default intensity was the max supported by the card, and perf is
not really better. I prefer to let it one under for cards with lower
memory (1GB)

Displayed data is the average of the last 50 scans in the 5 last minutes

Also move cuda common functions in a new file (cuda.cu)

Signed-off-by: Tanguy Pruvot <[email protected]>

sample usage made for blake and fugue (higher intensity for SM5.2)

add these to cuda_helper and clean unused code

Note: Heavy and Mjollnir are broken on linux (only)...

To check in the next version... 4 hours i try to fix that without
success. djm34 variant seems ok but also make a lot of rejects.

you can use PHP api-example.php as a json wrapper...

to be tested on windows...

Signed-off-by: Tanguy Pruvot <[email protected]>

PS: still under dev, wait the final 1.4.8 ;)

possible values :

 5000 or :5000 to use port 5000 (local only)
  0.0.0.0:5000 to allow connections from the network
127.0.0.1:4068 to only allow local connections (default)

Use -b 0 to disable the API system.

Signed-off-by: Tanguy Pruvot <[email protected]>

  Based on mwhite73 <[email protected]> implementation

  Linked to the api system

  Also fix Makefile to support standard c++ files
  This prevent nvcc use without device code

Signed-off-by: Tanguy Pruvot <[email protected]>

There was a different behavior on linux and visual studio

That was making it hard to link functions correctly

That remove some ifdef / extern "C" requirements

note about x86 releases, x86 nvml.dll is not installed on Windows x64!

nvml.dll doesnt exists for 32bit binaries! use nvapi to get infos

seems to have more/different features than NVML... like pstate etc..

This is nvapi r343 : https://developer.nvidia.com/nvapi

Signed-off-by: Tanguy Pruvot <[email protected]>

totals, readable uptime, autorefresh

handle "\r\n" and terminate connection if command is not known

windows only via nvapi, if nvml function is not supported

The previous way it was done was not correct (hashdone divided/2)

So add a factor 0.5 to the computed hashrate... to stay comparable

enhance multi-gpu stats and fix nvapi indexes

change syslog prefix to ccminer (cpuminer remains)

api 1.1 modified - not officially released yet

mapping by Name was a quick test...

renammed to "threads", stats was too much generic

But no way, even recent 346 driver series or 343. Seems limited to
Tesla products...

A dev version was released on http://cryptomining-blog.com/

Please update, the previous one has some bugs when using multiple
gpus and the API format has changed!

fix php warnings and enable cpuminer compat

Original Commit :
Removed sharedmem and reduced calculations with precalcing (ECHO hash).
750ti + 20KHASH(x11)

tpruvot notes:
Real change is more of 10 KH/s on stock clocks (but real)
launch bounds disabled, no perf increase with 64 registers

Previous echo commit was only increasing linux performance, and reducing
windows perf compared to the 1.4.9, this one seems to give at least
the 1.4.9 on windows, and the same on linux...

Shavite optimisation seems ok on both (use now 64 registers)

the launch_bounds will force the number of registers, so remove specific
Makefile rules on linux...

manual "cherry pick" with fixed line endings and some adaptations

This should not be used for all algos... not enabled yet

todo: multiple nounces or blake32 style checkup

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...

indeed, the pragma unroll doesnt always make things faster

asm part... to check later

in fact max exists in jansson includes (in tree only)

Add them to miner.h

seems to works in solo mode too...

note: anime wallet to check, different problem without reject-reason

tpruvot: i didnt pick the asm replace_hiword, slower on linux

allow pools to switch more coins without reconnect

Tested ok on a nicehash x11 pool

Signed-off-by: Tanguy Pruvot <[email protected]>

On maxwell compress1 and compress2 can be run in one run instead of two.(750TI + 20KHASH)

bug seen on windows with xpool.ca (x11)

There was sometimes problems with float values close to 0 (like 0.001)

and update the old var with active gpus if -d is set

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

use git "show -w <commithash>" to see changes

Duplicated functions in merged Cube+Luffa could be cross linked without

to fix histo api command without param

seen on yamp.com pools

subsys ids are wrong... so use generic one

Simple change which affect all algos based on SIMD512

fresh, qubit, s3, x11 to x17...

why was it even included here in the first place?
It's part of the CUDA Toolkit

to prevent config.h conflicts

remove vbios revision, goal was to find the vendor...

We have no way yet to identify a MSI 750 from a Gigabyte one...

Fix for the extranonce feature, can also make the miner more stable

will be used for external debugging...

move nvdriver to sys infos (unique on system)

prepare pool query

Sample with -i 18.5
  Adding 131072 threads to intensity 18, 393216 cuda threads

And with -i 19.5
  Adding 262144 threads to intensity 19, 786432 cuda threads

will allow kopiemtu livecd internal monitor to set it to cpuminer

The DLL exists for x64 targets but seems not loadable

The nvml.cpp code was wrote to support both NVAPI and NVML on windows
because both apis have unique "features". like Fan RPM vs Fan Percent

uids could be useful to create graphes from history data

Note: please do a clean build after this commit (changes in miner.h)

and report pstate in hwinfo only

The core problem was the cuda hefty Thread per block set to high
but took me several hours to find that...

btw... +25% in heavy 12500 with 256 threads per block... vs 128 & 512
if max reg count is set to 80...