Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

OpenCL: rocm shuffle #2447

Merged
merged 1 commit into from
Jun 9, 2019
Merged

OpenCL: rocm shuffle #2447

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions xmrstak/backend/amd/amd_gpu/gpu.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1280,17 +1280,17 @@ size_t XMRRunJob(GpuContext* ctx, cl_uint* HashOutput, const xmrstak_algo& miner
}
}

if((ret = clEnqueueNDRangeKernel(ctx->CommandQueues, Kernels[2], 2, Nonce, gthreads, lthreads, 0, NULL, NULL)) != CL_SUCCESS)
size_t NonceT[2] = {0, ctx->Nonce}, gthreadsT[2] = {8, g_thd}, lthreadsT[2] = {8 , w_size};
if((ret = clEnqueueNDRangeKernel(ctx->CommandQueues, Kernels[2], 2, NonceT, gthreadsT, lthreadsT, 0, NULL, NULL)) != CL_SUCCESS)
{
printer::inst()->print_msg(L1, "Error %s when calling clEnqueueNDRangeKernel for kernel %d.", err_to_str(ret), 2);
return ERR_OCL_API;
return ERR_OCL_API;
}

if(miner_algo != cryptonight_gpu)
{
for(int i = 0; i < 4; ++i)
{
size_t tmpNonce = ctx->Nonce;
if((ret = clEnqueueNDRangeKernel(ctx->CommandQueues, Kernels[i + 3], 1, &tmpNonce, &g_thd, &w_size, 0, NULL, NULL)) != CL_SUCCESS)
{
printer::inst()->print_msg(L1, "Error %s when calling clEnqueueNDRangeKernel for kernel %d.", err_to_str(ret), i + 3);
Expand Down
247 changes: 148 additions & 99 deletions xmrstak/backend/amd/amd_gpu/opencl/cryptonight.cl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -867,9 +867,14 @@ __kernel void JOIN(cn1,ALGO) (__global uint4 *Scratchpad, __global ulong *states
}

)==="
R"===(
R"===(
#if defined(__clang__)
# if __has_builtin(__builtin_amdgcn_ds_bpermute)
# define HAS_AMD_BPERMUTE 1
# endif
#endif

__attribute__((reqd_work_group_size(8, 8, 1)))
__attribute__((reqd_work_group_size(8, WORKSIZE, 1)))
__kernel void JOIN(cn2,ALGO) (__global uint4 *Scratchpad, __global ulong *states,

#if (ALGO == cryptonight_gpu)
Expand All @@ -878,88 +883,123 @@ __kernel void JOIN(cn2,ALGO) (__global uint4 *Scratchpad, __global ulong *states
__global uint *Branch0, __global uint *Branch1, __global uint *Branch2, __global uint *Branch3, uint Threads)
#endif
{
__local uint AES0[256], AES1[256], AES2[256], AES3[256];
uint ExpandedKey2[40];
uint4 text;
__local uint AES0[256], AES1[256], AES2[256], AES3[256];
uint ExpandedKey2[40];
uint4 text;

const uint gIdx = getIdx();
uint gIdx = get_global_id(1) - get_global_offset(1);
uint groupIdx = get_local_id(1);
uint lIdx = get_local_id(0);

for (int i = get_local_id(1) * 8 + get_local_id(0); i < 256; i += 8 * 8) {
const uint tmp = AES0_C[i];
AES0[i] = tmp;
AES1[i] = rotate(tmp, 8U);
AES2[i] = rotate(tmp, 16U);
AES3[i] = rotate(tmp, 24U);
}
for (int i = groupIdx * 8 + lIdx; i < 256; i += get_local_size(0) * get_local_size(1)) {
const uint tmp = AES0_C[i];
AES0[i] = tmp;
AES1[i] = rotate(tmp, 8U);
AES2[i] = rotate(tmp, 16U);
AES3[i] = rotate(tmp, 24U);
}

barrier(CLK_LOCAL_MEM_FENCE);
barrier(CLK_LOCAL_MEM_FENCE);

