libm: arm: Add arm specific floor() optimization
Add arm specific floor() implementation which avoids VMSR and VMRS
instructions.
Change-Id: Ibd4cd7147aa2f98c9b5bbaf74948843ea619dba4
diff --git a/libm/Android.mk b/libm/Android.mk
index dc6c704..ebc3c9f 100644
--- a/libm/Android.mk
+++ b/libm/Android.mk
@@ -130,7 +130,6 @@
upstream-freebsd/lib/msun/src/s_fdim.c \
upstream-freebsd/lib/msun/src/s_finite.c \
upstream-freebsd/lib/msun/src/s_finitef.c \
- upstream-freebsd/lib/msun/src/s_floor.c \
upstream-freebsd/lib/msun/src/s_floorf.c \
upstream-freebsd/lib/msun/src/s_fma.c \
upstream-freebsd/lib/msun/src/s_fmaf.c \
@@ -264,20 +263,39 @@
LOCAL_SRC_FILES_arm += \
arm/fenv.c \
+# s_floor.S requires neon instructions.
+ifdef TARGET_2ND_ARCH
+arch_variant := $(TARGET_2ND_ARCH_VARIANT)
+else
+arch_variant := $(TARGET_ARCH_VARIANT)
+endif
+
+# Use the C version on armv7-a since it doesn't support neon instructions.
+ifeq ($(arch_variant),armv7-a)
+LOCAL_SRC_FILES_arm += upstream-freebsd/lib/msun/src/s_floor.c
+else
+LOCAL_SRC_FILES_arm += arm/s_floor.S
+endif
+
LOCAL_SRC_FILES_arm64 += \
arm64/fenv.c \
+ upstream-freebsd/lib/msun/src/s_floor.c \
LOCAL_SRC_FILES_mips += \
mips/fenv.c \
+ upstream-freebsd/lib/msun/src/s_floor.c \
LOCAL_SRC_FILES_mips64 += \
mips/fenv.c \
+ upstream-freebsd/lib/msun/src/s_floor.c \
LOCAL_SRC_FILES_x86 += \
i387/fenv.c \
+ upstream-freebsd/lib/msun/src/s_floor.c \
LOCAL_SRC_FILES_x86_64 += \
amd64/fenv.c \
+ upstream-freebsd/lib/msun/src/s_floor.c \
LOCAL_C_INCLUDES_x86 += $(LOCAL_PATH)/i387
@@ -297,6 +315,9 @@
-Wno-unknown-pragmas \
-fvisibility=hidden \
+LOCAL_ASFLAGS := \
+ -Ibionic/libc \
+
# Workaround the GCC "(long)fn -> lfn" optimization bug which will result in
# self recursions for lrint, lrintf, and lrintl.
# BUG: 14225968
diff --git a/libm/arm/s_floor.S b/libm/arm/s_floor.S
new file mode 100644
index 0000000..4405358
--- /dev/null
+++ b/libm/arm/s_floor.S
@@ -0,0 +1,142 @@
+/*
+ * Copyright (c) 2013-2014, NVIDIA Corporation. All rights reserved.
+ * Johnny Qiu <joqiu@nvidia.com>
+ * Shu Zhang <chazhang@nvidia.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials provided
+ * with the distribution.
+ * * Neither the name of The Linux Foundation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED "AS IS" AND ANY EXPRESS OR IMPLIED
+ * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
+ * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
+ * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+ * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN
+ * IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <float.h>
+#include <private/bionic_asm.h>
+
+ENTRY(floor) /* x in r0, r1 */
+
+ and r3, r1, #0x80000000 /* sign(x) */
+ bic r1, r1, #0x80000000 /* x = abs(x) */
+
+ /* extract exp of x */
+ lsr r2, r1, #20
+ sub r2, r2, #0x3fc
+ subs r2, r2, #0x3 /* r2 <- exp */
+
+ /* |x| < 1.0? */
+ blt .Lx_lt_one
+
+ /* x < 0? */
+ cmp r3, #0
+ bne .Lclr_frac_neg
+
+ /* |x| <= 2^20? */
+ cmp r2, #20
+ ble .Lclr_frac_r1
+
+ /* |x| < 2^52? */
+ cmp r2, #52
+ blt .Lclr_frac_r0
+
+ /* return x */
+ bx lr
+
+.Lclr_frac_r1:
+ rsb r2, r2, #20
+ lsr r1, r1, r2
+ lsl r1, r1, r2
+ mov r0, #0
+ bx lr
+
+.Lclr_frac_r0:
+ rsb r2, r2, #52
+ lsr r0, r0, r2
+ lsl r0, r0, r2
+ bx lr
+
+.Lclr_frac_neg:
+ /* |x| <= 2^20? */
+ cmp r2, #20
+ ble .Lclr_frac_r1_neg
+
+ /* |x| < 2^52? */
+ cmp r2, #52
+ blt .Lclr_frac_r0_neg
+
+ /* return x */
+ orr r1, r1, #0x80000000
+ bx lr
+
+.Lclr_frac_r1_neg:
+ rsb r2, r2, #20
+ mov r3, #1
+ lsl r3, r3, r2
+ sub r3, r3, #1
+ and r3, r1, r3
+ orr r3, r3, r0
+ lsr r1, r1, r2
+ lsl r1, r1, r2
+ mov r0, #0
+ b .Lreturn_x_neg
+
+.Lclr_frac_r0_neg:
+ rsb r2, r2, #52
+ mov r3, #1
+ lsl r3, r3, r2
+ sub r3, r3, #1
+ and r3, r0, r3
+ lsr r0, r0, r2
+ lsl r0, r0, r2
+ b .Lreturn_x_neg
+
+.Lx_lt_one:
+ /* x == +-0? */
+ cmp r0, #0
+ cmpeq r1, #0
+ orreq r1, r1, r3
+ bxeq lr
+
+ /* (x > 0) ? 0 : -1 */
+ mov r1, #0x00100000
+ mov r0, #0
+ cmp r3, #0
+ movne r1, #0xc0000000
+ sub r1, r1, #0x00100000
+ bx lr
+
+.Lreturn_x_neg:
+ cmp r3, #0
+ orr r1, r1, #0x80000000
+ bxeq lr
+
+ vmov d16, r0, r1
+ vmov.f64 d18, #1.0
+ vsub.f64 d16, d16, d18
+ vmov r0, r1, d16
+ bx lr
+
+END(floor)
+
+#if LDBL_MANT_DIG == 53
+ .weak floorl
+ .equ floorl,floor
+#endif