core: fix use-after-scope when Mat::mul() is given a scalar #29447
### Summary
`cv::Mat::mul()` called with a scalar returns a `MatExpr` that reads a dead stack slot when it is evaluated. In a normal (non-instrumented) build this produces silently wrong values as soon as the slot is reused:
```cpp
static cv::MatExpr makeExpr(const cv::Mat& m)
{
return m.mul(7); // 7.0 is a temporary double in THIS frame
}
cv::Mat matrix(2, 3, CV_32FC1, cv::Scalar(3.0f));
cv::MatExpr expr = makeExpr(matrix);
// ... any further calls reuse the dead frame ...
cv::Mat result = expr; // observed: all 0, expected: all 21
```
Under AddressSanitizer this is the `stack-use-after-scope` reported in #23577, with the same stack trace (`cvt64s` -> `convertAndUnrollScalar` -> `arithm_op` -> `multiply` -> `MatOp_Bin::assign`).
Storing the expression is the documented lazy-evaluation usage of `MatExpr`; the argument is ordinary supported API usage (`mat.hpp` itself shows `Mat C = A.mul(5/B);`).
### Root cause
A scalar argument binds to `_InputArray(const double& val)`, which records the **address** of the temporary with kind `MATX`:
```cpp
inline _InputArray::_InputArray(const double& val)
{ init(FIXED_TYPE + FIXED_SIZE + MATX + CV_64F + ACCESS_READ, &val, Size(1,1)); }
```
`Mat::mul()` then parks `m.getMat()` inside the returned `MatExpr`. For `MATX` kind, `getMat_()` returns a non-owning, non-refcounted header over that stack memory (`return Mat(sz, flags, obj);`). The temporary dies at the end of the full expression, but the `MatExpr` keeps the header, and `MatOp_Bin::assign()` later feeds it to `cv::multiply()`. `Matx`/`Vec` arguments take the same path.
`Mat::mul()` is the only `MatExpr` factory in `matrix_expressions.cpp` that takes an `InputArray`; every other scalar operand there is stored by value in the `Scalar` member (`e.s`), so no other expression path can capture a stack pointer this way.
### Fix
Snapshot the operand with `clone()` unless it is a `Mat`/`UMat`, which keep the current zero-copy behaviour: their headers are refcounted and already safe to defer. Any other `InputArray` kind (a scalar, `Matx`, `Vec`, `std::vector`, an evaluated expression) is a potentially non-owning view, so it is copied once at expression construction, off any hot path.
### Test
Adds `Core_MatExpr.mul_scalar_use_after_scope_23577` to `modules/core/test/test_operations.cpp`. It builds the expression in a helper frame and overwrites the stack before evaluating; the helpers are called through volatile function pointers so they cannot be inlined, which makes the stale read deterministic. The test fails before the fix (result is all 0 instead of all 21) and passes after. It is self-contained: no opencv_extra data is needed.
Verified locally on macOS/AArch64 (Apple clang 17, Release): full `opencv_test_core` passes, and the AddressSanitizer reproducer from the issue is clean after the fix.
Fixes#23577.
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Self-contained accuracy regression test in `modules/core/test/test_operations.cpp`; no opencv_extra data required. No performance test: no existing perf test covers `Mat::mul` expression construction, and the copy happens once at expression construction, only for non-`Mat`/`UMat` operands.
- [ ] The feature is well documented and sample code can be built with the project CMake
N/A - bug fix, no new API.
core: Fix mul32f and addWeighted32f to use native f32 SIMD paths #29413
OpenCV Extra: https://github.com/opencv/opencv_extra/pull/1388
- add 32FC1 coverage to addWeighted benchmark
- avoid intermediate double for f32 variants of scaled multiply and addWeighted.
- Relax AddWeighted 32F test tolerance to match f32 FMA semantics.
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
core: SIMD optimizations for norm, distance and Hamming APIs. (Improves ORB & BRISK) #29335
Generic universal-intrinsic kernels (benefit all SIMD backends: NEON, AVX2, AVX-512, etc.), plus enabling wider dispatch for the norm module.
