Evidence index

Generated by docs/tools/gen_db_pages.py on every docs build. Do not edit.

Code comments do not carry measurements; they carry a pointer evidence: docs/<page>.md#<section> to the section that holds the measurement. This is the reverse map: for each cited section, every place in the code, tests and build that depends on it. Before changing or deleting a section, check what cites it here. .github/ci/check_evidence_anchors.py keeps every pointer resolvable.

710 pointers into 391 sections of 44 pages.

docs/algorithms/steqr.md

Section Cited by
#steqr-choosing-qr-versus-ql src/ops/syev/steqr/steqr_cta_device.hh:229
#steqr-the-relative-deflation-criterion src/ops/syev/steqr/steqr_cta_device.hh:207

docs/cpp-api.md

Section Cited by
#convergence-status-syev-syevx-gesvd-steqr-stedc include/batchlas/blas/extensions.hh:905
#row-major-source-data include/batchlas/blas/matrix.hh:1567
#trsm-alpha-moved-next-to-the-matrices include/batchlas/blas/functions/trsm.hh:115
include/batchlas/blas/functions/trsm.hh:160
#which-type-each-parameter-takes include/batchlas/blas/extensions.hh:2004

docs/design/api-conventions.md

Section Cited by
#api-conventions-cond-generators-default-to-cgs2 include/batchlas/blas/extra.hh:130
#api-conventions-generators-without-a-dispatch-overload include/batchlas/blas/extra.hh:220
#api-conventions-linalgsolve-keeps-its-transpose-branch include/batchlas/blas/linalg-ops.hh:373
#api-conventions-shape-checks-live-on-both-option-overloads include/batchlas/blas/options.hh:96
#api-conventions-the-bare-braces-potrf-trap include/batchlas/blas/options.hh:635
#api-conventions-the-usm-pointer-check include/batchlas/blas/queue-dispatch.hh:183
#api-conventions-the-variadic-dispatch-overload-and-its-requires-clause include/batchlas/blas/queue-dispatch.hh:270
#api-conventions-two-workspace-spellings-never-a-defaulted-span include/batchlas/blas/options.hh:264

docs/design/build-performance.md

Section Cited by
#build-performance-the-umbrella-header-excludes-device-code include/batchlas/blas/linalg.hh:13
include/batchlas/sycl_interop.hh:17

docs/design/device-group-blas.md

Section Cited by
#device-group-blas-register-and-local-memory-caveats include/batchlas/blas/device/detail/group_blas_subgroup_common.hh:137
#device-group-blas-the-3-d-launch-generic-fallback include/batchlas/blas/device/detail/group_blas_gemm.hh:407
include/batchlas/blas/device/detail/group_blas_rankk.hh:415
include/batchlas/blas/device/detail/group_blas_rankk.hh:450

docs/design/environment.md

Section Cited by
#environment-configure-sets-the-base-not-a-lock include/batchlas/settings.hh:516
src/util/settings.cc:71
#environment-files-written-from-the-environment include/batchlas/settings.hh:374
#environment-geometry-defaults-stay-at-the-call-site include/batchlas/settings.hh:247
#environment-knobs-that-override-an-explicit-argument include/batchlas/settings.hh:118
#environment-move-the-string-not-the-parser include/batchlas/settings.hh:16
#environment-parser-defects-recorded-not-fixed include/batchlas/settings.hh:217
include/batchlas/settings.hh:317
include/batchlas/settings.hh:360
#environment-sb2st-back-wave-fails-open include/batchlas/settings.hh:179
#environment-scopedenvvar include/batchlas/util/env.hh:103
#environment-the-shared-parsers-in-envhh include/batchlas/util/env.hh:28
#environment-the-unsafe-group-and-its-gate include/batchlas/settings.hh:451
#environment-two-knobs-that-look-unsafe-and-are-not include/batchlas/settings.hh:415
#environment-two-variables-for-one-quantity include/batchlas/settings.hh:265
#environment-what-a-settings-reload-does-not-cover include/batchlas/settings.hh:529

docs/design/error-model.md

Section Cited by
#error-model-the-three-tag-base-rules include/batchlas/error.hh:38

docs/design/flat-kernel-selection-phase3-plan.md

Section Cited by
(whole page) src/ops/gemm/choice.hh:162
src/ops/trsm/choice.hh:30

docs/design/flat-kernel-selection.md

Section Cited by
(whole page) docs/developer/documentation.md:209
src/ops/potrf/choice.hh:44
#flat-selection-tables-and-provenance tests/shipped_table_pick.hh:7
#phase-5-geqrf src/ops/geqrf/geqrf.hh:14

docs/design/gesvd.md

Section Cited by
#gesvd-design-vendor-binding-and-dispatch include/batchlas/blas/functions/gesvd.hh:206
#gesvd-design-why-bdsdc-goes-through-golub-kahan src/ops/gesvd/bdsdc/bdsdc.cc:5
#gesvd-design-why-one-sided-jacobi src/ops/gesvd/kernels/gesvdj_cta.cc:34
#gesvd-design-why-svdvectorsthin-exists include/batchlas/blas/enums.hh:219
#gesvdj_cta-global-power-of-two-scaling src/ops/gesvd/kernels/gesvdj_cta.cc:352
#gesvdj_cta-local-memory-budget-formula src/ops/gesvd/gesvd_native.hh:13
src/ops/gesvd/kernels/gesvdj_cta.cc:145
src/ops/gesvd/kernels/gesvdj_cta.cc:163
src/ops/gesvd/kernels/gesvdj_cta.cc:799
#gesvdj_cta-rank-deficiency-thin-and-m--n src/ops/gesvd/kernels/gesvdj_cta.cc:300
src/ops/gesvd/kernels/gesvdj_cta.cc:679
#gesvdj_cta-sweep-convergence-and-extraction-rules src/ops/gesvd/kernels/gesvdj_cta.cc:388
#gesvdj_cta-the-lane-equals-row-mapping-decision src/ops/gesvd/kernels/gesvdj_cta.cc:35
#gesvdj_cta-the-reduce-scatter-g3-trap src/ops/gesvd/kernels/gesvdj_cta.cc:440

docs/design/known-defects.md

Section Cited by
#defect-1-orthos-transposed-arm-builds-a-view-that-does-not-describe-the-memory src/ops/gemv/gemv.cc:52
#defect-11-native-gemm-reads-c-at-beta-zero src/ops/gemm/kernels/epilogue_linear.hh:12
#defect-13-complexdouble-cublas-calls-segfault-inside-cublaslt src/ops/gemv/gemv.cc:53
tests/gemv_tests.cc:1031
#defect-15-cusolver-gesvdjbatched-faults-on-values-only-non-square-input tools/tune/gesvd_spec.cc:88
#defect-17-cusparse-spmm-shapes-refused-in-can_run-and-the-alignment-waiver tools/tune/spmm_spec.cc:92
#defect-3-lanczos-issues-a-two-column-multiply-and-consumes-one-column src/ops/lanczos/lanczos.cc:61
#known-defects-the-cta-sytrd-lower-path src/ops/syev/kernels/syev_cta.cc:130
#linalgqr-returns-a-wrong-qr-after-an-earlier-call-in-the-process include/batchlas/blas/linalg-ops.hh:564
#lobpcg-picked-a-spurious-zero-ritz-value-from-a-null-p-column src/ops/syevx/kernels/syevx_lobpcg.cc:1062
tests/syev_tests.cc:289
tests/syevx_tests.cc:1172

