XeLabs/sonic
2
0
Fork 0
mirror of https://github.com/ii64/sonic.git synced 2026-06-25 02:46:43 +08:00
Code Issues Projects Releases Wiki Activity

fix: check prefix in native float parsing (#86)

Browse source
This commit is contained in:
liu 2021-09-06 16:01:42 +08:00 committed by GitHub
parent fa5f5026d9
commit 365da9c1a8
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
8 changed files with 629 additions and 635 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

518
internal/native/avx/native_amd64.s
View file

@ -283,7 +283,7 @@ LBB2_8:
LONG $0x4fdd6941; WORD $0x1293; BYTE $0x00 // imull $1217359, %r13d, %ebx
MOVQ R13, AX
SHLQ $4, AX
LONG $0x9f0d8d48; WORD $0x0070; BYTE $0x00 // leaq $28831(%rip), %rcx /* _DOUBLE_POW5_INV_SPLIT(%rip) */
LONG $0x8b0d8d48; WORD $0x0070; BYTE $0x00 // leaq $28811(%rip), %rcx /* _DOUBLE_POW5_INV_SPLIT(%rip) */
MOVQ R8, DI
ORQ $2, DI
MOVQ 0(AX)(CX*1), R10
@ -371,7 +371,7 @@ LBB2_15:
SHRL $19, DI
MOVLQSX AX, SI
SHLQ $4, SI
LONG $0xc70d8d4c; WORD $0x0084; BYTE $0x00 // leaq $33991(%rip), %r9 /* _DOUBLE_POW5_SPLIT(%rip) */
LONG $0xb30d8d4c; WORD $0x0084; BYTE $0x00 // leaq $33971(%rip), %r9 /* _DOUBLE_POW5_SPLIT(%rip) */
MOVQ R8, BX
ORQ $2, BX
MOVQ 0(SI)(R9*1), R15
@ -774,7 +774,7 @@ LBB2_65:
LEAQ 1(R13), BX
MOVQ BX, SI
MOVL R15, DX
LONG $0x0039dae8; BYTE $0x00 // callq _print_mantissa
LONG $0x0039a8e8; BYTE $0x00 // callq _print_mantissa
MOVB 1(R13), AX
MOVB AX, 0(R13)
MOVL $1, CX
@ -803,7 +803,7 @@ LBB2_70:
LEAL 0(DX)(DX*1), CX
LEAL 0(CX)(CX*4), CX
SUBL CX, AX
LONG $0xbd0d8d48; WORD $0x0093; BYTE $0x00 // leaq $37821(%rip), %rcx /* _Digits(%rip) */
LONG $0xa90d8d48; WORD $0x0093; BYTE $0x00 // leaq $37801(%rip), %rcx /* _Digits(%rip) */
MOVWLZX 0(CX)(DX*2), CX
MOVL BX, DX
MOVW CX, 0(R13)(DX*1)
@ -827,7 +827,7 @@ LBB2_73:
MOVL BX, SI
ADDQ R13, SI
MOVL R15, DX
LONG $0x003936e8; BYTE $0x00 // callq _print_mantissa
LONG $0x003904e8; BYTE $0x00 // callq _print_mantissa
TESTL R12, R12
JE LBB2_84
LEAL 0(R12)(BX*1), AX
@ -858,7 +858,7 @@ LBB2_78:
CMPL AX, $10
JL LBB2_86
WORD $0x9848 // cltq
LONG $0x140d8d48; WORD $0x0093; BYTE $0x00 // leaq $37652(%rip), %rcx /* _Digits(%rip) */
