[rocm-libraries] ROCm/rocm-libraries#8142 (commit b78e456)

fix(rocthrust): fix SetBytesProcessed() calls

## Motivation

There are three spots in rocThrust that currently wrongly use
`state.SetBytesProcessed(0);`, rather than passing it an actual value.

## Technical Details

Those three spots were updated to
`state.SetBytesProcessed(state.iterations() * needles * sizeof(T));`, to
make them consistent with the
`state.SetItemsProcessed(state.iterations() * needles);` under them.

The bandwidth calculation introduced here focuses strictly on the query
footprint (`needles * sizeof(T)`) to unblock `bytes_per_second` graphing
(which currently flatlines at 0). While this represents a lower bound of
the actual global memory traffic, it provides an immediate baseline for
throughput tracking. Any future refinements to capture the exact
algorithmic memory footprint will simply scale the resulting graph by a
predictable constant factor.

## Test Plan

Run the three updated benchmarks.

## Test Result

I have confirmed they now output proper `bytes_per_second` values,
rather than:
```json
"bytes_per_second": 0.0000000000000000e+00,
"gpu_noise": 7.1034145212086564e-02,
"items_per_second": 1.6391124340294927e+07
```

## Submission Checklist

- [x] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

Author: MyNameIsTrez

benchmark/bench/vectorized_search/basic.cu

@@ -1,6 +1,6 @@
 /******************************************************************************
  * Copyright (c) 2011-2023, NVIDIA CORPORATION.  All rights reserved.
- * Modifications Copyright (c) 2024-2025, Advanced Micro Devices, Inc.  All rights reserved.
+ * Modifications Copyright (c) 2024-2026, Advanced Micro Devices, Inc.  All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
@@ -90,7 +90,7 @@ void run_benchmark(
     gpu_times.push_back(duration);
   }
 
-  state.SetBytesProcessed(0);
+  state.SetBytesProcessed(state.iterations() * needles * sizeof(T));
   state.SetItemsProcessed(state.iterations() * needles);
 
   const double gpu_cv         = bench_utils::StatisticsCV(gpu_times);

benchmark/bench/vectorized_search/lower_bound.cu

@@ -1,6 +1,6 @@
 /******************************************************************************
  * Copyright (c) 2011-2023, NVIDIA CORPORATION.  All rights reserved.
- * Modifications Copyright (c) 2024-2025, Advanced Micro Devices, Inc.  All rights reserved.
+ * Modifications Copyright (c) 2024-2026, Advanced Micro Devices, Inc.  All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
@@ -90,7 +90,7 @@ void run_benchmark(
     gpu_times.push_back(duration);
   }
 
-  state.SetBytesProcessed(0);
+  state.SetBytesProcessed(state.iterations() * needles * sizeof(T));
   state.SetItemsProcessed(state.iterations() * needles);
 
   const double gpu_cv         = bench_utils::StatisticsCV(gpu_times);

benchmark/bench/vectorized_search/upper_bound.cu

@@ -1,6 +1,6 @@
 /******************************************************************************
  * Copyright (c) 2011-2023, NVIDIA CORPORATION.  All rights reserved.
- * Modifications Copyright (c) 2024-2025, Advanced Micro Devices, Inc.  All rights reserved.
+ * Modifications Copyright (c) 2024-2026, Advanced Micro Devices, Inc.  All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
@@ -90,7 +90,7 @@ void run_benchmark(
     gpu_times.push_back(duration);
   }
 
-  state.SetBytesProcessed(0);
+  state.SetBytesProcessed(state.iterations() * needles * sizeof(T));
   state.SetItemsProcessed(state.iterations() * needles);
 
   const double gpu_cv         = bench_utils::StatisticsCV(gpu_times);