Optimize for Speed [-O] can not work

[LinMaris]

I can execute swift -O xx.swift by ternimal, and get correct result.
but when I follow this method, I configured my Xcode project.
Once I execute my project, I will get an error.

let x = Tensor<Float>([[1, 2], [3, 4]])
print(x)

Whether or not I set this, Optimize for Speed [-O] , TensorFlow can't work.

[dan-zheng]

Which macOS toolchain did you download? I see eGPUOverrides in the error message, which makes me a bit suspicious.

[LinMaris]

I create a empty macOS project, and run it with default toolchain

[default] Unable to load Info.plist exceptions (eGPUOverrides)

It still prints in my console.

[rxwei]

It appears that you used a Swift for TensorFlow development snapshot (which is correct), not the default toolchain.

Let me try to reproduce this and get back to you.

[LinMaris]

@rxwei
Firstly, No matter which toolchain I choose, this log still exist

[default] Unable to load Info.plist exceptions (eGPUOverrides)

Secondly, I choose this chain Swift for TensorFlow development snapshot 208-09-17, run this code. that will crash.

let x = Tensor<Float>([[1, 2], [3, 4]])
print(x)

[rxwei]

I'm unable to reproduce this error. Here's how I set up a "Command Line Tool" project in Xcode in project settings:

• Set "Runpath Search Paths" to
  /Library/Developer/Toolchains/swift-tensorflow-DEVELOPMENT-2018-09-17-a.xctoolchain/usr/lib/swift/macosx
• Set "Library Search Paths" to
  /Library/Developer/Toolchains/swift-tensorflow-DEVELOPMENT-2018-09-17-a.xctoolchain/usr/lib/swift/macosx
• Set "Swift Compiler - Code Generation" - "Optimization Level" to "Optimize for Speed [-O]"

Then I can build and run a trivial program in main.swift:

import TensorFlow

var x = Tensor<Float>([[1, 2], [3, 4]])

for _ in 1...5 {
  x += x • x
}

print(x)

[LinMaris]

@rxwei you are right, I set up a "Command Line Tool" project in Xcode and that works, NO Crash

I set up the "Cocoa app" project before. that crash. :)

thanks