Mac support #17

psychocrypt · 2017-04-08T21:23:19Z

This pull requst integrate the changes of #10

Changes

include OpenCL/cl.j on Mac
add option to emulate amd_bfe and amd_bitalign

My changes on top of kth5's pull requst

remove the not existent pragma #pragma "Emulation *"
optimize amd_bfe (remove check for the edge case)

Tests

runtime test linux (native amd intrinsics)
runtime test linux (emulated amd bfe/bit_align)
@kth5 Mac test
please do not merge before merge master back to dev #39 this branch needs a rebase

psychocrypt · 2017-04-08T21:24:38Z

@fireice-uk I need to check my changes without on linux while using the emulated functions instead of the amd intrinsic.

@kth5 Could you please run this branch on our Mac. btw: the fix for the worksize is not included in this branch and is not needed for a validation test

fireice-uk · 2017-04-08T21:56:53Z

@psychocrypt Let me know when the code is ready to be merged (I don't think those tick widgets will ping me).

kth5 · 2017-04-09T05:55:36Z

I'll be running an hour worth later today and report back. So far the miner performs the same as with my PR alone. Still waiting for shares to get an average idea.

kth5 · 2017-04-10T02:41:17Z

@psychocrypt Had it running for around an hour now and due to a still slow hashrate of just 60H/s on a R9 M395X 4GB in my iMac, it only came to find a few shares but seems ok:

There are still GPU compute errors but to me it seems these may be expected? Someone with more indepth knowledge of the algorithms involved may have to have a look at the emulated amd_bfe and amd_bitalign to make sure I guess.

psychocrypt · 2017-04-10T14:39:15Z

I can reproduce a low rate of wrong shares under linux if I used the emulated functions. I will investigate time to find the issue as soon as the nvidia miner is released.

kth5 · 2017-04-10T16:18:30Z

If it helps, I'll extend my offer for an account on a recent Mac with AMD graphics. Just let me know so we can schedule.

Apart from that, even if the emulated functions aren't accurate enough. It seems to me that under 100H/s is very low since I can get around the same with a much lower end R5 240 on Ubuntu.

psychocrypt · 2017-04-10T16:23:21Z

I can reproduce a low rate of wrong shares under linux if I used the emulated functions. I will investigate time to find the issue as soon as the nvidia miner is released.
After I found the bug I will rebase against #13 this should fix your low hashrate. You need to increase the worksize to some differ from 1.

richard-underwood · 2017-06-02T12:22:33Z

I'm seeing up to 50% rejected submissions - could the lack of casts in the amd_bitalign mean it's doing a signed shift when it shouldn't be?

psychocrypt · 2017-06-02T21:35:09Z

@richard-underwood You pointed to the correct place. It was no issue with signed the function need a 64bit cast and than back to 32bit. Also the parentheses was wrongly placed.

I fixed it in the last commit, it is running on my system with 100% valid hashes.
Could you please test the latest changes.

psychocrypt · 2017-06-02T21:41:45Z

@kth5 Could you please try the latest version of the PR.

richard-underwood · 2017-06-05T13:31:36Z

Interestingly, I still get GPU compute errors.

psychocrypt · 2017-06-05T13:42:04Z

@richard-underwood could you please ckeck by hand that you are using the latest chages.

richard-underwood · 2017-06-05T14:17:42Z

I've just fetched it again with:

git clone https://github.com/fireice-uk/xmr-stak-amd.git
git checkout dev
wget 'https://github.com/fireice-uk/xmr-stak-amd/pull/17.diff'
patch -p1 < 17.diff
cmake -DOpenSSL_ENABLE=OFF -DMICROHTTPD_ENABLE=OFF .
make

This is still throwing errors:

RESULT REPORT
Difficulty       : 1000
Good results     : 14 / 17 (82.4 %)
Avg result time  : 3.1 sec
Pool-side hashes : 14000

Error details:
| Count | Error text                       | Last seen           |
|     3 | [GPU COMPUTE ERROR]              | 2017-06-05 15:15:17 |

snazzybunny · 2017-06-13T00:21:37Z

My macbook pro with a amd 460 gpu seems to freeze when running richard-underwood's instructions. I have to do a hard reboot. Also, it only shows that I have a platform index of 0. Using platform index = 1 does not allow the miner to start - I think it only sees one opencl device instead of two.

