Skip to content

Metal: Support vector-valued intrinsics. #515

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
maleadt merged 1 commit into from
Sep 14, 2023
Merged

Metal: Support vector-valued intrinsics. #515

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 16 additions & 11 deletions src/metal.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -880,18 +880,23 @@ function lower_llvm_intrinsics!(@nospecialize(job::CompilerJob), fun::LLVM.Funct

# determine type of the intrinsic
typ = value_type(call)
if typ isa LLVM.IntegerType
fn *= signed::Bool ? ".s" : ".u"
fn *= ".$(width(typ))"
elseif typ == LLVM.HalfType()
fn *= ".f16"
elseif typ == LLVM.FloatType()
fn *= ".f32"
elseif typ == LLVM.DoubleType()
fn *= ".f64"
else
error("Unsupported intrinsic type: $typ")
function type_suffix(typ)
# XXX: can't we use LLVM to do this kind of mangling?
if typ isa LLVM.IntegerType
(signed::Bool ? "s" : "u") * "$(width(typ))"
elseif typ == LLVM.HalfType()
"f16"
elseif typ == LLVM.FloatType()
"f32"
elseif typ == LLVM.DoubleType()
"f64"
elseif typ isa LLVM.VectorType
"v$(size(typ))$(type_suffix(eltype(typ)))"
else
error("Unsupported intrinsic type: $typ")
end
end
fn *= "." * type_suffix(typ)

new_intr = LLVM.Function(mod, fn, call_ft)
@dispose builder=IRBuilder() begin
Expand Down
8 changes: 8 additions & 0 deletions test/metal_tests.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,6 +77,14 @@ end
@test !occursin(r"call i32 @julia.air.thread_position_in_threadgroup.i32", ir)
end

@testset "vector intrinsics" begin
foo(x, y) = ccall("llvm.smax.v2i64", llvmcall, NTuple{2, VecElement{Int64}},
(NTuple{2, VecElement{Int64}}, NTuple{2, VecElement{Int64}}), x, y)

ir = sprint(io->Metal.code_llvm(io, foo, (NTuple{2, VecElement{Int64}}, NTuple{2, VecElement{Int64}})))
@test occursin("air.max.v2s64", ir)
end

end

end
Expand Down