Merge pull request #73 from marktwtn/hash-count

Calculate hash count for each PoW

Makefile

@@ -47,6 +47,10 @@ ifeq ("$(BUILD_JNI)","1")
 include mk/java.mk
 endif
 
+ifeq ("$(BUILD_STAT)","1")
+CFLAGS += -DENABLE_STAT
+endif
+
 TESTS = \
 	trinary \
 	curl \

README.md

@@ -26,6 +26,7 @@ Reference Implementation (IRI). Additionally, dcurl also supports the FPGA-accel
                      downloading from [latest JAVA source](https://github.com/DLTcollab/iri).
     - ``BUILD_COMPAT``: build extra cCurl compatible interface.
     - ``BUILD_FPGA_ACCEL``: build the interface interacting with the Cyclone V FPGA based accelerator. Verified on DE10-nano board and Arrow SoCKit board.
+    - ``BUILD_STAT``: show the statistics of the PoW information.
 * Alternatively, you can specify conditional build as following:
 ```shell
 $ make BUILD_GPU=0 BUILD_JNI=1 BUILD_AVX=1
@@ -43,13 +44,15 @@ $ make BUILD_GPU=1 check
         [ Verified ]
 *** Validating build/test-curl ***
         [ Verified ]
-*** Validating build/test-multi_pow_cpu ***
-        [ Verified ]
-*** Validating build/test-pow_sse ***
+*** Validating build/test-durl ***
+[dcurl] Implementation GPU (OpenCL) is initialized successfully
         [ Verified ]
-*** Validating build/test-pow_cl ***
+*** Validating build/test-multi_pow ***
         [ Verified ]
-*** Validating build/test-multi_pow_gpu ***
+*** Validating build/test-pow ***
+GPU - OpenCL
+[dcurl] Implementation GPU (OpenCL) is initialized successfully
+Success.
         [ Verified ]
 ```
 
@@ -64,9 +67,41 @@ $ make BUILD_AVX=1 check
         [ Verified ]
 *** Validating build/test-curl ***
         [ Verified ]
+*** Validating build/test-dcurl ***
+[dcurl] Implementation CPU (Intel AVX) is initialized successfully
+        [ Verified ]
 *** Validating build/test-multi_pow_cpu ***
         [ Verified ]
-*** Validating build/test-pow_avx ***
+*** Validating build/test-pow ***
+CPU - AVX
+[dcurl] Implementation CPU (Intel AVX) is initialized successfully
+Success.
+        [ Verified ]
+```
+
+* Test with AVX and show the PoW statistics
+```shell
+$ make BUILD_AVX=1 BUILD_STAT=1 check
+```
+
+* Expected Results
+```
+*** Validating build/test-trinary ***
+        [ Verified ]
+*** Validating build/test-curl ***
+        [ Verified ]
+*** Validating build/test-dcurl ***
+[dcurl] Implementation CPU (Intel AVX) is initialized successfully
+        [ Verified ]
+*** Validating build/test-multi_pow_cpu ***
+        [ Verified ]
+*** Validating build/test-pow ***
+CPU - AVX
+[dcurl] Implementation CPU (Intel AVX) is initialized successfully
+Hash count: 7396100
+PoW execution time: 3 sec
+Hash rate: 2465366.667 kH/sec
+Success.
         [ Verified ]
 ```
 
@@ -83,11 +118,19 @@ root@lampa:~/dcurl# make BUILD_FPGA_ACCEL=1 check
         [ Verified ]
 *** Validating build/test-curl ***
         [ Verified ]
-*** Validating build/test-pow_c ***
+*** Validating build/test-dcurl ***
+[dcurl] Implementation CPU (Intel SSE) is initialized successfully
+[dcurl] Implementation FPGA is initialized successfully
         [ Verified ]
-*** Validating build/test-multi_pow_cpu ***
+*** Validating build/test-multi_pow ***
         [ Verified ]
-*** Validating build/test-pow_fpga_accel *** 
+*** Validating build/test-pow ***
+CPU - SSE
+[dcurl] Implementation CPU (Intel SSE) is initialized successfully
+Success.
+FPGA
+[dcurl] Implementation FPGA is initialized successfully
+Success.
         [ Verified ] 
 ```
 

