diff --git a/Makefile b/Makefile
--- a/Makefile
+++ b/Makefile
@@ -469,7 +469,7 @@ endif
 export KERNELDOC
 
 KBUILD_USERHOSTCFLAGS := -Wall -Wmissing-prototypes -Wstrict-prototypes \
-			 -O2 -fomit-frame-pointer -std=gnu11
+			 -O3 -fomit-frame-pointer -std=gnu11
 KBUILD_USERCFLAGS  := $(KBUILD_USERHOSTCFLAGS) $(USERCFLAGS)
 KBUILD_USERLDFLAGS := $(USERLDFLAGS)

@@ -513,7 +513,7 @@ endif
 
 KBUILD_HOSTCFLAGS   := $(KBUILD_USERHOSTCFLAGS) $(HOST_LFS_CFLAGS) \
 		       $(HOSTCFLAGS) -I $(srctree)/scripts/include
-KBUILD_HOSTCXXFLAGS := -Wall -O2 $(HOST_LFS_CFLAGS) $(HOSTCXXFLAGS) \
+KBUILD_HOSTCXXFLAGS := -Wall -O3 $(HOST_LFS_CFLAGS) $(HOSTCXXFLAGS) \
 		       -I $(srctree)/scripts/include
 KBUILD_HOSTRUSTFLAGS := $(rust_common_flags) -O -Cstrip=debuginfo \
 			-Zallow-features=
@@ -931,24 +931,33 @@ KBUILD_CFLAGS	+= -fno-delete-null-pointer-checks
 
 ifdef CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE
 KBUILD_CFLAGS += -O2
+else ifdef CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE_O3
+KBUILD_CFLAGS += -O3
+KBUILD_RUSTFLAGS += -Copt-level=3
 # LLVM's Polly for high-level loop and data-locality optimizer based on integer
 # polyhedra to analyze and optimize the memory access pattern.
 KBUILD_CFLAGS += $(call cc-option,-mllvm=-polly -mllvm=-polly-ast-use-context \
 		 -mllvm=-polly-invariant-load-hoisting -mllvm=-polly-loopfusion-greedy \
 		 -mllvm=-polly-run-inliner -mllvm=-polly-vectorizer=stripmine)
 ifdef CONFIG_LD_DEAD_CODE_DATA_ELIMINATION
 KBUILD_CFLAGS += $(call cc-option,-mllvm=-polly-run-dce)
 endif
 # Enable software pipelining (Swing Modulo Scheduling) for innermost loops.
 KBUILD_CFLAGS += $(call cc-option,-fmodulo-sched -fmodulo-sched-allow-regmoves) \
 		 $(call cc-option,-mllvm=-enable-pipeliner)
 KBUILD_RUSTFLAGS += -Copt-level=2
 else ifdef CONFIG_CC_OPTIMIZE_FOR_SIZE
 KBUILD_CFLAGS += -Os
 KBUILD_CFLAGS += $(call cc-option,-mllvm=-enable-pipeliner-opt-size)
 KBUILD_RUSTFLAGS += -Copt-level=s
 endif
 
+# Perform swing modulo scheduling immediately before the first scheduling pass.
+# This pass looks at innermost loops and reorders their instructions by
+# overlapping different iterations.
+KBUILD_CFLAGS += $(call cc-option,-fmodulo-sched -fmodulo-sched-allow-regmoves -fivopts)
+KBUILD_CFLAGS += $(call cc-option,-mllvm -enable-pipeliner)
+
 # Always set `debug-assertions` and `overflow-checks` because their default
 # depends on `opt-level` and `debug-assertions`, respectively.
 KBUILD_RUSTFLAGS += -Cdebug-assertions=$(if $(CONFIG_RUST_DEBUG_ASSERTIONS),y,n)
@@ -1180,11 +1205,6 @@ KBUILD_CFLAGS	+= -fno-strict-overflow
 # Make sure -fstack-check isn't enabled (like gentoo apparently did)
 KBUILD_CFLAGS  += -fno-stack-check
 
-# conserve stack if available
-ifdef CONFIG_CC_IS_GCC
-KBUILD_CFLAGS   += -fconserve-stack
-endif
-
 # Ensure compilers do not transform certain loops into calls to wcslen()
 KBUILD_CFLAGS += -fno-builtin-wcslen
 
diff --git a/arch/x86/Makefile b/arch/x86/Makefile
index 598f17810..a2ef052ce 100644
--- a/arch/x86/Makefile
+++ b/arch/x86/Makefile
@@ -79,7 +79,7 @@ export BITS
 #
 #    https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53383
 #
-KBUILD_CFLAGS += -mno-sse -mno-mmx -mno-sse2 -mno-3dnow -mno-avx -mno-avx2 -mno-sse4a
+KBUILD_CFLAGS += -mno-sse -mno-mmx -mno-sse2 -mno-3dnow -mno-avx -mno-sse4a -mno-avx2 -fno-tree-vectorize
 KBUILD_RUSTFLAGS += --target=$(objtree)/scripts/target.json
 KBUILD_RUSTFLAGS += -Ctarget-feature=-sse,-sse2,-sse3,-ssse3,-sse4.1,-sse4.2,-avx,-avx2
 
diff --git a/init/Kconfig b/init/Kconfig
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1638,6 +1646,12 @@ config CC_OPTIMIZE_FOR_PERFORMANCE
 	  with the "-O2" compiler flag for best performance and most
 	  helpful compile-time warnings.
 
+config CC_OPTIMIZE_FOR_PERFORMANCE_O3
+	bool "Optimize more for performance (-O3)"
+	help
+	  Choosing this option will pass "-O3" to your compiler to optimize
+	  the kernel yet more for performance.
+
 config CC_OPTIMIZE_FOR_SIZE
 	bool "Optimize for size (-Os)"
 	help