diff options
| author | Georgios Pinitas <georgios.pinitas@arm.com> | 2019-09-12 12:26:11 +0100 |
|---|---|---|
| committer | Michele Di Giorgio <michele.digiorgio@arm.com> | 2019-09-12 15:26:15 +0000 |
| commit | c370c754ebd133cfed19ad1773ef540e48bd4240 (patch) | |
| tree | be15b445b1da42d2995874a4b6bf29c3cba2fd88 | |
| parent | 9285adb5ac8e28a9cc82ce708bb2975dc5a074dd (diff) | |
| download | ComputeLibrary-c370c754ebd133cfed19ad1773ef540e48bd4240.tar.gz | |
COMPMID-2570: Add hybrid precision flag in GEMMKernelInfo.
Adds fp_mixed_precision flag in GEMMKernelInfo in order to indicate use
of wider accumulators for FP16 workloads where supported.
Change-Id: I8d69bc62d945fd40af77395ff673d07880b7aeb6
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1914
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
| -rw-r--r-- | arm_compute/core/KernelDescriptors.h | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/arm_compute/core/KernelDescriptors.h b/arm_compute/core/KernelDescriptors.h index 3affc30f71..215f0f1651 100644 --- a/arm_compute/core/KernelDescriptors.h +++ b/arm_compute/core/KernelDescriptors.h @@ -60,6 +60,7 @@ struct GEMMKernelInfo unsigned int depth_output_gemm3d{ 0 }; /**< Depth of the output tensor in case is reinterpreted as 3D */ bool reinterpret_input_as_3d{ false }; /**< Flag used to reinterpret the input as 3D */ bool broadcast_bias{ false }; /**< Flag used to broadcase the bias addition */ + bool fp_mixed_precision{ false }; /**< Flag used to indicate wider accumulators (32 bit instead of 16 for FP16). */ ActivationLayerInfo activation_info{}; /**< Activation function to perform after the matrix multiplication */ }; |