From dafe7cef978278f987a628b2195a9f6654ef0a34 Mon Sep 17 00:00:00 2001 From: Varun Jaiswal <96684656+varun-jaiswal17@users.noreply.github.com> Date: Wed, 27 May 2026 15:43:26 +0530 Subject: [PATCH] Merge pull request #29156 from varun-jaiswal17:fix/msvc-windows MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit fix MSVC error: use name constexp for alignas array size in INT8 kernel #29156 ### Problem Windows CI fails with MSVC error C2131 in `conv2_int8_kernels.simd.hpp`: error C2131: expression did not evaluate to a constant failure was caused by a read of a variable outside its lifetime see usage of 'this' Two array declarations inside `parallel_for_` lambdas used `constexpr` variables defined inside the lambda body as array sizes: alignas(32) int8_t wbuf[128 * K0]; // K0 defined inside lambda alignas(32) int32_t sumbuf[SPAT_BLOCK_SIZE * K0]; // both defined inside lambda ### Fix Move the combined size constants to function scope (before the lambda), where they are true compile-time constants with no `this` involvement: constexpr int WBUF_SIZE = 128 * 8; // 128 * K0 constexpr int SUMBUF_SIZE = 8 * 8; // SPAT_BLOCK_SIZE * K0 ### Related Fixes Windows CI failure introduced by #29126. ### 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 --- .../dnn/src/int8layers/conv2_int8_kernels.simd.hpp | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) diff --git a/modules/dnn/src/int8layers/conv2_int8_kernels.simd.hpp b/modules/dnn/src/int8layers/conv2_int8_kernels.simd.hpp index 93c652aca8..58fee00b73 100644 --- a/modules/dnn/src/int8layers/conv2_int8_kernels.simd.hpp +++ b/modules/dnn/src/int8layers/conv2_int8_kernels.simd.hpp @@ -1163,6 +1163,9 @@ static void convInt8BlockDepthwise(const void* inp_, const void* residual_, ((planeblocks_ + nthreads * 4 - 1) / (nthreads * 4) + SB_TILE - 1) & ~(SB_TILE - 1)); int ntiles = (planeblocks_ + TILE - 1) / TILE; int total_tasks = N * K1 * ntiles; +#if CV_AVX2 && (defined(__x86_64__) || defined(_M_X64)) + constexpr int WBUF_SIZE = 128 * 8; // 128 * K0, K0=8; named for MSVC +#endif parallel_for_(Range(0, total_tasks), [&](const Range& range) { int H = H_, W = W_; @@ -1193,9 +1196,8 @@ static void convInt8BlockDepthwise(const void* inp_, const void* residual_, if (ksize <= 128) { int n_g = g_end - g_start; constexpr int SB = 8; - // Pre-gather weights for this group block into wbuf[ksize * K0]. - alignas(32) int8_t wbuf[128 * K0]; + alignas(32) int8_t wbuf[WBUF_SIZE]; for (int i = 0; i < ksize; i++) { for (int j = 0; j < n_g; j++) { int g = g_start + j; @@ -1584,13 +1586,16 @@ void convInt8Block(const void* inp_, const void* residual_, } int total_blocks = N * ngroups * Kblk; +#if CV_AVX2 + constexpr int SUMBUF_SIZE = 8 * 8; // SPAT_BLOCK_SIZE * K0, both=8; named for MSVC +#endif parallel_for_(Range(0, total_blocks), [&](const Range& range) { constexpr int C0 = 8, K0 = 8; constexpr int C0shift = 3; #if CV_AVX2 constexpr int SPAT_BLOCK_SIZE = 8; - alignas(32) int32_t sumbuf[SPAT_BLOCK_SIZE * K0]; + alignas(32) int32_t sumbuf[SUMBUF_SIZE]; #endif const uint8_t xor_val = inputIsU8 ? 0x80 : 0;