Execute gfx1201 K enc2 and generalize layer reference diagnostics

CMakeLists.txt

@@ -29,19 +29,24 @@ add_custom_command(OUTPUT "${wmma_object}"
     -o "${wmma_object}"
   DEPENDS tools/windows/wmma_gfx1201.hip kernels/common/wmma_backend.h VERBATIM)
 add_custom_target(d4r_wmma_code_object ALL DEPENDS "${wmma_object}")
-# Compile-only K port smoke test. This object is intentionally not installed as
-# a native override until its transformer outputs pass the K reference/replay.
-set(k_enc1_object "${CMAKE_CURRENT_BINARY_DIR}/dltss_pwin_enc1_layer_gfx1201.hsaco")
-add_custom_command(OUTPUT "${k_enc1_object}"
-  COMMAND "${D4R_HIP_CLANG}" -x hip -std=c++17 --offload-arch=gfx1201
-    --offload-device-only --no-gpu-bundle-output -mno-wavefrontsize64
-    -nogpuinc -nogpulib -O2 -DD4R_DEVICE_ONLY_MINIMAL
-    "${CMAKE_CURRENT_SOURCE_DIR}/kernels/k/dltss_pwin_enc1_layer.hip"
-    -o "${k_enc1_object}"
-  DEPENDS kernels/k/dltss_pwin_enc1_layer.hip kernels/k/pwin_layer.h
-    kernels/k/pwin_common.h kernels/common/wmma_backend.h
-    kernels/common/hip_device_minimal.h VERBATIM)
-add_custom_target(d4r_k_gfx1201_compile ALL DEPENDS "${k_enc1_object}")
+# These K objects stay in experimental/k, outside the ZLUDA native override
+# directory, until each layer and then the complete network pass replay.
+set(k_objects)
+foreach(layer enc1 enc2)
+  set(k_object "${CMAKE_CURRENT_BINARY_DIR}/dltss_pwin_${layer}_layer_gfx1201.hsaco")
+  add_custom_command(OUTPUT "${k_object}"
+    COMMAND "${D4R_HIP_CLANG}" -x hip -std=c++17 --offload-arch=gfx1201
+      --offload-device-only --no-gpu-bundle-output -mno-wavefrontsize64
+      -nogpuinc -nogpulib -O2 -DD4R_DEVICE_ONLY_MINIMAL
+      "${CMAKE_CURRENT_SOURCE_DIR}/kernels/k/dltss_pwin_${layer}_layer.hip"
+      -o "${k_object}"
+    DEPENDS kernels/k/dltss_pwin_${layer}_layer.hip kernels/k/pwin_layer.h
+      kernels/k/pwin_common.h kernels/common/wmma_backend.h
+      kernels/common/hip_device_minimal.h VERBATIM)
+  list(APPEND k_objects "${k_object}")
+  install(FILES "${k_object}" DESTINATION experimental/k)
+endforeach()
+add_custom_target(d4r_k_gfx1201_compile ALL DEPENDS ${k_objects})
 foreach(probe hip_gfx1201_probe gfx12_wmma_probe k_module_probe cuda_driver_probe d3d12_hip_interop_probe hip_stream_lifecycle_probe ngx_cuda_init_probe)
   add_executable(d4r_${probe} tools/windows/${probe}.cpp)
   target_compile_definitions(d4r_${probe} PRIVATE WIN32_LEAN_AND_MEAN NOMINMAX __HIP_PLATFORM_AMD__)
@@ -97,9 +102,8 @@ add_dependencies(d4r_gfx12_wmma_probe d4r_wmma_code_object)
 add_dependencies(d4r_k_module_probe d4r_k_gfx1201_compile)
 install(FILES "${probe_object}" DESTINATION bin)
 install(FILES "${wmma_object}" DESTINATION bin)
-install(FILES "${k_enc1_object}" DESTINATION experimental/k)
 install(FILES scripts/windows/test-windows-rdna4.ps1 DESTINATION .)
-install(FILES tools/windows/k_enc1_validate.py DESTINATION .)
+install(FILES tools/windows/k_layer_validate.py DESTINATION .)
 install(FILES kernels/tools/pwin_model.py DESTINATION reference)
 install(FILES docs/windows-rdna4-port.md DESTINATION docs)
 include(CTest)
@@ -109,9 +113,13 @@ set_tests_properties(hip_gfx1201 PROPERTIES TIMEOUT 90 LABELS "hardware;gfx1201"
 add_test(NAME gfx12_wmma COMMAND d4r_gfx12_wmma_probe
   --hip-root "${D4R_HIP_ROOT}" --module "${wmma_object}" --iterations 16)
 set_tests_properties(gfx12_wmma PROPERTIES TIMEOUT 90 LABELS "hardware;gfx1201;wmma")
-add_test(NAME k_enc1_execution COMMAND d4r_k_module_probe
-  --hip-root "${D4R_HIP_ROOT}" --module "${k_enc1_object}")
-set_tests_properties(k_enc1_execution PROPERTIES TIMEOUT 90 LABELS "hardware;gfx1201;k_execution")
+foreach(layer enc1 enc2)
+  add_test(NAME k_${layer}_execution COMMAND d4r_k_module_probe
+    --hip-root "${D4R_HIP_ROOT}"
+    --module "${CMAKE_CURRENT_BINARY_DIR}/dltss_pwin_${layer}_layer_gfx1201.hsaco"
+    --kernel-name ${layer})
+  set_tests_properties(k_${layer}_execution PROPERTIES TIMEOUT 90 LABELS "hardware;gfx1201;k_execution")
+endforeach()
 add_test(NAME d3d12_hip_interop COMMAND d4r_d3d12_hip_interop_probe
   --hip-root "${D4R_HIP_ROOT}" --module "${probe_object}" --iterations 32)
 set_tests_properties(d3d12_hip_interop PROPERTIES TIMEOUT 120 LABELS "hardware;interop")

README.md

@@ -8,8 +8,8 @@ It is an experiment. It has been tested on one GPU in a handful of games, it dep
 
 **Windows / RDNA4 development:** the `windows-rdna4` branch adds native Windows
 gfx1201 HIP/ZLUDA diagnostics, validated D3D12/HIP interop, an NGX init probe,
-and a CPU-checked gfx12 WMMA backend. The first K `enc1` kernel executes on
-gfx1201 and matches the existing numpy model on a synthetic nonzero fixture.
+and a CPU-checked gfx12 WMMA backend. K `enc1` and `enc2` execute on
+gfx1201 and match the existing numpy model on synthetic nonzero fixtures.
 The remaining K/M kernels and Windows game runtime are still in progress.
 Build commands, hardware results and remaining
 porting gates are in [docs/windows-rdna4-port.md](docs/windows-rdna4-port.md).

docs/windows-rdna4-port.md

@@ -7,8 +7,8 @@ This is a work log for an implementation in progress, not a claim of DLSS suppor
 ## Milestone and gates
 
 Current milestone: M1/M2/M4/M5 passed; NGX init + DLSS CUDA capability discovery
-passed with the user's DLLs. M6 is in progress: the first K `enc1` prep and
-transformer run on gfx1201 and pass a synthetic nonzero numpy reference check.
+passed with the user's DLLs. M6 is in progress: K `enc1` and `enc2` prep and
+transformers run on gfx1201 and pass synthetic nonzero numpy reference checks.
 Do not integrate NGX until the integer PTX workload is stable on the real GPU.
 
 | Gate | Status |
@@ -19,7 +19,7 @@ Do not integrate NGX until the integer PTX workload is stable on the real GPU.
 | M3: Windows NGX initialization and simple DLSS path | Init + SR capabilities PASS; Evaluate/transformer pending |
 | M4: D3D12 / HIP external memory and fence round trip | PASS with TheRock; stable 7.2 has mapped-view leak |
 | M5: independently validated gfx12 WMMA backend | PASS: raw + legacy adapter, 16 cases, max abs/relative error 0 |
-| M6: K layers, full transformer and image validation | Partial: `enc1` executes, nonzero fixture matches reference; other layers/full network pending |
+| M6: K layers, full transformer and image validation | Partial: `enc1`/`enc2` execute and match nonzero references; other layers/full network pending |
 | M7: M FP16-equivalent baseline and full transformer | Pending |
 | M8: standalone Windows D3D12 NGX harness | Pending |
 | M9: OptiScaler integration, profiling, installation | Pending |
@@ -137,13 +137,13 @@ different inputs; max absolute and relative errors were both zero.
