try to fix 1.7 #6
Conversation
Codecov Report
@@ Coverage Diff @@
## master #6 +/- ##
==========================================
- Coverage 86.41% 85.43% -0.98%
==========================================
Files 8 9 +1
Lines 471 515 +44
==========================================
+ Hits 407 440 +33
- Misses 64 75 +11
Continue to review full report at Codecov.
|
|
somehow the following case doesn't work anymore code_ircode(sin, (Float64, ); interp=NativeInterpreter()) |
|
@aviatesk mind take a second look if the fix is correct? |
| @@ -0,0 +1,3 @@ | |||
| @static if VERSION < v"1.7-" | |||
There was a problem hiding this comment.
I'm not sure if I can get YaoCompiler working again this weekend, so I think I'll probably try to keep the first version compatible with 1.6 then deprecate it
| @@ -48,24 +48,39 @@ Return the `IRCode` object along with inferred return type. | |||
| function code_ircode_by_mi(f, mi::MethodInstance; world=get_world_counter(), interp=NativeInterpreter(world)) | |||
| # passes | ||
| ir = f(ir, opt) | ||
| opt.src.inferred = true | ||
| ir = f(ir, sv) |
There was a problem hiding this comment.
@aviatesk f is used here, it kinda allow custom IRCode transform from this reflection function since the run_passes function is not called here and is replaced by the default_julia_passes function, so that I can put my own passes using the do ... end syntax to test with
|
I guess the rest looks good then? I'll merge this first. |
No description provided.