19 Pull requests merged by 13 people

• code-gen: improved tail loop and edge tile of swizzled A

  #1421 merged Dec 20, 2024

• Fix incorrect type casting for alpha and beta in f16 compute type

  #1462 merged Dec 20, 2024

• Support UseSgprForGRO for dtva

  #1437 merged Dec 19, 2024

• Update gfx942 NT/TN/NN FP8/BF8/BF16 Equality

  #1463 merged Dec 19, 2024

• gfx942_80cu BBS NN NT TN Tuning Release

  #1459 merged Dec 18, 2024

• Revert "Fix incorrect type casting for alpha and beta in f16 compute type"

  #1461 merged Dec 18, 2024

• Fix invalid string printed when running hipblaslt-test

  #1428 merged Dec 17, 2024

• Fix incorrect type casting for alpha and beta in f16 compute type

  #1454 merged Dec 17, 2024

• Update solutions for hstu bmm 3 sizes.

  #1453 merged Dec 17, 2024

• Update gfx942_80cu NT/TN/NN f16/f32 Equality

  #1452 merged Dec 17, 2024

• gfx942 38cu HSS/BSS NN TN NT grid tune

  #1448 merged Dec 16, 2024

• gfx942 38cu SGEMM NN TN NT grid tune

  #1439 merged Dec 16, 2024

• gfx942 38cu F8HS NN TN NT grid tune

  #1440 merged Dec 16, 2024

• Modified the Sparse Test in Tensile Lite

  #1450 merged Dec 16, 2024

• [Hotfix] correct occupancy calculation

  #1451 merged Dec 16, 2024

• Patch client writer

  #1447 merged Dec 16, 2024

• Bump rocm-docs-core from 1.10.0 to 1.11.0 in /docs/sphinx

  #1427 merged Dec 14, 2024

• Remove functions for generating file paths

  #1402 merged Dec 13, 2024

• Update gfx942 BBS NN/NT/TN Equality/GridBased yamls for 1204 bbs bmm dynamic

  #1446 merged Dec 13, 2024

17 Pull requests opened by 14 people

• correct USE_HIP_FP8_DEF define condition

  #1455 opened Dec 16, 2024

• Fix garbage value of bias_type

  #1457 opened Dec 17, 2024

• Bump rocm-docs-core from 1.11.0 to 1.12.0 in /docs/sphinx

  #1458 opened Dec 17, 2024

• Use archVGPR when accVGPR is not enough.

  #1460 opened Dec 18, 2024

• gfx12 - enlarging tolerance is not needed

  #1464 opened Dec 18, 2024

• Add BBS support for find_exact.py

  #1465 opened Dec 19, 2024

• Create README.md for TensileLite

  #1466 opened Dec 19, 2024

• Update BBS NN/NT/TN equality tuning for gfx942_80cu

  #1467 opened Dec 19, 2024

• Update README.md

  #1468 opened Dec 19, 2024

• Provide unsurprising behavior for tensile threads

  #1469 opened Dec 19, 2024

• Aquavanjaram 20CU equality GEMMM tuning updates for TF32 NN and TN data type

  #1470 opened Dec 19, 2024

• Changes to exclude StreamK by default

  #1471 opened Dec 19, 2024

• Build prototype

  #1472 opened Dec 19, 2024

• Tune Aquavanjaram942 Grid sizes for BBS & HHS NN

  #1473 opened Dec 19, 2024

• Gfx942 80cu grid based and equality tuning for HHS NN

  #1474 opened Dec 19, 2024

• Update rocm-docs-core to 1.12.0

  #1475 opened Dec 19, 2024

• update 942 7 range F8HS Gridebase

  #1476 opened Dec 20, 2024

1 Issue closed by 1 person

• [Issue]: "unused" dependencies in requirements.txt

  #1441 closed Dec 16, 2024

12 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• [Issue]: [Documentation] Document gfx908 support

  #1299 commented on Dec 19, 2024 • 0 new comments

• [Issue]: how to make sure the gemm operation using hipblasLtMatmul()

  #1442 commented on Dec 16, 2024 • 0 new comments

• dot2 fp16 mac kernel for gfx942

  #1258 commented on Dec 14, 2024 • 0 new comments

• Added multiple devices support for matrix transform

  #1338 commented on Dec 17, 2024 • 0 new comments

• Remove merge-files option

  #1407 commented on Dec 19, 2024 • 0 new comments

• Update gfx942_80cu TN/NN f16 gridbased

  #1419 commented on Dec 20, 2024 • 0 new comments

• Update gfx942_80cu BF16 TN equality

  #1420 commented on Dec 20, 2024 • 0 new comments

• Optimize generator

  #1424 commented on Dec 19, 2024 • 0 new comments

• Add reject states for failing streamk params

  #1425 commented on Dec 20, 2024 • 0 new comments

• Support conjugate-transpose as equivalent to transpose

  #1429 commented on Dec 19, 2024 • 0 new comments

• [Draft] Distributed tuning

  #1435 commented on Dec 19, 2024 • 0 new comments

• Tcl cleanup

  #1444 commented on Dec 19, 2024 • 0 new comments