#if (ALGO == cryptonight_gpu || ALGO == cryptonight_heavy || ALGO == cryptonight_haven || ALGO == cryptonight_bittube2 || ALGO == cryptonight_superfast)
__local uint4 xin1[8][8];
__local uint4 xin2[8][8];
__local uint4 xin1[WORKSIZE][8];
__local uint4 xin2[WORKSIZE][8];
#endif

#if(COMP_MODE==1)
// do not use early return here
if(gIdx < Threads)
// do not use early return here
if(gIdx < Threads)
#endif
{
states += 25 * gIdx;
{
states += 25 * gIdx;
#if(STRIDED_INDEX==0)
Scratchpad += gIdx * (MEMORY >> 4);
Scratchpad += gIdx * (MEMORY >> 4);
#elif(STRIDED_INDEX==1)
Scratchpad += gIdx;
Scratchpad += gIdx;
#elif(STRIDED_INDEX==2)
Scratchpad += (gIdx / WORKSIZE) * (MEMORY >> 4) * WORKSIZE + MEM_CHUNK * (gIdx % WORKSIZE);
Scratchpad += (gIdx / WORKSIZE) * (MEMORY >> 4) * WORKSIZE + MEM_CHUNK * (gIdx % WORKSIZE);
#elif(STRIDED_INDEX==3)
Scratchpad += (gIdx / WORKSIZE) * (MEMORY >> 4) * WORKSIZE + (gIdx % WORKSIZE);
Scratchpad += (gIdx / WORKSIZE) * (MEMORY >> 4) * WORKSIZE + (gIdx % WORKSIZE);
#endif

#if defined(__Tahiti__) || defined(__Pitcairn__)
#if defined(__Tahiti__) || defined(__Pitcairn__)

for(int i = 0; i < 4; ++i) ((ulong *)ExpandedKey2)[i] = states[i + 4];
text = vload4(get_local_id(1) + 4, (__global uint *)states);
for(int i = 0; i < 4; ++i) ((ulong *)ExpandedKey2)[i] = states[i + 4];
text = vload4(lIdx + 4, (__global uint *)states);

#else
#else
text = vload4(lIdx + 4, (__global uint *)states);
((uint8 *)ExpandedKey2)[0] = vload8(1, (__global uint *)states);

text = vload4(get_local_id(1) + 4, (__global uint *)states);
((uint8 *)ExpandedKey2)[0] = vload8(1, (__global uint *)states);
#endif

#endif

AESExpandKey256(ExpandedKey2);
}
AESExpandKey256(ExpandedKey2);
}

barrier(CLK_LOCAL_MEM_FENCE);
barrier(CLK_LOCAL_MEM_FENCE);

#if (ALGO == cryptonight_gpu || ALGO == cryptonight_heavy || ALGO == cryptonight_haven || ALGO == cryptonight_bittube2 || ALGO == cryptonight_superfast)
__local uint4* xin1_store = &xin1[get_local_id(1)][get_local_id(0)];
__local uint4* xin1_load = &xin1[(get_local_id(1) + 1) % 8][get_local_id(0)];
__local uint4* xin2_store = &xin2[get_local_id(1)][get_local_id(0)];
__local uint4* xin2_load = &xin2[(get_local_id(1) + 1) % 8][get_local_id(0)];
*xin2_store = (uint4)(0, 0, 0, 0);
# if (HAS_AMD_BPERMUTE == 1)
int lane = (groupIdx * 8 + ((lIdx + 1) % 8)) << 2;
uint4 tmp = (uint4)(0, 0, 0, 0);
# else
__local uint4* xin1_store = &xin1[groupIdx][lIdx];
__local uint4* xin1_load = &xin1[groupIdx][(lIdx + 1) % 8];
__local uint4* xin2_store = &xin2[groupIdx][lIdx];
__local uint4* xin2_load = &xin2[groupIdx][(lIdx + 1) % 8];
*xin2_store = (uint4)(0, 0, 0, 0);
# endif
#endif

#if(COMP_MODE == 1)
// do not use early return here
if (gIdx < Threads)
// do not use early return here
if (gIdx < Threads)
#endif
{
{

