error: Explicit load/store type does not match pointee type of pointer operand

[plasmapower]

error: Explicit load/store type does not match pointee type of pointer operand

[PlasmaPower]

Interesting. I haven't tested this on AMD GPUs yet. Do you know if Nano OpenCL work generation works on your machine?

[BitDesert]

What's the best way to test this? A quick transaction to the Desktop Wallet?

[PlasmaPower]

Yeah if you first enable opencl in your config (%HOME%\RaiBlocks\config.json IIRC).

[BitDesert]

From the Logfile:

[2018-03-27 22:31:18.656529]: OpenCL found 1 platforms and 2 devices
Platform: 0
FULL_PROFILE
OpenCL 2.1 AMD-APP (2527.10)
AMD Accelerated Parallel Processing
Advanced Micro Devices, Inc.
cl_khr_icd cl_khr_d3d10_sharing cl_khr_d3d11_sharing cl_khr_dx9_media_sharing cl_amd_event_callback cl_amd_offline_devices 
Device: 0
	Ellesmere
	Advanced Micro Devices, Inc.
	FULL_PROFILE
	GPU
	Compiler available: true
	Compute units available: 36
Device: 1
	Intel(R) Core(TM) i5-4460  CPU @ 3.20GHz
	GenuineIntel
	FULL_PROFILE
	CPU
	Compiler available: true
	Compute units available: 4

[2018-03-27 22:31:19.826846]: Node starting, version: 11.0
[2018-03-27 22:31:19.826846]: Work pool running 4 threads
[2018-03-27 22:31:19.826846]: Starting bootstrap attempt
...
[2018-03-27 22:35:20.876058]: Work generation complete: 310199 us

But I can't see any big load to the GPU.

[PlasmaPower]

It seems like that's working. I'll look into what could cause the nano-vanity error.

[BitDesert]

Small addition:

For short prefixes it works sometimes:

nano-vanity.exe 1aaa -g
Estimated attempts needed: 65536

Found matching account!
Private Key: DE7ED1CAE21841914A971A6CBEA7FABBB891C35B185A55985C0D1173262CD864
Account:     xrb_1aaa9fkhorpkbqpatk14jn5pxzhndxzcmgnxykagz41wcype4c46agxm4hrz

...and immediately afterwards it doesn't:

nano-vanity.exe 1aaa -g
Estimated attempts needed: 65536
error: Explicit load/store type does not match pointee type of pointer operand

Shorter prefixes work almost everytime, but longer ones never.

[PlasmaPower]

Could you try adding -t 0 to disable the CPU? I think the CPU is just finding a solution before the GPU starts up.

[BitDesert]

You're right, now it doesn't work at all. If you need any more debugging info just let me know.

[PlasmaPower]

This looks like a compiler problem, or at least a compiler incompatibility between Nvidia and AMD.

Given how many of these problems I've seen, I'm looking into using SPIR-V (an intermediate representation that would lower the chance of cross platform issues). That'd probably go along with #12.

[lukefor]

Managed to get this error sorted (see PR #15 )

Really shit error from AMD (no line number), took a bit of binary searching to find the cause!

[PlasmaPower]

Yeah, at least NVIDIA gives you the line number! Thanks a lot for finding it @lukefor.