- hal::normHamming: cached-pointer dispatch (resolve once, no per-call dispatch chain or trace region) + vector popcount path. cv::norm(NORM_HAMMING) and binary-descriptor matching (BFMatcher ORB/BRISK/FREAK via cv::batchDistance).
- hal::normL2Sqr_ / normL1_: direct inlinable kernels with single-vector tail (cv::batchDistance / BFMatcher float L2/L1, cv::kmeans).
- cv::norm masked NORM_INF: deinterleave SIMD for multichannel + back-step tail.
- cv::norm(src1, src2, type, mask): SIMD masked norm-diff; INF is one templated kernel for all element types, plus uchar L1/L2 and int L1 kernels.
- Enable AVX512_SKX/AVX512_ICL dispatch for the norm module.
- features2d: add BFMatcher knnMatch perf tests (float L2/L1, binary Hamming).
- ORB and BRISK performance improved
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
Optimized gemm implementation with Universal SIMD #29242
- vectorized GEMMSingleMul and GEMMBlockMul
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
core: fix Unicode temp path handling on Windows (fix#21960) #28935
## Summary
Fixes getCacheDirectoryForDownloads and related temp file functions failing when the user home directory contains Unicode characters like C:\\Users\\テスト\\AppData\\Local\\Temp.
## Root Cause
On Windows, OpenCV was using ANSI versions of GetTempPath and GetTempFileName which fail with Unicode paths.
## Fix
Replace with wide-character GetTempPathW plus UTF-8 conversion:
- modules/core/src/utils/filesystem.cpp - getCacheDirectory for cache paths
- modules/core/src/system.cpp - temporary file creation
- modules/ts/src/ts_gtest.cpp - test stream capture
## Testing
The fix has been verified compiles. Manual testing on Windows with Unicode username required.
## Risk
Low: Uses same output format, just different encoding path. Existing ASCII functionality preserved.
---
Fixesopencv/opencv#21960
fix MSVC warning for minMaxIdx_simd #29379
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
Use signmask+popcount only in AVX-512 dispatch units; keep the legacy
batched SIMD kernels for AVX2, NEON, and LASX to avoid regressions on
narrow SIMD widths.
Fast-path transposeND for identity and 2D transpose orders #29172
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
<!-- Note!!! If you are an automated agent, we have a special process for you: add 🤖🤖🤖 to the end of the PR title. -->
This PR adds fast paths to cv::transposeND() for two common cases:
- identity permutation: dispatch to copyTo()
- 2D permutation {1, 0}: dispatch to transpose()
All other permutations continue to use the existing generic ND implementation.
Why:
transposeND() currently falls back to the generic memcpy/index-update loop even for the common 2D transpose case, while OpenCV already has an optimized transpose() path. Reusing that path avoids unnecessary index arithmetic and improves performance for 2D inputs passed through transposeND().
Performance:
[BinaryOpTest.transposeND/21 (1920x1080, 8UC3): 28.21 ms -> 0.66 ms (-97.7%)]
[BinaryOpTest.transposeND/22 (1920x1080, 8UC4): 37.61 ms -> 0.88 ms (-97.7%)]
[BinaryOpTest.transposeND/29 (1920x1080, 32FC1): 10.26 ms -> 0.75 ms (-92.7%)]
Full transposeND perf subset: 6760 ms -> 592 ms (-91.2%)
core: Vectorize meanStdDev #29339
- Add SIMD SumSqr_SIMD reductions (previously scalar) plus AVX-512 dispatch for the mean TU.
- Uses universal intrinsics, so all SIMD backends benefit.
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
FileNode::operator double() and operator float() read an INT node via readInt()
(32-bit), truncating values above INT_MAX -- e.g. an integer 6662329666 from an
externally-produced json/yaml/xml is read back as -1927604926. The node stores
the value as int64 (operator int64_t() already reads it correctly via readLong),
so use readLong() for the floating-point conversions too. Values that fit in
int32 are unchanged (sign-extended); larger ones are now correct.
Reader side of #29363 (the writer side was #29364).
