Read the blog post at FEX-Emu's Site!
Time certainly goes by quicker than you'd expect. We even managed to skip last months release because we were busy doing other things. Let's take that
as an example and fly through the changes that we did over these past two months!
Optimizations and fixes for 256-bit SVE2 hardware
While this hardware doesn't exist yet, we know it is an inevitability that it will at some point. While we switched gears a couple years ago to
implement AVX using 128-bit operations, we never removed this code. In-fact we've been spending a bunch of effort on it fixing bugs and optimizing it
so that once hardware ships it won't be broken. We have now validated extensively that all AVX instructions now zero extend their results as expected
and optimized a bunch of the instructions so they generate faster code. We think it's now at a point that in the common case the code will be faster
than our 128-bit emulation, but there is definitely still some more work todo. SVE2.1 provides a significant improvement to how shuffles operate and
we haven't gotten those implemented yet. Because there aren't any 256-bit SVE2 hardware on the market, we might even require SVE2.1 or SVE2.2 for this
class of hardware. We'll of course still maintain our 128-bit path for lower-end hardware of course.
Various JIT fixes/changes
Once again, too many to go through individually, let's throw it in a list. Most of these are just bug fixes but there are a handful of optimizations
as well.
- Fixed corruption with back to back PMOVMSKB instructions and full SMC detection
- Fixes Vivado in this situation
- Handle incorrect LOCK prefixed instructions correctly
- Allow larger CPU context state
- Fixes compilation error with musl
- Fixes vsyscall page tracking
- Was accidentally a NOEXEC page
- Only use DC ZVA for Ampere when clearing AVX state
- It's the same or slower on other hardware
- Fix zero extension in VCVTPS2PH
- Fixes CRC32 with high 8-bit registers
- Fixes 64-bit LODS instruction with address size override
- Fix Mafia 3 in Arm64ec Proton
- Fix Bioshock and other 32-bit games that disable DEP on Arm64ec Proton
- Requires bleeding-edge Proton or WINE
- Optimize x87 FYL2X/FPREM/FPREM1
- Fix incorrect RSP update on 16-bit LEAVE instruction
- Handle some new CPL0 only instructions correctly
- Fix JIT allocations ending up in 32-bit VA space
- Fixes some 32-bit games under Proton
Implement support for a unixlib under Proton/WINE
For a long time FEX's WINE support works by building all of FEX as a DLL file which WINE loads at runtime. This has worked for us pretty well but
constrained some of our design decisions at times, causing us to take less optimal or hacky paths. We also have some upcoming changes that would be
dramatically harder to implement without a unixlib so we decided to start using one. Now FEX for WINE ships two DLL files, and two unixlib SO files.
Right now we are duplicating some code between the unixlib and the FEX WINE DLL file to make sure everything works correctly but within a few months
we are likely to remove the duplicated code and rely entirely on the unixlib. Currently all the functionality provided is optional, but we would
recommend using it because we may eventually not have it be optional.
Partial support for CUDA thunking
Some people were asking for CUDA thunking on the DGX Spark and we decided to implement partial support for it. It's not 100% coverage but depending
on the workload it can potentially work. We know today that if you try to execute applications with the static CUDA runtime linked, it won't work
well. Other than that give it a try and you might be surprised at what works.
Raw Changes
FEX Release FEX-2607
-
AVX
- Remove unnecessary moves from VPSHUF{D, HW, LW} (a4f89b7)
- Reduce codegen for 256-bit VMOVMSKPD/VMOVMSKPS (394a6f2)
- Handle two field insertions in VPERMQ (83a989c)
- Slightly trim codegen for VPSADW 256-bit case (9740f48)
- Handle trivial UZP/ZIP operations in VPERMQ (c85e426)
- Handle trivial cases better for VDPPS (3470dd1)
- Handle easily broadcastable permutations in VPERMQ (7ae55d7)
- Skip identity insertions in VPERMQ (d5be15c)
- Handle transpose cases in VPERMQ (ec1b24d)
- Remove unnecessary dup if sources are the same in SHUFOpImpl (d93997c)
- Shave some moves off 256-bit VPALIGNR (ab4e0f6)
- Reduce moves in 256-bit VPSHUFB (fda023e)
- Wire up helper to VPERMILPD (5881266)
- Wire up lane helper for VPERMILPS imm variant (b846cb6)
