Skip to content

build: raise x86-64 baseline from SSE2 to SSE4.2 #4

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
bruno-dasilva wants to merge 1 commit into
base: master
Choose a base branch
from
Open

build: raise x86-64 baseline from SSE2 to SSE4.2 #4

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 14 additions & 3 deletions CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -486,9 +486,12 @@ if (CMAKE_COMPILER_IS_GNUCXX)
# worth to test if sync
#set(FALLBACK_SSE_FLAGS "${FALLBACK_SSE_FLAGS} -mpopcnt -mlzcnt -mabm")

# disable rest
# SSE4.2 baseline. Requires Nehalem (2008) / Bulldozer (2011).
set(FALLBACK_SSE_FLAGS "${FALLBACK_SSE_FLAGS} -msse3 -mssse3 -msse4 -msse4.1 -msse4.2")

# disable rest — AVX/FMA remain banned: FMA contraction changes FP bit patterns and desyncs.
#set(FALLBACK_SSE_FLAGS "${FALLBACK_SSE_FLAGS} -mno-3dnow") tests showed it might sync
set(FALLBACK_SSE_FLAGS "${FALLBACK_SSE_FLAGS} -mno-sse3 -mno-ssse3 -mno-sse4.1 -mno-sse4.2 -mno-sse4 -mno-sse4a")
set(FALLBACK_SSE_FLAGS "${FALLBACK_SSE_FLAGS} -mno-sse4a")
set(FALLBACK_SSE_FLAGS "${FALLBACK_SSE_FLAGS} -mno-avx -mno-fma -mno-fma4 -mno-xop -mno-lwp")
set(FALLBACK_SSE_FLAGS "${FALLBACK_SSE_FLAGS} -mno-avx2")

Expand All @@ -505,6 +508,14 @@ if (CMAKE_COMPILER_IS_GNUCXX)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${SSE_FLAGS} -mfpmath=sse") #SSE_FLAGS are now user input only
endif ()
elseif (MSVC)
# MSVC stays on /arch:SSE2 even though GCC/Clang baseline is SSE4.2:
# MSVC has no SSE4-only switch — next step up is /arch:AVX which would
# enable AVX/FMA autovectorization and violate the sync-parity ban on
# FMA contraction. SSE4 intrinsics are still available on MSVC under
# /arch:SSE2 (unlike GCC), so source code can opt in explicitly. The
# practical cost is that MSVC builds don't autovectorize to SSE4.
# Note: shipped Windows artifacts use MinGW GCC, not MSVC; this branch
# is for local MSVC developer configures only.
set(MSVC_CXX_FLAGS "/arch:SSE2") #default, but still set explicitly
# silence warnings
add_definitions(-D_SILENCE_STDEXT_HASH_DEPRECATION_WARNINGS)
Expand All @@ -515,7 +526,7 @@ elseif ("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang" OR "${CMAKE_CXX_COMPILER_ID}
else()
# Clang on x86_64: use SSE like GCC
if (NOT MARCH_FLAG OR MARCH_FLAG STREQUAL "" OR MARCH_FLAG STREQUAL "generic")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -msse -mfpmath=sse")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -msse -msse3 -mssse3 -msse4 -msse4.1 -msse4.2 -mfpmath=sse")
else()
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=${MARCH_FLAG} -mtune=${MARCH_FLAG}")
endif()
Expand Down
5 changes: 3 additions & 2 deletions rts/lib/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,8 +119,9 @@ if(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64|arm64|armv8")
# ARM64: enable native NEON implementation (well-tested in simdjson)
set(SIMDJSON_IMPLEMENTATION "arm64" CACHE STRING "forced by Recoil build env" FORCE)
else()
# x86: use fallback only (original behavior)
set(SIMDJSON_IMPLEMENTATION "fallback" CACHE STRING "forced by Recoil build env" FORCE)
# x86: SSE4.2 is the engine baseline (see CMakeLists.txt FALLBACK_SSE_FLAGS),
# so use simdjson's "westmere" implementation (SSE4.2 + CLMUL).
set(SIMDJSON_IMPLEMENTATION "westmere" CACHE STRING "forced by Recoil build env" FORCE)
endif()

set(SIMDJSON_ENABLE_THREADS OFF CACHE BOOL "forced by Recoil build env" FORCE)
Expand Down