Skip to content

parity tool: a new-kernel mismatch must fail the run; dispatch comment fallback was stale #64

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Niko1221 merged 1 commit into from
Sep 28, 2026
Merged

parity tool: a new-kernel mismatch must fail the run; dispatch comment fallback was stale #64

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions src/kernels/cpu/native_expert.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -79,8 +79,9 @@ void native_gu_rows(const NativeFmt& f, const uint8_t* blob, const void* const*
int r0, int r1) {
// the multi-token kernels decode the weights once for all tokens: 2.0-2.4x ggml-cpu at three tokens, no faster
// at one (all are bound by the codebook lookups, ~5 GB/s per core), measured by native_expert_parity. AVX-512
// first, then the AVX-2 one (Zen 2/3, Intel 12th-14th gen); STRATA_NO_IQ512 / STRATA_NO_IQ256 fall back to
// ggml-cpu's single-token vec_dot.
// first, then the AVX-2 one (Zen 2/3, Intel 12th-14th gen). STRATA_NO_IQ512 drops an AVX-512 CPU to the
// AVX-2 kernel, STRATA_NO_IQ256 drops the AVX-2 kernel; ggml-cpu's single-token vec_dot is reached only with
// both set (and on a CPU without AVX-512, STRATA_NO_IQ512 changes nothing).
static const bool avx512 = cpu_avx512_ok() && std::getenv("STRATA_NO_IQ512") == nullptr;
static const bool avx2 = std::getenv("STRATA_NO_IQ256") == nullptr;
if (nt >= 2 && iq512_supported(f.gu_type)) { // one token: ggml-cpu is as fast or faster
Expand Down
21 changes: 16 additions & 5 deletions src/kernels/native_expert_parity.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,8 +132,14 @@ int main(int argc, char** argv) {
float* gp[NT];
for (int k = 0; k < NT; ++k) gp[k] = g.data() + k * FF;
(is512 ? cpu::iq512_rows : cpu::iq256_rows)(f.gu_type, blob.data(), f.gu_row, (int) H, a, NT, gp, 0, (int) FF);
const double rg = rel(g, gref);
std::printf(" %s %s gate rows vs ggml vec_dot: rel %.2e\n",
ggml_type_name((ggml_type) f.gu_type), tag, rel(g, gref));
ggml_type_name((ggml_type) f.gu_type), tag, rg);
if (rg > 1e-5) { // ggml's own vec_dot is the reference: above 1e-5 it is a real mismatch,
std::printf(" %s %s gate rows MISMATCH (rel %.2e > 1e-5)\n",
ggml_type_name((ggml_type) f.gu_type), tag, rg); // not rounding (measured 3e-8)
++failures;
}
float* f1[1] = {ffp[0]};
const int it = 50;
const auto gu = is512 ? cpu::iq512_gu_rows : cpu::iq256_gu_rows;
Expand Down Expand Up @@ -177,16 +183,21 @@ int main(int argc, char** argv) {
if (f.d_type == 20) {
// (b3) the IQ4_NL multi-token AVX-2 kernel the pool now uses for IQ4_NL down projections,
// against ggml-cpu's single-token vec_dot on the SAME Q8_0 activations (h), plus timing.
std::vector<float> alt((size_t) NT * H), ref((size_t) NT * H);
std::vector<float> alt((size_t) NT * H), refd((size_t) NT * H);
float* altp[NT];
for (int k = 0; k < NT; ++k) altp[k] = alt.data() + k * H;
cpu::iq4nl256_down_rows(blob.data() + f.down_off, f.d_row, (int) FF, hp, NT, altp, 0, (int) H);
const auto* tdc = ggml_get_type_traits_cpu((ggml_type) f.d_type);
for (int k = 0; k < NT; ++k)
for (int64_t r = 0; r < H; ++r)
tdc->vec_dot((int) FF, ref.data() + k * H + r, 0,
tdc->vec_dot((int) FF, refd.data() + k * H + r, 0,
blob.data() + f.down_off + (size_t) r * f.d_row, 0, hp[k], 0, 1);
std::printf(" iq4_nl AVX-2 multi-token down vs ggml down: rel %.2e\n", rel(alt, ref));
const double rdn = rel(alt, refd);
std::printf(" iq4_nl AVX-2 multi-token down vs ggml down: rel %.2e\n", rdn);
if (rdn > 1e-5) { // measured 1e-7; same reasoning as the gate/up rows above
std::printf(" iq4_nl down MISMATCH (rel %.2e > 1e-5)\n", rdn);
++failures;
}
const int it = 50;
const auto t0 = std::chrono::steady_clock::now();
for (int i = 0; i < it; ++i)
Expand All @@ -195,7 +206,7 @@ int main(int argc, char** argv) {
for (int i = 0; i < it; ++i)
for (int k = 0; k < NT; ++k)
for (int64_t r = 0; r < H; ++r)
tdc->vec_dot((int) FF, ref.data() + k * H + r, 0,
tdc->vec_dot((int) FF, refd.data() + k * H + r, 0,
blob.data() + f.down_off + (size_t) r * f.d_row, 0, hp[k], 0, 1);
const auto t2 = std::chrono::steady_clock::now();
const double usn = std::chrono::duration<double, std::micro>(t1 - t0).count() / it;
Expand Down