richard-underwood · 2017-06-13T10:12:10Z

I'm running on an iMac with one GPU - I have no idea what happens on a laptop which switches between integrated and discrete GPU, but I can take a look later.

richard-underwood · 2017-06-13T10:43:55Z

Is there documentation of the protocol/algorithms used anywhere? I intentionally broke SKEIN_ROT() and the result was pretty much the same percentage of failed hashes.

kth5 · 2017-06-13T11:04:24Z

@psychocrypt Did a run of the PR a while ago, here are the results:

This is on a MacBook Pro mid-2015 with a R9 M370X 2GB on Sierra.

gregordoltar · 2017-06-13T12:25:08Z

I am experiencing same problem as snazzybunny. But this happens on my iMac Sierra, AMD Radeon r9 m390 2gb.

Binary was compiled as:

git clone https://github.com/fireice-uk/xmr-stak-amd.git
git checkout dev
wget 'https://github.com/fireice-uk/xmr-stak-amd/pull/17.diff'
patch -p1 < 17.diff
cmake -DOpenSSL_ENABLE=OFF -DMICROHTTPD_ENABLE=OFF .
make

I had to modify my config.txt because some devices were not found:


-"gpu_thread_num" : 6,
+"gpu_thread_num" : 1,
 "gpu_threads_conf" : [ 
        { "index" : 0, "intensity" : 1000, "worksize" : 8, "affine_to_cpu" : false },
-       { "index" : 1, "intensity" : 1000, "worksize" : 8, "affine_to_cpu" : false },
-       { "index" : 2, "intensity" : 1000, "worksize" : 8, "affine_to_cpu" : false },
-       { "index" : 3, "intensity" : 1000, "worksize" : 8, "affine_to_cpu" : false },
-       { "index" : 4, "intensity" : 1000, "worksize" : 8, "affine_to_cpu" : false },
-       { "index" : 5, "intensity" : 1000, "worksize" : 8, "affine_to_cpu" : false },
 ],

I run binary as normal user (not a root). The system does not crash. I can still move the mouse but UI is not responsive.

ArtSabintsev · 2017-06-15T23:05:42Z

Mac user here. Installation instructions available, or do i simply run Make?

ArtSabintsev · 2017-06-16T00:41:02Z

Alright, got Cmake to work using:

cmake -DOPENSSL_ROOT_DIR=/usr/local/opt/openssl .

However, running make throws the following error on @psychocrypt's branch:

Arthur: ~/Documents/xmr-stak-amd 
(master)$ make
Scanning dependencies of target xmr-stak-amd
[  5%] Building C object CMakeFiles/xmr-stak-amd.dir/crypto/c_blake256.c.o
clang: warning: argument unused during compilation: '-s' [-Wunused-command-line-argument]
[ 11%] Building C object CMakeFiles/xmr-stak-amd.dir/crypto/c_groestl.c.o
clang: warning: argument unused during compilation: '-s' [-Wunused-command-line-argument]
[ 16%] Building C object CMakeFiles/xmr-stak-amd.dir/crypto/c_jh.c.o
clang: warning: argument unused during compilation: '-s' [-Wunused-command-line-argument]
[ 22%] Building C object CMakeFiles/xmr-stak-amd.dir/crypto/c_keccak.c.o
clang: warning: argument unused during compilation: '-s' [-Wunused-command-line-argument]
[ 27%] Building C object CMakeFiles/xmr-stak-amd.dir/crypto/c_skein.c.o
clang: warning: argument unused during compilation: '-s' [-Wunused-command-line-argument]
[ 33%] Building C object CMakeFiles/xmr-stak-amd.dir/crypto/soft_aes.c.o
clang: warning: argument unused during compilation: '-s' [-Wunused-command-line-argument]
[ 38%] Building CXX object CMakeFiles/xmr-stak-amd.dir/crypto/cryptonight_common.cpp.o
clang: warning: argument unused during compilation: '-s' [-Wunused-command-line-argument]
[ 44%] Building C object CMakeFiles/xmr-stak-amd.dir/amd_gpu/gpu.c.o
clang: warning: argument unused during compilation: '-s' [-Wunused-command-line-argument]
In file included from /Users/Arthur/Documents/xmr-stak-amd/amd_gpu/gpu.c:42:
/Users/Arthur/Documents/xmr-stak-amd/amd_gpu/gpu.h:3:10: fatal error: 'CL/cl.h' file not found
#include <CL/cl.h>
         ^
1 error generated.
make[2]: *** [CMakeFiles/xmr-stak-amd.dir/amd_gpu/gpu.c.o] Error 1
make[1]: *** [CMakeFiles/xmr-stak-amd.dir/all] Error 2
make: *** [all] Error 2

