summaryrefslogtreecommitdiffstats
path: root/xmrstak/backend/amd/minethd.cpp
blob: b1cc1f6b2540a96c42290b6c507ed5253b662f5b (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
/*
  * This program is free software: you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
  * the Free Software Foundation, either version 3 of the License, or
  * any later version.
  *
  * This program is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  * GNU General Public License for more details.
  *
  * You should have received a copy of the GNU General Public License
  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  *
  * Additional permission under GNU GPL version 3 section 7
  *
  * If you modify this Program, or any covered work, by linking or combining
  * it with OpenSSL (or a modified version of that library), containing parts
  * covered by the terms of OpenSSL License and SSLeay License, the licensors
  * of this Program grant you additional permission to convey the resulting work.
  *
  */

#include "minethd.hpp"
#include "autoAdjust.hpp"
#include "amd_gpu/gpu.hpp"

#include "xmrstak/backend/cpu/crypto/cryptonight_altivec.h"
#include "xmrstak/backend/cpu/crypto/cryptonight.h"
#include "xmrstak/misc/configEditor.hpp"
#include "xmrstak/misc/console.hpp"
#include "xmrstak/backend/cpu/minethd.hpp"
#include "xmrstak/jconf.hpp"
#include "xmrstak/misc/executor.hpp"
#include "xmrstak/misc/environment.hpp"
#include "xmrstak/params.hpp"
#include "xmrstak/backend/cpu/hwlocMemory.hpp"

#include <assert.h>
#include <cmath>
#include <chrono>
#include <thread>
#include <vector>

namespace xmrstak
{
namespace amd
{

minethd::minethd(miner_work& pWork, size_t iNo, GpuContext* ctx, const jconf::thd_cfg cfg)
{
	this->backendType = iBackend::AMD;
	oWork = pWork;
	bQuit = 0;
	iThreadNo = (uint8_t)iNo;
	iJobNo = 0;
	iHashCount = 0;
	iTimestamp = 0;
	pGpuCtx = ctx;
	this->affinity = cfg.cpu_aff;

	std::unique_lock<std::mutex> lck(thd_aff_set);
	std::future<void> order_guard = order_fix.get_future();

	oWorkThd = std::thread(&minethd::work_main, this);

	order_guard.wait();

	if(affinity >= 0) //-1 means no affinity
		if(!cpu::minethd::thd_setaffinity(oWorkThd.native_handle(), affinity))
			printer::inst()->print_msg(L1, "WARNING setting affinity failed.");
}

extern "C"  {
#ifdef WIN32
__declspec(dllexport) 
#endif
std::vector<iBackend*>* xmrstak_start_backend(uint32_t threadOffset, miner_work& pWork, environment& env)
{
	environment::inst(&env);
	return amd::minethd::thread_starter(threadOffset, pWork);
}
} // extern "C"

bool minethd::init_gpus()
{
	size_t i, n = jconf::inst()->GetThreadCount();

	printer::inst()->print_msg(L1, "Compiling code and initializing GPUs. This will take a while...");
	vGpuData.resize(n);

	jconf::thd_cfg cfg;
	for(i = 0; i < n; i++)
	{
		jconf::inst()->GetThreadConfig(i, cfg);
		vGpuData[i].deviceIdx = cfg.index;
		vGpuData[i].rawIntensity = cfg.intensity;
		vGpuData[i].workSize = cfg.w_size;
		vGpuData[i].stridedIndex = cfg.stridedIndex;
		vGpuData[i].memChunk = cfg.memChunk;
		vGpuData[i].compMode = cfg.compMode;
	}

	return InitOpenCL(vGpuData.data(), n, jconf::inst()->GetPlatformIdx()) == ERR_SUCCESS;
}

std::vector<GpuContext> minethd::vGpuData;

std::vector<iBackend*>* minethd::thread_starter(uint32_t threadOffset, miner_work& pWork)
{
	std::vector<iBackend*>* pvThreads = new std::vector<iBackend*>();

	if(!configEditor::file_exist(params::inst().configFileAMD))
	{
		autoAdjust adjust;
		if(!adjust.printConfig())
			return pvThreads;
	}

	if(!jconf::inst()->parse_config())
	{
		win_exit();
	}

	// \ todo get device count and exit if no opencl device 

	if(!init_gpus())
	{
		printer::inst()->print_msg(L1, "WARNING: AMD device not found");
		return pvThreads;
	}

	size_t i, n = jconf::inst()->GetThreadCount();
	pvThreads->reserve(n);

	jconf::thd_cfg cfg;
	for (i = 0; i < n; i++)
	{
		jconf::inst()->GetThreadConfig(i, cfg);

		const std::string backendName = xmrstak::params::inst().openCLVendor;
		
		if(cfg.cpu_aff >= 0)
		{
#if defined(__APPLE__)
			printer::inst()->print_msg(L1, "WARNING on macOS thread affinity is only advisory.");
#endif

			printer::inst()->print_msg(L1, "Starting %s GPU (OpenCL) thread %d, affinity: %d.", backendName.c_str(), i, (int)cfg.cpu_aff);
		}
		else
			printer::inst()->print_msg(L1, "Starting %s GPU (OpenCL) thread %d, no affinity.", backendName.c_str(), i);

		minethd* thd = new minethd(pWork, i + threadOffset, &vGpuData[i], cfg);
		pvThreads->push_back(thd);
	}

	return pvThreads;
}

void minethd::switch_work(miner_work& pWork)
{
	// iConsumeCnt is a basic lock-like polling mechanism just in case we happen to push work
	// faster than threads can consume them. This should never happen in real life.
	// Pool cant physically send jobs faster than every 250ms or so due to net latency.