LONG $0x000d8d48; WORD $0x0093; BYTE $0x00 // leaq $37632(%rip), %rcx /* _Digits(%rip) */
MOVWLZX 0(CX)(AX*2), AX
MOVL BX, CX
MOVW AX, 0(R13)(CX*1)
@ -1060,7 +1060,7 @@ LBB2_109:
ADDQ R13, SI
MOVL R15, DX
WORD $0xf8c5; BYTE $0x77 // vzeroupper
LONG $0x0035bce8; BYTE $0x00 // callq _print_mantissa
LONG $0x00358ae8; BYTE $0x00 // callq _print_mantissa
ADDL BX, R15
MOVL R15, BX
@ -1142,7 +1142,7 @@ _u64toa:
ADDQ AX, AX
CMPL SI, $1000
JB LBB4_3
LONG $0x370d8d48; WORD $0x008f; BYTE $0x00 // leaq $36663(%rip), %rcx /* _Digits(%rip) */
LONG $0x230d8d48; WORD $0x008f; BYTE $0x00 // leaq $36643(%rip), %rcx /* _Digits(%rip) */
MOVB 0(DX)(CX*1), CX
MOVB CX, 0(DI)
MOVL $1, CX
@ -1156,14 +1156,14 @@ LBB4_3:
LBB4_4:
MOVWLZX DX, DX
ORQ $1, DX
LONG $0x16358d48; WORD $0x008f; BYTE $0x00 // leaq $36630(%rip), %rsi /* _Digits(%rip) */
LONG $0x02358d48; WORD $0x008f; BYTE $0x00 // leaq $36610(%rip), %rsi /* _Digits(%rip) */
MOVB 0(DX)(SI*1), DX
MOVL CX, SI
INCL CX
MOVB DX, 0(DI)(SI*1)
LBB4_6:
LONG $0x05158d48; WORD $0x008f; BYTE $0x00 // leaq $36613(%rip), %rdx /* _Digits(%rip) */
LONG $0xf1158d48; WORD $0x008e; BYTE $0x00 // leaq $36593(%rip), %rdx /* _Digits(%rip) */
MOVB 0(AX)(DX*1), DX
MOVL CX, SI
INCL CX
@ -1172,7 +1172,7 @@ LBB4_6:
LBB4_7:
MOVWLZX AX, AX
ORQ $1, AX
LONG $0xed158d48; WORD $0x008e; BYTE $0x00 // leaq $36589(%rip), %rdx /* _Digits(%rip) */
LONG $0xd9158d48; WORD $0x008e; BYTE $0x00 // leaq $36569(%rip), %rdx /* _Digits(%rip) */
MOVB 0(AX)(DX*1), AX
MOVL CX, DX
INCL CX
@ -1219,7 +1219,7 @@ LBB4_8:
ADDQ R11, R11
CMPL SI, $10000000
JB LBB4_11
LONG $0x56058d48; WORD $0x008e; BYTE $0x00 // leaq $36438(%rip), %rax /* _Digits(%rip) */
LONG $0x42058d48; WORD $0x008e; BYTE $0x00 // leaq $36418(%rip), %rax /* _Digits(%rip) */
MOVB 0(R10)(AX*1), AX
MOVB AX, 0(DI)
MOVL $1, CX
@ -1233,14 +1233,14 @@ LBB4_11:
LBB4_12:
MOVL R10, AX
ORQ $1, AX
LONG $0x31358d48; WORD $0x008e; BYTE $0x00 // leaq $36401(%rip), %rsi /* _Digits(%rip) */
LONG $0x1d358d48; WORD $0x008e; BYTE $0x00 // leaq $36381(%rip), %rsi /* _Digits(%rip) */
MOVB 0(AX)(SI*1), AX
MOVL CX, SI
INCL CX
MOVB AX, 0(DI)(SI*1)
LBB4_14:
