[RFC 09/15] temp: remove various library optimisations
Ben Dooks
ben.dooks at codethink.co.uk
Fri Dec 20 07:57:55 PST 2024
These either need fixing or checking for big endian.
- memove is deifentyl not working
- ignore memset and memcpy optimisation for now
- uaccess code needs fixing
---
arch/riscv/lib/memcpy.S | 22 +++++++++++++++++++++-
arch/riscv/lib/memmove.S | 2 +-
arch/riscv/lib/memset.S | 1 +
arch/riscv/lib/strlen.S | 2 +-
arch/riscv/lib/uaccess.S | 7 +++++--
5 files changed, 29 insertions(+), 5 deletions(-)
diff --git a/arch/riscv/lib/memcpy.S b/arch/riscv/lib/memcpy.S
index 44e009ec5fef..b51380f06204 100644
--- a/arch/riscv/lib/memcpy.S
+++ b/arch/riscv/lib/memcpy.S
@@ -7,12 +7,15 @@
#include <asm/asm.h>
/* void *memcpy(void *, const void *, size_t) */
-SYM_FUNC_START(__memcpy)
+SYM_FUNC_START(__memcpy1)
move t6, a0 /* Preserve return value */
/* Defer to byte-oriented copy for small sizes */
sltiu a3, a2, 128
+ j 4f /* for now just always use bytes */
+
bnez a3, 4f
+
/* Use word-oriented copy only if low-order bits match */
andi a3, t6, SZREG-1
andi a4, a1, SZREG-1
@@ -87,6 +90,7 @@ SYM_FUNC_START(__memcpy)
or a5, a5, a3
andi a5, a5, 3
bnez a5, 5f
+ j 5f /* skip word */
7:
lw a4, 0(a1)
addi a1, a1, 4
@@ -104,6 +108,22 @@ SYM_FUNC_START(__memcpy)
bltu a1, a3, 5b
6:
ret
+
+SYM_FUNC_START(__memcpy)
+ move t6, a0 /* Preserve return value */
+ beqz a2, 6f
+ add a3, a1, a2
+
+5:
+ lb a4, 0(a1)
+ addi a1, a1, 1
+ sb a4, 0(t6)
+ addi t6, t6, 1
+ bltu a1, a3, 5b
+6:
+ ret
+
+
SYM_FUNC_END(__memcpy)
SYM_FUNC_ALIAS_WEAK(memcpy, __memcpy)
SYM_FUNC_ALIAS(__pi_memcpy, __memcpy)
diff --git a/arch/riscv/lib/memmove.S b/arch/riscv/lib/memmove.S
index cb3e2e7ef0ba..c51475e4f3ce 100644
--- a/arch/riscv/lib/memmove.S
+++ b/arch/riscv/lib/memmove.S
@@ -60,7 +60,7 @@ SYM_FUNC_START(__memmove)
*/
andi t0, a2, -(2 * SZREG)
beqz t0, .Lbyte_copy
-
+ j .Lbyte_copy
/*
* Now solve for t5 and t6.
*/
diff --git a/arch/riscv/lib/memset.S b/arch/riscv/lib/memset.S
index da23b8347e2d..a3cd79cb33b4 100644
--- a/arch/riscv/lib/memset.S
+++ b/arch/riscv/lib/memset.S
@@ -14,6 +14,7 @@ SYM_FUNC_START(__memset)
/* Defer to byte-oriented fill for small sizes */
sltiu a3, a2, 16
bnez a3, 4f
+ j 4f /* disabel optimised for now */
/*
* Round to nearest XLEN-aligned address
diff --git a/arch/riscv/lib/strlen.S b/arch/riscv/lib/strlen.S
index 962983b73251..bea650fd24af 100644
--- a/arch/riscv/lib/strlen.S
+++ b/arch/riscv/lib/strlen.S
@@ -8,7 +8,7 @@
/* int strlen(const char *s) */
SYM_FUNC_START(strlen)
- ALTERNATIVE("nop", "j strlen_zbb", 0, RISCV_ISA_EXT_ZBB, CONFIG_RISCV_ISA_ZBB)
+ /*ALTERNATIVE("nop", "j strlen_zbb", 0, RISCV_ISA_EXT_ZBB, CONFIG_RISCV_ISA_ZBB)*/
/*
* Returns
diff --git a/arch/riscv/lib/uaccess.S b/arch/riscv/lib/uaccess.S
index 6a9f116bb545..3d7da86277bb 100644
--- a/arch/riscv/lib/uaccess.S
+++ b/arch/riscv/lib/uaccess.S
@@ -46,7 +46,8 @@ SYM_FUNC_START(fallback_scalar_usercopy)
*/
li a3, 9*SZREG-1 /* size must >= (word_copy stride + SZREG-1) */
bltu a2, a3, .Lbyte_copy_tail
-
+ j .Lbyte_copy_tail
+
/*
* Copy first bytes until dst is aligned to word boundary.
* a0 - start of dst
@@ -73,7 +74,9 @@ SYM_FUNC_START(fallback_scalar_usercopy)
*/
/* a1 - start of src */
andi a3, a1, SZREG-1
- bnez a3, .Lshift_copy
+ /* bnez a3, .Lshift_copy */
+ /* for now, ignore shift copy until fixed */
+ bnez a3, .Lbyte_copy_tail
.Lword_copy:
/*
--
2.37.2.352.g3c44437643
More information about the linux-riscv
mailing list