andre/cosmopolitan
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Get codebase completely working with LLVM

Browse Source 
You can now build Cosmopolitan with Clang:

    make -j8 MODE=llvm
    o/llvm/examples/hello.com

The assembler and linker code is now friendly to LLVM too.
So it's not needed to configure Clang to use binutils under
the hood. If you love LLVM then you can now use pure LLVM.
This commit is contained in:
Justine Tunney
2021-02-08 09:19:00 -08:00
parent 0e36cb3ac4
commit e75ffde09e
4528 changed files with 7776 additions and 11640 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
libc/nexgen32e/djbsort-avx2.S
View File

@@ -1,12 +1,12 @@
#include "libc/macros.h"
.source __FILE__
/ D.J. Bernstein's outrageously fast integer sorting algorithm.
/
/ @param rdi is int32 array
/ @param rsi is number of elements in rdi
/ @note public domain
/ @see en.wikipedia.org/wiki/Sorting_network
// D.J. Bernstein's outrageously fast integer sorting algorithm.
//
// @param rdi is int32 array
// @param rsi is number of elements in rdi
// @note public domain
// @see en.wikipedia.org/wiki/Sorting_network
djbsort_avx2:
push %rbp
mov %rsp,%rbp
@@ -1115,7 +1115,7 @@ int32_sort_2power:
vpxor 32(%rdi),%ymm0,%ymm2
vpxor (%rdi),%ymm0,%ymm0
vmovdqa .LC1(%rip),%ymm4
cmp $0,-116(%rbp)
cmpl $0,-116(%rbp)
vpunpckldq %ymm2,%ymm0,%ymm1
vpunpckhdq %ymm2,%ymm0,%ymm0
vpunpcklqdq %ymm0,%ymm1,%ymm3
@@ -1179,7 +1179,7 @@ int32_sort_2power:
mov $16,%esi
mov %r13,%rdi
call int32_sort_2power
cmp $0,-116(%rbp)
cmpl $0,-116(%rbp)
vmovdqu (%r12),%ymm4
vmovdqu 32(%r12),%ymm1
vmovdqu 64(%r12),%ymm2
@@ -1750,7 +1750,7 @@ int32_sort_2power:
vpunpckhdq 160(%rax),%ymm7,%ymm0
vpunpcklqdq %ymm2,%ymm12,%ymm8
vpunpcklqdq %ymm4,%ymm6,%ymm9
cmp $0,-116(%rbp)
cmpl $0,-116(%rbp)
vmovdqu 192(%rax),%ymm7
vpunpckhqdq %ymm2,%ymm12,%ymm12
vpunpckhqdq %ymm4,%ymm6,%ymm4
@@ -1837,7 +1837,7 @@ int32_sort_2power:
vmovdqu %ymm2,-64(%rax)
vmovdqu %ymm0,-32(%rax)
jmp .L85
.L142: cmp $32,-112(%rbp)
.L142: cmpq $32,-112(%rbp)
jne .L94
.L93: mov -112(%rbp),%rcx
sar $2,%rcx
@@ -1871,7 +1871,7 @@ int32_sort_2power:
cmp %rax,%r15
jg .L92
sarq $3,-112(%rbp)
.L89: cmp $127,-112(%rbp)
.L89: cmpq $127,-112(%rbp)
jle .L142
jmp .L93
.L92: cmp -136(%rbp),%rax
@@ -1925,7 +1925,7 @@ int32_sort_2power:
add %rdx,-136(%rbp)
jmp .L90
.L145: sarq $2,-112(%rbp)
.L94: cmp $15,-112(%rbp)
.L94: cmpq $15,-112(%rbp)
jle .L144
mov -112(%rbp),%rcx
xor %esi,%esi
@@ -1962,7 +1962,7 @@ int32_sort_2power:
.L146: add %rdx,%rsi
add %rdx,%rcx
jmp .L95
.L144: cmp $8,-112(%rbp)
.L144: cmpq $8,-112(%rbp)
je .L111
.L102: mov -152(%rbp),%rdx
add -128(%rbp),%rdx
@@ -1997,7 +1997,7 @@ int32_sort_2power:
vpmaxsd %ymm3,%ymm5,%ymm3
vpminsd (%rdi),%ymm7,%ymm1
vpminsd %ymm2,%ymm4,%ymm10
cmp $0,-116(%rbp)
cmpl $0,-116(%rbp)
vpmaxsd (%rdi),%ymm7,%ymm0
vmovdqu (%rsi),%ymm7
vpmaxsd %ymm2,%ymm4,%ymm2