feat(KERNEL-QUANT-CIQ-GEMM-ROCM): land the W1 keep-quant providers on kROCM - #2782
Open
ghazni101 wants to merge 4 commits into
Open
feat(KERNEL-QUANT-CIQ-GEMM-ROCM): land the W1 keep-quant providers on kROCM#2782ghazni101 wants to merge 4 commits into
ghazni101 wants to merge 4 commits into
Conversation
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 3, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 3, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
This was referenced Sep 3, 2026
ghazni101
force-pushed
the
row/GFX1100-TG200-F1
branch
from
September 3, 2026 19:24
dc036a4 to
29eebfc
Compare
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 3, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 3, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
5 tasks
This was referenced Sep 4, 2026
ghazni101
force-pushed
the
row/GFX1100-TG200-F1
branch
from
September 4, 2026 07:34
29eebfc to
73478f6
Compare
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 4, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 4, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
Adds the VT_GEMV_MMVQ=1 opt-in K-quant decode GEMV arm for MatmulBTQuant, bit-exact vs the CPU oracle. The arm folds activation quant into the MMVQ GEMV prologue (deleting the standalone QuantizeQ8KK launch) and widens the gate to engine dtypes (bf16/f16 activations, bf16/f32 outputs). Sub-levers: - lever-B1: VT_GEMV_MMVQ_FOLD_MAX makes the fold crossover tunable at runtime - lever-B2: VT_SKINNY_BF16=1 f32-out decode-skinny arm for GDN BA projections - repair: m-gates the whole dispatch and makes the GEMV bit-equal to baseline - repair-2: host-side dispatch-route counters + F1/F2 routing-witness gates - lever-B2 test: red-first f32-out decode-skinny gate, true-unset routing window Architecture: F1 moved the live MatmulBTQuantKernelRocm to rocm_quant_dot.hip (anonymous namespace, internal linkage). T4a's MMVQ arm lives in rocm_grouped_gemm.hip's version (external linkage, renamed to *Gdn). This PR adds delegation: rocm_quant_dot.hip forwards Q4_K/Q5_K/Q6_K calls to the Gdn version, preserving F1's IQ-type providers while activating T4a's MMVQ arm. The default path (VT_GEMV_MMVQ unset) is byte-unchanged from F1. The arm is opt-in and validated by test_rocm_quant_dot (6/6 cases, 719 assertions) and test_rocm_skinny_f32 (2/2 cases, 51 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Rebased onto the external-contributor landing branch (staging tip b9f2ef4), which already carries F1 (mudler#2782) and its grouped-Q8_0 repair (mudler#2927). This version carries NONE of the removals the previous base commit 51f5222 made: the ten-row kCases table, kMaxNmseErr/nmse_ref_max and the three F1 provider cases are restored beside this arm's kKQuantCases (mudler#2938); Dp4a keeps the __ockl_sdot4 hardware dot (mudler#2939); the documented VT_ROCM_Q8K_BLOCK selector (SelectQ8KQuantArm/LaunchQ8KQuantizer), the mudler#2472 cooperative gfx1100 default (QuantizeQ8KCooperativeK) and the VT_ROCM_Q6K_SMALL_PRIVATE A/B arm are restored with their witness helpers; the shared bench-evidence file keeps lever B1's section 14 record, whose truncation this branch had carried. Depends on mudler#2782 (F1 keep-quant GEMM infra). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP] Assisted-by: AGENT:OMEN-ALPHA [OMP]
ghazni101
added a commit
to ghazni101/vllm.cpp
that referenced
this pull request
Sep 6, 2026
…Norm epilogue Lever-C adds an opt-in fused norm-quant epilogue (VT_NORM_QUANT_FUSED=1): RmsNormRowKernel emits the row's Q8_K blocks alongside its normal output, and MatmulBTQuant's K-quant branch skips the standalone QuantizeQ8KK when the consuming activation matches the producer token. Byte-identical to the standalone path by construction (shared QuantQ8KSBlock body). New files: - src/vt/rocm/rocm_act_quant.h: shared Q8_K quant-block body - src/vt/rocm/rocm_norm_quant_bridge.h: producer-consumer token contract Also fixes T4a routing counter placement (moved outside anonymous namespace for external linkage) and restores VT_GEMV_MMVQ_FOLD_MAX env var reading that was lost during cherry-pick conflict resolution. The default path (VT_NORM_QUANT_FUSED unset) is byte-unchanged. Validated by test_rocm_quant_dot (12/12 cases, 797 assertions). Token-identical to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0. Depends on mudler#2782 (F1) and mudler#2790 (T4a). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
The grouped arm delegates Q4_K/Q5_K/Q6_K to MatmulBTQuantGroupedKernelRocmGdn but not Q8_0, and Q8_0 has no arm in rocm_quant_dot.hip either, so a grouped Q8_0 expert GEMM throws on a path main serves. Add Q8_0 to the delegation list and a test case that fails when it is dropped. Addressing localai-org-maint-bot review on mudler#2782 (issue mudler#2927). FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: AGENT:GLM-5-2 [OMP]
Contributor
Author
|
Rebased onto The base is now FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true |
Contributor
|
The current
The landed #3001 fused provider also calls the old external grouped symbol twice at The grouped Q8_0 delegation omission is repaired in this head and should no longer be reported as missing. The current test source implies three cases and 145,303 assertions if every device loop runs; that is a derived expected count, not an execution result. The maintainer’s GLM hardware result also corrects the old blanket reachability objection. That opt-in production path reached the grouped provider and refused a dtype. The actual refused dtype remains unidentified, and the printed list includes Q4_0 and MXFP4, which this patch does not add. Please qualify the claimed fix with the exact dtype and consumer. #3019 and #2942 retain the separate admission work. The gfx1100 WMMA requirement remains separate. This patch adds no RDNA3 WMMA tile; the inherited matrix path admits gfx1200/gfx1201. #2109 closed through #2991 on 6 September, and that landed spec explicitly leaves RDNA3 to separate work. No device or performance acceptance is claimed by this audit. |
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
4 participants
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.
Closes #2781.
Row:
GFX1100-TG200Lands the ROCm keep-quant GEMM infrastructure: the
kROCMprovider forOpId::kMatmulBTQuantandOpId::kMatmulBTQuantGrouped. The provider doesthe dequant-in-kernel dot product on-device, keeping the weight in its
block-quantized form through the GEMM rather than expanding to bf16 at load
time.
VT_GGUF_KEEP_QUANT=0restores load-time expansion.The new provider lives in
src/vt/rocm/rocm_quant_dot.hipand registersitself for
kROCM. The oldMatmulBTQuantKernelRocmregistration inrocm_ops.hip(which routed to the limitedrocm_grouped_gemm.hipQ8_0/Q4_K/Q5_K/Q6_K-only dispatch) is removed so the new provider takes
effect.
Gates on gfx1100 / ROCm 7.14.0:
test_rocm_quant_dot132,094 assertionsgreen across all ten encodings (decode through prefill shapes, broadcast
and per-row grouped arms over a poisoned output buffer). Token-identical
to upstream baseline on Qwen3.5-4B Q4_K, 32-token greedy decode, seed 0.
This is the foundation PR for the GFX1100-TG200 decode campaign. Most
lever PRs depend on the keep-quant GEMM infrastructure it introduces.
FOLLOWING_AGENTS_PROTOCOL
Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: AGENT:GLM-5-2 [OMP]