docs/design/matrix-model.md

Section Cited by
#matrix-model-the-element-address-and-its-64-bit-batch-term include/batchlas/blas/matrix.hh:171
#matrix-model-the-pointer-array-and-sliced-views include/batchlas/blas/matrix.hh:920
src/ops/syev/sytrd/band_reduction.cc:597

docs/design/runtime-internals.md

Section Cited by
#runtime-internals-explicit-instantiation-macros src/util/template-instantiations.hh:11
#runtime-internals-matrix-layout-checks src/matrix.cc:1736
src/matrix.cc:1949
src/matrix.cc:229
src/matrix.cc:494
src/matrix.cc:867
#runtime-internals-namespace-placement-of-out-of-line-definitions src/queue.hh:49
src/util/queue-impl.cc:17
src/util/sycl-util-impl.cc:27
#runtime-internals-queue-thread-ownership-and-the-last-event-holder src/queue.hh:250
src/queue.hh:37
#runtime-internals-symbol-visibility-for-private-headers src/queue.hh:21
src/util/internal-api.hh:3
src/util/kernel-trace.hh:30
src/util/sycl-util-impl.cc:245
#runtime-internals-the-host-blas-dgemm-health-check src/backends/netlib_lapack.cc:36
#runtime-internals-the-per-item-info-span-contract src/kernels/common/info_span.hh:11
src/linalg-impl.hh:650
src/linalg-impl.hh:668
#runtime-internals-the-per-queue-workspace-arena src/queue.hh:70
src/util/queue-impl.cc:159
src/util/queue-impl.cc:234
#runtime-internals-the-settings-loader src/util/queue-impl.cc:71
src/util/settings.cc:3
src/util/settings.cc:349
#runtime-internals-vendor-tus-instantiate-only-vendor-symbols src/backends/batch_launch.hh:8
src/backends/cusolver.cc:543
src/backends/netlib_lapack.cc:1353
src/backends/rocblas.cc:270
src/backends/rocsolver.cc:445
src/backends/rocsparse.cc:89

docs/design/sycl-implementations.md

Section Cited by
#sycl-impl-asynchronous-errors src/sycl/impl.hh:218
src/util/queue-impl.cc:172
tests/util_device_queue_tests.cc:312
#sycl-impl-device-target-dispatch src/kernels/common/sg_partition/sg_partition.hh:33
#sycl-impl-jit-ptx-check tests/CMakeLists.txt:579
#sycl-impl-runtime-interop-seam src/sycl/impl.hh:4
#sycl-impl-slm-budget cmake/BatchLASOptions.cmake:55
src/sycl/acpp_slm_optin.cc:6
src/sycl/local_mem.hh:5
src/util/resident_capacity.hh:106
tests/test_utils.hh:24

docs/design/syevx-range-selection.md

Section Cited by
#syevx-range-direct-and-directsubset-may-disagree-on-m-by-one src/ops/syevx/kernels/syevx_direct.cc:47
#syevx-range-iterative-paths-cannot-answer-an-interior-range src/ops/syevx/kernels/syevx_filtered.cc:95
src/ops/syevx/kernels/syevx_lobpcg.cc:215
#syevx-range-normalization-and-validation src/ops/syevx/syevx.cc:205
#syevx-range-stein-workspace-scaling python/batchlas/_api.py:1222
#syevx-range-steins-per-item-count-bound src/ops/syev/tridiag/stein.cc:299
#syevx-range-syevxselect-rather-than-eigenrangetype include/batchlas/blas/enums.hh:413
#syevx-range-the-host-reference-oracle python/tests/test_batchlas.py:554
#syevx-range-the-m-output-and-what-neigs-means src/ops/syevx/kernels/syevx_direct.cc:204
#syevx-range-the-stein-poisoned-tail-test tests/stein_tests.cc:211
#syevx-range-throw-do-not-degrade src/ops/syevx/syevx.cc:250
#syevx-range-why-the-thresholds-carry-over-unchanged benchmarks/syevx_benchmark.cc:414
src/ops/syevx/syevx.cc:313

docs/design/syevx.md

Section Cited by
#bisection-and-inverse-iteration-stebz-and-stein-build src/ops/syev/tridiag/stebz.cc:5
#chebyshev-davidson-consider-as-the-lobpcg-replacement-for-unpreconditioned-problems benchmarks/syevx_benchmark.cc:173
#syevx-direct-the-range-is-checked-at-its-own-entry-point src/ops/syevx/kernels/syevx_direct.cc:93
#syevx-directsubset-reported-as-converged-with-a-caveat src/ops/syevx/kernels/syevx_direct_subset.cc:83
#syevx-dispatcher-environment-override-versus-explicit-request src/ops/syevx/syevx.cc:161
src/ops/syevx/syevx.cc:334
#syevx-dispatcher-problem-arguments-are-validated-before-dispatch src/ops/syevx/syevx.cc:108
src/ops/syevx/syevx.cc:60
#syevx-dispatcher-sparse-input-and-an-explicit-filtered-request src/ops/syevx/syevx.cc:286
#syevx-dispatcher-which-preconditioner-is-legal-at-which-end src/ops/syevx/syevx.cc:71
#syevx-filtered-normalising-the-filter-and-capping-its-degree src/ops/syevx/kernels/syevx_filtered.cc:422
#syevx-filtered-the-problem-derived-degree src/ops/syevx/kernels/syevx_filtered.cc:477
src/ops/syevx/kernels/syevx_filtered.cc:58
#syevx-filtered-the-scaled-chebyshev-filter src/ops/syevx/kernels/syevx_filtered.cc:14
#syevx-for-batches-of-medium-matrices src/ops/syev/tridiag/stein.cc:7
#syevx-small-n-why-direct-is-the-answer src/ops/syevx/kernels/syevx_direct.cc:3
#syevx-the-flop-count-cost-model include/batchlas/util/minibench.hh:869
#syevx-why-the-preconditioner-environment-variable-is-only-a-default include/batchlas/blas/enums.hh:448

docs/design/symbol-visibility.md

Section Cited by
#symbol-visibility-binaryop-is-a-template-argument-enum-too include/batchlas/blas/linalg-ops.hh:53
#symbol-visibility-enums-used-as-template-arguments include/batchlas/blas/enums.hh:22
#symbol-visibility-event-carries-per-member-exports include/batchlas/util/sycl-device-queue.hh:278
#symbol-visibility-exception-typeinfo-must-be-exported include/batchlas/error.hh:22
#symbol-visibility-exported-inline-queue-members include/batchlas/util/sycl-device-queue.hh:373
#symbol-visibility-the-export-attribute-on-the-matrix-class-template include/batchlas/blas/matrix.hh:1397
#symbol-visibility-workspacelease-is-exported-at-class-level include/batchlas/util/workspace.hh:45

