Poll: AVX1 64 bits vs AVX1 32 bits
This poll is closed.
slower : - 10%
6.06%
2 6.06%
same : +/- 5%
27.27%
9 27.27%
faster : + 10%
33.33%
11 33.33%
much faster : + 20%
15.15%
5 15.15%
on fire : + 50%
18.18%
6 18.18%
Total 33 vote(s) 100%
* You voted for this item. [Show Results]

Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
GSDX: Software mode Wanna bet on 64 bits performance
#41
Hum, potentially there are some margin of improvements for newer CPU. Texture conversion is done with a gather operation.
So far it is emulated with 8 instructions (for 1 texel lookup, bilinear is 4)
AVX2 adds a native gather instruction. However it isn't fast on my CPU (haswell), but skylake got intesting number.
Haswell : 20 uops, 9 cycles of latency
broadwell: 10 uops, 6 cycles of latency
skylake: 4 uops, 4 cycles of latency.
Reply

Sponsored links

#42
The dream will be AVX512 support.
* All reads (texture/frame buffer/depth buffer) could be replaced with a couple of gather instructions. And writes (frame & depth buffer) by 2 scatter instructions
* The additional mask parameter will remove various opcode
* 32 registers more than enough to not bother with temporary variable
* (and why not 16 pixels processed by shader instead of 8 (AVX2))
Reply
#43
Is there any advance on 64 bits in Linux? I'd love to install it in my Arch without all the 32 bits dependencies Laugh
Reply
#44
You can install it. And it will work if you select the interpreter for EE/VU (yes it is slow).
Reply




Users browsing this thread: 1 Guest(s)