LONG $0x20058d48; WORD $0x008e; BYTE $0x00 // leaq $36384(%rip), %rax /* _Digits(%rip) */
LONG $0x0c058d48; WORD $0x008e; BYTE $0x00 // leaq $36364(%rip), %rax /* _Digits(%rip) */
MOVB 0(R9)(AX*1), AX
MOVL CX, SI
INCL CX
@ -1249,7 +1249,7 @@ LBB4_14:
LBB4_15:
MOVWLZX R9, AX
ORQ $1, AX
LONG $0x06358d48; WORD $0x008e; BYTE $0x00 // leaq $36358(%rip), %rsi /* _Digits(%rip) */
LONG $0xf2358d48; WORD $0x008d; BYTE $0x00 // leaq $36338(%rip), %rsi /* _Digits(%rip) */
MOVB 0(AX)(SI*1), AX
MOVL CX, DX
MOVB AX, 0(DI)(DX*1)
@ -1331,7 +1331,7 @@ LBB4_16:
MOVL $16, CX
SUBL AX, CX
SHLQ $4, AX
LONG $0x7b158d48; WORD $0x008d; BYTE $0x00 // leaq $36219(%rip), %rdx /* _VecShiftShuffles(%rip) */
LONG $0x67158d48; WORD $0x008d; BYTE $0x00 // leaq $36199(%rip), %rdx /* _VecShiftShuffles(%rip) */
LONG $0x0071e2c4; WORD $0x1004 // vpshufb (%rax,%rdx), %xmm1, %xmm0
LONG $0x077ffac5 // vmovdqu %xmm0, (%rdi)
MOVL CX, AX
@ -1357,7 +1357,7 @@ LBB4_20:
CMPL DX, $99
JA LBB4_22
MOVL DX, AX
LONG $0x5e0d8d48; WORD $0x008c; BYTE $0x00 // leaq $35934(%rip), %rcx /* _Digits(%rip) */
LONG $0x4a0d8d48; WORD $0x008c; BYTE $0x00 // leaq $35914(%rip), %rcx /* _Digits(%rip) */
MOVB 0(CX)(AX*2), DX
MOVB 1(CX)(AX*2), AX
MOVB DX, 0(DI)
@ -1382,7 +1382,7 @@ LBB4_22:
WORD $0xc96b; BYTE $0x64 // imull $100, %ecx, %ecx
SUBL CX, AX
MOVWLZX AX, AX
LONG $0x0d0d8d48; WORD $0x008c; BYTE $0x00 // leaq $35853(%rip), %rcx /* _Digits(%rip) */
LONG $0xf90d8d48; WORD $0x008b; BYTE $0x00 // leaq $35833(%rip), %rcx /* _Digits(%rip) */
MOVB 0(CX)(AX*2), DX
MOVB 1(CX)(AX*2), AX
MOVB DX, 1(DI)
@ -1394,7 +1394,7 @@ LBB4_24:
WORD $0xc86b; BYTE $0x64 // imull $100, %eax, %ecx
SUBL CX, DX
MOVWLZX AX, AX
LONG $0xea058d4c; WORD $0x008b; BYTE $0x00 // leaq $35818(%rip), %r8 /* _Digits(%rip) */
LONG $0xd6058d4c; WORD $0x008b; BYTE $0x00 // leaq $35798(%rip), %r8 /* _Digits(%rip) */
MOVB 0(R8)(AX*2), CX
MOVB 1(R8)(AX*2), AX
MOVB CX, 0(DI)
@ -1480,8 +1480,8 @@ _quote:
MOVQ CX, R10
MOVQ DX, CX
TESTB $1, R8
LONG $0xe9058d48; WORD $0x008b; BYTE $0x00 // leaq $35817(%rip), %rax /* __SingleQuoteTab(%rip) */
LONG $0xe2158d48; WORD $0x009b; BYTE $0x00 // leaq $39906(%rip), %rdx /* __DoubleQuoteTab(%rip) */
