Segfault using MArray

[lcw]

I am getting the following segfault when I try to use an MArray inside a CUDAnative kernel. I am using the current master version of CUDAnative. I have tried both with and without @inbounds.

❯ julia --project=.
               _
   _       _ _(_)_     |  Documentation: https://docs.julialang.org
  (_)     | (_) (_)    |
   _ _   _| |_  __ _   |  Type "?" for help, "]?" for Pkg help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 1.2.0-DEV.670 (2019-04-10)
 _/ |\__'_|_|_|\__'_|  |  Commit 20499752ac (2 days old master)
|__/                   |

julia> using CUDAnative, StaticArrays

julia> function kernel()
         r = MArray{Tuple{5}, Float32}(undef)
         @inbounds r[1] = 0.0f0

         nothing
       end
kernel (generic function with 1 method)

julia> @cuda kernel()


signal (11): Segmentation fault
in expression starting at REPL[3]:1
get_specsig_function at /buildworker/worker/package_linux64/build/src/codegen.cpp:5344
emit_call_specfun_other at /buildworker/worker/package_linux64/build/src/codegen.cpp:3067
emit_invoke at /buildworker/worker/package_linux64/build/src/codegen.cpp:3218
emit_expr at /buildworker/worker/package_linux64/build/src/codegen.cpp:4025
emit_ssaval_assign at /buildworker/worker/package_linux64/build/src/codegen.cpp:3739
emit_stmtpos at /buildworker/worker/package_linux64/build/src/codegen.cpp:3931 [inlined]
emit_function at /buildworker/worker/package_linux64/build/src/codegen.cpp:6429
jl_get_llvmf_defn at /buildworker/worker/package_linux64/build/src/codegen.cpp:1538
compile_method_instance at /home/lucasw/.julia/packages/CUDAnative/goLI4/src/compiler/irgen.jl:97 [inlined]
macro expansion at /home/lucasw/.julia/packages/TimerOutputs/7zSea/src/TimerOutput.jl:216 [inlined]
irgen at /home/lucasw/.julia/packages/CUDAnative/goLI4/src/compiler/irgen.jl:110
#codegen#113 at /home/lucasw/.julia/packages/TimerOutputs/7zSea/src/TimerOutput.jl:216
#codegen at /home/lucasw/.julia/packages/CUDAnative/goLI4/src/compiler/driver.jl:0 [inlined]
#compile#112 at /home/lucasw/.julia/packages/CUDAnative/goLI4/src/compiler/driver.jl:38
#compile#111 at ./none:0 [inlined]
compile at /home/lucasw/.julia/packages/CUDAnative/goLI4/src/compiler/driver.jl:22 [inlined]
compile at /home/lucasw/.julia/packages/CUDAnative/goLI4/src/compiler/driver.jl:22 [inlined]
macro expansion at /home/lucasw/.julia/packages/CUDAnative/goLI4/src/execution.jl:378 [inlined]
#cufunction#146 at /home/lucasw/.julia/packages/CUDAnative/goLI4/src/execution.jl:347
jl_apply_generic at /buildworker/worker/package_linux64/build/src/gf.c:2197
cufunction at /home/lucasw/.julia/packages/CUDAnative/goLI4/src/execution.jl:347
jl_apply_generic at /buildworker/worker/package_linux64/build/src/gf.c:2197
do_call at /buildworker/worker/package_linux64/build/src/interpreter.c:323
eval_value at /buildworker/worker/package_linux64/build/src/interpreter.c:411
eval_body at /buildworker/worker/package_linux64/build/src/interpreter.c:635
jl_interpret_toplevel_thunk_callback at /buildworker/worker/package_linux64/build/src/interpreter.c:884
unknown function (ip: 0xfffffffffffffffe)
unknown function (ip: 0x7fcb6d2f398f)
unknown function (ip: 0x9)
jl_interpret_toplevel_thunk at /buildworker/worker/package_linux64/build/src/interpreter.c:893
jl_toplevel_eval_flex at /buildworker/worker/package_linux64/build/src/toplevel.c:797
jl_toplevel_eval_flex at /buildworker/worker/package_linux64/build/src/toplevel.c:746
jl_toplevel_eval_in at /buildworker/worker/package_linux64/build/src/toplevel.c:826
eval at ./boot.jl:330
jl_apply_generic at /buildworker/worker/package_linux64/build/src/gf.c:2191
eval_user_input at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.2/REPL/src/REPL.jl:86
macro expansion at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.2/REPL/src/REPL.jl:118 [inlined]
#26 at ./task.jl:268
jl_apply_generic at /buildworker/worker/package_linux64/build/src/gf.c:2191
jl_apply at /buildworker/worker/package_linux64/build/src/julia.h:1604 [inlined]
start_task at /buildworker/worker/package_linux64/build/src/task.c:583
unknown function (ip: 0xffffffffffffffff)
Allocations: 22287832 (Pool: 22283680; Big: 4152); GC: 48
fish: “julia --project=.” terminated by signal SIGSEGV (Address boundary error)

[vchuravy]

Can you check on current master? JuliaLang/julia#31687

[lcw]

I had separate issues with master so I tried the release-1.2 branch with that commit, i.e.,

git checkout release-1.2
git cherry-pick 8469856

and it seem to work! Hopefully we can get 8469856 back ported into Julia 1.2.

cc: @simonbyrne

[simonbyrne]

It's tagged for backport (JuliaLang/julia#31687), and is clearly a bug fix, so I imagine it should get merged.