mk/defs.mk

@@ -15,3 +15,6 @@ BUILD_COMPAT ?= 0
 
 # Build FPGA backend or not
 BUILD_FPGA_ACCEL ?= 0
+
+# Show the PoW-related statistic messages or not
+BUILD_STAT ?= 0

src/clcontext.h

@@ -41,6 +41,7 @@ typedef struct {
     cl_ulong max_memory;
     size_t num_work_group;
     KernelInfo kernel_info;
+    uint64_t hash_count;
 } CLContext;
 
 enum {

src/common.h

@@ -0,0 +1,11 @@
+#ifndef COMMON_H_
+#define COMMON_H_
+
+typedef struct _pow_info PoW_Info;
+
+struct _pow_info {
+    double time;
+    uint64_t hash_count;
+};
+
+#endif

src/constants.h

@@ -4,7 +4,6 @@
 #define MinTryteValue -13
 #define MaxTryteValue 13
 #define SignatureSize 6561
-#define HashSize 243
 #define Depth 3
 #define Radix 3
 

src/implcontext.c

@@ -64,3 +64,8 @@ int8_t *getPoWResult(ImplContext *impl_ctx, void *pow_ctx)
 {
     return impl_ctx->getPoWResult(pow_ctx);
 }
+
+void *getPoWInfo(ImplContext *impl_ctx, void *pow_ctx)
+{
+    return impl_ctx->getPoWInfo(pow_ctx);
+}

src/implcontext.h

@@ -25,6 +25,7 @@ struct _impl_context {
     void *(*getPoWContext)(ImplContext *impl_ctx, int8_t *trytes, int mwm);
     bool (*doThePoW)(void *pow_ctx);
     int8_t *(*getPoWResult)(void *pow_ctx);
+    void *(*getPoWInfo)(void *pow_ctx);
     bool (*freePoWContext)(ImplContext *impl_ctx, void *pow_ctx);
 
     /* Linked list */
@@ -40,5 +41,6 @@ void *getPoWContext(ImplContext *impl_ctx, int8_t *trytes, int mwm);
 bool doThePoW(ImplContext *impl_ctx, void *pow_ctx);
 bool freePoWContext(ImplContext *impl_ctx, void *pow_ctx);
 int8_t *getPoWResult(ImplContext *impl_ctx, void *pow_ctx);
+void *getPoWInfo(ImplContext *impl_ctx, void *pow_ctx);
 
 #endif

src/pow_avx.c

@@ -487,29 +487,29 @@ static int8_t *tx_to_cstate(Trytes_t *tx)
 {
     Trytes_t *inn = NULL;
     Trits_t *tr = NULL;
-    int8_t tyt[(transactionTrinarySize - HashSize) / 3] = {0};
+    int8_t tyt[TRANSACTION_TRYTES_LENGTH - HASH_TRYTES_LENGTH] = {0};
 
     Curl *c = initCurl();
     int8_t *c_state = (int8_t *) malloc(STATE_TRITS_LENGTH);
     if (!c || !c_state) goto fail;
 
-    /* Copy tx->data[:(transactionTrinarySize - HashSize) / 3] to tyt */
-    memcpy(tyt, tx->data, (transactionTrinarySize - HashSize) / 3);
+    /* Copy tx->data[:TRANSACTION_TRYTES_LENGTH - HASH_TRYTES_LENGTH] to tyt */
+    memcpy(tyt, tx->data, TRANSACTION_TRYTES_LENGTH - HASH_TRYTES_LENGTH);
 
-    inn = initTrytes(tyt, (transactionTrinarySize - HashSize) / 3);
+    inn = initTrytes(tyt, TRANSACTION_TRYTES_LENGTH - HASH_TRYTES_LENGTH);
     if (!inn) goto fail;
 
     Absorb(c, inn);
 
     tr = trits_from_trytes(tx);
     if (!tr) goto fail;
 
