Computing sum introduces true data dependency. This patch removes some
true data depdendencies, hence increases instruction level parallelism.
This patch brings up to 50% csum performance gain on Loongson 3a.
One example about how this patch works is in CSUM_BIGCHUNK1:
// ** original ** vs ** patch applied **
ADDC(sum, t0) ADDC(t0, t1)
ADDC(sum, t1) ADDC(t2, t3)
ADDC(sum, t2) ADDC(sum, t0)
ADDC(sum, t3) ADDC(sum, t2)
In the original implementation, each ADDC(sum, ...) depends on the sum
value updated by previous ADDC(as source operand).
With this patch applied, the first two ADDC operations are independent,
hence can be executed simultaneously if possible.
Another example is in the "copy and sum calculating chunk":
// ** original ** vs ** patch applied **
STORE(t0, UNIT(0) ... STORE(t0, UNIT(0) ...
ADDC(sum, t0) ADDC(t0, t1)
STORE(t1, UNIT(1) ... STORE(t1, UNIT(1) ...
ADDC(sum, t1) ADDC(sum, t0)
STORE(t2, UNIT(2) ... STORE(t2, UNIT(2) ...
ADDC(sum, t2) ADDC(t2, t3)
STORE(t3, UNIT(3) ... STORE(t3, UNIT(3) ...
ADDC(sum, t3) ADDC(sum, t2)
With this patch applied, ADDC and the **next next** ADDC are independent.
Signed-off-by: chenj <chenj@lemote.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9608/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
MIPS R6 dropped the unaligned load and store instructions so
we need to re-write this part of the code for R6 to store
one byte at a time.
Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com>
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
MIPS R6 does not support the unaligned load and store instructions
so we add a special MIPS R6 case to copy one byte at a time if we
need to read/write to unaligned memory addresses.
Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com>
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
The following instructions have been removed from MIPS R6
ulw, ulh, swl, lwr, lwl, swr.
However, all of them are used in the MIPS specific checksum implementation.
As a result of which, we will use the generic checksum on MIPS R6
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
The toolchain defines exactly one of __MIPSEB__ and
__MIPSEL__. As a result, simplify the ifdefery a little bit.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/8522/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Having #ifdefs just to guard comments is not really helpful
so drop them. Moreover, the code wasn't really reached anyway
since there is a #ifndef CONFIG_CPU_MIPSR2 on the top of the file.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/8513/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Commit cf62a8b813 ("MIPS: lib: memcpy: Use macro to build the
copy_user code") switched to a macro in order to build the memcpy
symbols in preparation for the EVA support. However, this commit
also removed the NOP instruction after the 'jr ra' when returning
back to the caller. This had no visible side-effects since the next
instruction was a load to the t0 register which was already in the
clobbered list, but it may have undesired effects in the future
if some other code is introduced in between the .Ldone and
the .Ll_exc_copy labels.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: <stable@vger.kernel.org> # v3.15+
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/8512/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Virtual page number of R3000 in entryhi is 20 bit from MSB. But in
dump_tlb(), the bit mask to read it from entryhi is 19 bit (0xffffe000).
The patch fixes that to 0xfffff000.
Signed-off-by: Isamu Mogi <isamu@leafytree.jp>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/8290/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
We were returning maxlen like the userland strnlen if no '\0' character
was encountered while the kernel version is expected to return a value
larger than maxlen. Fixed to return maxlen + 1.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
This small update to the previous fix to __delay removes a conditional
around the ABI-dependent subtraction operation within an inline asm in
favor to the standard <asm/asm.h> LONG_SUBU macro. No change in code
produced.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/6703/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Nobody is maintaining SMTC anymore and there also seems to be no userbase.
Which is a pity - the SMTC technology primarily developed by Kevin D.
Kissell <kevink@paralogos.com> is an ingenious demonstration for the MT
ASE's power and elegance.
Based on Markos Chandras <Markos.Chandras@imgtec.com> patch
https://patchwork.linux-mips.org/patch/6719/ which while very similar did
no longer apply cleanly when I tried to merge it plus some additional
post-SMTC cleanup - SMTC was a feature as tricky to remove as it was to
merge once upon a time.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
This change reverts most of commit
60724ca59e [MIPS: IP checksums: Remove
unncessary .set pseudos] that introduced warnings with the
CPU_DADDI_WORKAROUNDS option set:
arch/mips/lib/csum_partial.S: Assembler messages:
arch/mips/lib/csum_partial.S:467: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:467: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:467: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:467: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:467: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:467: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:467: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:467: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:467: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:467: Warning: used $3 with ".set at=$3"
[...]
arch/mips/lib/csum_partial.S:577: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:577: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:577: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:601: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:601: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:601: Warning: used $3 with ".set at=$3"
arch/mips/lib/csum_partial.S:601: Warning: used $3 with ".set at=$3"
[and so on, and so on...]