#if (ALGO == cryptonight_gpu || ALGO == cryptonight_heavy || ALGO == cryptonight_haven || ALGO == cryptonight_bittube2 || ALGO == cryptonight_superfast)

# if (HAS_AMD_BPERMUTE == 1)
#pragma unroll 2
for(int i = 0, i1 = lIdx; i < (MEMORY >> 7); ++i, i1 = (i1 + 16) % (MEMORY >> 4))
{
text ^= Scratchpad[IDX((uint)i1)];
text ^= tmp;

#pragma unroll 10
for(int j = 0; j < 10; ++j)
text = AES_Round(AES0, AES1, AES2, AES3, text, ((uint4 *)ExpandedKey2)[j]);

text.s0 ^= __builtin_amdgcn_ds_bpermute(lane, text.s0);
text.s1 ^= __builtin_amdgcn_ds_bpermute(lane, text.s1);
text.s2 ^= __builtin_amdgcn_ds_bpermute(lane, text.s2);
text.s3 ^= __builtin_amdgcn_ds_bpermute(lane, text.s3);
//__builtin_amdgcn_s_waitcnt(0);
text ^= Scratchpad[IDX((uint)i1 + 8u)];

#pragma unroll 10
for(int j = 0; j < 10; ++j)
text = AES_Round(AES0, AES1, AES2, AES3, text, ((uint4 *)ExpandedKey2)[j]);
tmp.s0 = __builtin_amdgcn_ds_bpermute(lane, text.s0);
tmp.s1 = __builtin_amdgcn_ds_bpermute(lane, text.s1);
tmp.s2 = __builtin_amdgcn_ds_bpermute(lane, text.s2);
tmp.s3 = __builtin_amdgcn_ds_bpermute(lane, text.s3);
//__builtin_amdgcn_s_waitcnt(0);
}

text ^= tmp;
# else

#pragma unroll 2
for(int i = 0, i1 = get_local_id(1); i < (MEMORY >> 7); ++i, i1 = (i1 + 16) % (MEMORY >> 4))
for(int i = 0, i1 = lIdx; i < (MEMORY >> 7); ++i, i1 = (i1 + 16) % (MEMORY >> 4))
{
text ^= Scratchpad[IDX((uint)i1)];
barrier(CLK_LOCAL_MEM_FENCE);
text ^= *xin2_load;

#pragma unroll 10
for(int j = 0; j < 10; ++j)
text = AES_Round(AES0, AES1, AES2, AES3, text, ((uint4 *)ExpandedKey2)[j]);

*xin1_store = text;

text ^= Scratchpad[IDX((uint)i1 + 8u)];
barrier(CLK_LOCAL_MEM_FENCE);
text ^= *xin1_load;
Expand All @@ -971,83 +1011,92 @@ __kernel void JOIN(cn2,ALGO) (__global uint4 *Scratchpad, __global ulong *states
*xin2_store = text;
}

barrier(CLK_LOCAL_MEM_FENCE);
text ^= *xin2_load;
barrier(CLK_LOCAL_MEM_FENCE);
text ^= *xin2_load;
# endif

#else
#pragma unroll 2
for (int i = 0; i < (MEMORY >> 7); ++i) {
text ^= Scratchpad[IDX((uint)((i << 3) + get_local_id(1)))];

#pragma unroll 10
for(int j = 0; j < 10; ++j)
text = AES_Round(AES0, AES1, AES2, AES3, text, ((uint4 *)ExpandedKey2)[j]);
}
#pragma unroll 2
for (int i = 0; i < (MEMORY >> 7); ++i)
{
text ^= Scratchpad[IDX((uint)((i << 3) + lIdx))];

#pragma unroll 10
for(int j = 0; j < 10; ++j)
text = AES_Round(AES0, AES1, AES2, AES3, text, ((uint4 *)ExpandedKey2)[j]);
}
#endif
}
}

