- --
Viewing Issue Advanced Details
ID | Category [?] | Severity [?] | Reproducibility | Date Submitted | Last Update |
---|---|---|---|---|---|
05627 | Speed | Trivial | Always | Jul 18, 2014, 10:41 | Jul 19, 2014, 18:57 |
Tester | haynor666 | View Status | Public | Platform | MAME (Official Binary) |
Assigned To | AWJ | Resolution | Fixed | OS | Windows Vista/7/8 (64-bit) |
Status [?] | Resolved | Driver | |||
Version | 0.153 | Fixed in Version | 0.154 | Build | 64-bit |
Fixed in Git Commit | Github Pull Request # | ||||
Summary | 05627: slyspy and clones: comparing to other games in this driver performance is low | ||||
Description | most of these games are around 1200% on 4670k but slyspy is about 350%. On older coreduo processors is hardly playable. | ||||
Steps To Reproduce | |||||
Additional Information | Might be intentional/not a bug | ||||
Github Commit | |||||
Flags | |||||
Regression Version | |||||
Affected Sets / Systems | slyspy and clones | ||||
Attached Files
|
|||||
Relationships
There are no relationship linked to this issue. |
Notes
8
No.10840
Haze Senior Tester
Jul 18, 2014, 11:33
edited on: Jul 18, 2014, 11:33 |
the protection device dynamically alters the memory map every single frame. doing this 'the proper way' in MAME is slow these days. doing it in a more hacky way (as it was a few years ago) would be faster, but the code wouldn't be as good. |
---|---|
No.10841
haynor666 Tester
Jul 18, 2014, 12:08
|
Ok, so it's not a mistake but intended change. Thanks Haze for explanation :) BTW. Other games are running much faster. Different protection scheme or hacks ? |
No.10842
AWJ Developer
Jul 18, 2014, 22:54
|
Modifying address maps dynamically is not "the proper way" to do things anymore, precisely because it's extremely slow. |
No.10844
Augusto Tester
Jul 19, 2014, 05:58
edited on: Jul 19, 2014, 10:48 |
Windows 8.1 64 - CPU AMD Athlon II X2 270 Dual Core 3.5 GHZ - 4 GB RAM - Geforce GT 640 with the games below. Sly Spy run 42% in some moments. Shadow Force just on game title is 57%. Any game using STV driver in recent MAME version is 25 to 30%, but the CPU is being used below of 50% in each core and in OS Power Management High Performance Profile was selected. |
No.10845
haynor666 Tester
Jul 19, 2014, 10:03
edited on: Jul 19, 2014, 11:33 |
Boulder Dash is also slower than the others, here is summary on 4670k: slyspy 347,48 bouldash 577,54 hbarrel 1053,29 baddudes 1058,72 birdtry 1127,86 midres 1228,72 hippodrm 1249,52 robocop 1263,05 EDIT. tested latest svn and I didn't see any visible performance improvement :/ |
No.10846
AWJ Developer
Jul 19, 2014, 13:03
edited on: Jul 19, 2014, 13:06 |
Really, no improvement? slyspy -bench 60 went from 300% to over 900% for me with r31317. bouldash is the same machine driver as slyspy so it should see a similar improvement. |
No.10847
haynor666 Tester
Jul 19, 2014, 13:18
edited on: Jul 19, 2014, 13:25 |
My mistake, instead hit latest svn mame I hit official 153 :/ I think I should delete all those svn builds and previous mame versions :] Right now with bench 240: slyspy; 1411.20% bouldash; 1369.85% Thanks for fix AWJ :)) |
No.10848
haynor666 Tester
Jul 19, 2014, 18:57
|
Augusto, there many drivers than works slow even if those are typical 2D games. Moreover, in the same driver performance between games can vary a lot. pgm: kovsh 260,65; killbldp 358,56; theglad 388,13; kov2 420,58; kov 685,54; espgal 758,10; ddp2 910,53 taito_f3: arkretrn 340,06; gunlock 553,25; arabianm 677,29; cv1k: mmpork 345,41; espgal2 509,50; dsmbl 730,95; ssv: wineag2 368,93; stmblade 602,55;mslider 1510,69; BTW. There is also one game with very low performance in dec8 driver - shackled 413,47 while ghostb 1374,3; srdarwin 2728,52 |