The warnings are benign and good code is produced regardless because no
macros that'd use the assembler's temporary register are involved, however
the `.set noat' directives removed by the commit referred are crucial to
guarantee this is still going to be the case after any changes in the
future. Therefore they need to be brought back to place which this
change does.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/6686/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
This corrects assembler warnings and broken code generated in
__strncpy_from_user_asm:
arch/mips/lib/strncpy_user.S: Assembler messages:
arch/mips/lib/strncpy_user.S:52: Warning: Macro instruction expanded into
multiple instructions in a branch delay slot
with the CPU_DADDI_WORKAROUNDS option set. The function schedules delay
slots manually where there is really no need to as GAS is happy to do it
all itself, so undo it all and remove `.set noreorder'.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/6685/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
With CPU_DADDI_WORKAROUNDS enabled __delay assembles with a macro in a
branch delay slot:
{standard input}: Assembler messages:
{standard input}:18: Warning: Macro instruction expanded into multiple
instructions in a branch delay slot
and broken code results:
0000000000000000 <__delay>:
0: 1480ffff bnez a0,0 <__delay>
4: 24010001 li at,1
8: 0081202f dsubu a0,a0,at
c: 03e00008 jr ra
10: 00000000 nop
14: 00000000 nop
Consequently the function loops indefinitely, showing up prominently as a
hang in the delay loop calibration at bootstrap.
This change corrects the problem by forcing the immediate 1 into a
register while keeping code produced identical where CPU_DADDI_WORKAROUNDS
is disabled.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/6669/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
In preparation for EVA support, we use a macro to build the
__csum_partial_copy_user main code so it can be shared across
multiple implementations. EVA uses the same code but it replaces
the load/store/prefetch instructions with the EVA specific ones
therefore using a macro avoids unnecessary code duplications.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Each load/store macro always adds an entry to the __ex_table
using the EXC macro. There are cases where a load instruction may
never fail such as when we are sure the load happens in the kernel
address space. Therefore, we merge these the EXC and LOADX/STOREX
macros into a single one. We also expand the argument list in the EXC
macro to make the macro more flexible. The extra 'type' argument is not
used by this commit, but it will be used when EVA support is added to
memcpy.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
The 'copy_user' symbol can be used to copy from or to
userland so we will use two different symbols for these
operations. This makes no difference in the existing code,
but when the core is operating in EVA mode, different instructions
need to be used to read and write to userland address space.
The old function has also been renamed to 'copy_kernel' to denote
that it is suitable for copy data to and from kernel space.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Build the __bzero function using the EVA load/store instructions
when operating in the EVA mode. This function is only used when
accessing user code so there is no need to build two distinct symbols
for user and kernel operations respectively.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Build the __bzero symbol using a macor. In EVA mode we will
need to use similar code to do the userspace load operations so
it is better if we use a macro to avoid code duplications.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Add copy_{to,from,in}_user when the CPU operates in EVA mode.
This is necessary so the EVA specific instructions can be used
to perform the virtual to physical translation for user space
addresses. We will use the non-EVA functions to read from kernel
if needed.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
The code can be shared between EVA and non-EVA configurations,
therefore use a macro to build it to avoid code duplications.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
In preparation for EVA support, the PREF macro is split into two
separate macros, PREFS and PREFD, for source and destination data
prefetching respectively.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Each load/store macro always adds an entry to the __ex_table
using the EXC macro. Therefore, these load/store macros are now merged
with the EXC one. The argument list is also expanded in order to make
the macro more flexible. The extra 'type' argument is not used by this
commit, but it will be used when the EVA support is added to the memcpy.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
In non-EVA mode, strncpy_from_user* aliases are used for the
strncpy_from_kernel* symbols since the code is identical. In EVA
mode, new strcpy_from_user* symbols are used which use the EVA
specific instructions to load values from userspace.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Build the __strncpy_from_user symbol using a macro. In EVA mode we will
need to use similar code to do the userspace load operations so
it is better if we use a macro to avoid code duplications.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
In non-EVA mode, strlen_user* aliases are used for the
strlen_kernel* symbols since the code is identical. In EVA
mode, new strlen_user* symbols are used which use the EVA
specific instructions to load values from userspace.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Build the __strlen_user symbol using a macro. In EVA mode we will
need to use similar code to do the userspace load operations so
it is better if we use a macro to avoid code duplications.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
In non-EVA mode, a strlen_user* alias is used for the
strlen_kernel* symbols since the code is identical. In EVA
mode, a new strlen_user* symbol is used which uses the EVA
specific instructions to load values from userspace.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Build the __strnlen_user symbol using a macro. In EVA mode we will
need to use similar code to do the userspace load operations so
it is better if we use a macro to avoid code duplications.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
None of these files are actually using any __init type directives
and hence don't need to include <linux/init.h>. Most are just a
left over from __devinit and __cpuinit removal, or simply due to
code getting copied from one driver to the next.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: John Crispin <blogic@openwrt.org>
Patchwork: http://patchwork.linux-mips.org/patch/6320/
commit 3747069b25e419f6b51395f48127e9812abc3596 upstream.
