Skip to content

Commit 3073525

Browse files
authored
Support GPUCompiler 0.12 (#145)
1 parent 031b345 commit 3073525

File tree

2 files changed

+7
-5
lines changed

2 files changed

+7
-5
lines changed

Project.toml

+2-2
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,7 @@
11
name = "AMDGPU"
22
uuid = "21141c5a-9bdb-4563-92ae-f87d6854732e"
33
authors = ["Julian P Samaroo <jpsamaroo@jpsamaroo.me>"]
4-
version = "0.2.7"
4+
version = "0.2.8"
55

66
[deps]
77
AbstractFFTs = "621f4979-c628-5d54-868e-fcf4e3e8185c"
@@ -27,7 +27,7 @@ Adapt = "3.0"
2727
BinaryProvider = "0.5"
2828
CEnum = "0.2, 0.3, 0.4"
2929
GPUArrays = "6"
30-
GPUCompiler = "0.11.5"
30+
GPUCompiler = "0.12"
3131
LLVM = "3"
3232
MacroTools = "0.5"
3333
Requires = "1"

src/execution.jl

+5-3
Original file line numberDiff line numberDiff line change
@@ -313,9 +313,11 @@ const rocfunction_cache = Dict{RuntimeDevice,Dict{UInt,Any}}()
313313
# compile to GCN
314314
function rocfunction_compile(@nospecialize(job::CompilerJob))
315315
# compile
316-
method_instance, world = GPUCompiler.emit_julia(job)
317-
ir, kernel = GPUCompiler.emit_llvm(job, method_instance, world)
318-
obj = GPUCompiler.emit_asm(job, ir, kernel; format=LLVM.API.LLVMObjectFile)
316+
method_instance, mi_meta = GPUCompiler.emit_julia(job)
317+
ir, ir_meta = GPUCompiler.emit_llvm(job, method_instance)
318+
kernel = ir_meta.entry
319+
320+
obj, obj_meta = GPUCompiler.emit_asm(job, ir; format=LLVM.API.LLVMObjectFile)
319321

320322
# find undefined globals and calculate sizes
321323
globals = map(gbl->Symbol(LLVM.name(gbl))=>llvmsize(eltype(llvmtype(gbl))),

0 commit comments

Comments
 (0)