docs: remove TPU / Cerebras / Metalium backends (removed from the compiler) - #148
Merged
Merged
Conversation
…piler)
The TPU, Cerebras (WSE-3), and Tenstorrent Metalium backends have been removed
from the compiler — the Tessera_{TPU,Cerebras,Metalium}_Backend dirs,
compiler/tpu_target.py, the tessera-lower-to-metalium pipeline, and
docs/metalium_kernel_inventory.md no longer exist, and the Python compiler emits
no target entries for them (generated dashboards already carry none). The
remaining doc/CLAUDE.md references were stale artifacts.
Scrubbed from CLAUDE.md (target-hardware list, IR-stack Target IR row, Phase 7/8
rows, source-location tables, Decisions #19/#20, the TPU/MXU constraint, the
lowering-pipeline + kernel-inventory tables) and from MASTER_AUDIT / APPLE_AUDIT
(cross-backend gate prose), tessera-api-reference + API_Reference_Index (drop the
"metalium" @jit string target), and programming-guide Ch6 (drop the Tenstorrent
bfp*/blockfp* gated-dtype prose; not in dtype.py).
Live targets are now NVIDIA / AMD ROCm / x86 (AMX/AVX512) / Apple (CPU/GPU), plus
the NV RubinCPX dialect. Kept: stablehlo_export (a real format-export feature,
not a TPU backend) and AMD mxfp* gated dtypes. Generated-doc drift gate + audit
tests green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The TPU, Cerebras (WSE-3), and Tenstorrent Metalium backends have been removed from the compiler —
Tessera_{TPU,Cerebras,Metalium}_Backend,compiler/tpu_target.py, thetessera-lower-to-metaliumpipeline, anddocs/metalium_kernel_inventory.mdno longer exist, and the Python compiler (backend_manifest/primitive_coverage) emits no target entries for them (the generated dashboards already carry none). The remaining references in CLAUDE.md and the docs were stale artifacts. This scrubs them.Scrubbed
tpu_target.py, the three backend dirs), Decisions Apple GPU Tier-2/3: reductions, native GQA, fused batched attention #19/Update apple_gpu_runtime.mm #20 (tessera_metaliumexample,"metalium"string), the TPU/MXU constraint bullet, the lowering-pipeline + kernel-inventory tables.NVIDIA/ROCm/Metalium→NVIDIA/ROCm)."metalium"@jitstring target.bfp*/blockfp*gated-dtype prose (doc-only; not indtype.py).Live targets now
NVIDIA / AMD ROCm / x86 (AMX/AVX512) / Apple (CPU/GPU), plus the NV RubinCPX dialect.
Deliberately kept
stablehlo_export(a real format-export feature intessera-translate/aot.py, independent of any TPU backend).mxfp*gated dtypes; JAX/XLA as reference vocabulary (Decision Apple GPU Tier-2: promote mla_decode to a real GPU kernel #23).Verification
🤖 Generated with Claude Code