-    /* Prepare an array storing tr[transactionTrinarySize - HashSize:] */
-    memcpy(c_state, tr->data + transactionTrinarySize - HashSize,
-           tr->len - (transactionTrinarySize - HashSize));
-    memcpy(c_state + tr->len - (transactionTrinarySize - HashSize),
-           c->state->data + tr->len - (transactionTrinarySize - HashSize),
-           c->state->len - tr->len + (transactionTrinarySize - HashSize));
+    /* Prepare an array storing tr[TRANSACTION_TRITS_LENGTH - HASH_TRITS_LENGTH:] */
+    memcpy(c_state, tr->data + TRANSACTION_TRITS_LENGTH - HASH_TRITS_LENGTH,
+           tr->len - (TRANSACTION_TRITS_LENGTH - HASH_TRITS_LENGTH));
+    memcpy(c_state + tr->len - (TRANSACTION_TRITS_LENGTH - HASH_TRITS_LENGTH),
+           c->state->data + tr->len - (TRANSACTION_TRITS_LENGTH - HASH_TRITS_LENGTH),
+           c->state->len - tr->len + (TRANSACTION_TRITS_LENGTH - HASH_TRITS_LENGTH));
 
     freeTrobject(inn);
     freeTrobject(tr);
@@ -525,22 +525,25 @@ static int8_t *tx_to_cstate(Trytes_t *tx)
 
 static void nonce_to_result(Trytes_t *tx, Trytes_t *nonce, int8_t *ret)
 {
-    int rst_len = tx->len - NonceTrinarySize / 3 + nonce->len;
+    int rst_len = tx->len - NONCE_TRYTES_LENGTH + nonce->len;
 
-    memcpy(ret, tx->data, tx->len - NonceTrinarySize / 3);
-    memcpy(ret + tx->len - NonceTrinarySize / 3, nonce->data,
-           rst_len - (tx->len - NonceTrinarySize / 3));
+    memcpy(ret, tx->data, tx->len - NONCE_TRYTES_LENGTH);
+    memcpy(ret + tx->len - NONCE_TRYTES_LENGTH, nonce->data,
+           rst_len - (tx->len - NONCE_TRYTES_LENGTH));
 }
 
 bool PowAVX(void *pow_ctx)
 {
     bool res = true;
     Trits_t *nonce_trit = NULL;
     Trytes_t *tx_tryte = NULL, *nonce_tryte = NULL;
+    time_t start_time, end_time;
 
     /* Initialize the context */
     PoW_AVX_Context *ctx = (PoW_AVX_Context *) pow_ctx;
     ctx->stopPoW = 0;
+    ctx->pow_info->time = 0;
+    ctx->pow_info->hash_count = 0;
     pthread_mutex_init(&ctx->lock, NULL);
     pthread_t *threads = ctx->threads;
     Pwork_struct *pitem = ctx->pitem;
@@ -556,6 +559,7 @@ bool PowAVX(void *pow_ctx)
         goto fail;
     }
 
+    time(&start_time);
     /* Prepare arguments for pthread */
     for (int i = 0; i < ctx->num_threads; i++) {
         pitem[i].mid = c_state;
@@ -573,9 +577,12 @@ bool PowAVX(void *pow_ctx)
         pthread_join(threads[i], NULL);
         if (pitem[i].n == -1)
             completedIndex = i;
+        ctx->pow_info->hash_count += (uint64_t) (pitem[i].ret >= 0 ? pitem[i].ret : -pitem[i].ret + 1);
     }
+    time(&end_time);
+    ctx->pow_info->time = difftime(end_time, start_time);
 