	while (globalStates::inst().iConsumeCnt.load(std::memory_order_seq_cst) < globalStates::inst().iThreadCount)
		std::this_thread::sleep_for(std::chrono::milliseconds(100));

	globalStates::inst().oGlobalWork = pWork;
	globalStates::inst().iConsumeCnt.store(0, std::memory_order_seq_cst);
	globalStates::inst().iGlobalJobNo++;
}

void minethd::consume_work()
{
	memcpy(&oWork, &globalStates::inst().oGlobalWork, sizeof(miner_work));
	iJobNo++;
	globalStates::inst().iConsumeCnt++;

}

void minethd::work_main()
{
	if(affinity >= 0) //-1 means no affinity
		bindMemoryToNUMANode(affinity);

	order_fix.set_value();
	std::unique_lock<std::mutex> lck(thd_aff_set);
	lck.release();
	std::this_thread::yield();

	uint64_t iCount = 0;
	cryptonight_ctx* cpu_ctx;
	cpu_ctx = cpu::minethd::minethd_alloc_ctx();
	
	// start with root algorithm and switch later if fork version is reached
	auto miner_algo = ::jconf::inst()->GetMiningAlgoRoot();
	cn_hash_fun hash_fun = cpu::minethd::func_selector(::jconf::inst()->HaveHardwareAes(), true /*bNoPrefetch*/, miner_algo);

	globalStates::inst().iConsumeCnt++;

	uint8_t version = 0;

	while (bQuit == 0)
	{
		if (oWork.bStall)
		{
			/* We are stalled here because the executor didn't find a job for us yet,
			 * either because of network latency, or a socket problem. Since we are
			 * raison d'etre of this software it us sensible to just wait until we have something
			 */

			while (globalStates::inst().iGlobalJobNo.load(std::memory_order_relaxed) == iJobNo)
				std::this_thread::sleep_for(std::chrono::milliseconds(100));

			consume_work();
			continue;
		}

		uint8_t new_version = oWork.getVersion();
		if(new_version != version)
		{
			if(new_version >= ::jconf::inst()->GetMiningForkVersion())
			{
				miner_algo = ::jconf::inst()->GetMiningAlgo();
				hash_fun = cpu::minethd::func_selector(::jconf::inst()->HaveHardwareAes(), true /*bNoPrefetch*/, miner_algo);
			}
			version = new_version;
		}

		uint32_t h_per_round = pGpuCtx->rawIntensity;
		size_t round_ctr = 0;

		assert(sizeof(job_result::sJobID) == sizeof(pool_job::sJobID));
		uint64_t target = oWork.iTarget;
		
		XMRSetJob(pGpuCtx, oWork.bWorkBlob, oWork.iWorkSize, target, miner_algo);

		if(oWork.bNiceHash)
			pGpuCtx->Nonce = *(uint32_t*)(oWork.bWorkBlob + 39);

		while(globalStates::inst().iGlobalJobNo.load(std::memory_order_relaxed) == iJobNo)
		{
			//Allocate a new nonce every 16 rounds
			if((round_ctr++ & 0xF) == 0)
			{
				globalStates::inst().calc_start_nonce(pGpuCtx->Nonce, oWork.bNiceHash, h_per_round * 16);
			}

			cl_uint results[0x100];
			memset(results,0,sizeof(cl_uint)*(0x100));

			XMRRunJob(pGpuCtx, results, miner_algo);

			for(size_t i = 0; i < results[0xFF]; i++)
			{
				uint8_t	bWorkBlob[112];
				uint8_t	bResult[32];

				memcpy(bWorkBlob, oWork.bWorkBlob, oWork.iWorkSize);
				memset(bResult, 0, sizeof(job_result::bResult));

				*(uint32_t*)(bWorkBlob + 39) = results[i];

				hash_fun(bWorkBlob, oWork.iWorkSize, bResult, cpu_ctx);
				if ( (*((uint64_t*)(bResult + 24))) < oWork.iTarget)
					executor::inst()->push_event(ex_event(job_result(oWork.sJobID, results[i], bResult, iThreadNo), oWork.iPoolId));
				else
					executor::inst()->push_event(ex_event("AMD Invalid Result", pGpuCtx->deviceIdx, oWork.iPoolId));
			}

			iCount += pGpuCtx->rawIntensity;
			uint64_t iStamp = get_timestamp_ms();
			iHashCount.store(iCount, std::memory_order_relaxed);
			iTimestamp.store(iStamp, std::memory_order_relaxed);
			std::this_thread::yield();
		}

		consume_work();
	}
}

} // namespace amd
} // namespace xmrstak
OpenPOWER on IntegriCloud