Enable SVE Support for IP Metric Computation in FP32 functions #1146
Conversation
|
Welcome @ThatikondV! It looks like this is your first PR to zilliztech/knowhere 🎉 |
|
/lgtm |
|
@ThatikondV We're checking the CI |
|
@ThatikondV could you please base to master branch? we've fixed an issue on CI. Thanks. |
Codecov ReportAll modified and coverable lines are covered by tests ✅
❌ Your project status has failed because the head coverage (0.00%) is below the target coverage (50.00%). You can increase the head coverage or adjust the target coverage. Additional details and impacted files@@ Coverage Diff @@
## main #1146 +/- ##
============================
============================
🚀 New features to boost your workflow:
|
|
@ThatikondV could you please fix the formatting by running clang-format or applying the following diff? diff --git a/src/simd/distances_sve.cc b/src/simd/distances_sve.cc
index 17dae74..94211f2 100644
--- a/src/simd/distances_sve.cc
+++ b/src/simd/distances_sve.cc
@@ -40,7 +40,7 @@ fvec_L2sqr_sve(const float* x, const float* y, size_t d) {
return svaddv_f32(svptrue_b32(), sum);
}
-float
+float
fvec_inner_product_sve(const float* x, const float* y, size_t d) {
svfloat32_t sum = svdup_f32(0.0f);
size_t i = 0;Thanks |
Signed-off-by: Thatikonda.Varunreddy <Thatikonda.Varunreddy@fujitsu.com>
|
@alexanderguzhva I've fixed the formatting issue using clang-format. Please take a look and let me know if anything else is required. |
|
/lgtm |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: alexanderguzhva, ThatikondV The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
Description:
This PR introduces SVE (Scalable Vector Extension) enablement for IP metric computation in FP32 function. These enhancements provide notable performance improvements over the existing NEON implementation, especially on ARM platforms with SVE capabilities.
Changes in This PR:
Benchmark Results:
Performance benchmarks were conducted comparing NEON and SVE on ARM. The table below summarizes execution time (in seconds) across supported algorithms:

Key observations:
/kind improvement
Fixes #782