This is an automated email from the ASF dual-hosted git repository.
chaokunyang pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/fory.git
The following commit(s) were added to refs/heads/main by this push:
new cf4439128 test(cpp): optimize float16 sign symmetry test for ASAN
(#3531)
cf4439128 is described below
commit cf4439128cf6b0d33d53d2e22ef43d8d7c47f07b
Author: Geethapranay1 <[email protected]>
AuthorDate: Tue Mar 31 07:50:17 2026 +0530
test(cpp): optimize float16 sign symmetry test for ASAN (#3531)
**Optimize float16 sign symmetry test to fix ASAN slow down**
## Why?
## What does this PR do?
The `SignSymmetryForNonNaNBitPatterns` test was doing a massive
brute-force scan across ~41k float32 patterns. Under ASAN, the heavy
memory instrumentation caused this single test to take ~246 seconds.
Instead of scanning the whole range, this rewrites the test to just
densely sample the actual critical boundaries (subnormals,
underflow/overflow thresholds, and normal bins). We get the exact same
edge-case coverage, but the test now finishes in just **~13ms**.
## Related issues
fix: #3524
## AI Contribution Checklist
- [ ] Substantial AI assistance was used in this PR: `yes` / `no`
- [ ] If `yes`, I included a completed [AI Contribution
Checklist](https://github.com/apache/fory/blob/main/AI_POLICY.md#9-contributor-checklist-for-ai-assisted-prs)
in this PR description and the required `AI Usage Disclosure`.
## Does this PR introduce any user-facing change?
- [ ] Does this PR introduce any public API change?
- [ ] Does this PR introduce any binary protocol compatibility change?
## Benchmark
normal build:
<img width="1125" height="650" alt="Screenshot From 2026-03-30 21-54-21"
src="https://github.com/user-attachments/assets/da74ac5a-8739-45eb-b385-30bc494ecebc";
/>
asan build:
<img width="1106" height="883" alt="Screenshot From 2026-03-30 21-55-36"
src="https://github.com/user-attachments/assets/43eed5c6-f5c3-4fc9-aed4-07d43a39b650";
/>
ASAN runtime down from 246s to 13ms!
---
cpp/fory/util/float16_test.cc | 28 +++++++++++++++++++++++-----
1 file changed, 23 insertions(+), 5 deletions(-)
diff --git a/cpp/fory/util/float16_test.cc b/cpp/fory/util/float16_test.cc
index f8b665152..c111cd17f 100644
--- a/cpp/fory/util/float16_test.cc
+++ b/cpp/fory/util/float16_test.cc
@@ -307,13 +307,31 @@ TEST(Float16FromFloatTest, IntegerAndUlpRegressionCases) {
}
TEST(Float16FromFloatTest, SignSymmetryForNonNaNBitPatterns) {
- for (uint32_t bits = 0; bits < 0xFFFFFFFFu; bits += 104729u) {
- const float value = bits_to_float(bits);
- if (std::isnan(value)) {
- continue;
+ auto test_range = [](uint32_t start, uint32_t end, uint32_t step) {
+ for (uint32_t bits = start; bits < end; bits += step) {
+ const float value = bits_to_float(bits);
+ if (!std::isnan(value)) {
+ ExpectSignSymmetry(value);
+ }
}
- ExpectSignSymmetry(value);
+ };
+
+ test_range(0x00000000u, 0x00800000u, 4099u);
+
+ test_range(0x33000000u, 0x38800000u, 1031u);
+ test_range(0x00800000u, 0x33000000u, 200003u);
+
+ for (uint32_t exp = 113; exp <= 142; ++exp) {
+ const uint32_t base = exp << 23;
+ test_range(base, base + 64u, 1u);
+ test_range(base + 0x007FFF00u, base + 0x00800000u, 1u);
}
+
+ test_range(0x38800000u, 0x47800000u, 50021u);
+
+ test_range(0x47800000u, 0x7F800000u, 100003u);
+
+ ExpectSignSymmetry(bits_to_float(0x7F800000u));
}
TEST(Float16Test, FromBitsRoundTrip) {
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
