github_mirrors/vectorscan
3
0
Fork 1
mirror of https://github.com/VectorCamp/vectorscan.git synced 2025-11-18 18:20:35 +03:00
Code Issues Packages Projects Releases Wiki Activity

Speed up truffle with 256b TBL instructions

Browse Source 
256b wide SVE vectors allow some simplification of truffle.
Up to 40% speedup on graviton3. Going from 12500 MB/s to 17000 MB/s
onhe microbenchmark.
SVE2 also offer this capability for 128b vector with a speedup around
25% compared to normal SVE

Add unit tests and benchmark for this wide variant

Signed-off-by: Yoan Picchi <yoan.picchi@arm.com>
This commit is contained in:
Yoan Picchi
2024-04-23 12:04:40 +00:00
parent b312112e87
commit 938c026256
23 changed files with 1125 additions and 78 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
src/hwlm/hwlm.c
View File

@@ -73,7 +73,12 @@ const u8 *run_hwlm_accel(const union AccelAux *aux, const u8 *ptr,
return shuftiExec(aux->shufti.lo, aux->shufti.hi, ptr, end);
case ACCEL_TRUFFLE:
DEBUG_PRINTF("truffle\n");
return truffleExec(aux->truffle.mask1, aux->truffle.mask2, ptr, end);
return truffleExec(aux->truffle.mask_lo, aux->truffle.mask_hi, ptr, end);
#ifdef CAN_USE_WIDE_TRUFFLE
case ACCEL_TRUFFLE_WIDE:
DEBUG_PRINTF("truffle wide\n");
return truffleExecWide(aux->truffle.mask, ptr, end);
#endif // CAN_USE_WIDE_TRUFFLE
default:
/* no acceleration, fall through and return current ptr */
DEBUG_PRINTF("no accel; %u\n", (int)aux->accel_type);