The __cpuinit type of throwaway sections might have made sense
some time ago when RAM was more constrained, but now the savings
do not offset the cost and complications. For example, the fix in
commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time")
is a good example of the nasty type of bugs that can be created
with improper use of the various __init prefixes.
After a discussion on LKML[1] it was decided that cpuinit should go
the way of devinit and be phased out. Once all the users are gone,
we can then finally remove the macros themselves from linux/init.h.
Note that some harmless section mismatch warnings may result, since
notify_cpu_starting() and cpu_up() are arch independent (kernel/cpu.c)
and are flagged as __cpuinit -- so if we remove the __cpuinit from
the arch specific callers, we will also get section mismatch warnings.
As an intermediate step, we intend to turn the linux/init.h cpuinit
related content into no-ops as early as possible, since that will get
rid of these warnings. In any case, they are temporary and harmless.
Here, we remove all the MIPS __cpuinit from C code and __CPUINIT
from asm files. MIPS is interesting in this respect, because there
are also uasm users hiding behind their own renamed versions of the
__cpuinit macros.
[1] https://lkml.org/lkml/2013/5/20/589
[ralf@linux-mips.org: Folded in Paul's followup fix.]
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/5494/
Patchwork: https://patchwork.linux-mips.org/patch/5495/
Patchwork: https://patchwork.linux-mips.org/patch/5509/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
This reverts commit d532f3d267.
The original commit has several problems:
1) Doesn't work with 64-bit kernels.
2) Calls TLBMISS_HANDLER_SETUP() before the code is generated.
3) Calls TLBMISS_HANDLER_SETUP() twice in per_cpu_trap_init() when
only one call is needed.
[ralf@linux-mips.org: Also revert the bits of the ASID patch which were
hidden in the KVM merge.]
Signed-off-by: David Daney <david.daney@cavium.com>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Cc: "Steven J. Hill" <Steven.Hill@imgtec.com>
Cc: David Daney <david.daney@cavium.com>
Patchwork: https://patchwork.linux-mips.org/patch/5242/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Optimise 'strnlen' to use microMIPS instructions and/or optimisations
for binary size reduction. When the microMIPS ISA is not being used,
the library function compiles to the original binary code.
Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com>
Optimise 'strlen' to use microMIPS instructions and/or optimisations
for binary size reduction. When the microMIPS ISA is not being used,
the library function compiles to the original binary code.
Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com>
Optimise 'strncpy' to use microMIPS instructions and/or optimisations
for binary size reduction. When the microMIPS ISA is not being used,
the library function compiles to the original binary code.
Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com>
Optimise 'memset' to use microMIPS instructions and/or optimisations
for binary size reduction. When the microMIPS ISA is not being used,
the library function compiles to the original binary code.
Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com>
Original patch by Ralf Baechle and removed by Harold Koerfgen
with commit f67e4ffc79905482c3b9b8c8dd65197bac7eb508. This
allows for more generic kernels since the size of the ASID
and corresponding masks can be determined at run-time. This
patch is also required for the new Aptiv cores and has been
tested on Malta and Malta Aptiv platforms.
[ralf@linux-mips.org: Added relevant part of fix
https://patchwork.linux-mips.org/patch/5213/]
Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The operations on the bitmap pointers are protected by "memory"
clobbering raw_local_irq_{save,restore}(), so there is no need for
volatile here. By removing the volatile we get better code generation
out of the compiler.
Signed-off-by: David Daney <david.daney@cavium.com>
Patchwork: http://patchwork.linux-mips.org/patch/4966/
Acked-by: John Crispin <blogic@openwrt.org>
commit 92d11594f6 (MIPS: Remove irqflags.h dependency from bitops.h)
factored some of the bitops code out into a separate file
(arch/mips/lib/bitops.c). Unfortunately the logic converting a bit
mask into a boolean result was lost in some of the functions. We had:
int res;
unsigned long shifted_result_bit;
.
.
.
res = shifted_result_bit;
return res;
Which truncates off the high 32 bits (thus yielding an incorrect
value) on 64-bit systems.
The manifestation of this is that a non-SMP 64-bit kernel will not
boot as the bitmap operations in bootmem.c are all screwed up.
Signed-off-by: David Daney <david.daney@cavium.com>
Cc: linux-mips@linux-mips.org
Cc: Jim Quinlan <jim2101024@gmail.com>
Cc: stable@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/4965/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Chen Jie