LONG $0xd5058d48; WORD $0x008b; BYTE $0x00 // leaq $35797(%rip), %rax /* __SingleQuoteTab(%rip) */
LONG $0xce158d48; WORD $0x009b; BYTE $0x00 // leaq $39886(%rip), %rdx /* __DoubleQuoteTab(%rip) */
LONG $0xd0440f48 // cmoveq %rax, %rdx
MOVQ CX, R8
MOVQ DI, AX
@ -1589,7 +1589,7 @@ LBB5_14:
TESTQ R10, R10
MOVQ -48(BP), CX
MOVQ -56(BP), DI
LONG $0x6e2d8d4c; WORD $0x008a; BYTE $0x00 // leaq $35438(%rip), %r13 /* __SingleQuoteTab(%rip) */
LONG $0x5a2d8d4c; WORD $0x008a; BYTE $0x00 // leaq $35418(%rip), %r13 /* __SingleQuoteTab(%rip) */
JLE LBB5_29
LBB5_16:
@ -1851,7 +1851,7 @@ _unquote:
MOVQ R8, -56(BP)
MOVL R8, R10
ANDL $1, R10
LONG $0xdb058d4c; WORD $0x00a7; BYTE $0x00 // leaq $42971(%rip), %r8 /* __UnquoteTab(%rip) */
LONG $0xc7058d4c; WORD $0x00a7; BYTE $0x00 // leaq $42951(%rip), %r8 /* __UnquoteTab(%rip) */
QUAD $0xffffffb5056ffac5 // vmovdqu $-75(%rip), %xmm0 /* LCPI6_0(%rip) */
MOVQ DI, R9
MOVQ SI, R14
@ -2367,64 +2367,61 @@ _atof_eisel_lemire64:
WORD $0x5441 // pushq %r12
BYTE $0x53 // pushq %rbx
BYTE $0x50 // pushq %rax
MOVL SI, R13
LEAL 348(R13), R14
XORL R12, R12
CMPL R14, $695
JA LBB7_10
MOVL SI, R12
LEAL 348(R12), BX
CMPL BX, $695
JA LBB7_1
MOVQ CX, R14
MOVL DX, R15
MOVQ DI, BX
MOVQ CX, -48(BP)
LONG $0x002771e8; BYTE $0x00 // callq _count_len_u64
MOVQ DI, R13
LONG $0x002743e8; BYTE $0x00 // callq _count_len_u64
MOVL AX, R8
MOVL AX, CX
NEGL CX
SHLQ CX, BX
MOVL R14, CX
SHLQ CX, R13
MOVL BX, CX
SHLQ $4, CX
LONG $0x970d8d4c; WORD $0x002a; BYTE $0x00 // leaq $10903(%rip), %r9 /* _POW10_M128_TAB(%rip) */
MOVQ BX, AX
MULQ 8(CX)(R9*1)
MOVQ AX, R10
LONG $0x88158d4c; WORD $0x002a; BYTE $0x00 // leaq $10888(%rip), %r10 /* _POW10_M128_TAB(%rip) */
MOVQ R13, AX
MULQ 8(CX)(R10*1)
MOVQ AX, R9
MOVQ DX, DI
ANDL $511, DX
MOVQ BX, CX
NOTQ CX
CMPQ AX, CX
JBE LBB7_6
MOVQ R13, SI
NOTQ SI
CMPQ AX, SI
JBE LBB7_8
CMPL DX, $511
JNE LBB7_6
MOVLQSX R13, SI
SHLQ $4, SI
MOVQ BX, AX
MULQ 5552(SI)(R9*1)
ADDQ DX, R10
JNE LBB7_8
MOVQ R13, AX
MULQ 0(CX)(R10*1)
ADDQ DX, R9
ADCQ $0, DI
MOVL DI, DX
ANDL $511, DX
CMPQ AX, CX
JBE LBB7_6
CMPQ R10, $-1
JNE LBB7_6
CMPQ AX, SI
JBE LBB7_8
CMPQ R9, $-1
JNE LBB7_8
CMPL DX, $511
JE LBB7_10
JE LBB7_1
LBB7_6:
LBB7_8:
MOVQ DI, AX
SHRQ $63, AX
LEAL 9(AX), CX
SHRQ CX, DI
ORQ R10, DX
JNE LBB7_8
ORQ R9, DX
JNE LBB7_11
MOVL DI, CX
ANDL $3, CX
CMPL CX, $1
JE LBB7_10
JE LBB7_1
LBB7_8:
LBB7_11:
MOVL $64, CX
SUBL R8, CX
LONG $0x6af56941; WORD $0x0352; BYTE $0x00 // imull $217706, %r13d, %esi
LONG $0x6af46941; WORD $0x0352; BYTE $0x00 // imull $217706, %r12d, %esi
SARL $16, SI
MOVLQSX CX, BX
MOVL DI, DX
@ -2440,7 +2437,22 @@ LBB7_8:
SBBQ $-1, AX
LEAQ -1(AX), SI
CMPQ SI, $2045
JA LBB7_10
JBE LBB7_13
LBB7_1:
XORL AX, AX
LBB7_14:
ADDQ $8, SP
BYTE $0x5b // popq %rbx
WORD $0x5c41 // popq %r12
WORD $0x5d41 // popq %r13
WORD $0x5e41 // popq %r14
WORD $0x5f41 // popq %r15
BYTE $0x5d // popq %rbp
RET
LBB7_13:
CMPQ CX, $1
MOVB $2, CX
SBBB $0, CX
@ -2453,20 +2465,9 @@ LBB7_8:
ORQ CX, AX
CMPL R15, $-1
LONG $0xc1450f48 // cmovneq %rcx, %rax
MOVQ -48(BP), CX
MOVQ AX, 0(CX)
MOVL $1, R12
LBB7_10:
MOVL R12, AX
ADDQ $8, SP
BYTE $0x5b // popq %rbx
WORD $0x5c41 // popq %r12
WORD $0x5d41 // popq %r13
WORD $0x5e41 // popq %r14
WORD $0x5f41 // popq %r15
BYTE $0x5d // popq %rbp
RET
MOVQ AX, 0(R14)
MOVB $1, AX
JMP LBB7_14
_decimal_to_f64:
BYTE $0x55 // pushq %rbp
@ -2502,7 +2503,7 @@ LBB8_5:
MOVQ BX, -48(BP)
JLE LBB8_12
XORL R12, R12
LONG $0x92358d4c; WORD $0x0054; BYTE $0x00 // leaq $21650(%rip), %r14 /* _POW_TAB(%rip) */
LONG $0x9d358d4c; WORD $0x0054; BYTE $0x00 // leaq $21661(%rip), %r14 /* _POW_TAB(%rip) */
JMP LBB8_8
LBB8_10:
@ -2514,7 +2515,7 @@ LBB8_10:
LBB8_11:
MOVQ R15, DI
MOVL BX, SI
LONG $0x00278ce8; BYTE $0x00 // callq _right_shift
LONG $0x00275ce8; BYTE $0x00 // callq _right_shift
LBB8_7:
ADDL BX, R12
@ -2531,7 +2532,7 @@ LBB8_8:
JMP LBB8_7
LBB8_12:
LONG $0x4b358d4c; WORD $0x0054; BYTE $0x00 // leaq $21579(%rip), %r14 /* _POW_TAB(%rip) */
LONG $0x56358d4c; WORD $0x0054; BYTE $0x00 // leaq $21590(%rip), %r14 /* _POW_TAB(%rip) */
JMP LBB8_14
LBB8_18:
@ -2542,7 +2543,7 @@ LBB8_18:
LBB8_20:
MOVQ R15, DI
MOVL BX, SI
LONG $0x0025c2e8; BYTE $0x00 // callq _left_shift
LONG $0x0025afe8; BYTE $0x00 // callq _left_shift
MOVL 804(R15), AX
LBB8_13:
@ -2582,7 +2583,7 @@ LBB8_21:
LBB8_25:
MOVQ R15, DI
MOVL $60, SI
LONG $0x0026dee8; BYTE $0x00 // callq _right_shift
LONG $0x0026aee8; BYTE $0x00 // callq _right_shift
ADDL $60, R12
CMPL R12, $-120
JL LBB8_25
@ -2608,7 +2609,7 @@ LBB8_31:
NEGL R12
MOVQ R15, DI
MOVL R12, SI
LONG $0x002698e8; BYTE $0x00 // callq _right_shift
LONG $0x002668e8; BYTE $0x00 // callq _right_shift
MOVL $-1022, R14
LBB8_32:
@ -2616,7 +2617,7 @@ LBB8_32:
JE LBB8_34
MOVQ R15, DI
MOVL $53, SI
LONG $0x0024f7e8; BYTE $0x00 // callq _left_shift
LONG $0x0024e4e8; BYTE $0x00 // callq _left_shift
LBB8_34:
MOVL 804(R15), AX
@ -2794,19 +2795,15 @@ _atof_native_decimal:
LONG $0x4511fcc5; BYTE $0xb8 // vmovups %ymm0, $-72(%rbp)
LONG $0xc057f8c5 // vxorps %xmm0, %xmm0, %xmm0
LONG $0x4511f8c5; BYTE $0xd8 // vmovups %xmm0, $-40(%rbp)
MOVB 0(DI), CX
MOVL $1, AX
CMPB CX, $43
JE LBB9_4
CMPB CX, $45
JNE LBB9_3
XORL AX, AX
CMPB 0(DI), $45
JNE LBB9_2
MOVL $1, -32(BP)
MOVL $1, AX
LBB9_4:
LBB9_2:
CMPL AX, SI
JGE LBB9_5
LBB9_6:
JGE LBB9_3
MOVL SI, R10
MOVL AX, AX
MOVLQSX SI, R11
@ -2814,9 +2811,9 @@ LBB9_6:
XORL R9, R9
XORL R14, R14
XORL R8, R8
JMP LBB9_7
JMP LBB9_5
LBB9_13:
LBB9_11:
MOVLQSX R14, CX
MOVB BX, -840(BP)(CX*1)
MOVL -40(BP), R9
@ -2824,129 +2821,124 @@ LBB9_13:
MOVL R9, -40(BP)
MOVL R9, R14
LBB9_17:
LBB9_15:
INCQ AX
CMPQ AX, R11
SETLT CX
CMPQ R10, AX
JE LBB9_18
JE LBB9_16
LBB9_7:
LBB9_5:
MOVBLZX 0(DI)(AX*1), BX
LEAL -48(BX), DX
CMPB DX, $9
JA LBB9_15
JA LBB9_13
CMPB BX, $48
JNE LBB9_12
JNE LBB9_10
TESTL R14, R14
JE LBB9_10
JE LBB9_8
CMPL R14, $800
JL LBB9_13
JMP LBB9_17
JL LBB9_11
JMP LBB9_15
LBB9_15:
LBB9_13:
CMPB BX, $46
JNE LBB9_19
JNE LBB9_17
MOVL R14, -36(BP)
MOVL $1, R8
JMP LBB9_17
LBB9_12:
CMPL R14, $799
JLE LBB9_13
MOVL $1, -28(BP)
JMP LBB9_17
JMP LBB9_15
LBB9_10:
CMPL R14, $799
JLE LBB9_11
MOVL $1, -28(BP)
JMP LBB9_15
LBB9_8:
DECL -36(BP)
XORL R14, R14
JMP LBB9_17
JMP LBB9_15
LBB9_3:
XORL AX, AX
CMPL AX, SI
JL LBB9_6
LBB9_5:
MOVL $0, -36(BP)
JMP LBB9_35
JMP LBB9_33
LBB9_18:
LBB9_16:
MOVL R10, AX
LBB9_19:
LBB9_17:
TESTL R8, R8
JE LBB9_20
JE LBB9_18
TESTB $1, CX
JNE LBB9_22
JMP LBB9_35
JNE LBB9_20
JMP LBB9_33
LBB9_20:
LBB9_18:
MOVL R9, -36(BP)
TESTB $1, CX
JE LBB9_35
JE LBB9_33
LBB9_22:
LBB9_20:
MOVL AX, DX
MOVB 0(DI)(DX*1), CX
ORB $32, CX
CMPB CX, $101
JNE LBB9_35
JNE LBB9_33