docs/design/tiered-tuning.md

Section Cited by
#engine-dominance-carry-forward-removed tests/tune_tests.cc:2242
tests/tune_tests.cc:3558
#engine-guard-retries-a-failed-nvidia-smi-query tests/tune_tests.cc:799
tools/tune/tune_core.hh:185
#engine-op-order-over-all-19-ops tools/tune/schedule.hh:59
#engine-persistent-workers-and-the-carve-out-audit tools/tune/cell_runner.hh:10
tools/tune/worker.hh:5
#engine-re-measuring-named-cells tests/tune_tests.cc:2539
tools/tune/schedule.hh:56
#engine-readiness-for-a-deep-run-over-all-19-ops tools/tune/race.cc:61
#engine-refinement-convergence-rules tools/tune/grid.cc:76
tools/tune/grid.hh:38
#engine-repeat-crashers-are-dropped tools/tune/worker.hh:79
#engine-the-ld-audit tools/tune/spec.hh:36
#engine-the-ledger-and-table-generation tools/tune/tune_core.hh:118
#engine-tiers-and-the-per-cell-algorithm tools/tune/schedule.hh:3

docs/design/tuning-data.md

Section Cited by
#tuning-data-kernel-hash tools/tune/tune_core.hh:135
#tuning-data-ledger-file tools/tune/ledger.hh:4

docs/design/vendor-independence.md

Section Cited by
#info-spans-on-syev-gesvd-and-steqr-forwarder-or-default include/batchlas/blas/extensions.hh:909
include/batchlas/blas/functions/gesvd.hh:144
include/batchlas/blas/functions/syev.hh:128
include/batchlas/blas/functions/syev.hh:92
#per-item-info-spans-for-potrf-getrf-and-getri include/batchlas/blas/functions/getrf.hh:108
include/batchlas/blas/functions/getri.hh:109
include/batchlas/blas/functions/potrf.hh:125
#positional-validators-reject-only-what-no-route-can-serve include/batchlas/blas/functions/geqrf.hh:55
include/batchlas/blas/functions/getrf.hh:57
include/batchlas/blas/functions/getri.hh:55
include/batchlas/blas/functions/getrs.hh:61
include/batchlas/blas/functions/orgqr.hh:53
include/batchlas/blas/functions/potrf.hh:55
#the-entry-point-facade include/batchlas/blas/functions/gemm.hh:82
include/batchlas/blas/functions/gemv.hh:79
include/batchlas/blas/functions/geqrf.hh:111
include/batchlas/blas/functions/getrf.hh:145
include/batchlas/blas/functions/getri.hh:147
include/batchlas/blas/functions/getrs.hh:125
include/batchlas/blas/functions/hemm.hh:91
include/batchlas/blas/functions/her2k.hh:93
include/batchlas/blas/functions/herk.hh:90
include/batchlas/blas/functions/orgqr.hh:108
include/batchlas/blas/functions/potrf.hh:153
include/batchlas/blas/functions/spmm.hh:121
include/batchlas/blas/functions/symm.hh:95
include/batchlas/blas/functions/syr2k.hh:93
include/batchlas/blas/functions/syrk.hh:89
include/batchlas/blas/functions/trmm.hh:99
include/batchlas/blas/functions/trsm.hh:177
#the-vendor-gate-why-the-tile-route-predicate-is-per-backend-and-scalar src/select/vendor.hh:102
#vendor-independence-headers-keyed-on-the-library-axis src/linalg-impl.hh:11
src/linalg-impl.hh:895

docs/design/verification.md

Section Cited by
#verification-no-fp64-device src/util/matrix_checks.cc:218

docs/design/workspace.md

Section Cited by
#workspace-bumpallocator-sizing-mode include/batchlas/util/mempool.hh:47
#workspace-forward-declarations-in-workspacehh include/batchlas/util/workspace.hh:14
#workspace-lvalue-only-lease-accessors include/batchlas/util/workspace.hh:89
#workspace-release-on-out-of-order-queues include/batchlas/util/workspace.hh:110
#workspace-the-bumpallocator-alignment-trap include/batchlas/util/mempool.hh:29
#workspace-the-reassignment-ratchet include/batchlas/util/workspace.hh:67

docs/developer/agent-guide.md

Section Cited by
(whole page) cmake/BatchLASCcache.cmake:16
tests/cmake/BatchLASTestGpus.cmake:9

docs/developer/documentation.md

Section Cited by
#building-the-site cmake/BatchLASDocs.cmake:3

docs/perf/blackwell.md

Section Cited by
(whole page) include/batchlas/util/sycl-device-queue.hh:168
#cfloat-spmm-under-the-precise-fp-model src/ops/spmm/kernels/spmm_native.cc:76
tests/spmm_tests.cc:940
#compute-capability-key tests/util_device_queue_tests.cc:255
#device-call-guard benchmarks/benchviz/store.py:189
scripts/check_device_calls.py:14
tests/CMakeLists.txt:555
#icpx-fast-fp-model cmake/BatchLASOptions.cmake:365
#lu-getrs-fused src/ops/getrs/kernels/getrs_fused.cc:96
#lu-getrs-fused-sm89-rows src/ops/getrs/kernels/getrs_fused.cc:68
#potrf-lpanel-vector-sb src/ops/potrf/kernels/potrf_lpanel_device.hh:37
src/ops/potrf/kernels/potrf_lpanel_device.hh:95
#trsm-sub-group-left-kernel src/ops/trsm/kernels/trsm_sg_left.cc:196
src/ops/trsm/kernels/trsm_sg_left.cc:61
src/ops/trsm/kernels/trsm_sg_left.cc:75
tests/trsm_tests.cc:728
#trsm-v1-ladder-cap src/ops/trsm/trsm_native.hh:34
tests/trsm_tests.cc:463
tests/trsm_tests.cc:919

docs/perf/dispatch.md

Section Cited by
#dispatch-buffer-size-queries-and-the-route-they-size src/backends/cusolver.cc:57
#dispatch-the-coverage-instrument src/select/coverage.cc:157
src/select/coverage.hh:6
#instrument-defects src/select/coverage.cc:38

docs/perf/gemm.md

Section Cited by
(whole page) src/ops/gemm/choice.hh:4
#choices-flat-selection-p34 src/ops/gemm/gemm.cc:90
#gemm-the-heterogeneous-batch-loop src/backends/gemm_heterogeneous.hh:7
#gemm-the-pod-device-scalar src/sycl/device_scalar.hh:135
src/sycl/device_scalar.hh:171
src/sycl/device_scalar.hh:6
src/util/matrix_checks.cc:60
#gemm-the-register-tiled-launcher-table src/ops/gemm/kernels/register_launchers.hh:13
#the-128x128-float-kernel src/ops/gemm/kernels/register_128x128.hh:18
#the-small-batched-kernel src/ops/gemm/kernels/small_batched.hh:6
#the-small-tiled-kernel src/ops/gemm/kernels/small_batched.hh:134
#the-staged-epilogue src/ops/gemm/kernels/register_128x128.hh:379
#the-strided-ld-defect-and-the-routing-fix src/ops/geqrf/kernels/geqrf_blocked.cc:250
src/ops/getrf/getrf.cc:71
src/ops/trsm/trsm.cc:72
#the-subgroup-workspace-budget cmake/BatchLASDetectSYCL.cmake:75
cmake/device_limits.h.in:13
#the-wide-scalar-kernel src/ops/gemm/kernels/register_64x64_k16_wide.hh:8
#wide-scalar-transposed-tiles src/ops/gemm/kernels/register_wide_transposed.hh:9

