We hereby present the source code used to evaluate TPC-H Query 01 optimized for CPU-GPU co-processing. (Paper accepted to ADMS/VLDB 2018)
| Implementation Flavor | Split CPU-GPU Computation | Filter Pushdown | Hash table Placement | Compression | Time (sec) |
|---|---|---|---|---|---|
| Global full | - | - | Global | - | 12.60 |
| In-register full | - | - | Register | - | 12.45 |
| Local full | - | - | Local | - | 12.40 |
| Local fp small | - | √ | Local | √ | 0.76 |
| In-register fp small | - | √ | Register | √ | 0.76 |
| Global fp small | - | √ | Global | √ | 0.76 |
| Global small | - | - | Global | √ | 0.74 |
| In-register small | - | - | Register | √ | 0.68 |
| Local small | - | - | Local | √ | 0.57 |
| Global sc small | √ | - | Global | √ | 0.51 |
| In-register sc small | √ | - | Register | √ | 0.43 |
| Local sc small | √ | - | Local | √ | 0.38 |
| SharedMemory sc small | √ | - | Shared | √ | 0.37 |
- CUDA v9.0 or later is recommended.
- A C++14-capable compiler compatible with your version of CUDA; only GCC has been tested.
- CMake v3.1 or later
- A Unix-like environment for the (simple) shell scripts; without it, you may need to perform a few tasks manually
- The cuda-api-wrappers library; it can be obtained
Assuming you've cloned into /path/to/tpchQ1:
- Configure the build and generate build files using
cmake /path/to/tpchQ1 - Build using either your default make'ing tool or with
cmake --build /path/to/tpchQ1; this will also generate input data for Scale Factor 1 (SF 1) - Optionally: Generate data for more scale factors - SF 10 and SF 100 - using
make -C /path/to/tpchQ1 data_table_sf10ormake -C /path/to/tpchQ1 data_table_sf100.
- When building, the binary
tpch_01is generated; - Invoke
bin/tpch_01in the directory where you've performed the build. That directory should have atpch/subdirectory with subdir for every scale factor with generated data. - You can use
scripts/genlineitem.shto manually generate a table for arbitrary scale factors, but you must place it under your build dir, undertpch/123.000000(for generated scale factor 123). Non-integral scale factors should work as well.
| Switch | Value range | Default value | Meaning |
|---|---|---|---|
| --apply-compression | N/A | (off) | Use the compression schemes described on the Wiki, to reduce the amount of data for transmission over PCI/e |
| --print-results | N/A | (off) | Print the computed aggregates to std::cout after every run. Useful for debugging result stability issues. |
| --use-filter-pushdown | N/A | (off) | Have the CPU check the TPC-H Q1 WHERE clause condition, passing only that result bit vector to the GPU. It's debatable whether this is actually a "push down" in the traditional sense of the term. |
| --use-coprocessing | N/A | (off) | Schedule some of the work to be done on the CPU and some on the GPU |
| --hash-table-placement | in-registers, local-mem, per-thread-shared-mem, global | in-registers | Memory space + granularity for the aggregation tables; see the paper itself or the code for an explanation of what this means. |
| --sf= | Integral or fractional number, limited precision | 1 | Which scale factor subdirectory to use (to look for the data table or cached column files). For sf 123.456789, data will be expected under tpch/123.456789 |
| --streams= | Positive integral value | 4 | The number of concurrent streams to use for scheduling GPU work. You should probably not change this. |
| --threads-per-block= | Positive integral number, preferably a multiple of 32 | 256 | Number of CUDA threads per block of a scheduled kernel of the computational work. |
| --tuples-per-thread=H | Positive integral number, preferably high | 1024 | The number of tuples each thread processes individually before merging results with other threads |
| --tuples-per-kernel= | Positive integral number, preferably a multiple of threads-per-block | 1024 | Every how many input tuples is a new kernel launched? |
- Paper Submission: Friday, 8 June, 2018 (Extended to 11 June, 2018)
- Notification of Acceptance: Friday, 29 June, 2018
- Camera-ready Submission: Friday, 20 July, 2018
- Workshop Date: Monday, 27 August, 2018
The query text and column information is on the Wiki. For further information about the benchmark it is part of, see the Transaction Processing Council's page for TPC-H