- Replace the fixed-128-bit per-depth minMaxIdx kernels with a single templated universal-intrinsic core
- added simd dispatch AVX-512 dispatch
- A dual-accumulator inner loop added
Uses universal intrinsics, so all SIMD backends benefit.
imgproc: optimized LUT and equalizeHist with SIMD #29250
- optimized lut with SIMD
- support equalizeHist with v_lut
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
Restricting the condition with _M_IX86/_M_X64 so it only applies to x86/x64 MSVC builds. MSVC ARM64 now falls through to the existing `#else` branch, which already has a portable CV_SIMD-based exp32f/exp64f implementation
**Performance Benchmarks:**
<img width="976" height="486" alt="image" src="https://github.com/user-attachments/assets/62daf2c3-34ac-4fc7-92d0-268073f746f3" />
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
More use of AutoBuffer #28907
When possible, AutoBuffer should be faster than std::vector<>, and should not be worse if it requires a heap allocation rather than a stack allocation.
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [X] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [X] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [ ] The feature is well documented and sample code can be built with the project CMake
core(opengl): fix memory leak in OpenCL extensions gathering #28981
## Problem
Fixes#28980
In `modules/core/src/opengl.cpp`, inside `initializeContextFromGL()`, a `char[]` buffer is allocated to query OpenCL device extensions:
```cpp
extensions = new char[extensionSize];
status = clGetDeviceInfo(..., extensions, &extensionSize);
if (status != CL_SUCCESS)
continue; // leaks `extensions`
```
When `clGetDeviceInfo()` fails, `continue` skips the corresponding `delete[]` on the success path, causing the allocated buffer to leak.
Additionally, the `catch (...)` block also bypasses cleanup, so any thrown exception leaks the buffer as well.
## Fix
Replace the raw `char*` allocation with `std::unique_ptr<char[]>`.
This ensures the buffer is automatically released on all exit paths, including:
- normal execution
- early `continue`
- exception handling paths
## Checklist
- [x] I agree to contribute to the project under the Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on code under GPL or another license incompatible with OpenCV.
- [x] The PR is proposed to the proper branch (`4.x`).
- [x] There is a reference to the original bug report and related work, if applicable.
- [x] Accuracy/performance tests and test data in `opencv_extra` are included, if applicable.
- [x] The feature/change is documented and sample code can be built with the project CMake configuration, if applicable.
core: fix inverted continuity check in cvReshapeMatND() #29132
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [√] I agree to contribute to the project under Apache 2 License.
- [√] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [ ] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [ ] The feature is well documented and sample code can be built with the project CMake
Better Durand-Kerner Initialization #29109
While investigating issue #23644, I have found [this paper](https://link.springer.com/article/10.1007/BF01935059) which presents a good initialization for the Durand-Kerner algorithm. Basically the idea is to put the initial points equidistantly on a circle on the complex plane. The radius of the circle is computed as
<img width="607" height="178" alt="image" src="https://github.com/user-attachments/assets/ea31b002-c924-4b93-9334-3e59597c896b" />
Note that the $a_i$ coefficients in that paper are reversed compared to OpenCV. That's where the `(n - i)` in the code comes from.
I have implemented just the mean of the $u_i$'s for the sake of simplicity. That's already enough to make the algorithm converge in all cases I have tested. I have used this to test for convergence for many polynomials of order 2 and 4 and coefficients of different magnitudes:
```cpp
TEST(Core_SolvePoly, large_test)
{
cv::Mat_<float> coefs3(1,3);
cv::Mat_<float> coefs5(1,5);
cv::Mat r;
double prec;
for (int c0 = -20; c0 <= 20; c0++)
{
coefs3.at<float>(0) = c0;
for (int c1 = -20; c1 <= 20; c1++)
{
coefs3.at<float>(1) = c1;
for (int c2 = -20; c2 <= 20; c2++)
{
coefs3.at<float>(2) = c2;
prec = cv::solvePoly(coefs3, r);
EXPECT_LE(prec, 1e-6);
}
}
}
for (int c0 = -10; c0 <= 10; c0++)
{
coefs5.at<float>(0) = c0;
for (int c1 = -10; c1 <= 10; c1++)
{
coefs5.at<float>(1) = c1;
for (int c2 = -10; c2 <= 10; c2++)
{
coefs5.at<float>(2) = c2;
for (int c3 = -10; c3 <= 10; c3++)
{
coefs5.at<float>(3) = c3;
for (int c4 = -10; c4 <= 10; c4++)
{
coefs5.at<float>(4) = c4;
prec = cv::solvePoly(coefs5, r);
EXPECT_LE(prec, 1e-2);
}
}
}
}
}
for (int i = -10; i < 10; i++)
{
coefs3.at<float>(0) = pow(2, i);
for (int j = -10; j < 10; j++)
{
coefs3.at<float>(1) = pow(2, j);
for (int k = -10; k < 10; k++)
{
coefs3.at<float>(2) = pow(2, k);
prec = cv::solvePoly(coefs3, r);
EXPECT_LE(prec, 1e-6);
}
}
}
}
```
This test passes, but I have not committed it because it runs for a couple of seconds.
This fixes#23644 and replaces #29055. I have checked #29055 and it does not pass the test above. It seems to be optimized to the precise polynomial of #23644.
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
core(rvv): fix v_matmul/v_matmuladd scalable semantics and expand lane-group test coverage #29080
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
<!-- Note!!! If you are an automated agent, we have a special process for you: add 🤖🤖🤖 to the end of the PR title. -->
## Platform
SpacemiT X60 (K1), 8-core RISC-V RVV 1.0, VLEN=256, 16GB RAM, OS: Bianbu Linux (kernel 6.6.63), GCC 13.2.0, Build: OpenCV 4.14.0-pre, Release, HAL: YES (RVV HAL 0.0.1)
## Motivation
OpenCV's Universal Intrinsics `v_matmul` and `v_matmuladd` have a semantic bug in the RVV scalable backend (`modules/core/include/opencv2/core/hal/intrin_rvv_scalable.hpp`).
The current implementation uses `v_extract_n(v, 0/1/2/3)` with hardcoded indices, assuming the vector holds exactly 4 float lanes (128-bit fixed). On hardware with VLEN=256 (e.g. SpacemiT K1 / BPI-F3), `v_float32` with LMUL=2 holds 16 lanes. As a result, lanes 4–15 silently reuse the inputs from lanes 0–3, producing wrong results.
OpenCV itself acknowledges this in `modules/core/src/matmul.simd.hpp`:
// v_matmuladd for RVV is 128-bit only but not scalable,
// this will fail the test Core_Transform.accuracy
The RVV scalable `transform_32f` path has been disabled because of this bug. However, the existing `TheTest<R>::test_matmul()` only checked the first 4-lane group (the outer loop was effectively hardcoded to `int i = 0`), so the bug was never caught by CI even on wide-vector backends.
## Modification
**Test fix** (`modules/core/test/test_intrin_utils.hpp`): Expanded `test_matmul()` to iterate over all 4-lane groups:
// Before (only checked lane group i=0)
int i = 0;
for (int j = i; j < i + 4; ++j) { ... }
// After (checks all lane groups)
for (int i = 0; i < VTraits<R>::vlanes(); i += 4)
{
for (int j = i; j < i + 4; ++j) { ... }
}
**Kernel fix** (`modules/core/include/opencv2/core/hal/intrin_rvv_scalable.hpp`): Rewrote `v_matmul` and `v_matmuladd` to process all 4-lane groups correctly. Each group of 4 lanes now independently computes the full matrix multiply using its own `v[i], v[i+1], v[i+2], v[i+3]` inputs. The `transform_32f` RVV path in `matmul.simd.hpp` remains disabled as the autovectorized path shows better performance on current hardware.
## Experiment 1: Bug reproduced on SpacemiT K1 (VLEN=256)
./opencv_test_core --gtest_filter="*intrin*"
Result: `hal_intrin128.float32x4_BASELINE` FAILED with 24 failures, all from lane groups i=4, i=8, i=12 (lanes 4–15).
Representative failures from `v_matmul` (line 1526):
i=4 j=4: actual=158 expected=56
i=4 j=5: actual=166.39999 expected=59.200001
i=8 j=8: actual=314.39999 expected=68.800003
i=12 j=12: actual=512.40002 expected=81.599998
Representative failures from `v_matmuladd` (line 1540):
i=4 j=4: actual=147.5 expected=51.5
i=8 j=8: actual=284.70001 expected=60.700001
i=12 j=12: actual=453.89999 expected=69.900002
Lane group i=0 (j=0..3) passed correctly — confirming the bug only affects lanes beyond the first 4, exactly as expected from the hardcoded `v_extract_n(v, 0/1/2/3)` implementation.
