Skip to content

Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.

C++

tests: iq_multi_parity under STRATA_KQ_SPLIT=1 also without V2K, with the fused gate/up SwiGLU, and with the sub-16 gate/up rows off

Niko12210e58a652fbaf
diff --git a/CMakeLists.txt b/CMakeLists.txt
index bf8b8bd79..e178f26e0 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -964,6 +964,14 @@ if(STRATA_ENABLE_CUDA OR STRATA_ENABLE_HIP)
target_link_libraries(iq_multi_parity PRIVATE strata_kernels ${_strata_gpu_runtime_target})
add_test(NAME iq_multi_parity COMMAND iq_multi_parity)
set_tests_properties(iq_multi_parity PROPERTIES ENVIRONMENT "STRATA_KQ_SPLIT=1") # the Q4_K / Q5_K / Q8_0 Split path against the old kernels
+ # the same Split path on the other routes it can meet: the generic multi kernels (no V2K) and the fused gate/up + SwiGLU kernel, and the sub-16-row
+ # gate/up route switched off (a K-quant row has 160 dots, not the 80 the sub-16 rows assume: STRATA_KQ_SPLIT=1 must stay bitwise on all of them)
+ add_test(NAME iq_multi_parity_kq_nov2k COMMAND iq_multi_parity)
+ set_tests_properties(iq_multi_parity_kq_nov2k PROPERTIES ENVIRONMENT "STRATA_KQ_SPLIT=1;STRATA_EXPERT_V2K=0")
+ add_test(NAME iq_multi_parity_kq_swiglu COMMAND iq_multi_parity)
+ set_tests_properties(iq_multi_parity_kq_swiglu PROPERTIES ENVIRONMENT "STRATA_KQ_SPLIT=1;STRATA_GU_SWIGLU=1")
+ add_test(NAME iq_multi_parity_kq_nosub16 COMMAND iq_multi_parity)
+ set_tests_properties(iq_multi_parity_kq_nosub16 PROPERTIES ENVIRONMENT "STRATA_KQ_SPLIT=1;STRATA_NO_SUB16_GU=1")
# the prompt path's DeltaNet recurrence: gdn_rec_kh_kernel against gdn_rec_cols_pipe_kernel, bitwise (GPU,
# synthetic, no model; --bench adds timings). NVIDIA only: cp.async and the CUDA occupancy API (not on gfx906)
if(STRATA_ENABLE_CUDA AND NOT STRATA_HIP_GFX906)