LBRYCommunity/cpuminer
1
0
Fork 0
Code Issues 3 Pull requests Projects Releases Packages Wiki Activity

Speed up generation of TX buffer in merkle root generation

Browse source 
The `gbt_work_decode` function builds a long string of all the
transactions (`work->txs`) when generating the merkle root. This
is done inefficiently using `strcat` on the beginning of the
buffer for each TX. `strcat` will take more time the longer the
string is.

To fix this, we keep track of the end of the end of the string and
use `strcpy` to directly write the current TX there.
This commit is contained in:
Axel Gembe 2020-11-10 15:39:30 +07:00
parent 173a497e60
commit 1abe346537
No known key found for this signature in database
GPG key ID: 1CB3AA2143109AAC
1 changed files with 7 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

10
cpu-miner.c
View file

@ -578,6 +578,7 @@ static bool gbt_work_decode(const json_t *val, struct work *work)
work->txs = malloc(2 * (n + cbtx_size + tx_size) + 1);
bin2hex(work->txs, txc_vi, n);
bin2hex(work->txs + 2*n, cbtx, cbtx_size);
char *txs_end = work->txs + strlen(work->txs);
/* generate merkle root */
merkle_tree = malloc(32 * ((1 + tx_count + 1) & ~1));
@ -585,7 +586,8 @@ static bool gbt_work_decode(const json_t *val, struct work *work)
for (i = 0; i < tx_count; i++) {
tmp = json_array_get(txa, i);
const char *tx_hex = json_string_value(json_object_get(tmp, "data"));
const int tx_size = tx_hex ? strlen(tx_hex) / 2 : 0;
const size_t tx_hex_len = tx_hex ? strlen(tx_hex) : 0;
const int tx_size = tx_hex_len / 2;
if (segwit) {
const char *txid = json_string_value(json_object_get(tmp, "txid"));
if (!txid || !hex2bin(merkle_tree[1 + i], txid, 32)) {
@ -603,8 +605,10 @@ static bool gbt_work_decode(const json_t *val, struct work *work)
sha256d(merkle_tree[1 + i], tx, tx_size);
free(tx);
}
if (!submit_coinbase)
strcat(work->txs, tx_hex);
if (!submit_coinbase) {
strcpy(txs_end, tx_hex);
txs_end += tx_hex_len;
}
}
n = 1 + tx_count;
while (n > 1) {