Differences From Artifact [2b14e6dad4]:
- File
src/invocation/apple-call-x86_64.S
— part of check-in
[780d096371]
at
2019-02-18 01:37:36
on branch trunk
— invocation: Make all offsets constants
This makes the assembly much more readable.
No difference between disassembled .o file before and after. (user: js, size: 2537) [annotate] [blame] [check-ins using]
To Artifact [c30e24b805]:
- File
src/invocation/apple-call-x86_64.S
— part of check-in
[a1da5c7b2d]
at
2019-02-18 20:55:15
on branch trunk
— Replace a few movdqa with movaps
movaps is smaller, and we don't know what kind of data is in there
anyway (and it's FP rather than integer in most cases anyway), so it
does not matter which move instruction we use. (user: js, size: 2537) [annotate] [blame] [check-ins using]
| ︙ | ︙ | |||
45 46 47 48 49 50 51 | pushq %r11 jmp Lfill_stack Lstack_filled: movb OFFSET_NUM_SSE_USED(%rdi), %al | | | | | | | | | | 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 | pushq %r11 jmp Lfill_stack Lstack_filled: movb OFFSET_NUM_SSE_USED(%rdi), %al movaps OFFSET_SSE_INOUT+112(%rdi), %xmm7 movaps OFFSET_SSE_INOUT+96(%rdi), %xmm6 movaps OFFSET_SSE_INOUT+80(%rdi), %xmm5 movaps OFFSET_SSE_INOUT+64(%rdi), %xmm4 movaps OFFSET_SSE_INOUT+48(%rdi), %xmm3 movaps OFFSET_SSE_INOUT+32(%rdi), %xmm2 movaps OFFSET_SSE_INOUT+16(%rdi), %xmm1 movaps OFFSET_SSE_INOUT(%rdi), %xmm0 movq OFFSET_GPR_IN+40(%rdi), %r9 movq OFFSET_GPR_IN+32(%rdi), %r8 movq OFFSET_GPR_IN+24(%rdi), %rcx movq OFFSET_GPR_IN+16(%rdi), %rdx movq OFFSET_GPR_IN+8(%rdi), %rsi |
| ︙ | ︙ | |||
78 79 80 81 82 83 84 | call _objc_msgSend Lafter_send: movq -8(%rbp), %rdi movq %rax, OFFSET_GPR_OUT(%rdi) movq %rdx, OFFSET_GPR_OUT+8(%rdi) | | | | 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 | call _objc_msgSend Lafter_send: movq -8(%rbp), %rdi movq %rax, OFFSET_GPR_OUT(%rdi) movq %rdx, OFFSET_GPR_OUT+8(%rdi) movaps %xmm0, OFFSET_SSE_INOUT(%rdi) movaps %xmm1, OFFSET_SSE_INOUT+16(%rdi) movb OFFSET_RETURN_TYPE(%rdi), %r11b cmpb $RETURN_TYPE_X87, %r11b je Lpop_long_double cmpb $RETURN_TYPE_COMPLEX_X87, %r11b |
| ︙ | ︙ |