docs/perf/gemv.md

Section Cited by
(whole page) src/ops/gemv/choice.hh:4
src/ops/gemv/gemv.cc:4
#blind-guards-found-and-closed tests/gemv_tests.cc:225
#gemv-correctness-findings tests/gemv_tests.cc:417
#kernel-hypotheses-refuted src/ops/gemv/kernels/gemv_native.cc:546
#the-body-4-gate src/ops/gemv/kernels/gemv_native.cc:76
tests/gemv_tests.cc:656
#the-body-5-gates src/ops/gemv/kernels/gemv_native.cc:105
src/ops/gemv/kernels/gemv_native.cc:97
tests/gemv_tests.cc:773
tests/gemv_tests.cc:952
#the-five-kernel-bodies src/ops/gemv/gemv_native.hh:4
src/ops/gemv/kernels/gemv_native.cc:3
#the-sub-route-gates include/batchlas/util/sycl-device-queue.hh:246
src/ops/gemv/kernels/gemv_native.cc:114
src/ops/gemv/kernels/gemv_native.cc:89

docs/perf/gesvd.md

Section Cited by
(whole page) src/ops/gesvd/choice.hh:4
#gesvd-bdsdc-null-space-repair-criteria-that-failed src/ops/gesvd/bdsdc/bdsdc.cc:232
#gesvd-bdsdc-the-repair-threshold-from-the-resolution-limit src/ops/gesvd/bdsdc/bdsdc.cc:256
#gesvd-bdsqr-the-missing-zero-shift-chase src/ops/gesvd/bdsqr/bdsqr.cc:170
#gesvd-defect-a-the-normal-equations-square-kappa src/ops/gesvd/kernels/gesvd_blocked.cc:918
tests/gesvdj_cta_tests.cc:5
#gesvd-defect-c-an-accuracy-harness-that-could-not-see-relative-error benchmarks/gesvd_blocked_acc.cc:238
benchmarks/gesvd_cta_acc.cc:238
benchmarks/gesvd_relacc.cc:5
#gesvd-defect-d-perf-sweeps-that-stopped-at-batch-64 benchmarks/gesvd_vendor_benchmark.cc:8
#gesvd-the-gebrd-panel-width-split-from-ormqr evaluation/tuning/generate_tuning_header.py:207
include/batchlas/tuning_params.hh:56
#gesvd-the-unblocked-gebrd-cliff src/ops/gesvd/kernels/gesvd_blocked.cc:40
#gesvd-tier-0-the-cusolver-gesvdjbatched-binding src/backends/cusolver.cc:345
#gesvd-tier-3-bdsdc-as-the-bidiagonal-solver include/batchlas/settings.hh:137
src/ops/gesvd/kernels/gesvd_blocked.cc:53
#gesvdj_cta-tier-2-preconditioning-tested-and-rejected src/ops/gesvd/kernels/gesvdj_cta.cc:372

docs/perf/iluk.md

Section Cited by
#iluk-host-numeric-phase-cost-findings src/ops/iluk/iluk.cc:152
src/ops/iluk/iluk.cc:369
src/ops/iluk/iluk.cc:409
#iluk-the-device-factorization src/ops/iluk/iluk.cc:577
src/ops/iluk/iluk.cc:898
#iluk-the-host-versus-device-crossover src/ops/iluk/iluk.cc:298

docs/perf/level3.md

Section Cited by
(whole page) src/ops/symm/choice.hh:4
src/ops/syr2k/choice.hh:4
src/ops/syrk/choice.hh:4
src/ops/trmm/choice.hh:4
#herk-and-her2k-the-gemm-plus-fold-crossovers src/backends/cublas.cc:264
src/expansion_budget.hh:73
#herk-on-the-gram-tile-kernel src/backends/cublas.cc:364
#level-3-generic-syrk-and-syr2k-fallbacks-fold-one-triangle src/ops/potrf/kernels/symmetric_product_fold.hh:7
#level-3-one-set-of-shape-validators src/backends/level3_shape.hh:14
#level-3-scratch-expansions-and-their-ceilings src/expansion_budget.hh:19
src/kernels/common/triangular_expand.hh:134
src/kernels/common/triangular_expand.hh:22
#level-3-the-complex-mac-and-return-by-value-rules src/kernels/common/triangular_tiles.hh:42
src/kernels/common/triangular_tiles.hh:72
#symm-and-hemm-expansion-crossover src/backends/cublas.cc:197
src/kernels/common/triangular_expand.hh:27
#syrk-and-syr2k-triangular-tiles-kernel-design src/kernels/common/triangular_tiles.hh:5
src/ops/syr2k/kernels/syr2k_triangular_tiles.hh:9
src/ops/syrk/kernels/syrk_triangular_tiles.hh:7
#syrk-gram-tiles-kernel-design src/ops/syrk/kernels/syrk_gram_tiles.hh:289
src/ops/syrk/kernels/syrk_gram_tiles.hh:74
src/ops/syrk/kernels/syrk_gram_tiles.hh:8
#the-band-split-syrk-bug src/ops/syrk/kernels/syrk_gram_tiles.hh:174
#trmm-choosing-the-row-tile-by-scalar-type src/ops/trmm/kernels/trmm_triangular_tiles.hh:317
#trmm-triangular-tiles-kernel-design src/ops/trmm/kernels/trmm_triangular_tiles.hh:57
src/ops/trmm/kernels/trmm_triangular_tiles.hh:7

