diff options
| author | Ma Ling <ling.ma@intel.com> | 2012-10-17 15:52:45 -0400 |
|---|---|---|
| committer | Ingo Molnar <mingo@kernel.org> | 2012-10-24 06:42:47 -0400 |
| commit | 269833bd5a0f4443873da358b71675a890b47c3c (patch) | |
| tree | 40d476ea896ded8c0957f20c77e583673f8d1d0b /arch/x86/lib | |
| parent | 0e9e3e306c7e472bdcffa34c4c4584301eda03b3 (diff) | |
x86/asm: Clean up copy_page_*() comments and code
Modern CPUs use fast-string instruction to accelerate copy
performance, by combining data into 128 bit chunks.
Modify comments and coding style to match it.
Signed-off-by: Ma Ling <ling.ma@intel.com>
Cc: iant@google.com
Link: http://lkml.kernel.org/r/1350503565-19167-1-git-send-email-ling.ma@intel.com
[ Cleaned up the clean up. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Diffstat (limited to 'arch/x86/lib')
| -rw-r--r-- | arch/x86/lib/copy_page_64.S | 120 |
1 files changed, 59 insertions, 61 deletions
diff --git a/arch/x86/lib/copy_page_64.S b/arch/x86/lib/copy_page_64.S index 6b34d04d096a..176cca67212b 100644 --- a/arch/x86/lib/copy_page_64.S +++ b/arch/x86/lib/copy_page_64.S | |||
| @@ -5,91 +5,89 @@ | |||
| 5 | #include <asm/alternative-asm.h> | 5 | #include <asm/alternative-asm.h> |
| 6 | 6 | ||
| 7 | ALIGN | 7 | ALIGN |
| 8 | copy_page_c: | 8 | copy_page_rep: |
| 9 | CFI_STARTPROC | 9 | CFI_STARTPROC |
| 10 | movl $4096/8,%ecx | 10 | movl $4096/8, %ecx |
| 11 | rep movsq | 11 | rep movsq |
| 12 | ret | 12 | ret |
| 13 | CFI_ENDPROC | 13 | CFI_ENDPROC |
| 14 | ENDPROC(copy_page_c) | 14 | ENDPROC(copy_page_rep) |
| 15 | 15 | ||
| 16 | /* Don't use streaming store because it's better when the target | 16 | /* |
| 17 | ends up in cache. */ | 17 | * Don't use streaming copy unless the CPU indicates X86_FEATURE_REP_GOOD. |
| 18 | 18 | * Could vary the prefetch distance based on SMP/UP. | |
| 19 | /* Could vary the prefetch distance based on SMP/UP */ | 19 | */ |
| 20 | 20 | ||
| 21 | ENTRY(copy_page) | 21 | ENTRY(copy_page) |
| 22 | CFI_STARTPROC | 22 | CFI_STARTPROC |
| 23 | subq $2*8,%rsp | 23 | subq $2*8, %rsp |
| 24 | CFI_ADJUST_CFA_OFFSET 2*8 | 24 | CFI_ADJUST_CFA_OFFSET 2*8 |
| 25 | movq %rbx,(%rsp) | 25 | movq %rbx, (%rsp) |
| 26 | CFI_REL_OFFSET rbx, 0 | 26 | CFI_REL_OFFSET rbx, 0 |
| 27 | movq %r12,1*8(%rsp) | 27 | movq %r12, 1*8(%rsp) |
| 28 | CFI_REL_OFFSET r12, 1*8 | 28 | CFI_REL_OFFSET r12, 1*8 |
| 29 | 29 | ||
| 30 | movl $(4096/64)-5,%ecx | 30 | movl $(4096/64)-5, %ecx |
| 31 | .p2align 4 | 31 | .p2align 4 |
| 32 | .Loop64: | 32 | .Loop64: |
| 33 | dec %rcx | 33 | dec %rcx |
| 34 | 34 | movq 0x8*0(%rsi), %rax | |
| 35 | movq (%rsi), %rax | 35 | movq 0x8*1(%rsi), %rbx |
| 36 | movq 8 (%rsi), %rbx | 36 | movq 0x8*2(%rsi), %rdx |
| 37 | movq 16 (%rsi), %rdx | 37 | movq 0x8*3(%rsi), %r8 |
| 38 | movq 24 (%rsi), %r8 | 38 | movq 0x8*4(%rsi), %r9 |
| 39 | movq 32 (%rsi), %r9 | 39 | movq 0x8*5(%rsi), %r10 |
| 40 | movq 40 (%rsi), %r10 | 40 | movq 0x8*6(%rsi), %r11 |
| 41 | movq 48 (%rsi), %r11 | 41 | movq 0x8*7(%rsi), %r12 |
