diff options
| author | SiCong Li <sicong.li@arm.com> | 2023-11-10 12:16:32 +0000 |
|---|---|---|
| committer | Anitha Raj <anitha.raj@arm.com> | 2023-11-10 14:44:07 +0000 |
| commit | 0e1ccebfd52248fd8ead2614eaf45828d1fab340 (patch) | |
| tree | defdbe671552dfb97336daaa9df0ccca9a95c643 /src/cpu/utils/CpuAuxTensorHandler.h | |
| parent | a63ece730acda74df26281e6341f6fedfb209554 (diff) | |
| download | ComputeLibrary-0e1ccebfd52248fd8ead2614eaf45828d1fab340.tar.gz | |
Fix CpuGemmConv2d int8 segfault
Bypass importation of memory of the original weights into the
reinterpreted_weights auxiliary tensor if other weight transformation
path is selected (which would've freed the original weights and its
tensor info)
Resolves COMPMID-6635
Signed-off-by: SiCong Li <sicong.li@arm.com>
Change-Id: Ib8a345c3ac542bc3745d6a67db822b55df37e827
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10698
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Anitha Raj <Anitha.Raj@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/cpu/utils/CpuAuxTensorHandler.h')
| -rw-r--r-- | src/cpu/utils/CpuAuxTensorHandler.h | 15 |
1 files changed, 10 insertions, 5 deletions
diff --git a/src/cpu/utils/CpuAuxTensorHandler.h b/src/cpu/utils/CpuAuxTensorHandler.h index 627216837b..0a39fdba81 100644 --- a/src/cpu/utils/CpuAuxTensorHandler.h +++ b/src/cpu/utils/CpuAuxTensorHandler.h @@ -74,15 +74,20 @@ public: /** Create a temporary handle to the original tensor with a new @ref TensorInfo * This is useful if we want to change a tensor's tensor info at run time without modifying the original tensor * - * @param[in] info New tensor info to "assign" to @p tensor - * @param[in] tensor Tensor to be assigned a new @ref TensorInfo + * @param[in] info New tensor info to "assign" to @p tensor + * @param[in] tensor Tensor to be assigned a new @ref TensorInfo + * @param[in] bypass_import Bypass importing @p tensor's memory into the handler */ - CpuAuxTensorHandler(TensorInfo &info, const ITensor &tensor) : _tensor() + CpuAuxTensorHandler(TensorInfo &info, const ITensor &tensor, bool bypass_import = false) : _tensor() { _tensor.allocator()->soft_init(info); - if (info.total_size() <= tensor.info()->total_size()) + if (!bypass_import) { - _tensor.allocator()->import_memory(tensor.buffer()); + ARM_COMPUTE_ERROR_ON(tensor.info() == nullptr); + if (info.total_size() <= tensor.info()->total_size()) + { + _tensor.allocator()->import_memory(tensor.buffer()); + } } } |