- Wire up lane helper for VSHUFPD/VSHUFPS (62ecdd6)
- Wire up lane helper for VPSHUFD/VPSHUFLW/VPSHUFHW (9f195ff)
- Reduce inserts in VBLENDPS/VPBLENDD/VPBLENDW (4eb5694)
- Remove unnecessary moves from PINSRX ops (5ad06b0)
-
Allocator
- Fix and optimize VA range detection (af4da43)
-
Arm64
- Fix byte-size handling in unaligned STLXR emulation (f66368b)
-
Arm64Emitter
- Tidy up load/stores in Push/PopCalleeSavedRegisters (500d237)
-
Build
- Enable ccache sloppiness for time macros (cae5da5)
-
CodeCache
- Support targeting WOW64/ARM64EC in FEXOfflineCompiler (08ed4fb)
- Resolve comment #5441 (comment) (df73e84)
- Implement lazy code loading (bb0d142)
-
Config
- Use more sensible default for portable config location (33f3b86)
-
Core
- Start a new block for the next op in full SMC check (b23fa30)
-
FEXBash
- Drop implicit -c and add colored PS1 (9be7d6d)
-
FEXCore
-
FEXGetConfig
-
FEXOfflineCompiler
-
FEXServerClient
- Workaround sun_path 108 byte limit (1db45e2)
-
Format
- Fix missed clang-format (5fd917e)
-
Frontend
- Fix vsyscall page tracking. (e02953d)
-
HostFeatures
-
IR
- Add constant for swapping midsections of 256-bit vectors around (110313e)
-
ImageTracker
- Support using image IDs as an extended volatile metadata key (ed724a6)
-
IntrusiveIRList
- Amend signature for PostRA() (b87ff1e)
-
JIT
-
LibraryForwarding
-
LinuxSyscalls
- add missing thread header (a1071ec)
-
OpcodeDispatcher
- Fix missing zeroing for vcvtps2ph (9f2e982)
- Fixes CRC32 with high 8-bit register (43bd243)
- Fix incorrect comment for BTOp. ZF must be preserved. (e19aa97)
- Fix typo in comment (7dc2dc8)
- Sanitize selectors for VBLENDPD/VPBLENDD (681c5e8)
- Eliminate redundant moves in VMOVHPOp (dd44bc8)
- Make use of Bind consistently (110c7cb)
- Move a few stray literal accesses to Literal() (09aa5ab)
- Fixes 64-bit LODs with address size override (ed6a178)
-
Passes
- Trim unnecessary forward declarations (280568d)
-
Proton
- Fixes Mafia 3 (34344e5)
-
Vector
-
VectorOps
-
WOW64
- Support disabling DEP (f5efdac)
-
Windows
-
Misc
- Re-optimize FYL2X for reduced precision x87 path (3d593ce)
- instcountci/VEX_map3: Add missing third param to VPBLENDD (d138c85)
- instcountci/VEX_map1: Remove obsolete comments (7e2d3b0)
- Do not exec FEX if it is a folder in FEXBash (848c4b2)
- Fix UAF of PS1 (4f995cb)
- [SVE256] Add fast paths for trivial VSHUFPD flags (0b0000, and 0b1111) (1619374)
- [SVE256] Remove unnecessary move in VCVTPS2PD (c13064e)
- [SVE256] Remove heavy handed moves from scalar compares (37e32fb)
- [SVE256] More comprehensively test SSE insertions for scalar comparisons (27acbba)
- [SVE256] Add more SSE scalar variant unit tests (6f33d2b)
- [SVE256] Handle SSE insertions for PCMPESTRM/PCMPISTRM ops (c5880e7)
- [SVE256] Handle SSE insertions for MOVQ2DQ (9d0c05d)
- [SVE256] Handle SSE insertions for EXTRQ/INSERTQ (f6a68cb)
- [SVE256] Handle SSE insertions for CVTPI2PD (462c785)
- Fix incorrect RSP update for 16bit leave (f547703)
- [SVE256] Handle SSE insertions for PMADDWD (ee4794c)
- [SVE256] Handle SSE insertions for CMPSD/CMPSS (3a23bb4)
- [SVE256] Handle SSE insertions for MOVSHDUP/MOVSLDUP (46ffb25)
- [SVE256] Handle SSE insertion for aligned/unaligned loads and non-temporal loads (069a402)
- [SVE256] Handle SSE insertions for MOVH(PD, PD, LPS) and MOVL(PD, PS, HPS) (3929d25)
- [SVE256] Handle SSE insertions for XOR special case (f98ac7f)
- [SVE256] Handle SSE insertions for INSERTPS, PSIGN, PINSR, and shuffles (2309910)
- [SVE256] Handle SSE insertions for shifts (adad3c2)
- (470aeab)
- [SVE256] Handle SSE insertions for more misc ops (99662b7)
- [SVE256] Handle SSE insertions for PHMINPOSUW, DPPD, and DPPS (1a606de)
- [SVE256] Handle SSE insertions for blends (223e0f4)
- (0b1f336)
- [SVE256] Handle more SSE insertions for some one-off instructions (d9ea665)
- [SVE256] Vector: Handle SSE insertion properly for various ALU operations (9d5494d)
- Drop unused CMakeSettings.json (97f1f47)
- set sysroot to X86_DEV_ROOTFS for guest toolchain (53c269e)
- New CPL0 instructions from #5510 but with unittests (1240a00)
- code-format-helper: More dependabot changes (0b871bf)
- Cherry-pick #5508 with instcountci changes (268081e)
- Windows additions for code caching (27324de)
- Library Forwarding: Various build system improvements (b4fe65f)
- JIT-inline FPREM/FPREM1 for reduced precision x87 path (0d72890)
-
arm64ec
-
instcountci
-
meta
- Add CONTRIBUTING.md (83f325d)
-
unittests