(I'm here because I saw a post on reddit saying that mac testers were needed. As a potential mac user, a README would be nice as well across all xmr-stak-* repos)

snazzybunny · 2017-06-16T00:55:38Z

I would try to add the dev patches as shown in the above instructions and then make.

ArtSabintsev · 2017-06-16T01:41:27Z

Thanks. Got it working, though I only have one GPU, so it took a while to get these results and unfreeze my computer:

Arthur: ~/Documents/xmr-stak-amd 
(dev)$ ./xmr-stak-amd 
[2017-06-15 21:08:32] : Compiling code and initializing GPUs. This will take a while...
[2017-06-15 21:08:32] : Device 0 work size 8 / 256.
-------------------------------------------------------------------
xmr-stak-amd 1.1.0-1.4.0-dev mining software, AMD Version.
AMD mining code was written by wolf9466.
Brought to you by fireice_uk under GPLv3.

Configurable dev donation level is set to 1.0 %

You can use following keys to display reports:
'h' - hashrate
'r' - results
'c' - connection
-------------------------------------------------------------------
[2017-06-15 21:08:36] : Starting GPU thread, no affinity.
[2017-06-15 21:08:36] : Connecting to pool monerohash.com:3333 ...
[2017-06-15 21:08:36] : Connected. Logging in...
[2017-06-15 21:08:36] : Difficulty changed. Now: 5000.
[2017-06-15 21:08:36] : New block detected.
[2017-06-15 21:09:51] : New block detected.
[2017-06-15 21:11:25] : Result accepted by the pool.
[2017-06-15 21:12:26] : New block detected.
[2017-06-15 21:13:49] : New block detected.
[2017-06-15 21:14:13] : New block detected.
[2017-06-15 21:16:27] : New block detected.
[2017-06-15 21:17:13] : New block detected.
[2017-06-15 21:17:57] : Result accepted by the pool.
[2017-06-15 21:17:58] : New block detected.
[2017-06-15 21:21:03] : New block detected.
[2017-06-15 21:21:46] : New block detected.
[2017-06-15 21:22:40] : Result accepted by the pool.
[2017-06-15 21:22:40] : New block detected.
[2017-06-15 21:24:03] : Result accepted by the pool.
[2017-06-15 21:25:28] : Result accepted by the pool.
[2017-06-15 21:25:56] : Result accepted by the pool.
[2017-06-15 21:26:24] : Result accepted by the pool.
[2017-06-15 21:27:38] : New block detected.
[2017-06-15 21:28:37] : New block detected.
[2017-06-15 21:29:12] : New block detected.
[2017-06-15 21:31:44] : New block detected.
[2017-06-15 21:36:25] : New block detected.
[2017-06-15 21:36:50] : New block detected.
[2017-06-15 21:39:00] : Result accepted by the pool.

Also adding visual proof.

snazzybunny · 2017-06-16T18:19:06Z

@ArtSabintsev Does it crash for you?

ArtSabintsev · 2017-06-16T18:24:24Z

@snazzybunny No, but again, I have one GPU on this machine, and it was pegged the entire time, so I only let it run for 20-25 minutes - worked without fail.

Panzerfather · 2017-06-17T22:59:45Z

@psychocrypt Please correct me if I'm wrong, but after building the AMD specific functions in code should this not also be working with Intel OpenCL drivers? If so, it simply doesn't. At least when I tested it with Intel 42xx Haswell CPUs running Windows it will produce 100% GPU compute errors. Maybe there is a rare compute condition on AMD Macs and a permanent condition on Intel GPUs?

Program was build with Intel's latest OpenCL SDK.