docs/perf/lu.md

Section Cited by
(whole page) src/ops/getrf/choice.hh:4
src/ops/getrf/getrf_native.hh:6
src/ops/getri/choice.hh:4
src/ops/getrs/choice.hh:4
src/ops/getrs/getrs_native.hh:4
tests/getrf_tests.cc:7
#armed-breaks tests/getrf_tests.cc:2993
tests/getrf_tests.cc:3133
tests/getrf_tests.cc:3224
tests/getrf_tests.cc:3344
tests/getrf_tests.cc:3391
#armed-breaks-p4 tests/getrf_tests.cc:1075
#d3-why-cdouble-stops-at-n16 src/ops/gesv/kernels/gesv_tiny.cc:61
src/ops/getrf/kernels/getrf_tiny.cc:63
#getrf-deferred-left-gather src/kernels/common/lu_laswp.hh:83
tests/getrf_tests.cc:857
#getrf-window-evidence src/ops/getrf/kernels/getrf_blocked.cc:4
src/ops/getrf/kernels/getrf_cta.cc:6
#getri-window-evidence src/ops/getri/getri_native.hh:4
src/ops/getri/kernels/getri_blocked.cc:5
#getrs-collapsed-permutation src/ops/getrs/getrs_native.hh:46
src/ops/getrs/kernels/getrs_native.cc:38
tests/getrf_tests.cc:1840
#getrs-composition-window-evidence src/ops/getrs/kernels/getrs_native.cc:5
#getrs-fused-window-evidence tests/getrf_tests.cc:2855
#lu-correctness-findings src/ops/getri/getri.cc:42
src/ops/getrs/kernels/getrs_fused.cc:859
tests/getrf_tests.cc:1721
tests/getri_candidates_tests.cc:535
#lu-negative-results src/ops/getrf/kernels/getrf_cta.cc:71
#lu-p2-the-register-bound-is-assumed-not-probed src/ops/gesv/kernels/gesv_tiny.cc:55
#lu-the-48-kb-launch-hole src/kernels/common/lu_laswp.hh:88
src/ops/getrf/getrf_native.hh:102
src/ops/getrf/kernels/getrf_cta.cc:60
src/ops/getrs/kernels/getrs_fused.cc:172
src/ops/getrs/kernels/getrs_native.cc:42
tests/getrf_tests.cc:572
#lu-the-occupancy-rule src/ops/getrf/getrf_native.hh:23
#lu-the-panel-leaf-is-not-the-tier-ceiling src/util/resident_capacity.hh:20
#lu-the-register-leaf-ab benchmarks/factor_bench.cc:194
src/ops/getrf/kernels/getrf_blocked.cc:85
src/ops/getrf/kernels/getrf_panel_reg.cc:4
tests/getrf_tests.cc:1481
#lu-the-tiny-tier src/ops/getrf/kernels/getrf_tiny.cc:3
#one-spelling-per-ceiling src/ops/getrf/getrf_native.hh:33
src/ops/getrf/kernels/getrf_cta.cc:34
#packed-resident-leaves tests/resident_capacity_tests.cc:306
#pad-rows-and-the-argmax-corrected src/kernels/common/tiny_device.hh:229
src/ops/getrf/kernels/getrf_panel_reg_device.hh:65
src/ops/getrf/kernels/getrf_tiny.cc:133
tests/getrf_tests.cc:3247
#the-column-bucket src/CMakeLists.txt:65
src/kernels/common/tiny_device.hh:152
src/ops/gesv/kernels/gesv_tiny.cc:376
src/ops/getrf/kernels/getrf_tiny.cc:117
src/ops/getrf/kernels/getrf_tiny.cc:271
#the-fused-gesv-tier include/batchlas/blas/functions/gesv.hh:9
src/kernels/common/solve_native.hh:5
src/ops/gesv/choice.hh:4
src/ops/gesv/kernels/gesv_tiny.cc:5
tests/gesv_tests.cc:5
#the-fused-narrow-rhs-getrs src/ops/getrs/kernels/getrs_fused.cc:5
tests/getrf_tests.cc:448
#the-host-dgetrf-oracle-is-broken-on-this-box include/batchlas/verify/reference.hh:106
include/batchlas/verify/reference.hh:42
tests/getrf_tests.cc:3524
#the-laswp-gather src/kernels/common/lu_laswp.hh:7
#the-local-memory-broadcast src/kernels/common/tiny_device.hh:158
src/ops/gesv/kernels/gesv_tiny.cc:80
src/ops/getrf/kernels/getrf_tiny.cc:73
tests/getrf_tests.cc:3316
tests/getrf_tests.cc:3624
#the-pivot-key-encoding src/kernels/common/tiny_device.hh:241
#the-pivot-margin-gate-on-elementwise-comparisons tests/getrf_tests.cc:3072
#the-register-cap-that-binds-is-per-sub-partition scripts/register_probe.sh:42
src/ops/geqrf/kernels/geqrf_tiny_device.hh:104
src/ops/getrf/kernels/getrf_panel_reg.cc:37
src/ops/getrs/kernels/getrs_fused.cc:156
src/ops/getrs/kernels/getrs_fused.cc:53
src/ops/posv/kernels/posv_tiny.cc:51
src/ops/trsm/kernels/trsm_native.cc:94
src/util/resident_capacity.hh:74
#the-register-panel-leaf src/ops/getrf/getrf_native.hh:117
src/ops/getrf/kernels/getrf_blocked.cc:170
src/ops/getrf/kernels/getrf_panel_reg.cc:6
src/ops/getrf/kernels/getrf_panel_reg_device.hh:8
#the-register-panel-leaf-register-probe src/ops/getrf/kernels/getrf_panel_reg.cc:46
src/ops/getrf/kernels/getrf_panel_reg.cc:59
src/ops/getrf/kernels/getrf_panel_reg_device.hh:25
tests/getrf_tests.cc:1403
#the-register-probe-and-the-unroll-that-decides-it scripts/register_probe.sh:70
src/ops/gesv/kernels/gesv_tiny.cc:151
src/ops/getrf/kernels/getrf_tiny.cc:127
#the-register-probe-at-wg--64 src/ops/getrf/kernels/getrf_tiny.cc:47
#the-rhs-pad-column-cost src/ops/gesv/kernels/gesv_tiny.cc:278
src/ops/gesv/kernels/gesv_tiny.cc:69
#the-right-hand-gather src/kernels/common/lu_laswp.hh:243
src/ops/getrf/kernels/getrf_blocked.cc:74
tests/getrf_tests.cc:930
#the-shape-of-the-getrf-device-code src/ops/getrf/getrf_cta_device.hh:52
src/ops/getrf/getrf_cta_device.hh:6
#the-work-group-ab src/ops/getrf/kernels/getrf_tiny.cc:42
#two-barriers-per-column src/ops/getrf/kernels/getrf_panel_reg_device.hh:127
#why-the-rank-1-update-starts-at-k--j--1 src/ops/getrf/kernels/getrf_panel_reg_device.hh:119
src/ops/getrf/kernels/getrf_tiny.cc:186
#why-the-tiny-vs-cta-element-bound-is-relative tests/getrf_tests.cc:3177

docs/perf/ortho.md

Section Cited by
#ortho-host-devices-force-householder src/ops/ortho/ortho.cc:71
#ortho-the-gram-matrix-through-syrk-per-precision src/ops/ortho/ortho.cc:163

docs/perf/potrf.md

Section Cited by
(whole page) src/ops/potrf/choice.hh:4
src/ops/potrf/kernels/potrf_cta.cc:5
src/ops/potrf/kernels/potrf_cta_device.hh:5
src/ops/potrf/potrf_native.hh:4
#armed-breaks-r9--lpanel tests/potrf_lpanel_cases.inc:113
tests/potrf_lpanel_cases.inc:163
#lpanel-register-gate src/ops/potrf/kernels/potrf_lpanel.cc:38
#nb-and-w src/ops/potrf/kernels/potrf_blocked.cc:42
#posv-selection-since-flat-kernel-selection-phase-3 src/ops/posv/choice.hh:5
#posv-the-backward-solve-costs-a-transpose src/ops/posv/kernels/posv_tiny.cc:9
#potrf-correctness-findings tests/potrf_tests.cc:6
#potrf-the-48-kb-launch-hole src/ops/potrf/potrf_slm_hole.hh:4
tests/potrf_tests.cc:257
#potrf-the-occupancy-rule src/ops/potrf/potrf_native.hh:45
tests/resident_capacity_tests.cc:4
#potrf-the-tiny-tier src/ops/potrf/kernels/potrf_tiny.cc:127
src/ops/potrf/kernels/potrf_tiny.cc:44
src/ops/potrf/kernels/potrf_tiny.cc:5
src/ops/potrf/kernels/potrf_tiny.cc:52
src/ops/potrf/potrf_native.hh:20
tests/potrf_tiny_cases.inc:6
#register-gate src/ops/potrf/kernels/potrf_cta.cc:38
src/ops/potrf/kernels/potrf_cta_device.hh:16
#the-blocked-driver tests/potrf_tests.cc:735
#the-contiguity-rule-and-the-synthetic-table tests/resident_capacity_tests.cc:51
#the-cta-comparison-is-elementwise-not-bit-exact tests/potrf_lpanel_cases.inc:206
tests/potrf_tiny_cases.inc:151
#the-fused-posv-tier include/batchlas/blas/functions/posv.hh:9
src/ops/posv/kernels/posv_tiny.cc:4
tests/posv_tests.cc:5
#the-g-packing-helper-and-its-three-limits tests/resident_capacity_tests.cc:134
#the-l-ladder src/ops/potrf/kernels/potrf_cta.cc:97
#the-lpanel-tier src/ops/potrf/kernels/potrf_lpanel.cc:5
src/ops/potrf/kernels/potrf_lpanel_device.hh:13
src/ops/potrf/potrf_native.hh:58
tests/potrf_lpanel_cases.inc:4
tests/potrf_lpanel_cases.inc:516
#the-occupancy-clamp-on-nb src/ops/potrf/kernels/potrf_blocked.cc:58
tests/potrf_tests.cc:206
#the-pivot-cell-war-race src/ops/potrf/kernels/potrf_lpanel_device.hh:150
#the-posv-local-memory-transpose src/ops/posv/kernels/posv_tiny.cc:376
src/ops/posv/kernels/posv_tiny.cc:74
#the-posv-tiny-tier-is-not-register-resident-for-cdouble src/ops/posv/kernels/posv_tiny.cc:48
#the-shared-helper-budget-slice-walk-and-pack src/util/resident_capacity.hh:5
#the-shared-tiny-tier-invariants src/kernels/common/tiny_device.hh:55
src/kernels/common/tiny_device.hh:9
#the-shipped-ceilings-pinned-in-both-scales tests/resident_capacity_tests.cc:174
#the-slm-budget-and-the-fit-ceilings src/ops/potrf/kernels/potrf_cta.cc:48
src/ops/potrf/kernels/potrf_cta.cc:57
tests/potrf_tests.cc:195
tests/potrf_tests.cc:697
#why-lpanel-may-pack-on-work-group-barriers src/ops/potrf/kernels/potrf_lpanel.cc:84

docs/perf/qr.md

Section Cited by
(whole page) src/ops/geqrf/choice.hh:4
src/ops/geqrf/kernels/geqrf_blocked.cc:4
src/ops/ormqr/choice.hh:4
tests/CMakeLists.txt:325
tests/geqrf_tests.cc:5
#a-residual-test-cannot-guard-a-convention src/ops/geqrf/geqrf_cta_device.hh:7
src/ops/geqrf/geqrf_panel_reg_device.hh:194
src/ops/geqrf/kernels/geqrf_tiny.cc:205
tests/geqrf_tests.cc:704
#block-width-evidence src/ops/geqrf/kernels/geqrf_blocked.cc:42
src/ops/orgqr/kernels/orgqr_blocked.cc:35
#break-sweeps-the-tiny-tier src/ops/geqrf/kernels/geqrf_tiny.cc:237
tests/geqrf_tests.cc:1356
tests/geqrf_tests.cc:1813
#cta-capacity src/ops/geqrf/kernels/geqrf_cta.cc:314
#larft_wyhh-pack_v-indexes-the-row-fastest src/kernels/common/larft_wy.hh:259
#larft_wyhh-usedevice-is-a-template-parameter src/kernels/common/larft_wy.hh:192
#larft_wyhh-why-the-wy-helpers-are-shared src/kernels/common/larft_wy.hh:7
#ormqr-one-route-resolution-for-the-call-and-its-size-query include/batchlas/blas/functions/ormqr.hh:115
#ormqr-wy-trmm-gate src/ops/ormqr/kernels/ormqr_blocked.cc:50
#qr-route-arms src/ops/geqrf/geqrf_native.hh:4
src/ops/geqrf/kernels/geqrf_cta.cc:5
src/ops/orgqr/kernels/orgqr_blocked.cc:101
#qr-the-occupancy-rule tests/geqrf_tests.cc:225
#qr-the-panel-leaf-is-not-the-tier-ceiling src/ops/geqrf/kernels/geqrf_cta.cc:350
#qr-the-register-leaf-ab include/batchlas/settings.hh:146
src/ops/geqrf/kernels/geqrf_blocked.cc:58
#qr-the-vendor-baseline src/ops/orgqr/kernels/orgqr_blocked.cc:2
#settling-the-target-on-the-tall-panels src/ops/geqrf/geqrf_native.hh:21
tests/geqrf_tests.cc:1073
#the-48-kib-launch-hole src/ops/geqrf/geqrf_cta_device.hh:156
src/ops/geqrf/geqrf_panel_reg_device.hh:115
src/ops/geqrf/kernels/geqrf_cta.cc:283
src/ops/geqrf/kernels/geqrf_cta.cc:42
tests/geqrf_tests.cc:340
#the-cdouble-n32-cell src/ops/geqrf/kernels/geqrf_tiny.cc:83
src/ops/geqrf/kernels/geqrf_tiny_device.hh:75
#the-chunk-width-and-the-one-cell-that-is-pinned src/ops/geqrf/kernels/geqrf_tiny_device.hh:159
src/ops/geqrf/kernels/geqrf_tiny_device.hh:191
src/ops/geqrf/kernels/geqrf_tiny_device.hh:63
#the-fixtures-tolerance-floor-and-why-it-is-new tests/geqrf_tests.cc:109
tests/geqrf_tests.cc:1590
tests/geqrf_tests.cc:1631
#the-height-ceiling-the-aot-probe-got-wrong src/ops/geqrf/geqrf_panel_reg_device.hh:89
tests/geqrf_tests.cc:2012
#the-launch-shape-64-work-items-not-128 src/ops/geqrf/kernels/geqrf_tiny_device.hh:139
#the-orgqr_buffer_size-latent-defect src/ops/orgqr/orgqr.cc:79
#the-panel-height-window src/ops/geqrf/geqrf_native.hh:69
src/ops/geqrf/geqrf_native.hh:81
src/ops/geqrf/geqrf_panel_reg_device.hh:106
src/ops/geqrf/kernels/geqrf_blocked.cc:228
src/ops/geqrf/kernels/geqrf_cta.cc:227
src/ops/geqrf/kernels/geqrf_cta.cc:390
src/ops/geqrf/kernels/geqrf_cta.cc:526
tests/geqrf_tests.cc:2200
#the-register-panel-leaf-width-and-height-table src/ops/geqrf/geqrf_panel_reg_device.hh:22
src/ops/geqrf/geqrf_panel_reg_device.hh:57
#the-register-panel-leaf-wp6--p5 src/ops/geqrf/geqrf_panel_reg_device.hh:6
tests/geqrf_tests.cc:1915
#the-shipped-orgqr-ceiling src/ops/orgqr/choice.hh:4
src/ops/orgqr/orgqr_native.hh:5
#the-stack-frame-is-the-gate-for-this-kernel-not-the-spill-counter src/ops/geqrf/geqrf_panel_reg_device.hh:143
src/ops/geqrf/kernels/geqrf_tiny.cc:163
#the-synthesis-pass-four-more-breaks tests/geqrf_tests.cc:1651
tests/geqrf_tests.cc:1714
tests/geqrf_tests.cc:1836
#the-tiny-ceiling-predicate src/ops/geqrf/geqrf_native.hh:37
#the-tiny-tier-register-table src/ops/geqrf/kernels/geqrf_tiny.cc:36
src/ops/geqrf/kernels/geqrf_tiny_device.hh:182
src/ops/geqrf/kernels/geqrf_tiny_device.hh:42
src/ops/geqrf/kernels/geqrf_tiny_device.hh:87
#the-tiny-tier-wp6--p1-square-n--32-in-registers src/ops/geqrf/kernels/geqrf_tiny.cc:3
#the-two-partition-butterflies src/ops/geqrf/kernels/geqrf_tiny.cc:174
src/ops/geqrf/kernels/geqrf_tiny_device.hh:121
src/ops/geqrf/kernels/geqrf_tiny_device.hh:21
src/ops/geqrf/kernels/geqrf_tiny_device.hh:50
#why-bit-identity-against-the-cta-route-is-not-the-padding-test tests/geqrf_tests.cc:1432
tests/geqrf_tests.cc:2046
#why-the-elementwise-scan-is-the-m-x-n-window-only tests/geqrf_tests.cc:951