#if (ALGO == cryptonight_gpu || ALGO == cryptonight_heavy || ALGO == cryptonight_haven || ALGO == cryptonight_bittube2 || ALGO == cryptonight_superfast)
/* Also left over threads performe this loop.
* The left over thread results will be ignored
*/
#pragma unroll 16
for(size_t i = 0; i < 16; i++)
{
#pragma unroll 10
for (int j = 0; j < 10; ++j) {
text = AES_Round(AES0, AES1, AES2, AES3, text, ((uint4 *)ExpandedKey2)[j]);
}

barrier(CLK_LOCAL_MEM_FENCE);
*xin1_store = text;
barrier(CLK_LOCAL_MEM_FENCE);
text ^= *xin1_load;
}
/* Also left over threads performe this loop.
* The left over thread results will be ignored
*/
#pragma unroll 16
for(size_t i = 0; i < 16; i++)
{
#pragma unroll 10
for (int j = 0; j < 10; ++j) {
text = AES_Round(AES0, AES1, AES2, AES3, text, ((uint4 *)ExpandedKey2)[j]);
}
#if (HAS_AMD_BPERMUTE == 1)
text.s0 ^= __builtin_amdgcn_ds_bpermute(lane, text.s0);
text.s1 ^= __builtin_amdgcn_ds_bpermute(lane, text.s1);
text.s2 ^= __builtin_amdgcn_ds_bpermute(lane, text.s2);
text.s3 ^= __builtin_amdgcn_ds_bpermute(lane, text.s3);
//__builtin_amdgcn_s_waitcnt(0);
#else
barrier(CLK_LOCAL_MEM_FENCE);
*xin1_store = text;
barrier(CLK_LOCAL_MEM_FENCE);
text ^= *xin1_load;
#endif
}
#endif

__local ulong State_buf[8 * 25];
__local ulong State_buf[8 * 25];
#if(COMP_MODE==1)
// do not use early return here
if(gIdx < Threads)
// do not use early return here
if(gIdx < Threads)
#endif
{
vstore2(as_ulong2(text), get_local_id(1) + 4, states);
}
{
vstore2(as_ulong2(text), lIdx + 4, states);
}

barrier(CLK_GLOBAL_MEM_FENCE);
barrier(CLK_GLOBAL_MEM_FENCE);

#if(COMP_MODE==1)
// do not use early return here
if(gIdx < Threads)
// do not use early return here
if(gIdx < Threads)
#endif
{
if(!get_local_id(1))
{
__local ulong* State = State_buf + get_local_id(0) * 25;
{
if(!lIdx)
{
__local ulong* State = State_buf + groupIdx * 25;

for(int i = 0; i < 25; ++i) State[i] = states[i];
for(int i = 0; i < 25; ++i) State[i] = states[i];

keccakf1600_2(State);
keccakf1600_2(State);

#if (ALGO == cryptonight_gpu)
if(State[3] <= Target)
{
ulong outIdx = atomic_inc(output + 0xFF);
if(outIdx < 0xFF)
output[outIdx] = get_global_id(0);
output[outIdx] = get_global_id(1);
}
#else
for(int i = 0; i < 25; ++i) states[i] = State[i];
for(int i = 0; i < 25; ++i) states[i] = State[i];

uint StateSwitch = State[0] & 3;
__global uint *destinationBranch1 = StateSwitch == 0 ? Branch0 : Branch1;
__global uint *destinationBranch2 = StateSwitch == 2 ? Branch2 : Branch3;
__global uint *destinationBranch = StateSwitch < 2 ? destinationBranch1 : destinationBranch2;
destinationBranch[atomic_inc(destinationBranch + Threads)] = gIdx;
uint StateSwitch = State[0] & 3;
__global uint *destinationBranch1 = StateSwitch == 0 ? Branch0 : Branch1;
__global uint *destinationBranch2 = StateSwitch == 2 ? Branch2 : Branch3;
__global uint *destinationBranch = StateSwitch < 2 ? destinationBranch1 : destinationBranch2;
destinationBranch[atomic_inc(destinationBranch + Threads)] = gIdx;
#endif
}
}
mem_fence(CLK_GLOBAL_MEM_FENCE);
}
}
mem_fence(CLK_GLOBAL_MEM_FENCE);
}

)==="
Expand Down