Edit: Just a side note which may help you to isolate the problem, I also noticed rare GPU compute errors on Windows with AMD Radeon 270/280 and AMD Radeon 480 cards on Windows/Linux. Just like around 1-2%, but not in every run of the miner. E.g. running miner 7 days without a compute error and right now, after running the same compilation after reboot for ~2 days 971/988 (98.3%) with 17 GPU compute errors. It seems that the problems were triggered from another (rare) case than from the new functions implemented via the new code.

psychocrypt · 2017-06-21T07:34:32Z

@Panzerfather I have not tested it on Intel, I can do it. If you get compute errors please post the output that we can see the kind of error.

@kth5 The error in the post is a network error and is independent of this PR.

- include `OpenCL/cl.j` on Mac - add option to emulate `amd_bfe` and `amd_bitalign`

- remove edge case check in emulated `amd_bfe` function - remove unknown `#pragma "Emulation *"`

- add clean inline functions for `amd_bitalign` - change `amd_bfe` implementation - add links to the original documentations from kronos for `amd_bitalign` and `amd_bfe`

psychocrypt · 2017-06-21T07:39:52Z

@fireice-uk Could you please merge this PR. If there is something not 100% well for MACOSX it is no issue because it is not effecting linux. This means we have no side effects to the current supported environments.

Panzerfather · 2017-06-21T23:27:14Z

@psychocrypt Off course, it will look just like that:

RESULT REPORT
Difficulty       : 1000
Good results     : 0 / 17 (0 %)
Avg result time  : 4.2 sec
Pool-side hashes : 0

Error details:
| Count | Error text                       | Last seen           |
|    17 | [GPU COMPUTE ERROR]              | 2017-06-22 22:34:15 |

I've ran other tests with more than 100 miscalculated hashes and 0 good results [all GPU compute errors] on 2 different Intel GPUs.

Is there an input and correct output hash example to work with for debugging proposes or do I have to setup a testing environment myself with a working card to locate the miscalculation on Intel cards? The testing network isn't an option because you have to know what the correct internal and result values should be.

fireice-uk · 2017-06-25T14:00:38Z

@psychocrypt ok - merging without much review, this one is on you =)

psychocrypt · 2017-06-25T16:57:44Z

@Panzerfather Which OS do you used and which gpu?

Panzerfather · 2017-06-25T20:58:16Z

@psychocrypt Just take a look #17 (comment) 😉

The 100% GPU compute errors are Intel NUC Systems with latest drivers. All Systems running at standard power voltage and are not overclocked.

psychocrypt · 2017-06-26T19:00:22Z

@Panzerfather Please switch to #56 for discussion about Intel GPU support

mihaiile · 2017-09-20T14:26:43Z

I was trying the latest dev branch as of today on iMac 2017 (has the Radeon Pro 580) but I get a LOT of GPU compute errors, I mean around maybe 40-60% is compute errors. :(

fireice-uk · 2017-09-20T16:56:02Z

Unfortunately most often those are due to the hardware - card is either too hot or too old.

psychocrypt added the enhancement label Apr 10, 2017

psychocrypt assigned fireice-uk Apr 10, 2017

fireice-uk mentioned this pull request May 26, 2017

Release roll #36

Closed

psychocrypt force-pushed the topic-appleOpenCLHeader branch from 6a4959e to 8dbe852 Compare June 2, 2017 21:29

psychocrypt mentioned this pull request Jun 2, 2017

More OSX glue #10

Closed

coruble mentioned this pull request Jun 19, 2017

Fatal error: 'CL/cl.h' file not found #52

Closed

kth5 and others added 3 commits June 21, 2017 09:35

add APPLE OpenCL support

8e26cad

- include `OpenCL/cl.j` on Mac - add option to emulate `amd_bfe` and `amd_bitalign`

cleanup and optimize

0ac7865

- remove edge case check in emulated `amd_bfe` function - remove unknown `#pragma "Emulation *"`

fix high rejected hashes

ec54b59

- add clean inline functions for `amd_bitalign` - change `amd_bfe` implementation - add links to the original documentations from kronos for `amd_bitalign` and `amd_bfe`