 `kernels/k/pwin_common.h` now calls the tested adapter. The Windows device-only
 compiler uses `kernels/common/hip_device_minimal.h`, exposing only public
 Clang AMDGPU work-item/grid/barrier builtins and HIP-compatible qualifiers;
-the ordinary Linux HIP include path remains. CMake builds `enc1` for gfx1201
+the ordinary Linux HIP include path remains. CMake builds `enc1`/`enc2` for gfx1201
 into `experimental/k/` and verifies that HIP loads both prep and transformer
 entry points. The `k_module_probe` now launches prep and transformer. The
 identity fixture verifies all 4096 full-resolution and 1024 merged FP16 values
 bitwise. A second fixture enables nonzero V projections, position-only
 attention, Wo, MLP/GELU and patch merge; it checks finite values and a patch
-identity, then saves all inputs/outputs for `k_enc1_validate.py`. That script
+identity, then saves all inputs/outputs for `k_layer_validate.py`. That script
 runs the existing `pwin_model.py` and measures max absolute/relative error and
 PSNR. The code object remains in `experimental/k/`, outside ZLUDA's override
 path, until the other K layers and real captured weights/activations pass.
@@ -259,7 +259,7 @@ One command adds the local NGX initialization gate after eight diagnostics:
 powershell -NoProfile -File scripts/windows/test-windows-rdna4.ps1 -RuntimeProfile therock -NgxCore "C:/Users/Administrator/d4r/_nvngx.dll" -DlssDll "C:/Users/Administrator/d4r/nvngx_dlss.dll"
 ```
 
-Success requires all nine tests, including `PASS K_REFERENCE` and
+Success with the two local NGX DLLs requires all eleven tests, including both `PASS K_REFERENCE` checks and
 `PASS NGX_INIT ... sr_available=1`. The NGX probe still reports
 `transformer_executed=0`: the standalone K launch has not been integrated into
 DLSS. On NGX failure, CUDA trace and NVAPI query
@@ -279,7 +279,7 @@ cmake --install build/windows-rdna4 --prefix dist/windows-rdna4-diagnostics
 The one-command diagnostic runner is also installed at
 `dist/windows-rdna4-diagnostics/test-windows-rdna4.ps1`. It requires the selected
 HIP SDK and ZLUDA directory; the developer setup auto-discovers local defaults.
-Success: exit 0, all five tests passed in `summary.json`, `PASS HIP`, `PASS CUDA`,
+Success: exit 0, every selected test passed in `summary.json`, `PASS HIP`, `PASS CUDA`,
 `PASS INTEROP_LIFETIME` and `PASS INTEROP` in stdout. Failure: send the **single printed ZIP path**; it contains
 stdout/stderr, timeout/exception exit code, DLL paths/versions/hashes, driver/OS
 inventory and a minidump for an unhandled exception. Upstream trace is attempted
@@ -351,6 +351,16 @@ Reference dependency: pinned local NumPy `2.4.6` on Python 3.11, installed by
 `setup-windows-tools.ps1` into `.tools/python/vendor` (not the game runtime).
 This is a synthetic one-layer check, not proof of complete K or DLSS output.
 
+2026-09-29 M6 second kernel: `enc2` exercises nonzero Q/K/V, learned attention
+and softmax, Wo, MLP/GELU and patch merge on gfx1201. Full output max absolute
+error `3.81469727e-06`, max relative error `0.00166893` (denominator floor
+`1e-3`), PSNR `130.39 dB`; merged output max absolute error `3.81469727e-06`,
+PSNR `132.49 dB`. All 4096 full and 1536 merged elements written, no NaN/Inf.
+Both layers pass 32 identity iterations and the nonzero numpy reference.
+Bundle `test-results/m6-k-enc2-final.zip`: 11/11 gates PASS. CTest: 8/8 PASS.
+The generic probe/reference runner selects `--kernel-name enc1|enc2`; both
+objects remain under `experimental/k` and are not used as DLSS overrides yet.
