Visual Studio 2022 17.10 brings new optimizations, intrinsics, features, and improvements to the MSVC backend. Check out the highlights below:

• Performance improvements and additional functionality for all architectures:
  • Improved SLP vectorizer. Now it runs on more functions, recognizes more operations, and may choose to use larger vector widths if allowed by the appropriate /arch flag. Add the /Qvec-report:1 switch and look for info C5003: block vectorized for more SLP vectorization.
  • Support __declspec(guard(overflow)) with signed operations for add/mul.
  • Improvements in Variable Visibility when Debugging.
  • Enabled code coverage of managed C++ binaries targeting .NET Core.
  • Increase debug information for symbol removal during /LTCG.
  • Significant improvements to ASAN code generation, fixing codegen bugs, and reducing false negatives. ASAN now emits more checks in the presence of inlining.
  • Made security improvements to the PDB handling code.
• New functionality and performance improvements on ARM64:
  • The new /feature (ARM64) flag allows you to enable one or more optional Arm A-Profile architecture features for a specified ARM64 extension. You can also enable ARM64 features by specifying them in the updated /arch ARM64 option. For example, if you’d like to enable feature FEAT_LSE (Large System Extensions), you can use both /feature:lse and /arch:armv8.0 or only use /arch:armv8.0+lse. Note that only certain ARM64 features are currently supported by MSVC. Along with adding the option to enable Arm A-Profile architecture features, the /arch ARM64 option now has the additional arguments for Armv8-A architecture extension version armv8.9 and Armv9-A architecture extensions armv9.0, armv9.1, armv9.2, armv9.3, and armv9.4. For more information about both options, please refer to our documentation for /feature (ARM64) and /arch (ARM64).
  • Added Arm64 intrinsics, and variants, for exclusive load/store and load/store for unprivileged registers, ldtr and sttr. Thanks to our friends at ARM.
  • Optimized Arm64 CRT functions to run faster and take less space
    • strstr/wcschr/wcsrchr/wcsstr
    • strchr, strrchr, and memchr, thanks to our friends at ARM.
  • Speed up intrinsic expansion by reducing intermediate steps
  • Support SVE2 instructions in the assembler and disassembler, thanks to our friends at ARM.
    • bit permutation
    • element move and broadcasting
    • constructive multiplication
    • polynomial and complex integer arithmetic/dot product/matrix multiplication/comparison
    • uniform/widening/narrowing/unary DSP operations
    • cryptography support
    • extended table lookup
    • non-temporal gather/scatter
    • floating-point extra conversion and floating-point widening multiply-accumulate
• Improvements to ARM64EC:
  • No longer generate mismatched thunks with different call signatures
  • Now considers x64 functions during long-branch optimizations /OPT:LBR
  • /map now contains Arm64 native symbols
• Performance improvements and additional functionality for x86 and x64:
  • Vectorizer optimized to run in additional scenarios, thanks to our friends at AMD.
  • Optimize FMA generation with /favor:ATOM, thanks to our friends at Intel.
  • New functionality on x64:
    • Support the vectorization of the multiplication of 8-bit types for AVX512, thanks to our friends at Intel.
    • Support AVX2 and AVX512 variable bit shift instructions, thanks to our friends at AMD.
    • Our friends at Intel added intrinsics for Intel Lion Cove to enable:
      • AVX-VNNI-INT16
        • VPDPW[SU,US,UU]D[,S]
      • TSE
        • PBNDKB
      • SHA512
        • VSHA512MSG1, VSHA512MSG2, VSHA512RNDS2
      • SM3
        • VSM3MSG1, VMS3MSG2, VSM3RNDS2
      • SM4
        • VSM4KEY4, VSM4RNDS4

Do you want to experience the new improvements in the C++ backend? Please download the latest Visual Studio 2022 and give it a try! Any feedback is welcome. We can be reached via the comments below, Developer Community, X (@VisualC), or email at visualcpp@microsoft.com.

Stay tuned for more information on updates to the latest Visual Studio.