Apply correct, optimized mfpu compiler flag for ARMv7-A Cortex CPUs

For those ARMv7-A Cortex CPUs that can handle VFPv4 floating point, We can set "-mfpu=neon-vfpv4" instead of generic "-mfpu=neon" to gain extra performance improvements. References : - GCC : https://gcc.gnu.org/onlinedocs/gcc/ARM-Options.html - Cortex A15 : http://www.arm.com/products/processors/cortex-a/cortex-a15.php - Cortex A9 : http://www.arm.com/products/processors/cortex-a/cortex-a9.php - Cortex A8 : http://www.arm.com/products/processors/cortex-a/cortex-a8.php - Cortex A7 : http://www.arm.com/products/processors/cortex-a/cortex-a7.php - Cortex A5 : http://www.arm.com/products/processors/cortex-a/cortex-a5.php Currently prebuilt clang doesn't support neon-vfpv4 flag so fall back to neon Also added support for cortex-a5 arch Change-Id: I91893789ed8edabf3767e1782e494b81158332bb Signed-off-by: Park Ju Hyung <[email protected]> Signed-off-by: Maxime Poulain <[email protected]>
ubports · Aug 16, 2015 · 8dd9ea1 · 8dd9ea1
1 parent ba7e53f
commit 8dd9ea1
Show file tree

Hide file tree

Showing 2 changed files with 13 additions and 9 deletions.
diff --git a/core/clang/arm.mk b/core/clang/arm.mk
@@ -28,5 +28,6 @@ define subst-clang-incompatible-arm-flags
   $(subst -march=armv5te,-march=armv5t,\
   $(subst -march=armv5e,-march=armv5,\
   $(subst -mcpu=cortex-a15,-march=armv7-a,\
-  $(1))))
+  $(subst -mfpu=neon-vfpv4,-mfpu=neon,\
+  $(1)))))
 endef
diff --git a/core/combo/arch/arm/armv7-a-neon.mk b/core/combo/arch/arm/armv7-a-neon.mk
@@ -6,27 +6,30 @@ ARCH_ARM_HAVE_VFP               := true
 ARCH_ARM_HAVE_VFP_D32           := true
 ARCH_ARM_HAVE_NEON              := true
 
-ifneq (,$(filter cortex-a15 krait denver,$(TARGET_$(combo_2nd_arch_prefix)CPU_VARIANT)))
-	arch_variant_cflags := -mcpu=cortex-a15
+ifneq (,$(filter cortex-a15 denver krait,$(TARGET_$(combo_2nd_arch_prefix)CPU_VARIANT)))
+	arch_variant_cflags := -mcpu=cortex-a15 -mfpu=neon-vfpv4
 else
 ifeq ($(strip $(TARGET_$(combo_2nd_arch_prefix)CPU_VARIANT)),cortex-a9)
-	arch_variant_cflags := -mcpu=cortex-a9
+	arch_variant_cflags := -mcpu=cortex-a9 -mfpu=neon
 else
 ifneq (,$(filter cortex-a8 scorpion,$(TARGET_$(combo_2nd_arch_prefix)CPU_VARIANT)))
-	arch_variant_cflags := -mcpu=cortex-a8
+	arch_variant_cflags := -mcpu=cortex-a8 -mfpu=neon
 else
 ifeq ($(strip $(TARGET_$(combo_2nd_arch_prefix)CPU_VARIANT)),cortex-a7)
-	arch_variant_cflags := -mcpu=cortex-a7
+	arch_variant_cflags := -mcpu=cortex-a7 -mfpu=neon-vfpv4
 else
-	arch_variant_cflags := -march=armv7-a
+ifeq ($(strip $(TARGET_$(combo_2nd_arch_prefix)CPU_VARIANT)),cortex-a5)
+	arch_variant_cflags := -mcpu=cortex-a7 -mfpu=neon-vfpv4
+else
+	arch_variant_cflags := -march=armv7-a -mfpu=neon
+endif
 endif
 endif
 endif
 endif
 
 arch_variant_cflags += \
-    -mfloat-abi=softfp \
-    -mfpu=neon
+    -mfloat-abi=softfp
 
 arch_variant_ldflags := \
 	-Wl,--fix-cortex-a8