Remove pushes from memsets (krait/cortex-a9).
On the path that only uses r0 in both the krait and cortex-a9
memset, remove the push and use r3 instead.
In addition, for cortex-a9, remove the artificial function since
it's not needed since dwarf unwinding is now supported on arm.
Change-Id: Icd39cfb6b8350f44368e022063cd97a6b60d46da
diff --git a/libc/arch-arm/cortex-a9/bionic/memset.S b/libc/arch-arm/cortex-a9/bionic/memset.S
index 48ba815..b39fcc4 100644
--- a/libc/arch-arm/cortex-a9/bionic/memset.S
+++ b/libc/arch-arm/cortex-a9/bionic/memset.S
@@ -69,12 +69,9 @@
ENTRY(memset)
// The neon memset only wins for less than 132.
cmp r2, #132
- bhi __memset_large_copy
+ bhi .L_memset_large_copy
- stmfd sp!, {r0}
- .cfi_def_cfa_offset 4
- .cfi_rel_offset r0, 0
-
+ mov r3, r0
vdup.8 q0, r1
/* make sure we have at least 32 bytes to write */
@@ -84,7 +81,7 @@
1: /* The main loop writes 32 bytes at a time */
subs r2, r2, #32
- vst1.8 {d0 - d3}, [r0]!
+ vst1.8 {d0 - d3}, [r3]!
bhs 1b
2: /* less than 32 left */
@@ -93,22 +90,20 @@
beq 3f
// writes 16 bytes, 128-bits aligned
- vst1.8 {d0, d1}, [r0]!
+ vst1.8 {d0, d1}, [r3]!
3: /* write up to 15-bytes (count in r2) */
movs ip, r2, lsl #29
bcc 1f
- vst1.8 {d0}, [r0]!
+ vst1.8 {d0}, [r3]!
1: bge 2f
- vst1.32 {d0[0]}, [r0]!
+ vst1.32 {d0[0]}, [r3]!
2: movs ip, r2, lsl #31
- strbmi r1, [r0], #1
- strbcs r1, [r0], #1
- strbcs r1, [r0], #1
- ldmfd sp!, {r0}
+ strbmi r1, [r3], #1
+ strbcs r1, [r3], #1
+ strbcs r1, [r3], #1
bx lr
-END(memset)
-ENTRY_PRIVATE(__memset_large_copy)
+.L_memset_large_copy:
/* compute the offset to align the destination
* offset = (4-(src&3))&3 = -src & 3
*/
@@ -180,7 +175,7 @@
movs r2, r2, lsl #2
strbcs r1, [r0]
ldmfd sp!, {r0, r4-r7, pc}
-END(__memset_large_copy)
+END(memset)
.data
error_string:
diff --git a/libc/arch-arm/krait/bionic/memset.S b/libc/arch-arm/krait/bionic/memset.S
index a4fbe17..ae05965 100644
--- a/libc/arch-arm/krait/bionic/memset.S
+++ b/libc/arch-arm/krait/bionic/memset.S
@@ -69,10 +69,7 @@
/* memset() returns its first argument. */
ENTRY(memset)
- stmfd sp!, {r0}
- .cfi_def_cfa_offset 4
- .cfi_rel_offset r0, 0
-
+ mov r3, r0
vdup.8 q0, r1
/* make sure we have at least 32 bytes to write */
@@ -82,7 +79,7 @@
1: /* The main loop writes 32 bytes at a time */
subs r2, r2, #32
- vst1.8 {d0 - d3}, [r0]!
+ vst1.8 {d0 - d3}, [r3]!
bhs 1b
2: /* less than 32 left */
@@ -91,18 +88,17 @@
beq 3f
// writes 16 bytes, 128-bits aligned
- vst1.8 {d0, d1}, [r0]!
+ vst1.8 {d0, d1}, [r3]!
3: /* write up to 15-bytes (count in r2) */
movs ip, r2, lsl #29
bcc 1f
- vst1.8 {d0}, [r0]!
+ vst1.8 {d0}, [r3]!
1: bge 2f
- vst1.32 {d0[0]}, [r0]!
+ vst1.32 {d0[0]}, [r3]!
2: movs ip, r2, lsl #31
- strbmi r1, [r0], #1
- strbcs r1, [r0], #1
- strbcs r1, [r0], #1
- ldmfd sp!, {r0}
+ strbmi r1, [r3], #1
+ strbcs r1, [r3], #1
+ strbcs r1, [r3], #1
bx lr
END(memset)