik added inline comments to D142928: [mlir][sparse] add merger/topo sort support for slice-based affine sparse index codegen.
aartbik added a reviewer for D146110: [mlir][gpu][nvvm] fixed bug with literal for inline asm for mma instruction: guraypp.
Wed, Mar 15
Wed, Mar 15
aartbik updated the diff for D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs.
rebased with main, updated RUN command to get into actual running state
aartbik added a comment to D146110: [mlir][gpu][nvvm] fixed bug with literal for inline asm for mma instruction.
Tue, Mar 14
Tue, Mar 14
aartbik updated the diff for D146110: [mlir][gpu][nvvm] fixed bug with literal for inline asm for mma instruction.
typo
aartbik updated the diff for D146110: [mlir][gpu][nvvm] fixed bug with literal for inline asm for mma instruction.
typo
aartbik retitled D146110: [mlir][gpu][nvvm] fixed bug with literal for inline asm for mma instruction from [mlir][gpu][nvvm] fix bug with literal for inline asm for mma instruction to [mlir][gpu][nvvm] fixed bug with literal for inline asm for mma instruction.
aartbik added a comment to D146052: [mlir][Interfaces] Add ReifyShapeDimTypeInterface.
LG for sparse
aartbik added a comment to D146053: [mlir][tensor/memref] Replace various redundant helpers with `reifyShapeDims`.
LG for sparse with one nit
We should perhaps start thinking about moving all sorted related codegen to a more general place and in its own file.
(not know, but this is becoming a powerful utility that others may want to use too)
Mon, Mar 13
Mon, Mar 13
aartbik added a comment to D145756: [mlir][sparse] Cleaning up names in {Merger,LoopEmitter,CodegenEnv}.{h,cpp}.
are we really going to follow up on all the TODO's and notes?
is it really worth putting them *all* in our codebase?
Fri, Mar 10
Fri, Mar 10
aartbik added inline comments to D145824: [mlir][vector] Clarify OOB semantics for `gather` and `scatter`.
aartbik added inline comments to D141641: [mlir][sparse] extend storage specifier operations for slices..
aartbik added inline comments to D141532: [mlir][sparse] support dynamic sparse tensor slices..
Thu, Mar 9
Thu, Mar 9
aartbik updated the summary of D145708: [sparse][mlir] test transposition on sorted COO.
aartbik committed rGe1b3c5c403b8: [sparse][mlir] test transposition on sorted COO (authored by aartbik).
[sparse][mlir] test transposition on sorted COO
aartbik updated the diff for D145708: [sparse][mlir] test transposition on sorted COO.
works for direct IR now
aartbik requested review of D145708: [sparse][mlir] test transposition on sorted COO.
aartbik added inline comments to D145532: [mlir][sparse] deduplicate non-unique coordinates when coiterating collapsed COO tensors..
Wed, Mar 8
Wed, Mar 8
Tue, Mar 7
Tue, Mar 7
aartbik added inline comments to D145518: [mlir][sparse] deduplicate non-unique coordinates when coiterating COO tensors.
aartbik added a comment to D145440: [mlir][sparse] Improve quick sort by using a loop to sort the bigger partition..
any perf results?
aartbik added a comment to D145442: [mlir][sparse] Improve sort operation by generating inlined code to compare values..
any perf result you can put in the description?
Mon, Mar 6
Mon, Mar 6
aartbik added a reviewer for D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs: christopherbate.
aartbik requested review of D145447: [mlir][sparse][gpu] end-to-end test for 2:4 sparsity on NVidia GPUs.
Fri, Mar 3
Fri, Mar 3
aartbik committed rG657f60a07b2d: [mlir][vector] add support for printing f16 and bf16 (authored by aartbik).
[mlir][vector] add support for printing f16 and bf16