Code:
fatal error: error in backend: Cannot select: 0x21baa46d3a0: i32,ch = AtomicLoadAdd
I wonder if I can do something similar to this fix...
https://github.com/CTSRD-CHERI/llvm/commit/68565aebafc1fb144d89a291120efd8a2061a847
Topic: Gateless Gate Sharp 1.3.8: 30Mh/s (Ethash) on RX 480! - page 160. (Read 214410 times)
Hmm... I'm seeing the exact same bug as Crimson 16.9.2... Very interesting.
I wonder if I can do something similar to this fix...
https://github.com/CTSRD-CHERI/llvm/commit/68565aebafc1fb144d89a291120efd8a2061a847
I wonder if I can do something similar to this fix...
https://github.com/CTSRD-CHERI/llvm/commit/68565aebafc1fb144d89a291120efd8a2061a847
Is it possible to solo mine Ethash coins with this miner?
I haven't tried that yet. You could try it yourself and let us know!
Is it possible to solo mine Ethash and Cryptonight coins with this miner?
What kind of speeds does GG give for XMR with RX 470/480?
I'm running GG XMR(0.1.3-pre1) on a mixed rig of 4*470 +1*370. It does 700-710h on 470 and 400+h on 370. stable for weeks (faster than 9.7 claymore, ans slightly less watts ~3-5%)470: 1920-1980&1150-1170 @ 900-915mv
370: 875 & 1375
--rawintensity 504,504,504,504,380 --worksize 4,4,4,4,10 --gpu-threads 2,2,2,2,2
EDIT: Something fishy is going on, here, consistently...
Hynix has 1500 straps and samsung 1625. Same voltage. Discovered accidentally while testing ZEC and leaving the voltages with no change.
Hynix behaves as "expected" - the higher mem gives higher hash, but samsung is different - it makes somehow very good use of the 1750(1625 timings)..
past 18 hours:
I just uploaded a new pre-release. ZEC mining on NVIDIA cards is fixed. It is also faster at 180 sol/s on stock GTX 1060 3GB. Enjoy!
https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre2
Thanks! Will try... https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre2
Now let's see if the new compiler can build the kernel for 7990...
I just uploaded a new pre-release. ZEC mining on NVIDIA cards is fixed. It is also faster at 180 sol/s on stock GTX 1060 3GB. Enjoy!
https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre2
Thanks.https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre2
Now it works with Nvidia.
200 H/s - 1070
160 H/s - 970
My bat:
Code:
@echo off
@set GPU_FORCE_64BIT_PTR 0
@set GPU_MAX_HEAP_SIZE 100
@set GPU_USE_SYNC_OBJECTS 1
@set GPU_MAX_ALLOC_PERCENT 100
@set GPU_SINGLE_ALLOC_PERCENT 100
gatelessgate.exe -k equihash -I 25 --gpu-threads 2 --gpu-platform 0 -o zec-eu.suprnova.cc:2142 -u xxxx.x -p xxx
pause
Looks OK to me.
ETH mining on NVIDIA cards needs much more work, so I will take care of it later.
I just uploaded a new pre-release. ZEC mining on NVIDIA cards is fixed. It is also faster at 180 sol/s on stock GTX 1060 3GB. Enjoy!
https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre2
Thanks.https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre2
Now it works with Nvidia.
200 H/s - 1070
160 H/s - 970
My bat:
Code:
@echo off
@set GPU_FORCE_64BIT_PTR 0
@set GPU_MAX_HEAP_SIZE 100
@set GPU_USE_SYNC_OBJECTS 1
@set GPU_MAX_ALLOC_PERCENT 100
@set GPU_SINGLE_ALLOC_PERCENT 100
gatelessgate.exe -k equihash -I 25 --gpu-threads 2 --gpu-platform 0 -o zec-eu.suprnova.cc:2142 -u xxxx.x -p xxx
pause
only zec? or also pascalcoin e eth
I just uploaded a new pre-release. ZEC mining on NVIDIA cards is fixed. It is also faster at 180 sol/s on stock GTX 1060 3GB. Enjoy!
https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre2
https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre2
I have compiled GG on Ubuntu several days ago and it worked fine with gtx1070 mining PascalCoin
I thought NVIDIA guys (and gals) were using sp_'s miner.
Anyway, let me take a look...
Anyway, let me take a look...
LoL I know it's carnival but don't be the Jecke (jerk)
I use gateless for ZEC and Pasc only as there are a ton of eth and XMR miners.
Can you dual mine ZEC and Pasc together?
I thought NVIDIA guys (and gals) were using sp_'s miner.
Anyway, let me take a look...
Anyway, let me take a look...
NVIDIA support may be broken.
I will take a look at it.
Indeed getting error with gatelessgate-0.1.3-pre1 and gtx 1060... I will take a look at it.
Here is an example of GCN inline assembly:
This code sends the value of nr_slots to all the lanes on the wavefront without using shared memory.
Code:
#ifdef __GCNMINC__
uint thread_index = 0;
if (get_local_id(0) % 64 == 0)
nr_slots = get_nr_slots(PARAM_K - 1, rowCountersSrc, assigned_row_index);
//barrier(CLK_LOCAL_MEM_FENCE);
__asm("ds_bpermute_b32 %0, %1, %2\r\n"
"s_waitcnt lgkmcnt(0)"
: "=v" (nr_slots) : "v" (thread_index), "0" (nr_slots));
This code sends the value of nr_slots to all the lanes on the wavefront without using shared memory.
I know, I know... In order for the miner to stay competitive, however, the implementation has to be vendor-specific, and I have to start somewhere. I plan to do the same for NVIDIA eventually, but I can only handle one thing at a time because my resources are quite limited.
Yeah zawawa you lost track of the NVIDIA compatibility completely since you try to make all the AMD fanboys happy.
But I saw this coming already month ago... so you should drop nvidia either or really make it work for this vendor too.
But as I see all your efforts about AMD GCN/AMD SDK and AMD specific OpenCL implementation you should drop NV to avoid unhappy NVIDIA fanboys like me.
But I saw this coming already month ago... so you should drop nvidia either or really make it work for this vendor too.
But as I see all your efforts about AMD GCN/AMD SDK and AMD specific OpenCL implementation you should drop NV to avoid unhappy NVIDIA fanboys like me.
Jump to: