Remove CONFIG_POWER4_ONLY, the option is badly named and only does two
things:
- It wraps the MMU segment table code. With feature fixups there is
little downside to compiling this in.
- It uses the newer mtocrf instruction in various assembly functions.
Instead of making this a compile option just do it at runtime via
a feature fixup.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Implement a POWER7 optimised copy_to_user/copy_from_user using VMX.
For large aligned copies this new loop is over 10% faster, and for
large unaligned copies it is over 200% faster.
If we take a fault we fall back to the old version, this keeps
things relatively simple and easy to verify.
On POWER7 unaligned stores rarely slow down - they only flush when
a store crosses a 4KB page boundary. Furthermore this flush is
handled completely in hardware and should be 20-30 cycles.
Unaligned loads on the other hand flush much more often - whenever
crossing a 128 byte cache line, or a 32 byte sector if either sector
is an L1 miss.
Considering this information we really want to get the loads aligned
and not worry about the alignment of the stores. Microbenchmarks
confirm that this approach is much faster than the current unaligned
copy loop that uses shifts and rotates to ensure both loads and
stores are aligned.
We also want to try and do the stores in cacheline aligned, cacheline
sized chunks. If the store queue is unable to merge an entire
cacheline of stores then the L2 cache will have to do a
read/modify/write. Even worse, we will serialise this with the stores
in the next iteration of the copy loop since both iterations hit
the same cacheline.
Based on this, the new loop does the following things:
1 - 127 bytes
Get the source 8 byte aligned and use 8 byte loads and stores. Pretty
boring and similar to how the current loop works.
128 - 4095 bytes
Get the source 8 byte aligned and use 8 byte loads and stores,
1 cacheline at a time. We aren't doing the stores in cacheline
aligned chunks so we will potentially serialise once per cacheline.
Even so it is much better than the loop we have today.
4096 - bytes
If both source and destination have the same alignment get them both
16 byte aligned, then get the destination cacheline aligned. Do
cacheline sized loads and stores using VMX.
If source and destination do not have the same alignment, we get the
destination cacheline aligned, and use permute to do aligned loads.
In both cases the VMX loop should be optimal - we always do aligned
loads and stores and are always doing stores in cacheline aligned,
cacheline sized chunks.
To be able to use VMX we must be careful about interrupts and
sleeping. We don't use the VMX loop when in an interrupt (which should
be rare anyway) and we wrap the VMX loop in disable/enable_pagefault
and fall back to the existing copy_tofrom_user loop if we do need to
sleep.
The VMX breakpoint of 4096 bytes was chosen using this microbenchmark:
http://ozlabs.org/~anton/junkcode/copy_to_user.c
Since we are using VMX and there is a cost to saving and restoring
the user VMX state there are two broad cases we need to benchmark:
- Best case - userspace never uses VMX
- Worst case - userspace always uses VMX
In reality a userspace process will sit somewhere between these two
extremes. Since we need to test both aligned and unaligned copies we
end up with 4 combinations. The point at which the VMX loop begins to
win is:
0% VMX
aligned 2048 bytes
unaligned 2048 bytes
100% VMX
aligned 16384 bytes
unaligned 8192 bytes
Considering this is a microbenchmark, the data is hot in cache and
the VMX loop has better store queue merging properties we set the
breakpoint to 4096 bytes, a little below the unaligned breakpoints.
Some future optimisations we can look at:
- Looking at the perf data, a significant part of the cost when a
task is always using VMX is the extra exception we take to restore
the VMX state. As such we should do something similar to the x86
optimisation that restores FPU state for heavy users. ie:
/*
* If the task has used fpu the last 5 timeslices, just do a full
* restore of the math state immediately to avoid the trap; the
* chances of needing FPU soon are obviously high now
*/
preload_fpu = tsk_used_math(next_p) && next_p->fpu_counter > 5;
and
/*
* fpu_counter contains the number of consecutive context switches
* that the FPU is used. If this is over a threshold, the lazy fpu
* saving becomes unlazy to save the trap. This is an unsigned char
* so that after 256 times the counter wraps and the behavior turns
* lazy again; this to deal with bursty apps that only use FPU for
* a short time
*/
- We could create a paca bit to mirror the VMX enabled MSR bit and check
that first, avoiding multiple calls to calling enable_kernel_altivec.
