2022 14d post processing reservations and results - Page 6

frmky · 2022-05-20, 03:14

Quote:

Originally Posted by RichD

71111_381 has built a matrix. Too big for my GPU. Should have factors in less than 60 hours.

Can you give it a try on your gpu with

Code:

./msieve -nc2 "skip_matbuild=1 use_managed=1" -g 0 -v

I'm curious if it will run, and if so how long it will take.

RichD · 2022-05-20, 12:11

Quote:

Originally Posted by frmky

Can you give it a try on your gpu with

Code:

./msieve -nc2 "skip_matbuild=1 use_managed=1" -g 0 -v

I'm curious if it will run, and if so how long it will take.

I still get OUT_OF_MEMORY. The size of the matrix is 7.56M.

RichD · 2022-05-20, 12:32

Quote:

Originally Posted by RichD

I still get OUT_OF_MEMORY. The size of the matrix is 7.56M.

Oops, I typed used_managed=1. Now running with ETA 28+ hours.

RichD · 2022-05-21, 19:33

Code:

p78 factor: 762750299372672355883268454137776253787176740120346983485993070985177851516039
p87 factor: 349377274288344974698603745794705474301449007747822341745620295139645750449066123036479

99M unique relations built a 7.56M matrix using TD=120.
Solve time about 28 hours on GTX 1660 (6GB), VBITS=256, use_managed=1.

Look like the GPU memory was a little tight.

Code:

commencing Lanczos iteration 
vector memory use: 1614.2 MB 
dense rows memory use: 230.6 MB 
sparse matrix memory use: 5497.9 MB 
memory use: 7342.8 MB 
Allocated 517.7 MB for SpMV library 
Allocated 517.7 MB for SpMV library

Log at: https://pastebin.com/azeWA2XU

frmky · 2022-05-21, 23:33

Yep. With the managed memory option, the program stores portions of the sparse matrix blocks in main memory if necessary and moves them to the GPU when they are needed in each iteration. This significantly increases traffic on the PCIe bus. The GPU spends much more time waiting for data, but it can still be faster than running on the CPU.

RichD · 2022-05-24, 19:52

Taking 126__127_7m1.

richs · 2022-05-25, 04:40

Taking 72223_245

richs · 2022-05-26, 20:02

Code:

p61 factor: 2102566791387466632448419655029903561085790971150084562135149
p89 factor: 61742469877737080442278649390467103244337617054848672724667378258844197673925682100250493

Approximately 4 hours on 6 threads of a Core i7-10750H with 16 GB memory for a 2.14M matrix at TD=130.

Log attached and at https://pastebin.com/ahqPPTaf

Factors reported to factordb and Studio Kamada.

RichD · 2022-05-27, 00:52

Code:

p73 factor: 1389370807537644850457220186716158435702624687245915808336684977801386389
p106 factor: 3876599714493456973044198680302223391853977361111696109004364392838186145973313523108830715065226722317349

110M unique relations built a 6.95M matrix using TD=112. (116 failed)
Solve time about 25.5 hours using GTX 1660 (6GB), --use_managed=1.

Log at: https://pastebin.com/aTmUympa

richs · 2022-05-27, 20:21

Taking 74441_229

richs · 2022-06-01, 03:45

Taking 44447_239

2022-05-27, 00:52	#64
RichD Sep 2008 Kansas 59×61 Posts	126__127_7m1 factored Code: p73 factor: 1389370807537644850457220186716158435702624687245915808336684977801386389 p106 factor: 3876599714493456973044198680302223391853977361111696109004364392838186145973313523108830715065226722317349 110M unique relations built a 6.95M matrix using TD=112. (116 failed) Solve time about 25.5 hours using GTX 1660 (6GB), --use_managed=1. Log at: https://pastebin.com/aTmUympa

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
2021 15e post processing reservations and results	swellman	NFS@Home	188	2021-12-30 21:10
2020 15e post processing reservations and results	swellman	NFS@Home	112	2020-12-29 22:58
2019 14e post processing reservations and results	swellman	NFS@Home	862	2019-12-31 10:51
2018 15e post-processing reservations and results	fivemack	NFS@Home	221	2019-01-04 13:08
2018 14e post processing reservations and results	pinhodecarlos	NFS@Home	551	2019-01-04 13:06

2022-05-21, 23:33	#60
frmky Jul 2003 So Cal 2,399 Posts	Yep. With the managed memory option, the program stores portions of the sparse matrix blocks in main memory if necessary and moves them to the GPU when they are needed in each iteration. This significantly increases traffic on the PCIe bus. The GPU spends much more time waiting for data, but it can still be faster than running on the CPU.

2022-05-24, 19:52	#61
RichD Sep 2008 Kansas 7017₈ Posts	Taking 126__127_7m1.

2022-05-25, 04:40	#62
richs "Rich" Aug 2002 Benicia, California 3·7·73 Posts	Taking 72223_245

2022-05-27, 20:21	#65
richs "Rich" Aug 2002 Benicia, California 3·7·73 Posts	Taking 74441_229

2022-06-01, 03:45	#66
richs "Rich" Aug 2002 Benicia, California 3×7×73 Posts	Taking 44447_239