+
 Stable HIP SDK 7.2 compiler caveat: its `-O2` code object for `enc1` produces
 NaN already on the identity fixture (first element `0xfe00` instead of
 `0xbc00`). Its `-O1` object also miscomputes the identity fixture; `-O0`

scripts/windows/test-windows-rdna4.ps1

@@ -101,7 +101,8 @@ try {
     $files = @((Join-Path $HipRoot 'bin/amdhip64_7.dll'), $comgr,
         (Join-Path $ZludaRoot 'nvcuda.dll'), (Join-Path $PackageRoot 'bin/probe_gfx1201.hsaco'),
         (Join-Path $PackageRoot 'bin/wmma_gfx1201.hsaco'),
-        (Join-Path $PackageRoot 'experimental/k/dltss_pwin_enc1_layer_gfx1201.hsaco'))
+        (Join-Path $PackageRoot 'experimental/k/dltss_pwin_enc1_layer_gfx1201.hsaco'),
+        (Join-Path $PackageRoot 'experimental/k/dltss_pwin_enc2_layer_gfx1201.hsaco'))
     if ($NgxCore -or $DlssDll) {
         if (!$NgxCore -or !$DlssDll) { throw 'Supply both -NgxCore and -DlssDll absolute paths.' }
         if (![IO.Path]::IsPathRooted($NgxCore) -or ![IO.Path]::IsPathRooted($DlssDll)) {
@@ -127,16 +128,19 @@ try {
         $wmmaOk = Invoke-Probe 'gfx12-wmma' (Join-Path $bin 'd4r_gfx12_wmma_probe.exe') @(
             '--hip-root', $HipRoot, '--module', (Join-Path $bin 'wmma_gfx1201.hsaco'), '--iterations', "$Iterations")
         if (!$wmmaOk) { throw 'gfx12 WMMA layout validation failed; refusing to mark K/M readiness.' }
-        $kFixture = Join-Path $OutputDirectory 'k-enc1-fixture'
-        $kModuleOk = Invoke-Probe 'k-enc1-gpu' (Join-Path $bin 'd4r_k_module_probe.exe') @(
-            '--hip-root', $HipRoot, '--module', (Join-Path $PackageRoot 'experimental/k/dltss_pwin_enc1_layer_gfx1201.hsaco'),
-            '--fixture-dir', $kFixture, '--iterations', "$Iterations")
-        if (!$kModuleOk) { throw 'K enc1 GPU execution failed; see the k-enc1-gpu logs.' }
         $pythonCommand = Get-Command python -ErrorAction SilentlyContinue
-        if (!$pythonCommand) { throw 'Python 3.11+ with NumPy 2.4.6 is needed for the K reference check; run setup-windows-tools.ps1.' }
-        $kReferenceOk = Invoke-Probe 'k-enc1-reference' $pythonCommand.Source @(
-            (Join-Path $PackageRoot 'k_enc1_validate.py'), '--fixture-dir', $kFixture)
-        if (!$kReferenceOk) { throw 'K enc1 output differs from pwin_model.py; see the k-enc1-reference logs.' }
+        if (!$pythonCommand) { throw 'Python 3.11+ with NumPy 2.4.6 is needed for K reference checks; run setup-windows-tools.ps1.' }
+        foreach ($layer in @('enc1', 'enc2')) {
+            $kFixture = Join-Path $OutputDirectory "k-$layer-fixture"
+            $kModuleOk = Invoke-Probe "k-$layer-gpu" (Join-Path $bin 'd4r_k_module_probe.exe') @(
+                '--hip-root', $HipRoot,
+                '--module', (Join-Path $PackageRoot "experimental/k/dltss_pwin_${layer}_layer_gfx1201.hsaco"),
+                '--kernel-name', $layer, '--fixture-dir', $kFixture, '--iterations', "$Iterations")
+            if (!$kModuleOk) { throw "K $layer GPU execution failed; see the k-$layer-gpu logs." }
+            $kReferenceOk = Invoke-Probe "k-$layer-reference" $pythonCommand.Source @(
+                (Join-Path $PackageRoot 'k_layer_validate.py'), '--kernel-name', $layer, '--fixture-dir', $kFixture)
+            if (!$kReferenceOk) { throw "K $layer output differs from pwin_model.py; see the k-$layer-reference logs." }
+        }
         $cudaOk = $true
         foreach ($context in @('primary', 'created')) {
             $ok = Invoke-Probe "cuda-$context" (Join-Path $bin 'd4r_cuda_driver_probe.exe') @(

tools/windows/diagnostic.h

@@ -90,6 +90,7 @@ struct Args {
     std::string nvapi_dll;
     std::string ngx_abi = "driver";
     std::string fixture_dir;
+    std::string kernel_name = "enc1";
     unsigned iterations = 32;
     int device = -1;
     Args(int argc, char** argv) {
@@ -108,6 +109,7 @@ struct Args {
             else if (key == "--nvapi-dll") nvapi_dll = value;
             else if (key == "--ngx-abi") ngx_abi = value;
             else if (key == "--fixture-dir") fixture_dir = value;
+            else if (key == "--kernel-name") kernel_name = value;
             else if (key == "--iterations") {
                 size_t end = 0;
                 iterations = static_cast<unsigned>(std::stoul(value, &end));

tools/windows/k_layer_validate.py

@@ -1,5 +1,5 @@
 #!/usr/bin/env python3
-"""Compare a real gfx1201 K enc1 launch with the existing numpy K reference."""