## Experiment 2: Both tests pass after fixing the kernel
./opencv_test_core --gtest_filter='hal_intrin128.float32x4_BASELINE'
[ OK ] hal_intrin128.float32x4_BASELINE (1859 ms)
[ PASSED ] 1 test.
./opencv_test_core --gtest_filter='Core_Transform.accuracy'
[ OK ] Core_Transform.accuracy (819 ms)
[ PASSED ] 1 test.
## Experiment 3: RVV transform path remains disabled (performance regression)
After re-enabling the RVV scalable `transform_32f` path experimentally, benchmarks showed a significant regression vs the compiler-autovectorized scalar path (CV_32FC3):
Size RVV path Scalar path Ratio
640x480 7.83 ms 1.48 ms 5.3x slower
1280x720 23.96 ms 5.17 ms 4.6x slower
1920x1080 53.76 ms 10.12 ms 5.3x slower
The compiler-autovectorized path outperforms the hand-written RVV kernel for this workload, consistent with the original comment in `matmul.simd.hpp`. The `transform_32f` RVV path is therefore kept disabled in this PR. The kernel fix to `v_matmul`/`v_matmuladd` remains necessary for correctness on wide-vector hardware, and the expanded test ensures the bug cannot regress silently in future.
core : add NEON intrinsics support for norm_mask function #28610
- This PR adds NEON intrinsics-based implementations for masked norm operations in norm.simd.hpp for ARM64 architecture.
- The optimized implementation uses ARM NEON intrinsics to accelerate masked norm computations (Infinity norm, L1 norm, and L2 norm) used by the norm function when a mask is provided.
- In the x64 architecture, masked norm operations benefit from IPP-based optimized implementations. However, on ARM64, the execution falls back to scalar implementations, which results in lower performance.
- To achieve performance parity with x64, NEON-based SIMD implementations have been added for ARM64.
- Additionally, scalar loop unrolling optimizations have been added for non-masked norm operations.
- After introducing these changes, masked norm operations showed significant performance improvements on ARM64 platforms, particularly for single-channel (cn=1) operations where NEON intrinsics provide the greatest benefit.
<img width="952" height="822" alt="image" src="https://github.com/user-attachments/assets/12d35f93-a316-4520-9d9c-12ce6b371ddb" />
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
The previous OpenCL implementation of UMat::diag() relied on a sequence of
operations involving buffer initialization followed by copy/transpose on
an aliased diag() view. Although these operations were enqueued on an
in-order command queue, this implicitly assumed memory visibility between
kernels operating on aliased regions of the same buffer.
According to the OpenCL specification, in-order command queues only guarantee
command scheduling order, while memory visibility between commands is only
established through explicit command-level synchronization points (e.g.
events, barriers, or clFinish).Under OpenCL’s relaxed memory model, such
assumptions are not guaranteed without an explicit command-level synchronization
point, and can lead to data races and incorrect results, especially for very
small matrices (e.g. 1x1). The issue is more likely to be exposed on Mesa-based drivers
when the GPU is running at lower frequencies (e.g. 500 MHz).
This change introduces a dedicated OpenCL kernel to construct the diagonal
matrix in a single kernel invocation, avoiding intermediate aliasing and
eliminating the need for implicit ordering assumptions. If the OpenCL path
is unavailable or unsupported, the implementation transparently falls back
to the CPU path.
Signed-off-by: jiajia Qian <jiajia.qian@nxp.com>
The minMaxIdx dispatch table in modules/core/src/minmax.cpp stored
seven differently typed functions through C-style casts to a single
MinMaxIdxFunc pointer type, which is undefined behaviour and trips
UBSan's -fsanitize=function for any CV_32F or CV_64F input. This
change switches the typedef and every depth-specific helper to take
void pointers for src, minval and maxval, with the original typed
pointer recovered at function entry. The dispatch table no longer
needs C-style casts and the int pointer punning at the call site
goes away. Fixes#28928.
