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++

ds4_iq.cu: the NUL character literal in dq_use_old_kernel had been written as a raw NUL byte by the merge script (nvcc: quoted string should contain at least one character); it is a backslash-zero escape again

Niko122120c2f42bf106
diff --git a/src/dsv41/ds4_iq.cu b/src/dsv41/ds4_iq.cu
index b0539d1dc..931e3d33b 100644
--- a/src/dsv41/ds4_iq.cu
+++ b/src/dsv41/ds4_iq.cu
@@ -1393,7 +1393,7 @@ __global__ void __launch_bounds__(kThreads) dq_rows_kernel(const uint8_t* __rest
// prompt 79.5 -> 88.5 tok/s with it, so a gfx11 device defaults to it. DS4_DQ_OLD=1 forces it anywhere, DS4_DQ_OLD=0 forces kernel2
// anywhere. The detection is the device's gcnArchName (HIP only; a CUDA build has no gfx11). Decided once per device.
static bool dq_use_old_kernel() {
- if (const char* e = std::getenv("DS4_DQ_OLD"); e != nullptr && *e != '') return std::atoi(e) != 0;
+ if (const char* e = std::getenv("DS4_DQ_OLD"); e != nullptr && *e != '\0') return std::atoi(e) != 0;
#if defined(STRATA_USE_HIP)
static std::atomic<int> cache[64]; // 0 unknown, 1 no, 2 yes
int dev = 0;