+"""Compare real gfx1201 K enc1/enc2 launches with the existing numpy reference."""
 import argparse
 import math
 import pathlib
@@ -57,21 +57,24 @@ def compare(name, actual, expected):
 def main():
     parser = argparse.ArgumentParser()
     parser.add_argument("--fixture-dir", required=True, type=pathlib.Path)
+    parser.add_argument("--kernel-name", choices=("enc1", "enc2"), default="enc1")
     args = parser.parse_args()
     folder = args.fixture_dir
     params = SyntheticParams(folder)
+    output_channels = 96 if args.kernel_name == "enc2" else 64
     reference_full, reference_merged = pwin_encoder(
-        params, 0, 0, H=2, C=64, COUT=64, posattn=True)
+        params, 0, 0, H=2, C=64, COUT=output_channels,
+        posattn=args.kernel_name == "enc1")
     actual_full = np.fromfile(folder / "full.bin", np.float16).reshape(64, 64)
-    actual_merged = np.fromfile(folder / "merged.bin", np.float16).reshape(16, 64)
+    actual_merged = np.fromfile(folder / "merged.bin", np.float16).reshape(16, output_channels)
     full_abs, full_psnr = compare("full", actual_full, reference_full)
     merged_abs, merged_psnr = compare("merged", actual_merged, reference_merged)
     # Initial acceptance for the synthetic nonzero FP16 fixture. Tighten this
     # after each K stage has been compared on captured DLSS weights/activations.
     if full_abs > 0.001 or full_psnr < 60.0 or merged_abs > 0.001 or merged_psnr < 60.0:
-        raise RuntimeError("K enc1 numerical mismatch against pwin_model.py")
-    print("PASS K_REFERENCE architecture=gfx1201 kernel=enc1 nonzero_weights=1 "
-          "transformer_executed=1 nan_inf=0")
+        raise RuntimeError(f"K {args.kernel_name} numerical mismatch against pwin_model.py")
+    print(f"PASS K_REFERENCE architecture=gfx1201 kernel={args.kernel_name} "
+          "nonzero_weights=1 transformer_executed=1 nan_inf=0")
 
 
 if __name__ == "__main__":

tools/windows/k_module_probe.cpp

@@ -43,24 +43,29 @@ int main(int argc, char** argv)
         HipApi hip(args.hip_root);
         hipDeviceProp_t properties{};
         hip.select_gfx1201(args.device, properties);
+        if (args.kernel_name != "enc1" && args.kernel_name != "enc2")
+            throw std::runtime_error("--kernel-name must be enc1 or enc2");
+        const bool enc2 = args.kernel_name == "enc2";
+        const std::string entry = "dltss_pwin_" + args.kernel_name + "_layer";
         hipModule_t module = nullptr;
-        hip.check(hip.hipModuleLoad(&module, args.module.c_str()), "hipModuleLoad(K enc1)");
+        hip.check(hip.hipModuleLoad(&module, args.module.c_str()), "hipModuleLoad(K)");
         struct Cleanup {
             HipApi& hip; hipModule_t module;
             ~Cleanup() { if (module) (void)hip.hipModuleUnload(module); }
         } cleanup{hip, module};
         hipFunction_t prep = nullptr, transformer = nullptr;
-        hip.check(hip.hipModuleGetFunction(&prep, module, "dltss_pwin_enc1_layer_prep"), "K enc1 prep");
-        hip.check(hip.hipModuleGetFunction(&transformer, module, "dltss_pwin_enc1_layer"), "K enc1 transformer");
+        hip.check(hip.hipModuleGetFunction(&prep, module, (entry + "_prep").c_str()), "K prep");
+        hip.check(hip.hipModuleGetFunction(&transformer, module, entry.c_str()), "K transformer");
         if (!prep || !transformer) throw std::runtime_error("Null K entry point");
 
         // Identity fixture: all projection, MLP and patch-merge weights are zero.
