diff options
Diffstat (limited to 'vere/ext/gmp/gen/x86_64-linux/mpn/mulmid_basecase.s')
| -rw-r--r-- | vere/ext/gmp/gen/x86_64-linux/mpn/mulmid_basecase.s | 573 |
1 files changed, 573 insertions, 0 deletions
diff --git a/vere/ext/gmp/gen/x86_64-linux/mpn/mulmid_basecase.s b/vere/ext/gmp/gen/x86_64-linux/mpn/mulmid_basecase.s new file mode 100644 index 0000000..b607e84 --- /dev/null +++ b/vere/ext/gmp/gen/x86_64-linux/mpn/mulmid_basecase.s @@ -0,0 +1,573 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + .text + .align 16, 0x90 + .globl __gmpn_mulmid_basecase + .type __gmpn_mulmid_basecase,@function + +__gmpn_mulmid_basecase: + + + + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + + mov %rcx, %r15 + + + lea 1(%rdx), %r13 + sub %r8, %r13 + + lea (%rdi,%r13,8), %rdi + + cmp $4, %r13 + jc .Ldiagonal + + lea (%rsi,%rdx,8), %rsi + + test $1, %r8 + jz .Lmul_2 + + + + +.Lmul_1: + mov %r13d, %ebx + + neg %r13 + mov (%rsi,%r13,8), %rax + mov (%r15), %r12 + mul %r12 + + and $-4, %r13 + mov %r13, %r11 + + and $3, %ebx + jz .Lmul_1_prologue_0 + cmp $2, %ebx + jc .Lmul_1_prologue_1 + jz .Lmul_1_prologue_2 + +.Lmul_1_prologue_3: + mov %rax, %r10 + mov %rdx, %rbx + lea .Laddmul_prologue_3(%rip), %r14 + jmp .Lmul_1_entry_3 + + .align 16, 0x90 +.Lmul_1_prologue_0: + mov %rax, %rbp + mov %rdx, %r10 + lea .Laddmul_prologue_0(%rip), %r14 + jmp .Lmul_1_entry_0 + + .align 16, 0x90 +.Lmul_1_prologue_1: + add $4, %r11 + mov %rax, %rcx + mov %rdx, %rbp + mov $0, %r10d + mov (%rsi,%r11,8), %rax + lea .Laddmul_prologue_1(%rip), %r14 + jmp .Lmul_1_entry_1 + + .align 16, 0x90 +.Lmul_1_prologue_2: + mov %rax, %rbx + mov %rdx, %rcx + mov 24(%rsi,%r11,8), %rax + mov $0, %ebp + mov $0, %r10d + lea .Laddmul_prologue_2(%rip), %r14 + jmp .Lmul_1_entry_2 + + + + + .align 16, 0x90 +.Lmul_1_top: + mov %rbx, -16(%rdi,%r11,8) + add %rax, %rcx + mov (%rsi,%r11,8), %rax + adc %rdx, %rbp +.Lmul_1_entry_1: + mov $0, %ebx + mul %r12 + mov %rcx, -8(%rdi,%r11,8) + add %rax, %rbp + adc %rdx, %r10 +.Lmul_1_entry_0: + mov 8(%rsi,%r11,8), %rax + mul %r12 + mov %rbp, (%rdi,%r11,8) + add %rax, %r10 + adc %rdx, %rbx +.Lmul_1_entry_3: + mov 16(%rsi,%r11,8), %rax + mul %r12 + mov %r10, 8(%rdi,%r11,8) + mov $0, %ebp + mov %rbp, %r10 + add %rax, %rbx + mov 24(%rsi,%r11,8), %rax + mov %rbp, %rcx + adc %rdx, %rcx +.Lmul_1_entry_2: + mul %r12 + add $4, %r11 + js .Lmul_1_top + + mov %rbx, -16(%rdi) + add %rax, %rcx + mov %rcx, -8(%rdi) + mov %rbp, 8(%rdi) + adc %rdx, %rbp + mov %rbp, (%rdi) + + dec %r8 + jz .Lret + + lea -8(%rsi), %rsi + lea 8(%r15), %r15 + + mov %r13, %r11 + mov (%r15), %r12 + mov 8(%r15), %r9 + + jmp *%r14 + + + + + .align 16, 0x90 +.Lmul_2: + mov %r13d, %ebx + + neg %r13 + mov -8(%rsi,%r13,8), %rax + mov (%r15), %r12 + mov 