That should help with iovec based system calls like readv.
- We could have two VMX breakpoints, one for when we know the user VMX
state is loaded into the registers and one when it isn't. This could
be a second bit in the paca so we can calculate the break points quickly.
- One suggestion from Ben was to save and restore the VSX registers
we use inline instead of using enable_kernel_altivec.
[BenH: Fixed a problem with preempt and fixed build without CONFIG_ALTIVEC]
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Here is a patch from Paul Mackerras that improves the ppc64 copy_tofrom_user.
The loop now does 32 bytes at a time and as well as pairing loads and stores.
A quick test case that reads 8kB over and over shows the improvement:
POWER6: 53% faster
POWER7: 51% faster
#define _XOPEN_SOURCE 500
#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <fcntl.h>
#include <sys/types.h>
#include <sys/stat.h>
#define BUFSIZE (8 * 1024)
#define ITERATIONS 10000000
int main()
{
char tmpfile[] = "/tmp/copy_to_user_testXXXXXX";
int fd;
char *buf[BUFSIZE];
unsigned long i;
fd = mkstemp(tmpfile);
if (fd < 0) {
perror("open");
exit(1);
}
if (write(fd, buf, BUFSIZE) != BUFSIZE) {
perror("open");
exit(1);
}
for (i = 0; i < 10000000; i++) {
if (pread(fd, buf, BUFSIZE, 0) != BUFSIZE) {
perror("pread");
exit(1);
}
}
unlink(tmpfile);
return 0;
}
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This fixes a regression introduced by commit
a4e22f02f5 ("powerpc: Update 64bit
__copy_tofrom_user() using CPU_FTR_UNALIGNED_LD_STD").
The same bug that existed in the 64bit memcpy() also exists here so fix
it here too. The fix is the same as that applied to memcpy() with the
addition of fixes for the exception handling code required for
__copy_tofrom_user().
This stops us reading beyond the end of the source region we were told
to copy.
Signed-off-by: Mark Nelson <markn@au1.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
In exactly the same way that we updated memcpy() with new feature
sections in commit 25d6e2d7c5 ("powerpc:
Update 64bit memcpy() using CPU_FTR_UNALIGNED_LD_STD"), we do the same
thing here for __copy_tofrom_user(). Once again this is purely a
performance tweak for Cell and Power6 - this has no effect on all the
other 64bit powerpc chips.
We can make these same changes to __copy_tofrom_user() because the
basic copy algorithm is the same as in memcpy() - this version just
has all the exception handling logic needed when copying to or from
userspace as well as a special case for copying whole 4K pages that
are page aligned.
CPU_FTR_UNALIGNED_LD_STD CPU was added in commit
4ec577a289 ("powerpc: Add new CPU
feature: CPU_FTR_UNALIGNED_LD_STD").
We also make the same simple one line change from cmpldi r1,... to
cmpldi cr1,... for consistency.
Signed-off-by: Mark Nelson <markn@au1.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
mtocrf is a faster single-field mtcrf (move to condition register
fields) instruction available in POWER4 and later processors. It can
make quite a difference in performance on some implementations, so use
it for CONFIG_POWER4_ONLY builds.
Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This patch removes all self references and fixes references to files
in the now defunct arch/ppc64 tree. I think this accomplises
everything wanted, though there might be a few references I missed.
Signed-off-by: Jon Mason <jdmason@us.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Adds a new CONFIG_PPC_64K_PAGES which, when enabled, changes the kernel
base page size to 64K. The resulting kernel still boots on any
hardware. On current machines with 4K pages support only, the kernel
will maintain 16 "subpages" for each 64K page transparently.
Note that while real 64K capable HW has been tested, the current patch
will not enable it yet as such hardware is not released yet, and I'm
still verifying with the firmware architects the proper to get the
information from the newer hypervisors.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This doesn't change any code, just renames things so we consistently
have foo_32.c and foo_64.c where we have separate 32- and 64-bit
versions.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Anton Blanchard