Character map

[MihaZupan]

A quick benchmark on the dataset I use (https://github.com/dotnet/docs/tree/master/docs/core) shows a 35.9 > 34.7 ms change, so ~ 3-3,5%.

The codegen diff for IndexOfOpeningCharacter is

IndexOfOpeningCharacter(System.String, Int32, Int32)
    L0000: push rdi
    L0001: push rsi
    L0002: push rbp
    L0003: push rbx
    L0004: sub rsp, 0x48
    L0008: vzeroupper
    L000b: mov rsi, rcx
    L000e: lea rdi, [rsp+0x20]
    L0013: mov ecx, 0xa
    L0018: xor eax, eax
    L001a: rep stosd
    L001c: mov rcx, rsi
    L001f: mov rax, 0xb55cfaa848e5
    L0029: mov [rsp+0x40], rax
    L002e: mov esi, r9d
    L0031: mov eax, r8d
    L0034: mov rdi, rcx
    L0037: lea rcx, [rdi+0x20]
    L003b: vmovdqu xmm0, [rcx]
    L003f: vmovdqu [rsp+0x20], xmm0
    L0045: test rdx, rdx
    L0048: jnz L004e
    L004a: xor ebx, ebx
    L004c: jmp L005c
    L004e: add rdx, 0xc
    L0052: mov [rsp+0x38], rdx
    L0057: mov rbx, [rsp+0x38]
    L005c: cmp qword [rdi+0x10], 0x0
    L0061: jnz L00ba
    L0063: cmp eax, esi
    L0065: jg L010b
    L006b: movsxd rcx, eax
    L006e: movzx ecx, word [rbx+rcx*2]
    L0072: cmp ecx, 0x80
    L0078: jge L0092
    L007a: lea rdx, [rsp+0x20]
    L007f: mov r8d, ecx
    L0082: sar r8d, 0x5
    L0086: movsxd r8, r8d
    L0089: mov edx, [rdx+r8*4]
    L008d: bt edx, ecx
    L0090: jb L009a
    L0092: inc eax
    L0094: cmp eax, esi
    L0096: jle L006b
    L0098: jmp L010b
    L009a: mov rcx, 0xb55cfaa848e5
    L00a4: cmp [rsp+0x40], rcx
    L00a9: jz L00b0
    L00ab: call 0x7ffa70dc0780
    L00b0: nop
    L00b1: add rsp, 0x48
    L00b5: pop rbx
    L00b6: pop rbp
    L00b7: pop rsi
    L00b8: pop rdi
    L00b9: ret
    L00ba: mov ebp, eax
    L00bc: cmp ebp, esi
    L00be: jg L010b
    L00c0: movsxd rcx, ebp
    L00c3: movzx edx, word [rbx+rcx*2]
    L00c7: cmp edx, 0x80
    L00cd: jl L00e8
    L00cf: mov rcx, [rdi+0x10]
    L00d3: mov eax, [rcx]
    L00d5: call System.Collections.Generic.Dictionary`2[[System.Char, System.Private.CoreLib],[System.__Canon, System.Private.CoreLib]].FindEntry(Char)
    L00da: test eax, eax
    L00dc: setge al
    L00df: movzx eax, al
    L00e2: test eax, eax
    L00e4: jz L0105
    L00e6: jmp L0137
    L00e8: lea rax, [rsp+0x20]
    L00ed: mov ecx, edx
    L00ef: sar ecx, 0x5
    L00f2: movsxd rcx, ecx
    L00f5: mov eax, [rax+rcx*4]
    L00f8: bt eax, edx
    L00fb: setb al
    L00fe: movzx eax, al
    L0101: test eax, eax
    L0103: jnz L0137
    L0105: inc ebp
    L0107: cmp ebp, esi
    L0109: jle L00c0
    L010b: xor eax, eax
    L010d: mov [rsp+0x38], rax
    L0112: mov eax, 0xffffffff
    L0117: mov rcx, 0xb55cfaa848e5
    L0121: cmp [rsp+0x40], rcx
    L0126: jz L012d
    L0128: call 0x7ffa70dc0780
    L012d: nop
    L012e: add rsp, 0x48
    L0132: pop rbx
    L0133: pop rbp
    L0134: pop rsi
    L0135: pop rdi
    L0136: ret
    L0137: mov eax, ebp
    L0139: mov rcx, 0xb55cfaa848e5
    L0143: cmp [rsp+0x40], rcx
    L0148: jz L014f
    L014a: call 0x7ffa70dc0780
    L014f: nop
    L0150: add rsp, 0x48
    L0154: pop rbx
    L0155: pop rbp
    L0156: pop rsi
    L0157: pop rdi
    L0158: ret

PR

IndexOfOpeningCharacter(System.String, Int32, Int32)
    L0000: cmp qword [rcx+0x10], 0x0
    L0005: jnz L0041
    L0007: mov eax, [rdx]
    L0009: add rdx, 0xc
    L000d: cmp r8d, r9d
    L0010: jg L0037
    L0012: movsxd rax, r8d
    L0015: movzx eax, word [rdx+rax*2]
    L0019: cmp rax, 0x80
    L001f: jae L002f
    L0021: lea r10, [rcx+0x20]
    L0025: cmp [r10], r10d
    L0028: cmp byte [r10+rax], 0x0
    L002d: jnz L003d
    L002f: inc r8d
    L0032: cmp r8d, r9d
    L0035: jle L0012
    L0037: mov eax, 0xffffffff
    L003c: ret
    L003d: mov eax, r8d
    L0040: ret
    L0041: mov rax, 0x7ffa18c500f8
    L004b: jmp rax

IndexOfOpeningCharacterNonAscii(System.String, Int32, Int32)
    L0000: push rdi
    L0001: push rsi
    L0002: push rbp
    L0003: push rbx
    L0004: sub rsp, 0x28
    L0008: mov rdi, rcx
    L000b: mov esi, r9d
    L000e: mov ecx, [rdx]
    L0010: add rdx, 0xc
    L0014: mov rbx, rdx
    L0017: mov ebp, r8d
    L001a: cmp ebp, esi
    L001c: jg L005c
    L001e: movsxd rcx, ebp
    L0021: movzx edx, word [rbx+rcx*2]
    L0025: cmp edx, 0x80
    L002b: jl L0046
    L002d: mov rcx, [rdi+0x10]
    L0031: mov eax, [rcx]
    L0033: call System.Collections.Generic.Dictionary`2[[System.UInt32, System.Private.CoreLib],[System.__Canon, System.Private.CoreLib]].FindEntry(UInt32)
    L0038: test eax, eax
    L003a: setge al
    L003d: movzx eax, al
    L0040: test eax, eax
    L0042: jz L0056
    L0044: jmp L006a
    L0046: lea rax, [rdi+0x20]
    L004a: cmp [rax], eax
    L004c: mov edx, edx
    L004e: movzx eax, byte [rax+rdx]
    L0052: test eax, eax
    L0054: jnz L006a
    L0056: inc ebp
    L0058: cmp ebp, esi
    L005a: jle L001e
    L005c: mov eax, 0xffffffff
    L0061: add rsp, 0x28
    L0065: pop rbx
    L0066: pop rbp
    L0067: pop rsi
    L0068: pop rdi
    L0069: ret
    L006a: mov eax, ebp
    L006c: add rsp, 0x28
    L0070: pop rbx
    L0071: pop rbp
    L0072: pop rsi
    L0073: pop rdi
    L0074: ret