Julia can already generate static code:
- http://juliacomputing.com/blog/2016/02/09/static-julia.html
- http://juliacomputing.com/blog/2016/03/10/j2c-announcement.html
However, Julia's @code_llvm output is not meant to be compilable. What are
you trying to do?
On Tue, Oct 25, 2016 at 1:11 PM, <iceblue2...@gmail.com> wrote:
> Thanks, is it possible for julia to avoid generating jl_value_t
>
> When I use a scalar Int64 julia generates code with i64 as input and no
> jl_value_t appears. is it possible to do the same with arrays?
> Array{Float32,2} genrates a jl_value_t? Is there an array type that can do
> this? For example fixed arrays?
>
> Finally, is this features planned to be supported if julia statically
> compiles?
>
> Thanks
>
> On Tuesday, October 25, 2016 at 9:23:02 AM UTC-7, Yichao Yu wrote:
>
>> On Mon, Oct 24, 2016 at 8:15 PM, <icebl...@gmail.com> wrote:
>> > Hi All,
>> >
>> > I am trying to call llvm ir generated from julia.
>> >
>> > Here is my function:
>> >
>> > function incr(a::Array{Int64}) a+1; end
>> >
>> >
>> > thus:
>> >
>> > @code_llvm(incr([1 2 5 6 7]))
>> >
>> > returns:
>> >
>> > define %jl_value_t* @julia_incr_62127(%jl_value_t*) #0 {
>> > top:
>> > %1 = call %jl_value_t* @"julia_.+_62128"(%jl_value_t* %0, i64 1) #0
>> > ret %jl_value_t* %1
>> > }
>> >
>> >
>> >
>> > Thus to call it using llvm I try :
>> >
>> > Base.llvmcall(("""declare %jl_value_t* @julia_incr_62127(%jl_value_t*)""",
>>
>> > """top:
>> > %1 = call %jl_value_t* @"julia_.+_62128"(%jl_value_t* %0, i64
>> 1) #0
>> > ret %jl_value_t* %1
>> > """),Vector{Int64},Tuple{Array{Int64}},[1 2 5 6 7])
>> >
>> >
>> >
>> >
>> > and I get the error:
>> >
>> > ERROR: error compiling anonymous: Failed to parse LLVM Assembly:
>> > julia: llvmcall:2:9: error: use of undefined type named 'jl_value_t'
>> > declare %jl_value_t* @julia_incr_62127(%jl_value_t*)
>> >
>> >
>> > I m not sure how to define jl_value_t.
>>
>> This is not supported.
>> You also shouldn't call julia generated functions with `llvmcall`.
>>
>> >
>> >
>> > any help?
>> >
>> > Thanks
>>
>
