cuda-ab: victory-series coverage battery (8 new clients) + soak runner list

victory_saxpy, victory_reduce (shared+atomics), victory_matmul (tiled
shared), victory_strided, victory_memset2d (pitched), victory_atomics,
victory_streams_events (3 streams + events), victory_math (mixed
int/float/double). Each verifies via readback and prints STATUS OK|FAIL.
soak.sh now runs the full 14-client set (seq or sustained).

1 files changed, 121 insertions, 0 deletions

diff --git a/tools/cuda-ab/vecadd_loop.cu b/tools/cuda-ab/vecadd_loop.cu
new file mode 100644
index 0000000..bde3d05
--- /dev/null
+++ b/tools/cuda-ab/vecadd_loop.cu
@@ -0,0 +1,121 @@
+// SPDX-License-Identifier: BSD-3-Clause
+/*
+ * vecadd_loop.cu - 100 repeated vectorAdd launches, soak-style.
+ *
+ * The same three device buffers are reused for 100 identical vadd launches:
+ * the point is repetition and stability, not new data. Every 10th launch the
+ * result is copied back and verified element by element against the
+ * host-computed a+b, so a silent corruption that only appears after many
+ * launches is caught.
+ *
+ * Output lines are stable and greppable:
+ *
+ *   DEVICE <name> cc=<major>.<minor> sm=<count> mem=<bytes>
+ *   VERIFY launch=<n> bad=<mismatches>
+ *   LOOP launches=100 bad=<mismatches>
+ *   STATUS OK|FAIL
+ *
+ * Build with a static CUDA runtime (--cudart=static) so the guest only needs
+ * the driver's libcuda.so, not a CUDA toolkit installation:
+ *
+ *   nvcc -O2 --cudart=static -o vecadd_loop vecadd_loop.cu
+ *
+ * Then, from the guest:  /tmp/vecadd_loop
+ */
+
+#include <cstdio>
+#include <cstdlib>
+#include <cstdint>
+#include <cuda_runtime.h>
+
+#define NLAUNCH 100
+#define VERIFY_EVERY 10
+
+#define CHECK(x)                                                         \
+	do {                                                             \
+		cudaError_t e_ = (x);                                    \
+		if (e_ != cudaSuccess) {                                 \
+			printf("CUDA_ERROR %s at line %d\n",             \
+			       cudaGetErrorString(e_), __LINE__);        \
+			printf("STATUS FAIL\n");                         \
+			return 1;                                        \
+		}                                                        \
+	} while (0)
+
+__global__ void vadd(const float *a, const float *b, float *c, int n)
+{
+	int i = blockIdx.x * blockDim.x + threadIdx.x;
+
+	if (i < n)
+		c[i] = a[i] + b[i];
+}
+
+int main(void)
+{
+	cudaDeviceProp prop;
+	const int n = 1 << 15;
+	size_t bytes = (size_t)n * sizeof(float);
+	float *h_a, *h_b, *h_c;
+	float *d_a, *d_b, *d_c;
+	int bad = 0, i, launch;
+
+	setvbuf(stdout, NULL, _IONBF, 0);
+
+	CHECK(cudaGetDeviceProperties(&prop, 0));
+	printf("DEVICE %s cc=%d.%d sm=%d mem=%llu\n", prop.name,
+	       prop.major, prop.minor, prop.multiProcessorCount,
+	       (unsigned long long)prop.totalGlobalMem);
+
+	h_a = (float *)malloc(bytes);
+	h_b = (float *)malloc(bytes);
+	h_c = (float *)malloc(bytes);
+	if (!h_a || !h_b || !h_c)
+		return 1;
+
+	for (i = 0; i < n; i++) {
+		h_a[i] = (float)(i % 1000) * 0.5f;
+		h_b[i] = (float)(i % 997) * 0.25f;
+	}
+
+	CHECK(cudaMalloc(&d_a, bytes));
+	CHECK(cudaMalloc(&d_b, bytes));
+	CHECK(cudaMalloc(&d_c, bytes));
+	CHECK(cudaMemcpy(d_a, h_a, bytes, cudaMemcpyHostToDevice));
+	CHECK(cudaMemcpy(d_b, h_b, bytes, cudaMemcpyHostToDevice));
+
+	for (launch = 1; launch <= NLAUNCH; launch++) {
+		vadd<<<(n + 255) / 256, 256>>>(d_a, d_b, d_c, n);
+		CHECK(cudaGetLastError());
+		if (launch % VERIFY_EVERY != 0)
+			continue;
+		CHECK(cudaDeviceSynchronize());
+		CHECK(cudaMemcpy(h_c, d_c, bytes, cudaMemcpyDeviceToHost));
+		{
+			int vbad = 0;
+
+			for (i = 0; i < n; i++) {
+				if (h_c[i] != h_a[i] + h_b[i])
+					vbad++;
+			}
+			printf("VERIFY launch=%d bad=%d\n", launch, vbad);
+			bad += vbad;
+		}
+	}
+	CHECK(cudaDeviceSynchronize());
+
+	printf("LOOP launches=%d bad=%d\n", NLAUNCH, bad);
+
+	if (bad != 0) {
+		printf("STATUS FAIL\n");
+		return 1;
+	}
+
+	printf("STATUS OK\n");
+	CHECK(cudaFree(d_a));
+	CHECK(cudaFree(d_b));
+	CHECK(cudaFree(d_c));
+	free(h_a);
+	free(h_b);
+	free(h_c);
+	return 0;
+}