code wiki / q8

topic: q8

12 modules sharing the q8 name family (derived from the tree's prefix discipline).

The 'q8' topic family in the Nishi sovereign ecosystem focuses on quantization and dequantization operations for efficient tensor processing. Modules like nx_q8_0_dot handle fused integer dot products with dequantization, while nx_q8_0_from_f32_kat ensures roundtrip accuracy between F32 and Q8_0. nx_q8_0_to_f32 provides precise dequantization to IEEE 754 binary32, supporting verification and simulation across the ecosystem.

auto-narrated by the local model from this topic's module headers; links verified against the wiki index.

narrated overview -- maintained by the narration lane, module links verified against this wiki.

moduledescriptionlinesfuncs
nx_q8_0_dot.nxsovereign Q8_0 dequant + fused integer dot, verified on a REAL Z-Image DiT block.693
nx_q8_0_from_f32.nxquantize F32 -> Q8_0 blocks (the inverse of893
nx_q8_0_from_f32_kat.nxroundtrip KAT: F32 -> Q8_0 -> F32 within a quant371
nx_q8_0_from_q5_0.nxre-quantize a Q5_0 tensor to Q8_0 IN MEMORY.582
nx_q8_0_linear_verify.nxverify a full Q8_0 linear at REAL DiT scale vs a Python ggml golden.833
nx_q8_0_simd_gate.nxdoes the SIMD __f32_i8dot32 fused Q8_0 dequant-dot1648
nx_q8_0_to_f32.nxbits-up Q8_0 dequantization to IEEE 754 binary32.542
nx_q8_coldwarm.nxdecides the post-futex decode-matmul direction:965
nx_q8_matmul_micro.nxDECISIVE micro-probe of the REAL production Q8_01465
nx_q8_mix_ab.nxsame-process, same-load A/B of the REAL per-layer block1284
nx_q8_scaling.nxpool SCALING probe: same-process ST vs MT(pool) cold Q81044
nx_q8_st_bw.nxSINGLE-THREAD sovereign Q8 matmul bandwidth, to split the654