| 42 | movq 56 (%rsi), %r12 | ||
| 43 | 42 | ||
| 44 | prefetcht0 5*64(%rsi) | 43 | prefetcht0 5*64(%rsi) |
| 45 | 44 | ||
| 46 | movq %rax, (%rdi) | 45 | movq %rax, 0x8*0(%rdi) |
| 47 | movq %rbx, 8 (%rdi) | 46 | movq %rbx, 0x8*1(%rdi) |
| 48 | movq %rdx, 16 (%rdi) | 47 | movq %rdx, 0x8*2(%rdi) |
| 49 | movq %r8, 24 (%rdi) | 48 | movq %r8, 0x8*3(%rdi) |
| 50 | movq %r9, 32 (%rdi) | 49 | movq %r9, 0x8*4(%rdi) |
| 51 | movq %r10, 40 (%rdi) | 50 | movq %r10, 0x8*5(%rdi) |
| 52 | movq %r11, 48 (%rdi) | 51 | movq %r11, 0x8*6(%rdi) |
| 53 | movq %r12, 56 (%rdi) | 52 | movq %r12, 0x8*7(%rdi) |
| 54 | 53 | ||
| 55 | leaq 64 (%rsi), %rsi | 54 | leaq 64 (%rsi), %rsi |
| 56 | leaq 64 (%rdi), %rdi | 55 | leaq 64 (%rdi), %rdi |
| 57 | 56 | ||
| 58 | jnz .Loop64 | 57 | jnz .Loop64 |
| 59 | 58 | ||
| 60 | movl $5,%ecx | 59 | movl $5, %ecx |
| 61 | .p2align 4 | 60 | .p2align 4 |
| 62 | .Loop2: | 61 | .Loop2: |
| 63 | decl %ecx | 62 | decl %ecx |
| 64 | 63 | ||
| 65 | movq (%rsi), %rax | 64 | movq 0x8*0(%rsi), %rax |
| 66 | movq 8 (%rsi), %rbx | 65 | movq 0x8*1(%rsi), %rbx |
| 67 | movq 16 (%rsi), %rdx | 66 | movq 0x8*2(%rsi), %rdx |
| 68 | movq 24 (%rsi), %r8 | 67 | movq 0x8*3(%rsi), %r8 |
| 69 | movq 32 (%rsi), %r9 | 68 | movq 0x8*4(%rsi), %r9 |
| 70 | movq 40 (%rsi), %r10 | 69 | movq 0x8*5(%rsi), %r10 |
| 71 | movq 48 (%rsi), %r11 | 70 | movq 0x8*6(%rsi), %r11 |
| 72 | movq 56 (%rsi), %r12 | 71 | movq 0x8*7(%rsi), %r12 |
| 73 | 72 | ||
| 74 | movq %rax, (%rdi) | 73 | movq %rax, 0x8*0(%rdi) |
| 75 | movq %rbx, 8 (%rdi) | 74 | movq %rbx, 0x8*1(%rdi) |
| 76 | movq %rdx, 16 (%rdi) | 75 | movq %rdx, 0x8*2(%rdi) |
| 77 | movq %r8, 24 (%rdi) | 76 | movq %r8, 0x8*3(%rdi) |
| 78 | movq %r9, 32 (%rdi) | 77 | movq %r9, 0x8*4(%rdi) |
| 79 | movq %r10, 40 (%rdi) | 78 | movq %r10, 0x8*5(%rdi) |
| 80 | movq %r11, 48 (%rdi) | 79 | movq %r11, 0x8*6(%rdi) |
| 81 | movq %r12, 56 (%rdi) | 80 | movq %r12, 0x8*7(%rdi) |
| 82 | 81 | ||
| 83 | leaq 64(%rdi),%rdi | 82 | leaq 64(%rdi), %rdi |
| 84 | leaq 64(%rsi),%rsi | 83 | leaq 64(%rsi), %rsi |
| 85 | |||
| 86 | jnz .Loop2 | 84 | jnz .Loop2 |
| 87 | 85 | ||
| 88 | movq (%rsp),%rbx | 86 | movq (%rsp), %rbx |
| 89 | CFI_RESTORE rbx | 87 | CFI_RESTORE rbx |
| 90 | movq 1*8(%rsp),%r12 | 88 | movq 1*8(%rsp), %r12 |
| 91 | CFI_RESTORE r12 | 89 | CFI_RESTORE r12 |
| 92 | addq $2*8,%rsp | 90 | addq $2*8, %rsp |
| 93 | CFI_ADJUST_CFA_OFFSET -2*8 | 91 | CFI_ADJUST_CFA_OFFSET -2*8 |
| 94 | ret | 92 | ret |
| 95 | .Lcopy_page_end: | 93 | .Lcopy_page_end: |
| @@ -103,7 +101,7 @@ ENDPROC(copy_page) | |||
| 103 | 101 | ||
| 104 | .section .altinstr_replacement,"ax" | 102 | .section .altinstr_replacement,"ax" |
| 105 | 1: .byte 0xeb /* jmp <disp8> */ | 103 | 1: .byte 0xeb /* jmp <disp8> */ |
| 106 | .byte (copy_page_c - copy_page) - (2f - 1b) /* offset */ | 104 | .byte (copy_page_rep - copy_page) - (2f - 1b) /* offset */ |
| 107 | 2: | 105 | 2: |
| 108 | .previous | 106 | .previous |
| 109 | .section .altinstructions,"a" | 107 | .section .altinstructions,"a" |