355b835ae0
benchmark: enhance the mem leak detection
...
reduce "false" warnings, and ignore unrelated/small ones <= 1 MB
On windows the gpu memory can be allocated by other processes
+ some cleanup in algos... (free/gpulog)
9 years ago
5bf1f98200
various fixes for SM 2.1 and the benchmark
...
X11+ algos and quark are not compatible for the moment
but these ones are :
Benchmark results for Gigabyte GTX 460 (SM 2.1 / 1 GB):
blakecoin : 159090.5 kH/s, 1 MB, 1048576 thr.
blake : 70208.9 kH/s, 1 MB, 1048576 thr.
bmw : 122802.6 kH/s, 65 MB, 2097152 thr.
deep : 3533.6 kH/s, 33 MB, 524288 thr.
fugue256 : 43177.9 kH/s, 17 MB, 524288 thr.
heavy : 4118.2 kH/s, 147 MB, 524032 thr.
keccak : 18673.1 kH/s, 129 MB, 2097152 thr.
luffa : 28816.0 kH/s, 257 MB, 4194304 thr.
lyra2 : 213.7 kH/s, 570 MB, 65536 thr.
mjollnir : 3895.6 kH/s, 147 MB, 524032 thr.
nist5 : 1101.4 kH/s, 67 MB, 1048576 thr.
penta : 501.6 kH/s, 21 MB, 327680 thr.
skein : 5432.4 kH/s, 65 MB, 1048576 thr.
skein2 : 6788.9 kH/s, 33 MB, 524288 thr.
whirlpool : 688.5 kH/s, 33 MB, 524288 thr.
zr5 : 122.5 kH/s, 86 MB, 262144 thr.
9 years ago
9dfa757dc7
warn on cuda errors + various small changes
...
The full benchmark can now be launched with "ccminer --benchmark"
add a new helper function which log a warning with last cuda error
(not shown with the quiet option) : CUDA_LOG_ERROR();
it can be used where miner.h is included (.c/.cpp/.cu)
fix x14 (in ccminer.cpp), a break was missing in switch..case
9 years ago
d195f2e8a2
intensity: do not reduce throughput before init
...
Else the memory allocated could be less than required later
btw, use the new "cuda" function to apply intensity/throughput
9 years ago
ab5cc7162e
refactor: create bench.cpp and algos.h
...
Also enhance multi-thread benchmark synchro. with pthread barriers
9 years ago
c2214091ae
benchmark: free last memory leaks on algo switch
...
remains my original lyra2 implementation to fix... (cuda_lyra2.cu)
I guess some kind of memory overflow force the driver to allocate
memory... but was unable to free it without device reset.
9 years ago
4e1e03b891
benchmark: store all algos results + cuda fixes
...
Note: lyra2, lyra2v2 and script seems to have problems
to coexist with other algos... to run after some of them...
moved lyra2 first and skip scrypt/jane for the moment...
Only stored in memory for now.. to display a table after the bench
ccminer -a auto --benchmark
Results may be exported later to a json file...
9 years ago
922c2a5cd7
algos: free allocated mem for algo switch
...
All can be freed propertly now, except script (reset) and lyra2 (leak)
9 years ago
ee93927fac
diff: use the new function in all algos
9 years ago
5f12943de5
whirlpool: add algo free function + vstudio
9 years ago
b641bfdf8b
diff: rename functions like cpuminer-multi
...
more proper, intuitive...
9 years ago
3f589cc4db
restore the whirlpool algo
9 years ago
e1c4b3042c
algos: add functions to free allocated resources
...
Will be used later for algo switching
not really tested yet...
9 years ago
5308898d1c
start v1.7, apply new prototypes to all algos
9 years ago
3788fef9fd
Restore whirlpoolx algo
10 years ago
42bcb91ca0
x11: update sp luffa/cube to get closer x11 speeds..
...
i had to clean it... lot of unused defines...
10 years ago
298573b45d
Drop Whirlpool and whirpoolx algos, no more used...
...