docs/perf/small-n-baseline.md

Section Cited by
#geqrf src/ops/geqrf/kernels/geqrf_blocked.cc:116
#the-arms-list benchmarks/factor_bench.cc:290
#warm-up-order-and-the-variance-gate benchmarks/factor_bench.cc:492
tools/tune/cell_runner.hh:131

docs/perf/spmm.md

Section Cited by
(whole page) src/ops/spmm/choice.hh:4
#measurement-harness-and-hygiene benchmarks/spmm_benchmark.cc:6
#spmm-correctness-findings tests/spmm_tests.cc:27
#spmm-the-kernel-contract src/ops/spmm/kernels/spmm_native.cc:3
src/ops/spmm/spmm_native.hh:6
#supports-and-what-is-deliberately-not-in-it src/ops/spmm/spmm_native.hh:24
#the-eleventh-blind-guard tests/spmm_tests.cc:160
tests/spmm_tests.cc:569
#the-gather-window src/ops/spmm/kernels/spmm_native.cc:49
#three-vendor-defects-found-here-and-fixed src/backends/cusparse.cc:28
src/backends/cusparse.cc:47
tests/spmm_tests.cc:533

docs/perf/stedc.md

Section Cited by
#stedc-convergence-reporting-through-info src/ops/syev/stedc/stedc.cc:28
src/ops/syev/stedc/stedc_internal.hh:11
src/ops/syev/stedc/stedc_secular.hh:11
#stedc-current-tuning-values evaluation/tuning/generate_tuning_header.py:265
evaluation/tuning/generate_tuning_header.py:273
include/batchlas/tuning_params.hh:114
include/batchlas/tuning_params.hh:122
#stedc-deflation-aware-back-transform-gemm src/ops/syev/stedc/stedc.cc:454
#stedc-eigenvalues-only-still-builds-eigenvectors src/ops/syev/stedc/stedc.cc:534
#stedc-open-debts src/ops/syev/stedc/stedc_secular.cc:685
#stedc-partition-parallel-rescale-and-normalize src/ops/syev/stedc/stedc_merge_kernels.hh:51
#stedc-the-absolute-deflation-tolerance src/ops/syev/stedc/stedc.cc:207
src/ops/syev/stedc/stedc.cc:376
#stedc-the-dd--1-secular-solve-fault src/ops/syev/stedc/stedc_merge_cta.cc:437
#stedc-the-deflation-gemm-threshold src/ops/syev/stedc/stedc.cc:70
#stedc-the-leaf-cap-at-the-sub-group-width src/ops/syev/stedc/stedc_levels_plan.hh:27
#stedc-the-level-synchronous-driver src/ops/syev/stedc/stedc.cc:103
src/ops/syev/stedc/stedc.cc:591
src/ops/syev/stedc/stedc.cc:789
src/ops/syev/stedc/stedc.cc:840
#stedc-the-rejected-private-pole-fast-path src/ops/syev/stedc/stedc_merge_kernels.cc:60
#stedc-the-work-group-multiplier-after-the-barriers-went src/ops/syev/stedc/stedc_merge_cta.cc:37

docs/perf/steqr.md

Section Cited by
#cta-steqr-chase-micro-structure-decisions src/ops/syev/steqr/steqr_cta_device.hh:126
src/ops/syev/steqr/steqr_cta_device.hh:24
src/ops/syev/steqr/steqr_cta_device.hh:361
src/ops/syev/steqr/steqr_cta_device.hh:6
src/ops/syev/steqr/steqr_cta_device.hh:618
#full-warp-partition src/kernels/common/sg_partition/backend_nvptx.hh:13
src/kernels/common/sg_partition/sg_partition.hh:18
src/ops/syev/steqr/steqr_cta_device.hh:980
#lockstep-flat-solver src/ops/syev/steqr/steqr_cta_device.hh:689
src/ops/syev/steqr/steqr_cta_device.hh:969
#work-group-multiplier src/ops/syev/kernels/syev_cta_fused.cc:39
src/ops/syev/steqr/steqr_cta.cc:32

docs/perf/syev.md

Section Cited by
(whole page) src/ops/syev/choice.hh:4
#syev-complex-panel-width-in-the-256-to-512-bucket src/ops/syev/kernels/syev_blocked.cc:57
#syev-her2k-trailing-update-for-complex-float-wp3 src/ops/syev/sytrd/sytrd_blocked.cc:714
#syev-latrd-grid-gate-confirmed-in-eigenvector-mode include/batchlas/settings.hh:256
#syev-open-debt-stebz-reports-no-convergence-status src/ops/syev/kernels/syev_two_stage.cc:253
#syev-stebz-values-only-in-the-blocked-solver-wp1 src/ops/syev/kernels/syev_blocked.cc:229
#syev-the-2026-08-03-small-n-bake-off src/ops/syev/kernels/syev_jacobi_cta.cc:32
#syev-the-2026-08-07-constant-retune evaluation/tuning/generate_tuning_header.py:217
include/batchlas/tuning_params.hh:66
#syev-the-fused-cta-kernel-design src/ops/syev/kernels/syev_cta_fused.cc:267
src/ops/syev/kernels/syev_cta_fused.cc:52
src/ops/syev/kernels/syev_cta_fused.cc:95
#syev-the-jacobi-kernel-design-and-its-accuracy-argument src/ops/syev/kernels/syev_jacobi_cta.cc:31
#syev-the-two-stage-band-width-kd src/kernels/common/two_stage_common.hh:32
src/kernels/common/two_stage_common.hh:49
#syev-the-upper-to-lower-mirror-for-lower-only-providers src/ops/syev/uplo_mirror.hh:4

