checkpoint/restore problem

[YixinSong-e]

When I use the test script for c/r.
The output is

[New Thread 0x7f37c96c5000 (LWP 38397)]
[New Thread 0x7f37a12e0000 (LWP 38398)]
[Detaching after fork from child process 38399]
[Thread 0x7f37c96c5000 (LWP 38397) exited]

Thread 1 "kernel.testapp" received signal SIGURG, Urgent I/O condition.
[Switching focus to CUDA kernel 0, grid 1, block (0,0,0), thread (0,0,0), device 0, sm 0, warp 0, lane 0]                                                                                    
0x00007f37a2fae0a0 in kernel(unsigned short*, unsigned short*, unsigned short*, char, short, int, long long)<<<(1,1,1),(32,1,1)>>> ()                                                        
[DEBUG] gdb_init out.
Cuda api initialized and attached!
got API
Device "NVIDIA A100-SXM4-80GB":
        index: 0
        type: "GA100GL-A"
        SM type: "sm_80"
        lanes: 32
        predicates 8
        registers: 255
        SMs: 108
        warps: 64

checkpointing kernel with name: "_Z6kernelPtS_S_csix"
stack-size: 336, param-addr: 352, param-size: 40, param-num: 7
SM 0: 1 - 0000000000000000000000000000000000000000000000000000000000000001
03/12/23 10:42:43.888585   DEBUG: relative 6a0, virtual 7f37a2fae0a0    in /cricket-cr.c(670)
SM 0 warp 0 (active): 55555555 - 01010101010101010101010101010101
SM 0 warp 0 (valid): ffffffff - 11111111111111111111111111111111
03/12/23 10:42:43.888637   DEBUG: function "_Z6kernelPtS_S_csix" has no room (0 slots)  in /cricket-cr.c(831)                                                                                
03/12/23 10:42:43.888647   ERROR: There is no room in the top level function (i.e. the kernel). This kernel can thus never be restored!        in /cricket-cr.c(835)                         
cricket-checkpoint: could not make checkpointable.

Thread 1 "kernel.testapp" received signal SIGURG, Urgent I/O condition.
0x00007ffed552baea in clock_gettime ()
[Inferior 1 (process 38378) detached]

[n-eiling]

This is related to GPU kernel checkpointing. This is currently not working - but we are working on repairing this. (see #19 )
Note that most likely you don't need this. The code in the cpu subdirectory can also do checkpoints.