Disable batch benchmarks for DeviceTransform (#7450)

bernhardmgruber · web-flow · commit f8b36e648875 · 2026-02-02T20:05:57.000+01:00
diff --git a/cub/benchmarks/bench/transform/common.h b/cub/benchmarks/bench/transform/common.h
@@ -78,7 +78,7 @@ void bench_transform(nvbench::state& state,
                      OffsetT num_items,
                      TransformOp transform_op)
 {
-  state.exec(nvbench::exec_tag::gpu, [&](const nvbench::launch& launch) {
+  state.exec(nvbench::exec_tag::gpu | nvbench::exec_tag::no_batch, [&](const nvbench::launch& launch) {
     cub::detail::transform::dispatch<cub::detail::transform::requires_stable_address::no>(
       inputs,
       output,

Original file line number	Diff line number	Diff line change
`@@ -78,7 +78,7 @@ void bench_transform(nvbench::state& state,`
`78`	`78`	`OffsetT num_items,`
`79`	`79`	`TransformOp transform_op)`
`80`	`80`	`{`
`81`		`- state.exec(nvbench::exec_tag::gpu, [&](const nvbench::launch& launch) {`
	`81`	`+ state.exec(nvbench::exec_tag::gpu \| nvbench::exec_tag::no_batch, [&](const nvbench::launch& launch) {`
`82`	`82`	`cub::detail::transform::dispatch<cub::detail::transform::requires_stable_address::no>(`
`83`	`83`	`inputs,`
`84`	`84`	`output,`