psychocrypt force-pushed the topic-appleOpenCLHeader branch from 8dbe852 to ec54b59 Compare June 21, 2017 07:35

fireice-uk merged commit 3848669 into fireice-uk:dev Jun 25, 2017

psychocrypt deleted the topic-appleOpenCLHeader branch June 25, 2017 16:56

psychocrypt mentioned this pull request Jun 26, 2017

Intel GPU support #56

Open

lamba84 mentioned this pull request Oct 22, 2017

Error UNKNOWN_ERROR when calling clGetPlatformIDs for number of platforms. #115

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Mac support #17

Mac support #17

psychocrypt commented Apr 8, 2017 •

edited

psychocrypt commented Apr 8, 2017 •

edited

fireice-uk commented Apr 8, 2017

kth5 commented Apr 9, 2017

kth5 commented Apr 10, 2017

psychocrypt commented Apr 10, 2017

kth5 commented Apr 10, 2017 •

edited

psychocrypt commented Apr 10, 2017

richard-underwood commented Jun 2, 2017

psychocrypt commented Jun 2, 2017

psychocrypt commented Jun 2, 2017

richard-underwood commented Jun 5, 2017

psychocrypt commented Jun 5, 2017

richard-underwood commented Jun 5, 2017

snazzybunny commented Jun 13, 2017

richard-underwood commented Jun 13, 2017

richard-underwood commented Jun 13, 2017

kth5 commented Jun 13, 2017

gregordoltar commented Jun 13, 2017

ArtSabintsev commented Jun 15, 2017

ArtSabintsev commented Jun 16, 2017

snazzybunny commented Jun 16, 2017 •

edited

ArtSabintsev commented Jun 16, 2017

snazzybunny commented Jun 16, 2017

ArtSabintsev commented Jun 16, 2017

Panzerfather commented Jun 17, 2017 •

edited

psychocrypt commented Jun 21, 2017

psychocrypt commented Jun 21, 2017

Panzerfather commented Jun 21, 2017

fireice-uk commented Jun 25, 2017

psychocrypt commented Jun 25, 2017

Panzerfather commented Jun 25, 2017

psychocrypt commented Jun 26, 2017

mihaiile commented Sep 20, 2017

fireice-uk commented Sep 20, 2017

Mac support #17

Mac support #17

Conversation

psychocrypt commented Apr 8, 2017 • edited

Changes

My changes on top of kth5's pull requst

Tests

psychocrypt commented Apr 8, 2017 • edited

fireice-uk commented Apr 8, 2017

kth5 commented Apr 9, 2017

kth5 commented Apr 10, 2017

psychocrypt commented Apr 10, 2017

kth5 commented Apr 10, 2017 • edited

psychocrypt commented Apr 10, 2017

richard-underwood commented Jun 2, 2017

psychocrypt commented Jun 2, 2017

psychocrypt commented Jun 2, 2017

richard-underwood commented Jun 5, 2017

psychocrypt commented Jun 5, 2017

richard-underwood commented Jun 5, 2017

snazzybunny commented Jun 13, 2017

richard-underwood commented Jun 13, 2017

richard-underwood commented Jun 13, 2017

kth5 commented Jun 13, 2017

gregordoltar commented Jun 13, 2017

ArtSabintsev commented Jun 15, 2017

ArtSabintsev commented Jun 16, 2017

snazzybunny commented Jun 16, 2017 • edited

ArtSabintsev commented Jun 16, 2017

snazzybunny commented Jun 16, 2017

ArtSabintsev commented Jun 16, 2017

Panzerfather commented Jun 17, 2017 • edited

psychocrypt commented Jun 21, 2017

psychocrypt commented Jun 21, 2017

Panzerfather commented Jun 21, 2017

fireice-uk commented Jun 25, 2017

psychocrypt commented Jun 25, 2017

Panzerfather commented Jun 25, 2017

psychocrypt commented Jun 26, 2017

mihaiile commented Sep 20, 2017

fireice-uk commented Sep 20, 2017

psychocrypt commented Apr 8, 2017 •

edited

psychocrypt commented Apr 8, 2017 •

edited

kth5 commented Apr 10, 2017 •

edited

snazzybunny commented Jun 16, 2017 •

edited

Panzerfather commented Jun 17, 2017 •

edited