IVGCVSW-7880 Add check for FP16 backend support
* Check if preferred backends have FP16 support before enable fp16-turbo-mode
* Unit tests
* Replaced global gpuAccCapabilities with getter method construction
* Replaced deprecated function call in SL shim
Signed-off-by: Narumol Prangnawarat <narumol.prangnawarat@arm.com>
Signed-off-by: Ryan OShea <ryan.oshea3@arm.com>
Change-Id: If29b62b330ca8987de8acf6408db11daf25ca0b5
diff --git a/tests/ExecuteNetwork/ExecuteNetworkProgramOptions.cpp b/tests/ExecuteNetwork/ExecuteNetworkProgramOptions.cpp
index 7b55b28..4066cdc 100644
--- a/tests/ExecuteNetwork/ExecuteNetworkProgramOptions.cpp
+++ b/tests/ExecuteNetwork/ExecuteNetworkProgramOptions.cpp
@@ -410,7 +410,9 @@
("fp16-turbo-mode",
"If this option is enabled, FP32 layers, "
- "weights and biases will be converted to FP16 where the backend supports it",
+ "weights and biases will be converted to FP16 where the backend supports it. "
+ "If the first preferred backend does not have FP16 support, this option will be disabled. "
+ "If the value of converted FP16 is infinity, round to the closest finite FP16 value.",
cxxopts::value<bool>(m_ExNetParams.m_EnableFp16TurboMode)
->default_value("false")->implicit_value("true"))