Row: `GFX1100-TG200`
T27 adds the `VT_QUANT_Q8K_WARP=1` opt-in warp-cooperative `QuantizeQ8KKWarpCoop` kernel for the K-quant activation quantization. The warp-cooperative variant uses 16 super-blocks per block (vs 128 threads per block in the standard kernel) and is byte-identical to the standard `QuantizeQ8KK`. Measured +2.06% decode throughput improvement.
The default path (`VT_QUANT_Q8K_WARP` unset) is byte-unchanged.
Depends on #2782 (F1), #2790 (T4a), and #2792 (lever-C).
Row: `GFX1100-TG200`
T27 adds the `VT_QUANT_Q8K_WARP=1` opt-in warp-cooperative `QuantizeQ8KKWarpCoop` kernel for the K-quant activation quantization. The warp-cooperative variant uses 16 super-blocks per block (vs 128 threads per block in the standard kernel) and is byte-identical to the standard `QuantizeQ8KK`. Measured +2.06% decode throughput improvement.
The default path (`VT_QUANT_Q8K_WARP` unset) is byte-unchanged.
Depends on #2782 (F1), #2790 (T4a), and #2792 (lever-C).