docs/perf/syevx.md

Section Cited by
#lobpcg-column-reversal-kernels-removed src/ops/syevx/kernels/syevx_lobpcg.cc:33
#lobpcg-device-staged-instrumentation src/ops/syevx/kernels/syevx_lobpcg.cc:728
src/ops/syevx/kernels/syevx_lobpcg.cc:98
tests/syevx_tests.cc:794
#lobpcg-guard-vectors-by-default benchmarks/syevx_benchmark.cc:224
#lobpcg-host-synchronization-only-at-convergence-checks src/ops/syevx/kernels/syevx_filtered.cc:16
src/ops/syevx/kernels/syevx_lobpcg.cc:772
src/ops/syevx/kernels/syevx_lobpcg.cc:92
#lobpcg-jacobi-preconditioners include/batchlas/blas/enums.hh:447
tests/syevx_tests.cc:997
#lobpcg-soft-locking-by-column-masking src/ops/syevx/kernels/syevx_lobpcg.cc:144
src/ops/syevx/kernels/syevx_lobpcg.cc:856
tests/syevx_tests.cc:1351
#lobpcg-the-accuracy-pin-on-the-projected-syev src/ops/syevx/kernels/syevx_lobpcg.cc:179
tests/syevx_tests.cc:1205
#lobpcg-x-only-fill-and-power-iteration-start src/ops/syevx/kernels/syevx_lobpcg.cc:517
src/ops/syevx/kernels/syevx_lobpcg.cc:59
#syevx-automatic-chebyshev-filter-degree include/batchlas/settings.hh:202
src/ops/syevx/kernels/syevx_filtered.cc:117
src/ops/syevx/kernels/syevx_filtered.cc:54
#syevx-directsubset-batch-crossover-with-eigenvectors src/ops/syevx/syevx.cc:312
#syevx-routing-thresholds-as-they-stand src/ops/syevx/kernels/syevx_direct.cc:4
src/ops/syevx/syevx.cc:4
#syevx-the-diagonal-boost-gap-sweep-and-lobpcg-stagnation benchmarks/syevx_benchmark.cc:399
#syevx-the-direct-baseline-was-never-cusolver src/ops/syevx/syevx.cc:37
#syevx-the-first-gpu-crossover-measurement-superseded-baseline benchmarks/syevx_benchmark.cc:62
src/ops/syevx/syevx.cc:320
#syevx-the-parallel-gershgorin-bounds-kernel src/ops/syevx/kernels/syevx_filtered.cc:199
#syevx-tier-1-stebz-and-stein src/ops/syev/tridiag/stein.cc:6
#syevx-tier-2-directsubset src/ops/syevx/kernels/syevx_direct_subset.cc:12
#syevx-tier-3-filtered src/ops/syevx/kernels/syevx_filtered.cc:3

docs/perf/sytrd.md

Section Cited by
#sytrd-complex-multiply-in-the-latrd-symv src/ops/syev/sytrd/latrd_lower_panel.cc:106
src/ops/syev/sytrd/latrd_lower_panel.cc:73
#sytrd-her2k-scratch-fit-headroom src/ops/syev/sytrd/sytrd_blocked.cc:765
#sytrd-no-symmetrize-after-the-rank-2k-update src/ops/syev/sytrd/sytrd_blocked.cc:786
#sytrd-open-debts src/ops/syev/sytrd/sytrd_sb2st_cta.cc:66
#sytrd-rejected-latrd-symv-loop-shapes src/ops/syev/sytrd/latrd_lower_panel.cc:371
#sytrd-sb2st-back-transform-sub-groups-per-work-group src/ops/syev/sytrd/sytrd_sb2st_hh.cc:762
#sytrd-sb2st-back-transform-tile-width src/ops/syev/sytrd/sytrd_sb2st_hh.cc:798
#sytrd-sb2st-per-type-wave-geometry src/ops/syev/sytrd/sytrd_sb2st_hh.cc:674
#sytrd-the-dense-to-band-ormqr-block-width-hint src/ops/syev/sytrd/sytrd_sy2sb.cc:36
#sytrd-the-householder-chase-against-the-givens-chase src/kernels/common/two_stage_common.hh:40
src/ops/syev/kernels/syev_two_stage.cc:169
src/ops/syev/sytrd/sytrd_sb2st_hh.hh:6
#sytrd-the-householder-chase-work-group-shape src/ops/syev/sytrd/sytrd_sb2st_hh.cc:97
#sytrd-the-latrd-grid-kernel-and-its-co-residency-cap src/ops/syev/sytrd/latrd_lower_panel.cc:159
src/ops/syev/sytrd/latrd_lower_panel.cc:485
#sytrd-the-latrd-grid-path-threshold src/ops/syev/sytrd/latrd_lower_panel.cc:51
#sytrd-the-latrd-implementation-selector src/ops/syev/sytrd/latrd_lower_panel.cc:36
src/ops/syev/sytrd/sytrd_blocked.cc:40
#sytrd-the-q2-back-transform-design src/ops/syev/kernels/syev_two_stage.cc:275
src/ops/syev/sytrd/sytrd_sb2st_hh.cc:442
src/ops/syev/sytrd/sytrd_sb2st_hh.cc:462
src/ops/syev/sytrd/sytrd_sb2st_hh.cc:570
#sytrd-the-rank-2k-trailing-update-in-the-blocked-reduction src/ops/syev/sytrd/sytrd_blocked.cc:713
#sytrd-the-sb2st-tuning-precedence-trap src/ops/syev/sytrd/sytrd_sb2st_hh.cc:746
#sytrd-the-sb2st-wave-knob-spelling-history src/ops/syev/sytrd/sytrd_sb2st_hh.cc:699
src/ops/syev/sytrd/sytrd_sb2st_hh.cc:732
#sytrd-the-vacuous-latrd-grid-groups-ab src/ops/syev/sytrd/latrd_lower_panel.cc:1016

docs/perf/trsm.md

Section Cited by
(whole page) src/ops/trsm/choice.hh:4
#design-v1-v2-and-the-canonical-fold src/ops/trsm/trsm_canonical.hh:5
tests/trsm_tests.cc:222
#rejected-outer_nb-of-128-for-sideright src/ops/trsm/kernels/trsm_native.cc:348
#rejected-the-n64-cta-bucket src/ops/trsm/kernels/trsm_native.cc:40
#the-bucket-ladder-that-truncated src/ops/trsm/kernels/trsm_native.cc:31
tests/trsm_tests.cc:367
#the-final-grid-after-the-routed-trailing-gemm src/ops/trsm/trsm_native.hh:64
#the-measured-grid benchmarks/trsm_benchmark.cc:43
#the-missing-group-barrier src/ops/trsm/kernels/trsm_native.cc:184
tests/trsm_tests.cc:452
#the-register-gate-and-the-cta-capacity src/ops/trsm/kernels/trsm_native.cc:494
tests/trsm_tests.cc:356
#the-sideleft-staging-tile src/ops/trsm/kernels/trsm_native.cc:51
#the-two-level-blocked-driver src/ops/trsm/kernels/trsm_native.cc:344
tests/trsm_tests.cc:618
#tuning-knobs-and-environment src/ops/trsm/kernels/trsm_native.cc:352
tests/trsm_tests.cc:657

docs/perf/tuning.md

Section Cited by
#tuning-knobs-that-shadow-each-other evaluation/tuning/generate_tuning_header.py:232
include/batchlas/tuning_params.hh:81
#tuning-regenerating-the-header evaluation/tuning/generate_tuning_header.py:167
include/batchlas/settings.hh:347
include/batchlas/tuning_params.hh:17
#tuning-runtime-overrides-of-the-constants evaluation/tuning/generate_tuning_header.py:187
include/batchlas/tuning_params.hh:36