- Replace unsafe pointer arithmetic and direct buffer modification with std::string methods.
- Update documentation to clarify that the last digit is used as compression level and truncated from the actual filename.
- Add test cases for .gz and .gz0-9
Previously, OpenCL in-place flip kernels (rows/cols/both) could produce
incorrect results compared to the CPU implementation on strict OpenCL
drivers such as Mesa. The kernels relied on implicit load–load–store–store
(LLSS) ordering when src and dst alias, which is not guaranteed by the
OpenCL memory model and may be reordered.
Some vendor drivers happened to preserve the expected ordering, masking
the issue, but Mesa correctly exposes the undefined behavior.
This change introduces dedicated in-place flip kernels that:
- Explicitly detect in-place execution (src == dst)
- Stage data through local memory tiles
- Enforce correct ordering with work-group barriers
- Avoid global memory read/write aliasing hazards
The non in-place path is unchanged.
With this fix, OpenCL in-place flip produces correct and consistent results
across drivers, matches CPU behavior, and complies with the OpenCL memory
model.
Signed-off-by: jiajia Qian <jiajia.qian@nxp.com>
core(opencl): fix inplace transpose race by enforcing LLSS ordering via local barrier #28686
The former inplace transpose implementation allowed a reordering of global-memory operations across work-items. Specifically, the intended LLSS (Load–Load–Store–Store) access pattern could be reordered by the GPU into LSLS (Load–Store–Load–Store), causing partially written tiles to be observed by other work-items and producing incorrect output.
This patch introduces a tiled LDS-based algorithm and adds an explicit:
barrier(CLK_LOCAL_MEM_FENCE);
between the load and store phases.
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [ ] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [ ] The feature is well documented and sample code can be built with the project CMake
core: add NEON implementation for rotate function #28609
- This PR adds a NEON intrinsics-based implementation for the rotate function in matrix_transform.cpp for Windows-ARM64.
- The optimized implementation uses ARM NEON intrinsics to accelerate the internal transpose step used by the rotate function.
- In the x64 architecture, the rotate operation benefits from IPP-based optimized implementations. However, on ARM64, the execution falls back to the scalar implementation, which results in lower performance.
- To achieve performance parity with x64, a NEON-based SIMD implementation has been added for ARM64.
- After introducing these changes, the rotate function showed noticeable performance improvements on ARM64 platforms.
<img width="1009" height="817" alt="image" src="https://github.com/user-attachments/assets/8bec0041-b19c-4fc8-9103-532746224515" />
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
Optimized flip Horizontal #28614
- Refactor flipHoriz implementation.
- Optimizations for horizontal image flipping added.
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
core: fix heap-buffer-overflow in YAML parseKey for empty keys #28620
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [ ] The feature is well documented and sample code can be built with the project CMake
### Description
Fixes https://github.com/opencv/opencv/issues/28619
Moves the "empty key" check before the backward do-while scan in `YAMLParser::parseKey()`.
**Problem:** When parsing a YAML mapping with an empty key (e.g. `: 10` at column 0), `endptr == ptr` after the forward scan finds `:`. The do-while loop `do c = *--endptr; while(c == ' ')` always executes at least once, so it decrements `endptr` to `ptr-1` and reads one byte before the heap allocation (ASan: heap-buffer-overflow READ of size 1).
**Fix:** Check `endptr == ptr` before entering the backward loop. If the key is empty, raise `CV_PARSE_ERROR_CPP("An empty key")` immediately without the OOB read.
This contribution was developed with AI assistance (Claude Code).
Alexander Smorkalov
MAAZIZ Adel Ayoub
Alexander Smorkalov
Madan mohan Manokar
Mulham Fetna
胡晨宇
Tonu Samuel
uwezkhan
Pratham Kumar
Pierre Chatelier
Srujan rai
Tiansuanyu
Adrian Kretz
kevinylin88
jiajia Qian
Samaresh Kumar Singh
Kumataro
kjg0724
Lurie97
Vincent Rabaud
Abhishek Gola
usernotfound-101
Matt Van Horn