GOSTSec
/
ccminer-gostd-lite
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
Browse Source

skein: restore previous x11 speed

master
Tanguy Pruvot 10 years ago
parent
6d74256c2d
commit
37395eefe4
3 changed files with 12 additions and 16 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 2
      configure.ac
  2. 24
      quark/cuda_skein512.cu
  3. 2
      skein.cu

2
configure.ac
Unescape View File

@ -1,4 +1,4 @@
AC_INIT([ccminer], [1.6]) AC_INIT([ccminer], [1.6.1-git])
AC_PREREQ([2.59c]) AC_PREREQ([2.59c])
AC_CANONICAL_SYSTEM AC_CANONICAL_SYSTEM

24
quark/cuda_skein512.cu
Unescape View File

@ -358,14 +358,13 @@ void quark_skein512_gpu_hash_64(uint32_t threads, uint32_t startNounce, uint64_t
if (thread < threads) if (thread < threads)
{ {
// Skein // Skein
uint2 p[8];
uint2 h0, h1, h2, h3, h4, h5, h6, h7, h8; uint2 h0, h1, h2, h3, h4, h5, h6, h7, h8;
uint2 t0, t1, t2; uint2 t0, t1, t2;
uint32_t nounce = (g_nonceVector != NULL) ? g_nonceVector[thread] : (startNounce + thread); uint32_t nounce = (g_nonceVector != NULL) ? g_nonceVector[thread] : (startNounce + thread);
int hashPosition = nounce - startNounce; uint32_t hashPosition = nounce - startNounce;
uint64_t *inpHash = &g_hash[8 * hashPosition]; uint64_t *inpHash = &g_hash[hashPosition * 8U];
// Initialisierung // Initialisierung
h0 = vectorize(0x4903ADFF749C51CEull); h0 = vectorize(0x4903ADFF749C51CEull);
@ -377,6 +376,7 @@ void quark_skein512_gpu_hash_64(uint32_t threads, uint32_t startNounce, uint64_t
h6 = vectorize(0x991112C71A75B523ull); h6 = vectorize(0x991112C71A75B523ull);
h7 = vectorize(0xAE18A40B660FCC33ull); h7 = vectorize(0xAE18A40B660FCC33ull);
uint2 p[8];
// 1. Runde -> etype = 480, ptr = 64, bcount = 0, data = msg // 1. Runde -> etype = 480, ptr = 64, bcount = 0, data = msg
#pragma unroll 8 #pragma unroll 8
for (int i = 0; i < 8; i++) for (int i = 0; i < 8; i++)
@ -417,7 +417,7 @@ void quark_skein512_gpu_hash_64(uint32_t threads, uint32_t startNounce, uint64_t
// 2. Runde -> etype = 510, ptr = 8, bcount = 0, data = 0 // 2. Runde -> etype = 510, ptr = 8, bcount = 0, data = 0
#pragma unroll 8 #pragma unroll 8
for(int i=0; i<8; i++) for(int i=0; i<8; i++)
p[i] = make_uint2(0,0); p[i] = vectorize(0ull);
t0 = vectorize(8); // ptr t0 = vectorize(8); // ptr
t1 = vectorize(510ull << 55); // etype t1 = vectorize(510ull << 55); // etype
@ -442,12 +442,11 @@ void quark_skein512_gpu_hash_64(uint32_t threads, uint32_t startNounce, uint64_t
TFBIG_4o_UI2(17); TFBIG_4o_UI2(17);
TFBIG_ADDKEY_UI2(p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7], h, t, 18); TFBIG_ADDKEY_UI2(p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7], h, t, 18);
// fertig // output
uint2 *outpHash = (uint2*) (&g_hash[hashPosition * 8]); uint64_t *outpHash = &g_hash[hashPosition * 8U];
#pragma unroll 8 #pragma unroll 8
for(int i=0; i<8; i++) for(int i=0; i<8; i++)
outpHash[i] = p[i]; outpHash[i] = devectorize(p[i]);
} }
} }
@ -467,7 +466,7 @@ void quark_skein512_gpu_hash_64_v30(uint32_t threads, uint32_t startNounce, uint
int hashPosition = nounce - startNounce; int hashPosition = nounce - startNounce;
uint64_t *inpHash = &g_hash[8 * hashPosition]; uint64_t *inpHash = &g_hash[8 * hashPosition];
// Initialisierung // Init
h0 = 0x4903ADFF749C51CEull; h0 = 0x4903ADFF749C51CEull;
h1 = 0x0D95DE399746DF03ull; h1 = 0x0D95DE399746DF03ull;
h2 = 0x8FD1934127C79BCEull; h2 = 0x8FD1934127C79BCEull;
@ -517,7 +516,7 @@ void quark_skein512_gpu_hash_64_v30(uint32_t threads, uint32_t startNounce, uint
// 2. Runde -> etype = 510, ptr = 8, bcount = 0, data = 0 // 2. Runde -> etype = 510, ptr = 8, bcount = 0, data = 0
#pragma unroll 8 #pragma unroll 8
for(int i=0; i<8; i++) for(int i=0; i<8; i++)
p[i] = 0; p[i] = 0ull;
t0 = 8; // ptr t0 = 8; // ptr
t1 = 510ull << 55; // etype t1 = 510ull << 55; // etype
@ -574,9 +573,6 @@ void skein512_gpu_hash_close(uint32_t threads, uint32_t startNounce, uint64_t *g
TFBIG_KINIT_UI2(h0, h1, h2, h3, h4, h5, h6, h7, h8, t0, t1, t2); TFBIG_KINIT_UI2(h0, h1, h2, h3, h4, h5, h6, h7, h8, t0, t1, t2);
uint2 p[8] = { 0 }; uint2 p[8] = { 0 };
//#pragma unroll 8
//for (int i = 0; i<8; i++)
// p[i] = make_uint2(0, 0);
TFBIG_4e_UI2(0); TFBIG_4e_UI2(0);
TFBIG_4o_UI2(1); TFBIG_4o_UI2(1);
@ -674,7 +670,7 @@ void skein512_gpu_hash_80(uint32_t threads, uint32_t startNounce, uint64_t *outp
#pragma unroll #pragma unroll
for (int i = 2; i < 8; i++) for (int i = 2; i < 8; i++)
p[i] = make_uint2(0,0); p[i] = vectorize(0ull);
t0 = vectorize(0x50ull); // SPH_T64(bcount << 6) + (sph_u64)(extra); t0 = vectorize(0x50ull); // SPH_T64(bcount << 6) + (sph_u64)(extra);
t1 = vectorize(0xB000000000000000ull); // (bcount >> 58) + ((sph_u64)(etype) << 55); t1 = vectorize(0xB000000000000000ull); // (bcount >> 58) + ((sph_u64)(etype) << 55);

2
skein.cu
Unescape View File

@ -412,7 +412,7 @@ extern "C" int scanhash_skeincoin(int thr_id, uint32_t *pdata,
throughput = min(throughput, (max_nonce - first_nonce)); throughput = min(throughput, (max_nonce - first_nonce));
if (opt_benchmark) if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x0FFF; ((uint32_t*)ptarget)[7] = 0x07;
if (!init[thr_id]) if (!init[thr_id])
{ {

Write Preview
Loading…
Cancel
Save