704776027b
gsdx linux: update gui to add a HPO v2 checkbox
...
Sort std hack & upscaling hack
2016-11-29 17:22:02 +01:00
a95adcb999
gsdx-ogl: correct texture coordinate by 0.5 when vertex position is -0.5
...
Avoid ghosting in Captain tsubasa
2016-11-29 17:22:02 +01:00
61a7c747e1
gsdx-ogl: alternate implementation of half pixel offset
...
The previous implementation of HPO adds an offset on vertex position. It
doesn't always work beside it moves the rendering window.
The new implementation will add a texture offset so that instead to sample
the middle of the GS texel, we will sample the middle of the real texture texel.
It must be manually enabled with
* UserHacks_HalfPixelOffset_New = 1 (keep a small offset as intended by GS effect)
* UserHacks_HalfPixelOffset_New = 2 (no offset)
v2: always apply a 0.5 offset in case of float coordinates (Tales of Abyss)
Might break other games but few of them uses float coordinates to read
back the target
2016-11-29 17:22:02 +01:00
c2229e3c0b
gsdx-ogl: add a texture offset uniform parameter to vertex shader
...
It would be used for a new implementation of the half pixel offset hack
Hopefully it doesn't badly impact the perf on low end iGPU
2016-11-29 17:22:02 +01:00
f6cad2235b
gsdx: defer GSScanlineConstantData init
...
Avoid AVX instruction in the middle
Issue #1677
2016-11-28 19:40:25 +01:00
c9db1c6c4b
vtune: plug PCSX2 core + add missing profiling (VU/VIF/TLB)
...
Doesn't fully work yet
* Unknown stack frame
* Outside any known module
Potential root cause:
* Nvidia driver
* VU code as ebp is required for emulation so likely no frame
2016-11-28 19:07:04 +01:00
e4516ac9b8
cmake: add extra SSE4 and AVX2 build of GSdx when DISABLE_ADVANCE_SIMD is enabled
...
It will provide a speed boost on distribution that only enable SSE2
2016-11-25 16:35:40 +01:00
8431299b92
gsdx sw: port code to the new constant object
2016-11-24 23:03:26 +01:00
3b5bc9c38d
gsdx sw: create a constant buffer
...
* Use POD type to avoid SSE/AVX compilation dependency
* global object to reduce cache miss
* dynamically object so give a chance to allocate below 2GB (allow x64
optimization)
2016-11-24 23:03:26 +01:00
0f5529be18
gsdx sw: s/g_cpu/m_cpu/
2016-11-24 23:03:25 +01:00
c3e38e46c7
gsdx sw x64: disable mipmap support on AVX
...
Until it is implemented (might never happen)
2016-11-24 23:03:25 +01:00
608bb5ccb2
gsdx sw x64: add AVX2 implementation for VS
...
FS was copied from 32 bits (require massive update)
2016-11-24 23:03:25 +01:00
15220c386a
gsdx sw x64: setup prim miss some optimizations
2016-11-24 23:03:25 +01:00
e3bfa2be88
gsdx sw: factorize common draw scanline code
...
Ymm inherite from Xmm so it is useless to duplicate the code
Add a parameter to alltrue to test the good register
2016-11-24 23:03:25 +01:00
211c7745de
gsdx: don't try to correct depth in primitive trace
...
Avoid to go above the maximum size allowed by the format
Issue #1674
2016-11-24 22:24:00 +01:00
417d0a3606
GSdx Merge Circuit: Fix regressions
...
Fixes screen shaking in Tenchu: Wrath Of Heaven, and graphical issues in
NASCAR 09.
2016-11-23 22:14:44 +01:00
0d275868a5
gsdx x64: quick fix for windows
...
Until we got a full implementation
2016-11-21 18:18:09 +01:00
37379d5d1d
gsdx sw x64: shuffle memory allocation
...
tex address is a3
vm address is a1
Could help to avoid REX prefix
Reduce prologue/epilogue register copy
Byte code size 41893 => 38912 (on my testcase)
2016-11-20 20:19:13 +01:00
b76305a0d2
gsdx sw x64: keep the copy of top in the stack for dthe
...
An extra stack access by pixel won't impact perf (only 16 bits game with dithering)
Allow to save 2 registers (a1/a3)
2016-11-20 20:19:13 +01:00
994274623b
gsdx sw x64: use rip addressing on draw scan line
...
byte code 41997 => 41893
However it will allow to save 2 registers as future optimization
2016-11-20 20:19:12 +01:00
7c06e87d59
gsdx sw x64: use rip addressing on setup prim
...
byte code: 9017 => 8736
Save a register
2016-11-20 20:19:12 +01:00
923c297dfc
gsdx sw: vinsert128 opcode require an XMM register
...