Vanillacoin is now PoS only
10 years ago
2f541065fb
cuda_helper: rename correctly hiword/loword functions
10 years ago
2113be6eec
blake80: some changes and launch bounds, no perf changes
10 years ago
3d3f2e2cb5
warnings: use the right device id (device_map[thr_id])
10 years ago
5988e945ef
some cleanup and longpoll changes
10 years ago
ae8e863591
remove uint32_t cast
10 years ago
d6a421ae10
whirlpoolx: windows stay fluent with intensity 20
10 years ago
ebd23bcc66
whirlpoolx: real fix for multi gpus
...
Main problem was the arrays allocations which should be made per cpu
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
10 years ago
10d4f4a96a
whirlpoolx: try to fix linux hashrates
...
bug reported by PVMining with a lot of cards (not seen here)
10 years ago
f8f46bfe5b
whirlpoolx: fix for multi-gpu support
...
add also cputest entry and do some cleanup...
10 years ago
0019936328
Import Whirlpoolx algo from alexis78 repo
...
remove trailing spaces and move to x15 folder
10 years ago
e6112e878d
cleanup: use unsigned throughput parameters
...
Yes, its a big commit, was waiting 1.6 to do that...
Sorry for your possible merge issues ;)
10 years ago
26b51a557b
Allow different intensity per device
...
and clean the old variables, no more required
10 years ago
2a5233f56e
api: report throughput when default
10 years ago
cafd4477d7
Handle a maximum of 16 gpus (vs 8 before)
...
Some cards have 2 gpus on board...
10 years ago
9b5d3bf528
Fix whirl algo
10 years ago
c3bdb623e8
Check and submit multiple nonces in one loop
...
Added to most algos, checkhash function scans a big range
and can find multiple nonces at once if the difficulty is low.
Stop ignoring them, submit second one if found...
Clean the draft code for rc=2 implemented for blake and pentablake
btw... fix the reduced displayed hashrate when a nonce is found...
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
10 years ago
118a6be361
checkhash: simplify the common function
...
use klaus trivial function, the old code has always been a bit weird..
split cuda_check_cpu_hash_64 in two functions, keep old for branched stuff
10 years ago
cc9df2d361
use ROTL32 in x14_shabal
...
no impacts on perf but easier to read
10 years ago
8ad180cc70
various small changes
...
heavy: reduce by 256 threads default intensity to all -i 20
cuda: put static thread init bools outside the code (made once)
api: fix nvml header to build without
10 years ago
6ae28162db
various extern cleanup + api history uids and gpu SM
...
uids could be useful to create graphes from history data
Note: please do a clean build after this commit (changes in miner.h)
10 years ago
f0d91ab8a6
Luffa and simd merged to one kernal.
...
Small echo rewrite. +10KHASH on the 650(compute 3.0)
tpruvot: add Linux Makefile - Force to 80 registers (else -30KH/s)
Note : the hashrate seems more constant with this change
10 years ago
26b9fe3586
faster x15, +23KH or 4ms on whirpool (30ms vs 34ms)
...
tpruvot: i didnt pick the asm replace_hiword, slower on linux
10 years ago
73f22b237a
Prepare trap of hardware/mem failures
10 years ago
fe4ad36b73
intensity: sign warnings fixes min(i,u)
10 years ago
c859041993
quark/blake512 opt. pointed by sp without asm
...
indeed, the pragma unroll doesnt always make things faster
asm part... to check later
10 years ago
438308b3a2
Rework benchmark mode and min/max range
...
Was maybe my fault, but the benchmark mode was
always recomputing from nonce 0.
Also fix blake if -d 1 is used (one thread but second gpu)
stats: do not use thread id as key, prefer gpu id...
10 years ago
b128312efb
cuda: store device SM in a global var
...
sample usage made for blake and fugue (higher intensity for SM5.2)
add these to cuda_helper and clean unused code
10 years ago
11c5ec810d
Handle intensity param in all algos
...
and add a check related to start/max nounce params
10 years ago
7cc5222394
Move common check_cpu functions to root
10 years ago
95ac1d0f19
x11: adapt some blake 256 opts to 512 one
...
blake512: for the moment 6.2ms vs 7.12 before (+10%)
10 years ago
10314d844f
whirlpool: remove dead code, win 2ms in final hash
11 years ago
64e8cd3f98
add x17 algo, cleaned djm34 commit
...
todo: visual studio...
11 years ago
5cf024cc5d
whirl: add missing merkle for some pools
11 years ago
Tanguy Pruvot