perf(GFX1100-TG200): T9 cooperative gated norm arm - #2875

Open

ghazni101 wants to merge 15 commits into

mudler:mainfrom

ghazni101:row/GFX1100-TG200-T9

Commits on Sep 6, 2026

feat(KERNEL-QUANT-CIQ-GEMM-ROCM): land the W1 keep-quant providers on kROCM
ghazni101
committed
7f58bc0
fix(KERNEL-QUANT-CIQ-GEMM-ROCM): keep grouped Q8_0 on the kROCM provider

mudler
authored and
ghazni101
committed
7906b33
Revert "fix(KERNEL-QUANT-CIQ-GEMM-ROCM): keep grouped Q8_0 on the kROCM provider"

mudler
authored and
ghazni101
committed
eccafd1
fix(KERNEL-QUANT-CIQ-GEMM-ROCM): keep grouped Q8_0 on the kROCM provider

mudler
authored and
ghazni101
committed
937fcba
spec(BACKEND-ROCM): define Qwen3.5 gfx1100 numerical gate

VikashLoomba
authored and
ghazni101
committed
d7ada59
spec(BACKEND-ROCM): repair Qwen3.5 numerical plan

VikashLoomba
authored and
ghazni101
committed
836941e
spec(BACKEND-ROCM): correct Qwen3.5 capture contracts

VikashLoomba
authored and
ghazni101
committed
e426348
perf(GFX1100-TG200): T4a MMVQ K-quant decode GEMV arm
ghazni101
committed
794b16c
feat(GFX1100-TG200): lever-C fuses Q8_K activation quant into the RmsNorm epilogue
ghazni101
committed
6274512
T27: warp-cooperative QuantizeQ8KK for decode (+2.06%, byte-identical)
ghazni101
committed
d8266f5
feat(GFX1100-TG200): T24 LDS-buffered quant epilogue in RmsNormRowCoopKernel
ghazni101
committed
85cd2df
feat(GFX1100-TG200): T21 keep-quant for V-head row-permuted GDN projections
ghazni101
committed
91b5394
fix(rocm): gate the coop rmsnorm vector body to bf16 gamma, seed quant amax at 0
ghazni101
committed
db3b1b9
T25: keep ssm_out as Q5_K with runtime input permutation
ghazni101
committed
d8d0786
perf(GFX1100-TG200): T9 gives the gated norm a per-row block
ghazni101
committed
bbad875

Navigation Menu

perf(GFX1100-TG200): T9 cooperative gated norm arm - #2875

perf(GFX1100-TG200): T9 cooperative gated norm arm#2875
ghazni101 wants to merge 15 commits into
mudler:mainfrom
ghazni101:row/GFX1100-TG200-T9

Commits on Sep 6, 2026

feat(KERNEL-QUANT-CIQ-GEMM-ROCM): land the W1 keep-quant providers on kROCM

fix(KERNEL-QUANT-CIQ-GEMM-ROCM): keep grouped Q8_0 on the kROCM provider

Revert "fix(KERNEL-QUANT-CIQ-GEMM-ROCM): keep grouped Q8_0 on the kROCM provider"

fix(KERNEL-QUANT-CIQ-GEMM-ROCM): keep grouped Q8_0 on the kROCM provider

spec(BACKEND-ROCM): define Qwen3.5 gfx1100 numerical gate

spec(BACKEND-ROCM): repair Qwen3.5 numerical plan

spec(BACKEND-ROCM): correct Qwen3.5 capture contracts

perf(GFX1100-TG200): T4a MMVQ K-quant decode GEMV arm

feat(GFX1100-TG200): lever-C fuses Q8_K activation quant into the RmsNorm epilogue

T27: warp-cooperative QuantizeQ8KK for decode (+2.06%, byte-identical)

feat(GFX1100-TG200): T24 LDS-buffered quant epilogue in RmsNormRowCoopKernel

feat(GFX1100-TG200): T21 keep-quant for V-head row-permuted GDN projections

fix(rocm): gate the coop rmsnorm vector body to bf16 gamma, seed quant amax at 0

T25: keep ssm_out as Q5_K with runtime input permutation

perf(GFX1100-TG200): T9 gives the gated norm a per-row block