         // G1/G2 are one. The full transformer must return its input through its
         // residual path and the merged output must be zero. Every stage launches.
         constexpr size_t full_elements = 8u * 8u * 64u;
-        constexpr size_t merged_elements = 4u * 4u * 64u;
-        constexpr size_t weight_bytes = 149504u;
-        constexpr size_t prep_items = 256u * 16u + 1024u;
+        const size_t output_channels = enc2 ? 96u : 64u;
+        const size_t merged_elements = 4u * 4u * output_channels;
+        const size_t weight_bytes = enc2 ? 166080u : 149504u;
+        const size_t prep_items = (enc2 ? 288u : 256u) * 16u + 1024u;
         std::vector<uint8_t> weights(weight_bytes, 0);
         auto set_half = [&](size_t byte, uint16_t bits) {
             weights.at(byte) = static_cast<uint8_t>(bits);
@@ -119,8 +124,8 @@ int main(int argc, char** argv)
                     return 5;
                 }
         }
-        std::printf("PASS K_IDENTITY architecture=gfx1201 iterations=%u full_elements=%zu merged_elements=%zu transformer_executed=1 nonzero_weights=0\n",
-            args.iterations, full_elements, merged_elements);
+        std::printf("PASS K_IDENTITY architecture=gfx1201 kernel=%s iterations=%u full_elements=%zu merged_elements=%zu transformer_executed=1 nonzero_weights=0\n",
+            args.kernel_name.c_str(), args.iterations, full_elements, merged_elements);
 
         // Exercise nonzero V projection, position-only attention and Wo GEMMs.
         // The generated data is saved for comparison with pwin_model.py.
@@ -135,6 +140,12 @@ int main(int argc, char** argv)
         constexpr size_t qkv = 128u, head = 12288u, bias = 24704u, wo = 41216u;
         constexpr size_t w1 = 49664u, w2 = 82944u, pm = 115712u;
         for (unsigned k = 0; k < 32u; ++k) {
+            if (enc2) {
+                set_matrix(qkv, k, k, 1024u);                           // head 0 Q
+                set_matrix(qkv + 2048u, k, k, 1024u);                   // head 0 K
+                set_matrix(qkv + head + 2048u * 3u, k, k, 1024u);      // head 1 Q
+                set_matrix(qkv + head + 2048u * 4u, k, k, 1024u);      // head 1 K
+            }
             set_matrix(qkv + 2048u * 2u, k, k, 1024u);                   // head 0 V <- input 0:32
             set_matrix(qkv + head + 2048u * 5u, k, k, 1024u);          // head 1 V <- input 32:64
             set_matrix(wo, k, k, 1024u);                               // head 0 Wo -> output 0:32
@@ -181,9 +192,10 @@ int main(int argc, char** argv)
         for (unsigned mt = 0; mt < 16u; ++mt) {
             const unsigned my = mt >> 2, mx = mt & 3u;
             const size_t full_base = (8u * (2u * my) + 2u * mx) * 64u;
-            for (unsigned channel = 0; channel < 64u; ++channel) {
-                const size_t i = mt * 64u + channel;
-                if (merged[i] != full[full_base + channel])
+            for (unsigned channel = 0; channel < output_channels; ++channel) {
+                const size_t i = mt * output_channels + channel;
+                const uint16_t expected = channel < 64u ? full[full_base + channel] : 0u;
+                if (merged[i] != expected)
                     throw std::runtime_error("K patch merge identity mismatch at " + std::to_string(i));
             }
         }
@@ -201,8 +213,8 @@ int main(int argc, char** argv)
             save("full.bin", full.data(), full.size() * sizeof(uint16_t));
             save("merged.bin", merged.data(), merged.size() * sizeof(uint16_t));
         }
-        std::printf("PASS K_NONZERO architecture=gfx1201 changed_elements=%zu full_elements=%zu merged_elements=%zu fixture_saved=%d numerical_reference_pending=1\n",
-            changed, full.size(), merged.size(), !args.fixture_dir.empty());
+        std::printf("PASS K_NONZERO architecture=gfx1201 kernel=%s changed_elements=%zu full_elements=%zu merged_elements=%zu fixture_saved=%d numerical_reference_pending=1\n",
+            args.kernel_name.c_str(), changed, full.size(), merged.size(), !args.fixture_dir.empty());
         return 0;
     } catch (const std::exception& error) {
         std::fprintf(stderr, "FAIL K_MODULE %s\n", error.what());