| Age | Commit message (Collapse) | Author | Lines | |
|---|---|---|---|---|
| 2024-07-23 | BS baseline: Add memcpy variant | Birte Kristina Friesel | -42/+153 | |
| 2024-07-10 | BS: No --resume | Birte Kristina Friesel | -12/+12 | |
| 2024-07-10 | BS hetsim: adjust for tinos | Birte Kristina Friesel | -4/+4 | |
| 2024-07-08 | BS: nits | Birte Kristina Friesel | -9/+13 | |
| 2024-07-05 | BS NMC: log NUMA data | Birte Kristina Friesel | -12/+68 | |
| 2024-07-04 | BS: nits | Birte Kristina Friesel | -4/+4 | |
| 2024-07-04 | BS: configurable input size | Birte Kristina Friesel | -3/+7 | |
| 2024-07-04 | BS: Add separate files for HBM and NMC benchmarks | Birte Kristina Friesel | -40/+152 | |
| 2024-07-04 | BS: Remove unused SDK_SINGLETHREADED ; compile baseline with -march=native | Birte Kristina Friesel | -15/+8 | |
| 2024-07-04 | BS: Add NUMA variant for HBM | Birte Kristina Friesel | -15/+123 | |
| 2024-03-13 | BS: throughput is based on num queries, not input size | Birte Kristina Friesel | -13/+13 | |
| 2024-03-08 | BS: avoid double free | Birte Kristina Friesel | -0/+2 | |
| 2024-03-08 | BS: handle alloc/load/free overhead as parameters; vary them for fgbs24a | Birte Kristina Friesel | -4/+11 | |
| 2024-02-28 | more run-fgbs24a | Birte Kristina Friesel | -0/+28 | |
| 2023-12-01 | BS: add overhead measurements | Birte Kristina Friesel | -46/+144 | |
| 2023-08-03 | run-opti: PrIM always uses four threads for CPU baseline | Birte Kristina Friesel | -2/+2 | |
| 2023-07-24 | BS: adjust #DPUs for tinos | Birte Friesel | -2/+3 | |
| 2023-07-24 | BFS, BS: remove duplicate configurations | Birte Friesel | -2/+2 | |
| 2023-07-06 | run-strong-full: add 2544 DPUs (maximum possible on tinos @ UOS) | Birte Friesel | -1/+2 | |
| 2023-06-01 | BS: updates and comments | Daniel Friesel | -18/+33 | |
| 2023-05-31 | BS: add 256 and 512 DPUs to benchmarks | Daniel Friesel | -2/+4 | |
| 2023-05-30 | BS: switch to new dfatool format; add -O and reproduction scripts | Daniel Friesel | -15/+92 | |
| 2023-05-25 | port BS NMC to dfatool | Daniel Friesel | -45/+89 | |
| 2023-05-25 | port BS CPU to dfatool | Daniel Friesel | -6/+70 | |
| 2021-12-23 | bs_refactor: Remove printf for logging | Yun-Ze Li | -8/+2 | |
| Signed-off-by: Yun-Ze Li <p76091292@gs.ncku.edu.tw> | ||||
| 2021-12-23 | bs_refactor: Remove redundant lines | Yun-Ze Li | -7/+4 | |
| This commit reduce redundant lines for calculating `current_mram_block_addrA`. Signed-off-by: Yun-Ze Li <p76091292@gs.ncku.edu.tw> | ||||
| 2021-12-23 | bs_bug_fix: Modify boundary case handlings in BS | Yun-Ze Li | -41/+38 | |
| Current boundary case handling may discard some numbers obliged to be compared with `searching_for`, which in my opinion can result in false positives (BS fails to identify the number which is indeed in the input array). This commit changed the boundary check condition to `if(current_mram_block_addr_A < start_mram_block_addr_A + BLOCK_SIZE)`, where the expression returns true if and only if the length of the range [start_mram_block_addr_A, end_mram_block_addr_A) has become smaller than 2*BLOCK_SIZE. When this happens, we can then finalize the BS result by checking if `searching_for` exists within [start_mram_block_addr_A, end_mram_block_addr_A) without overlooking any number that should be checked. Signed-off-by: Yun-Ze Li <p76091292@gs.ncku.edu.tw> | ||||
| 2021-12-23 | bs_bug_fix: Make MRAM reads 8-byte aligned | Yun-Ze Li | -0/+15 | |
| Current implementation does not guarantee that `current_mram_block_addr_A` is 8-byte aligned before using it as the start address of `mram_read`s. This commit makes `current_mram_block_addr_A` 8-byte aligned whenever we try to use it for a MRAM read by `current_mram_block_addr_A &= WORD_MASK`, which will clear the unaligned bytes. Signed-off-by: Yun-Ze Li <p76091292@gs.ncku.edu.tw> | ||||
| 2021-06-16 | PrIM -- first commit | Juan Gomez Luna | -0/+1007 | |
 |
index : prim-benchmarks | |
| Processing-in-Memory Benchmarks | derf |
| summaryrefslogtreecommitdiff |