Skip to content

Pull requests: ROCm/hip-ep

Pull request creation is restricted
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort

Pull requests list

feat(MIOpen): Use Find API for Convolution
#715 opened Aug 14, 2026 by ilango100 Contributor Loading…
1 of 4 tasks
Feature/dynamicdispatch npu backend large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#714 opened Aug 14, 2026 by bcrane-amd Contributor Draft
feat(hipsr): require host/device memory space on hipsr op tensor operands large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#713 opened Aug 14, 2026 by amd-genmingz Collaborator Draft
3 tasks
testing PR700
#708 opened Aug 13, 2026 by BoarQing Contributor Loading…
feat(gqa): resolve autotune configs from an offline LUT large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#701 opened Aug 12, 2026 by ZhanxingPu Contributor Loading…
fix(loop): support shape-changing loop carriers large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#689 opened Aug 10, 2026 by fhanuman Collaborator 6/12 Draft
fix(shape): canonicalize host reshape provenance large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#681 opened Aug 7, 2026 by fhanuman Collaborator 3/12 Draft
fix(compiler): reuse late strided promotion temporaries
#676 opened Aug 6, 2026 by fhanuman Collaborator Draft
4 tasks done
perf(gqa): halve decode KV traffic and unify the dispatch path (up to 1.9x fp16, 2.3x INT8) large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#675 opened Aug 6, 2026 by ZhanxingPu Contributor Loading…
4 tasks done
ci: pin OGA to 0.15.1 and the rebased AMDGPU integration PR
#667 opened Aug 5, 2026 by zz002 Collaborator Draft
6 of 8 tasks
perf(matmul_nbits): lower the WMMA cutoff to M=8 (-2.3% TTFT on gpt-oss-20b 16k)
#661 opened Aug 4, 2026 by BoarQing Contributor Loading…
4 tasks done
fix(hip): harden shape refinement failure handling
#647 opened Aug 3, 2026 by fhanuman Collaborator 12/12 Draft
refactor(hip): enforce explicit DPS shape contracts locally large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#646 opened Aug 3, 2026 by fhanuman Collaborator 11/12 Draft
fix(hip): share attention and normalization shape contracts large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#645 opened Aug 3, 2026 by fhanuman Collaborator 10/12 Draft
fix(hip): share Gather and tensor-shape contracts large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#644 opened Aug 3, 2026 by fhanuman Collaborator 9/12 Draft
fix(hip): share Conv and Pool shape contracts large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#643 opened Aug 3, 2026 by fhanuman Collaborator 8/12 Draft
feat(hip): add payload-driven shapes and bulk readback large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#642 opened Aug 3, 2026 by fhanuman Collaborator 7/12 Draft
fix(hip): unify broadcast and reduction shape contracts large-pr Soft size threshold (>500 LOC or >10 files). Reviewer signal, not a block.
#641 opened Aug 3, 2026 by fhanuman Collaborator 5/12 Draft
ProTip! Exclude everything labeled bug with -label:bug.