Reported as an error on lastest Xbyak
2016-11-20 20:19:12 +01:00
e674518c4e
Update from xbyak 4.84 to 5.11
...
* bin2hex.h is removed
* vptest/vpblendvb YMM support integrated upsteam
* better support of rip for 64 bits
* AVX512 support (only miss the CPU now)
Local change: add BSD3 clause
2016-11-20 13:21:11 +01:00
681c09f25c
gsdx sw: catch xbyak exception
2016-11-20 12:58:07 +01:00
0fd5346860
gsdx x64 linux: only mmap in lower 2GB code segment
...
Allow to use rip addressing
2016-11-20 10:38:15 +01:00
2252ba1bb7
gsdx: add the detected ISA on the name
...
As a SSE build could potentially uses AVX operations.
2016-11-20 10:38:15 +01:00
ef25502491
gsdx build: don't exclude AVX files.
...
Thanks for the patch :)
2016-11-19 17:01:36 +01:00
cc6d193e1d
gsdx: Relax SSE/AVX constraint on 64 bits
...
The JIT will automatically select the best ISA (only AVX1 so far)
2016-11-19 17:01:29 +01:00
8fd46e96aa
gsdx sw JIT: dynamically select ISA for SetupPrim
2016-11-19 17:00:33 +01:00
574a2c774e
gsdx sw JIT: dynamically select between AVX1 and SSE code path (scanline)
2016-11-19 17:00:33 +01:00
6b78b8f9ce
gsdx sw JIT: dynamically select SSE41 at runtime even on SSE2 build (scanline)
...
It won't give the full SSE41 speed boost but it is better than nothing
2016-11-19 17:00:33 +01:00
322473c295
gsdx sw: add a code example for gather instruction
...
It will requires a generic (register naming) linear interpolation to use it properly
Gather instruction requires an extra mask register therefore all registers name will be shuffled
Perf wise, initial haswell implementation seems to be microcode emulated.
2016-11-19 17:00:33 +01:00
2e20693583
gsdx sw x64: restore read texel optimization
2016-11-19 17:00:33 +01:00
e728a14c19
gsdx sw: factorize color split in split16_2x8
2016-11-19 17:00:33 +01:00
d58e43edbf
gsdx linux: plug vtune as Windows
2016-11-19 17:00:32 +01:00
8abf242e2c
gsdx: small x64 printf warning fixes
2016-11-19 17:00:32 +01:00
051c5c4bf7
gsdx sw x64: small stack optimization on linux
...
mov with the stack pointer require less bytecode
2016-11-19 17:00:32 +01:00
141c9e9c86
gsdx sw x64: prefer faster 32 bits operation when possible
2016-11-19 17:00:32 +01:00
a281bda9a6
gsdx sw x64: port the scanline generator on AVX
...
Based on Gabest's work.
* Miss mipmap
Note: dithering info
It is a bit tricky as a2 on linux was rdx register which overlap with fzm (dh/dl)
It might require dedicated windows code
2016-11-19 17:00:32 +01:00
8e29e09943
gsdx sw x64: update setup prim generator x64 SSE&AVX
2016-11-19 17:00:32 +01:00
4a47224ac1
gsdx: define the linux x64 ABI
2016-11-19 17:00:32 +01:00
e31ce87bb3
gsdx: SW JIT debug helper
...
Allow to compare 32/64 bits (and all ISAs too)
Allow to breakpoint (int3)
Print selector info
Print size of buffer and start (disabled by default)
2016-11-19 17:00:32 +01:00
633f7a1db9
xbyak: add int3 instruction
...
Very useful to stop the JIT
2016-11-19 17:00:32 +01:00
43b4cfc215
gsdx: separate dump directory for 32/64 bits
2016-11-19 17:00:32 +01:00
82d12691e1
gsdx: properly check SSE support
...
1/ Check all "levels"
2/ requires AVX for 64 bits
2016-11-19 17:00:32 +01:00
49d5c4260f
gsdx state: post fix depth tracing
...
The main FindMinMax methods is perf critical so instead I created a separate function
to ensure the constness of the depth
Fix letter regression on Xenosaga3
2016-11-11 23:39:34 +01:00
1530effb29
gsdx: Catch bad_alloc exceptions
2016-11-09 22:41:02 +00:00
ae6f26f3ef
gsdx: Skip texture cache read if any dimension is 0
...
Fixes a crash at the PSX logo if either the DX9 or DX11 hardware
renderer is used.
2016-11-09 22:39:53 +00:00
31b32d3b11
gsdx linux: don't require to hit enter for Texture Offset text entry
...
close #1292
2016-11-08 20:28:42 +01:00
ed3052dea5
gsdx: remove various assertion
...
Useless or wrong assertion
2016-11-08 18:58:10 +01:00
Gregory Hainaut