1 files changed, 152 insertions, 0 deletions
diff --git a/zap/src/mem/utf8enc.S b/zap/src/mem/utf8enc.S
new file mode 100644
index 0000000..24e09d8
--- /dev/null
+++ b/zap/src/mem/utf8enc.S
@@ -0,0 +1,152 @@
+/*
+ Copyright 2022 Gabriel Jensen.
+ This Source Code Form is subject to the terms of the Mozilla Public License, v. 2.0.
+ If a copy of the MPL was not distributed with this file, You can obtain one at https://mozilla.org/MPL/2.0/.
+*/
+
+#include <zap/priv.h>
+
+.globl zap_utf8enc
+
+zap_utf8enc:
+
+ /*
+ zap_chr20 const * in
+ zap_chr8 * out
+ */
+
+#if defined(__amd64__)
+
+ # rdi: Current input codepoint.
+ # rsi: Current output octet.
+ # rax: Current codepoint.
+ # rdx: Temporary.
+
+ # Iterate over the input:
+.loop:
+
+ movl (%rdi),%eax
+
+ cmpl $0xFFFF,%eax
+ jg .oct4
+
+ cmpl $0x7FF,%eax
+ jg .oct3
+
+ cmpl $0x7F,%eax
+ jg .oct2 # Otherwise, only one octet is needed.
+
+ # One octet:
+.oct1:
+
+ # Octet #0:
+ movb %al,(%rsi) # No conversion needed:
+
+ incq %rsi
+
+ # Test if we have reached the null-terminator:
+ testb %al,%al
+ jz .done
+
+ jmp .cnt
+
+ # Two octets:
+.oct2:
+
+ /* Octet #0: */
+ movl %eax,%edx
+ shrl $0x6,%edx
+ orb $0xC0,%dl
+ movb %dl,(%rsi)
+
+ incq %rsi
+
+ # Octet #1:
+ movl %eax,%edx
+ andb $0x3F,%dl
+ orb $0x80,%dl
+ movb %dl,(%rsi)
+
+ incq %rsi
+
+ jmp .cnt
+
+ # Three octets:
+.oct3:
+
+ # Octet #0:
+ movl %eax,%edx
+ shrl $0xC,%edx
+ orb $0xE0,%dl
+ movb %dl,(%rsi)
+
+ incq %rsi
+
+ # Octet #1:
+ movl %eax,%edx
+ shrl $0x6,%edx
+ andb $0x3F,%dl
+ orb $0x80,%dl
+ movb %dl,(%rsi)
+
+ incq %rsi
+
+ # Octet #2:
+ movl %eax,%edx
+ andb $0x3F,%dl
+ orb $0x80,%dl
+ movb %dl,(%rsi)
+
+ incq %rsi
+
+ jmp .cnt
+
+ # Four octets:/
+.oct4:
+
+ # Octet #0:
+ movl %eax,%edx
+ shrl $0x12,%edx
+ orb $0xF0,%dl
+ movb %dl,(%rsi)
+
+ incq %rsi
+
+ # Octet #1:
+ movl %eax,%edx
+ shrl $0xC,%edx
+ andb $0x3F,%dl
+ orb $0x80,%dl
+ movb %dl,(%rsi)
+
+ incq %rsi
+
+ # Octet #2:
+ movl %eax,%edx
+ shrl $0x6,%edx
+ andb $0x3F,%dl
+ orb $0x80,%dl
+ movb %dl,(%rsi)
+
+ incq %rsi
+
+ # Octet #3:
+ movl %eax,%edx
+ andb $0x3F,%dl
+ orb $0x80,%dl
+ movb %dl,(%rsi)
+
+ incq %rsi
+
+ # Continue to the next codepoint:
+.cnt:
+
+ addq $0x4,%rdi
+ jmp .loop
+
+ # Done:
+.done:
+
+ ret
+ 
+#endif