summaryrefslogtreecommitdiff
path: root/tools/arch/x86/lib/memcpy_64.S
diff options
context:
space:
mode:
authorTakashi Iwai <tiwai@suse.de>2023-06-26 16:38:02 +0300
committerTakashi Iwai <tiwai@suse.de>2023-06-26 16:38:02 +0300
commitd6048fdc870240e5020343f8af0c825829c232bd (patch)
treed83ca76eaac5f8bf1c4c16f003aad64baa8fae62 /tools/arch/x86/lib/memcpy_64.S
parenta15b51375684c2bfa6017bb185139477e7a3b96c (diff)
parent2d0cad0473bd1ffbc5842be0b9f2546265acb011 (diff)
downloadlinux-d6048fdc870240e5020343f8af0c825829c232bd.tar.xz
Merge tag 'asoc-v6.5' of https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound into for-linus
ASoC: Updates for v6.5 A fairly quiet release from a core and framework point of view, but a very big one from the point of view of new drivers: - More refectoring from Morimoto-san, this time mainly around DAI links and how we control the ordering of trigger() callbacks. - Convert a lot of drivers to use maple tree based caches. - Lots of work on the x86 driver stack. - Compressed audio support for Qualcomm. - Support for AMD SoundWire, Analog Devices SSM3515, Google Chameleon, Ingenic X1000, Intel systems with various CODECs, Longsoon platforms, Maxim MAX98388, Mediatek MT8188, Nuvoton NAU8825C, NXP platforms with NAU8822, Qualcomm WSA884x, StarFive JH7110, Texas Instruments TAS2781.
Diffstat (limited to 'tools/arch/x86/lib/memcpy_64.S')
-rw-r--r--tools/arch/x86/lib/memcpy_64.S34
1 files changed, 10 insertions, 24 deletions
diff --git a/tools/arch/x86/lib/memcpy_64.S b/tools/arch/x86/lib/memcpy_64.S
index a91ac666f758..d055b82d22cc 100644
--- a/tools/arch/x86/lib/memcpy_64.S
+++ b/tools/arch/x86/lib/memcpy_64.S
@@ -10,13 +10,6 @@
.section .noinstr.text, "ax"
/*
- * We build a jump to memcpy_orig by default which gets NOPped out on
- * the majority of x86 CPUs which set REP_GOOD. In addition, CPUs which
- * have the enhanced REP MOVSB/STOSB feature (ERMS), change those NOPs
- * to a jmp to memcpy_erms which does the REP; MOVSB mem copy.
- */
-
-/*
* memcpy - Copy a memory block.
*
* Input:
@@ -26,17 +19,21 @@
*
* Output:
* rax original destination
+ *
+ * The FSRM alternative should be done inline (avoiding the call and
+ * the disgusting return handling), but that would require some help
+ * from the compiler for better calling conventions.
+ *
+ * The 'rep movsb' itself is small enough to replace the call, but the
+ * two register moves blow up the code. And one of them is "needed"
+ * only for the return value that is the same as the source input,
+ * which the compiler could/should do much better anyway.
*/
SYM_TYPED_FUNC_START(__memcpy)
- ALTERNATIVE_2 "jmp memcpy_orig", "", X86_FEATURE_REP_GOOD, \
- "jmp memcpy_erms", X86_FEATURE_ERMS
+ ALTERNATIVE "jmp memcpy_orig", "", X86_FEATURE_FSRM
movq %rdi, %rax
movq %rdx, %rcx
- shrq $3, %rcx
- andl $7, %edx
- rep movsq
- movl %edx, %ecx
rep movsb
RET
SYM_FUNC_END(__memcpy)
@@ -45,17 +42,6 @@ EXPORT_SYMBOL(__memcpy)
SYM_FUNC_ALIAS(memcpy, __memcpy)
EXPORT_SYMBOL(memcpy)
-/*
- * memcpy_erms() - enhanced fast string memcpy. This is faster and
- * simpler than memcpy. Use memcpy_erms when possible.
- */
-SYM_FUNC_START_LOCAL(memcpy_erms)
- movq %rdi, %rax
- movq %rdx, %rcx
- rep movsb
- RET
-SYM_FUNC_END(memcpy_erms)
-
SYM_FUNC_START_LOCAL(memcpy_orig)
movq %rdi, %rax