8(%r15), %r9 + mul %r9 + + and $-4, %r13 + mov %r13, %r11 + + and $3, %ebx + jz .Lmul_2_prologue_0 + cmp $2, %ebx + jc .Lmul_2_prologue_1 + jz .Lmul_2_prologue_2 + +.Lmul_2_prologue_3: + mov %rax, %rcx + mov %rdx, %rbp + lea .Laddmul_prologue_3(%rip), %r14 + jmp .Lmul_2_entry_3 + + .align 16, 0x90 +.Lmul_2_prologue_0: + mov %rax, %rbx + mov %rdx, %rcx + lea .Laddmul_prologue_0(%rip), %r14 + jmp .Lmul_2_entry_0 + + .align 16, 0x90 +.Lmul_2_prologue_1: + mov %rax, %r10 + mov %rdx, %rbx + mov $0, %ecx + lea .Laddmul_prologue_1(%rip), %r14 + jmp .Lmul_2_entry_1 + + .align 16, 0x90 +.Lmul_2_prologue_2: + mov %rax, %rbp + mov %rdx, %r10 + mov $0, %ebx + mov 16(%rsi,%r11,8), %rax + lea .Laddmul_prologue_2(%rip), %r14 + jmp .Lmul_2_entry_2 + + + + + .align 16, 0x90 +.Lmul_2_top: + mov -8(%rsi,%r11,8), %rax + mul %r9 + add %rax, %rbx + adc %rdx, %rcx +.Lmul_2_entry_0: + mov $0, %ebp + mov (%rsi,%r11,8), %rax + mul %r12 + add %rax, %rbx + mov (%rsi,%r11,8), %rax + adc %rdx, %rcx + adc $0, %ebp + mul %r9 + add %rax, %rcx + mov %rbx, (%rdi,%r11,8) + adc %rdx, %rbp +.Lmul_2_entry_3: + mov 8(%rsi,%r11,8), %rax + mul %r12 + mov $0, %r10d + add %rax, %rcx + adc %rdx, %rbp + mov $0, %ebx + adc $0, %r10d + mov 8(%rsi,%r11,8), %rax + mov %rcx, 8(%rdi,%r11,8) + mul %r9 + add %rax, %rbp + mov 16(%rsi,%r11,8), %rax + adc %rdx, %r10 +.Lmul_2_entry_2: + mov $0, %ecx + mul %r12 + add %rax, %rbp + mov 16(%rsi,%r11,8), %rax + adc %rdx, %r10 + adc $0, %ebx + mul %r9 + add %rax, %r10 + mov %rbp, 16(%rdi,%r11,8) + adc %rdx, %rbx +.Lmul_2_entry_1: + mov 24(%rsi,%r11,8), %rax + mul %r12 + add %rax, %r10 + adc %rdx, %rbx + adc $0, %ecx + add $4, %r11 + mov %r10, -8(%rdi,%r11,8) + jnz .Lmul_2_top + + mov %rbx, (%rdi) + mov %rcx, 8(%rdi) + + sub $2, %r8 + jz .Lret + + lea 16(%r15), %r15 + lea -16(%rsi), %rsi + + mov %r13, %r11 + mov (%r15), %r12 + mov 8(%r15), %r9 + + jmp *%r14 + + + + + .align 16, 0x90 +.Laddmul_prologue_0: + mov -8(%rsi,%r11,8), %rax + mul %r9 + mov %rax, %rcx + mov %rdx, %rbp + mov $0, %r10d + jmp .Laddmul_entry_0 + + .align 16, 0x90 +.Laddmul_prologue_1: + mov 16(%rsi,%r11,8), %rax + mul %r9 + mov %rax, %rbx + mov %rdx, %rcx + mov $0, %ebp + mov 24(%rsi,%r11,8), %rax + jmp .Laddmul_entry_1 + + .align 16, 0x90 +.Laddmul_prologue_2: + mov 8(%rsi,%r11,8), %rax + mul %r9 + mov %rax, %r10 + mov %rdx, %rbx + mov $0, %ecx + jmp .Laddmul_entry_2 + + .align 16, 0x90 +.Laddmul_prologue_3: + mov (%rsi,%r11,8), %rax + mul %r9 + mov %rax, %rbp + mov %rdx, %r10 + mov $0, %ebx + mov $0, %ecx + jmp .Laddmul_entry_3 + + + + .align 16, 0x90 +.Laddmul_top: + mov $0, %r10d + add %rax, %rbx + mov -8(%rsi,%r11,8), %rax + adc %rdx, %rcx + adc $0, %ebp + mul %r9 + add %rbx, -8(%rdi,%r11,8) + adc %rax, %rcx + adc %rdx, %rbp +.Laddmul_entry_0: + mov (%rsi,%r11,8), %rax + mul %r12 + add %rax, %rcx + mov (%rsi,%r11,8), %rax + adc %rdx, %rbp + adc $0, %r10d + mul %r9 + add %rcx, (%rdi,%r11,8) + mov $0, %ecx + adc %rax, %rbp + mov $0, %ebx + adc %rdx, %r10 +.Laddmul_entry_3: + mov 8(%rsi,%r11,8), %rax + mul %r12 + add %rax, %rbp + mov 8(%rsi,%r11,8), %rax + adc %rdx, %r10 + adc $0, %ebx + mul %r9 + add %rbp, 8(%rdi,%r11,8) + adc %rax, %r10 + adc %rdx, %rbx +.Laddmul_entry_2: + mov 16(%rsi,%r11,8), %rax + mul %r12 + add %rax, %r10 + mov 16(%rsi,%r11,8), %rax + adc %rdx, %rbx + adc $0, %ecx + mul %r9 + add %r10, 16(%rdi,%r11,8) + nop + adc %rax, %rbx + mov $0, %ebp + mov 24(%rsi,%r11,8), %rax + adc %rdx, %rcx +.Laddmul_entry_1: + mul %r12 + add $4, %r11 + jnz .Laddmul_top + + add %rax, %rbx + adc %rdx, %rcx + adc $0, %ebp + + add %rbx, -8(%rdi) + adc %rcx, (%rdi) + adc %rbp, 8(%rdi) + + sub $2, %r8 + jz .Lret + + lea 16(%r15), %r15 + lea -16(%rsi), %rsi + + mov %r13, %r11 + mov (%r15), %r12 + mov 8(%r15), %r9 + + jmp *%r14 + + + + + .align 16, 0x90 +.Ldiagonal: + xor %ebx, %ebx + xor %ecx, %ecx + xor %ebp, %ebp + + neg %r13 + + mov %r8d, %eax + and $3, %eax + jz .Ldiag_prologue_0 + cmp $2, %eax + jc .Ldiag_prologue_1 + jz .Ldiag_prologue_2 + +.Ldiag_prologue_3: + lea -8(%r15), %r15 + mov %r15, %r10 + add $1, %r8 + mov %r8, %r11 + lea .Ldiag_entry_3(%rip), %r14 + jmp .Ldiag_entry_3 + +.Ldiag_prologue_0: + mov %r15, %r10 + mov %r8, %r11 + lea 0(%rip), %r14 + mov -8(%rsi,%r11,8), %rax + jmp .Ldiag_entry_0 + +.Ldiag_prologue_1: + lea 8(%r15), %r15 + mov %r15, %r10 + add $3, %r8 + mov %r8, %r11 + lea 0(%rip), %r14 + mov -8(%r10), %rax + jmp .Ldiag_entry_1 + +.Ldiag_prologue_2: + lea -16(%r15), %r15 + mov %r15, %r10 + add $2, %r8 + mov %r8, %r11 + lea 0(%rip), %r14 + mov 16(%r10), %rax + jmp .Ldiag_entry_2 + + + + + .align 16, 0x90 +.Ldiag_top: + add %rax, %rbx + adc %rdx, %rcx + mov -8(%rsi,%r11,8), %rax + adc $0, %rbp +.Ldiag_entry_0: + mulq (%r10) + add %rax, %rbx + adc %rdx, %rcx + adc $0, %rbp +.Ldiag_entry_3: + mov -16(%rsi,%r11,8), %rax + mulq 8(%r10) + add %rax, %rbx + mov 16(%r10), %rax + adc %rdx, %rcx + adc $0, %rbp +.Ldiag_entry_2: + mulq -24(%rsi,%r11,8) + add %rax, %rbx + mov 24(%r10), %rax + adc %rdx, %rcx + lea 32(%r10), %r10 + adc $0, %rbp +.Ldiag_entry_1: + mulq -32(%rsi,%r11,8) + sub $4, %r11 + jnz .Ldiag_top + + add %rax, %rbx + adc %rdx, %rcx + adc $0, %rbp + + mov %rbx, (%rdi,%r13,8) + + inc %r13 + jz .Ldiag_end + + mov %r8, %r11 + mov %r15, %r10 + + lea 8(%rsi), %rsi + mov %rcx, %rbx + mov %rbp, %rcx + xor %ebp, %ebp + + jmp *%r14 + +.Ldiag_end: + mov %rcx, (%rdi) + mov %rbp, 8(%rdi) + +.Lret: pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + + ret + .size __gmpn_mulmid_basecase,.-__gmpn_mulmid_basecase |