-    nonce_trit = initTrits(nonce_array[completedIndex], NonceTrinarySize);
+    nonce_trit = initTrits(nonce_array[completedIndex], NONCE_TRITS_LENGTH);
     if (!nonce_trit) {
         res = false;
         goto fail;
@@ -611,16 +618,18 @@ static bool PoWAVX_Context_Initialize(ImplContext *impl_ctx)
     void *threads_chunk = malloc(impl_ctx->num_max_thread * sizeof(pthread_t) * nproc);
     void *pitem_chunk = malloc(impl_ctx->num_max_thread * sizeof(Pwork_struct) * nproc);
     void *nonce_ptr_chunk = malloc(impl_ctx->num_max_thread * sizeof(int8_t *) * nproc);
-    void *nonce_chunk = malloc(impl_ctx->num_max_thread * NonceTrinarySize * nproc);
-    if (!threads_chunk || !pitem_chunk || !nonce_ptr_chunk || !nonce_chunk) goto fail;
+    void *nonce_chunk = malloc(impl_ctx->num_max_thread * NONCE_TRITS_LENGTH * nproc);
+    void *pow_info_chunk = malloc(impl_ctx->num_max_thread * sizeof(PoW_Info));
+    if (!threads_chunk || !pitem_chunk || !nonce_ptr_chunk || !nonce_chunk || !pow_info_chunk) goto fail;
 
     for (int i = 0; i < impl_ctx->num_max_thread; i++) {
         ctx[i].threads = (pthread_t *) (threads_chunk + i * sizeof(pthread_t) * nproc);
         ctx[i].pitem = (Pwork_struct *) (pitem_chunk + i * sizeof(Pwork_struct) * nproc);
         ctx[i].nonce_array = (int8_t **) (nonce_ptr_chunk + i * sizeof(int8_t *) * nproc);
         for (int j = 0; j < nproc; j++)
-            ctx[i].nonce_array[j] = (int8_t *) (nonce_chunk + i * NonceTrinarySize * nproc +
-                                                j * NonceTrinarySize);
+            ctx[i].nonce_array[j] = (int8_t *) (nonce_chunk + i * NONCE_TRITS_LENGTH * nproc +
+                                                j * NONCE_TRITS_LENGTH);
+        ctx[i].pow_info = (PoW_Info *) (pow_info_chunk + i * sizeof(PoW_Info));
         ctx[i].num_threads = nproc;
         impl_ctx->bitmap = impl_ctx->bitmap << 1 | 0x1;
     }
@@ -634,6 +643,7 @@ static bool PoWAVX_Context_Initialize(ImplContext *impl_ctx)
     free(pitem_chunk);
     free(nonce_ptr_chunk);
     free(nonce_chunk);
+    free(pow_info_chunk);
     return false;
 }
 
@@ -644,6 +654,7 @@ static void PoWAVX_Context_Destroy(ImplContext *impl_ctx)
     free(ctx[0].pitem);
     free(ctx[0].nonce_array[0]);
     free(ctx[0].nonce_array);
+    free(ctx[0].pow_info);
     free(ctx);
 }
 
@@ -681,6 +692,11 @@ static int8_t *PoWAVX_getPoWResult(void *pow_ctx)
     return ret;
 }
 
+static void *PoWAVX_getPoWInfo(void *pow_ctx)
+{
+    return ((PoW_AVX_Context *) pow_ctx)->pow_info;
+}
+
 ImplContext PoWAVX_Context = {
     .context = NULL,
     .description = "CPU (Intel AVX)",
@@ -693,4 +709,5 @@ ImplContext PoWAVX_Context = {
     .freePoWContext = PoWAVX_freePoWContext,
     .doThePoW = PowAVX,
     .getPoWResult = PoWAVX_getPoWResult,
+    .getPoWInfo = PoWAVX_getPoWInfo,
 };

src/pow_avx.h

@@ -5,6 +5,7 @@
 #include <stdbool.h>
 #include <stdint.h>
 #include <pthread.h>
+#include "common.h"
 #include "constants.h"
 
 typedef struct _pwork_struct Pwork_struct;
@@ -35,6 +36,8 @@ struct _pow_avx_context {
     int8_t input_trytes[TRANSACTION_TRYTES_LENGTH]; /* 2673 */
     int8_t output_trytes[TRANSACTION_TRYTES_LENGTH]; /* 2673 */
     int mwm;
+    /* PoW-related information */
+    PoW_Info *pow_info;
 };
 
 bool PowAVX(void *pow_ctx);