MOVB 1(DI)(DX*1), BX
CMPB BX, $45
JE LBB9_27
JE LBB9_25
MOVL $1, R8
CMPB BX, $43
JNE LBB9_25
JNE LBB9_23
ADDL $2, AX
JMP LBB9_28
JMP LBB9_26
LBB9_27:
LBB9_25:
ADDL $2, AX
MOVL $-1, R8
LBB9_28:
LBB9_26:
MOVL AX, DX
XORL AX, AX
CMPL DX, SI
JL LBB9_30
JMP LBB9_34
JL LBB9_28
JMP LBB9_32
LBB9_25:
LBB9_23:
INCQ DX
XORL AX, AX
CMPL DX, SI
JGE LBB9_34
JGE LBB9_32
LBB9_30:
LBB9_28:
MOVLQSX DX, AX
SUBL DX, SI
ADDQ AX, DI
XORL DX, DX
XORL AX, AX
LBB9_31:
LBB9_29:
CMPL AX, $9999
JG LBB9_34
JG LBB9_32
MOVBLZX 0(DI)(DX*1), BX
LEAL -48(BX), CX
CMPB CX, $9
JA LBB9_34
JA LBB9_32
LEAL 0(AX)(AX*4), AX
LEAL -48(BX)(AX*2), AX
INCQ DX
CMPL SI, DX
JNE LBB9_31
JNE LBB9_29
LBB9_34:
LBB9_32:
IMULL R8, AX
ADDL AX, -36(BP)
LBB9_35:
LBB9_33:
LEAQ -840(BP), DI
LEAQ -24(BP), SI
WORD $0xf8c5; BYTE $0x77 // vzeroupper
LONG $0xfffaa1e8; BYTE $0xff // callq _decimal_to_f64
LONG $0xfffab4e8; BYTE $0xff // callq _decimal_to_f64
LONG $0x4510fbc5; BYTE $0xe8 // vmovsd $-24(%rbp), %xmm0
ADDQ $832, SP
BYTE $0x5b // popq %rbx
@ -4117,7 +4109,7 @@ LBB13_66:
CMPL BX, $23
JL LBB13_75
LEAL -22(BX), AX
LONG $0xe60d8d48; WORD $0x014e; BYTE $0x00 // leaq $85734(%rip), %rcx /* _P10_TAB(%rip) */
LONG $0x040d8d48; WORD $0x014f; BYTE $0x00 // leaq $85764(%rip), %rcx /* _P10_TAB(%rip) */
LONG $0x0459fbc5; BYTE $0xc1 // vmulsd (%rcx,%rax,8), %xmm0, %xmm0
MOVL $22, AX
JMP LBB13_76
@ -4136,7 +4128,7 @@ LBB13_72:
CMPL BX, $-22
JB LBB13_79
NEGL BX
LONG $0xad058d48; WORD $0x014e; BYTE $0x00 // leaq $85677(%rip), %rax /* _P10_TAB(%rip) */
LONG $0xcb058d48; WORD $0x014e; BYTE $0x00 // leaq $85707(%rip), %rax /* _P10_TAB(%rip) */
LONG $0x045efbc5; BYTE $0xd8 // vdivsd (%rax,%rbx,8), %xmm0, %xmm0
LONG $0x4511fbc5; BYTE $0xc8 // vmovsd %xmm0, $-56(%rbp)
JMP LBB13_85
@ -4157,7 +4149,7 @@ LBB13_76:
LONG $0xc82ef9c5 // vucomisd %xmm0, %xmm1
JA LBB13_79
MOVL AX, AX
LONG $0x690d8d48; WORD $0x014e; BYTE $0x00 // leaq $85609(%rip), %rcx /* _P10_TAB(%rip) */
LONG $0x870d8d48; WORD $0x014e; BYTE $0x00 // leaq $85639(%rip), %rcx /* _P10_TAB(%rip) */
LONG $0x0459fbc5; BYTE $0xc1 // vmulsd (%rcx,%rax,8), %xmm0, %xmm0
LONG $0x4511fbc5; BYTE $0xc8 // vmovsd %xmm0, $-56(%rbp)
JMP LBB13_85
@ -4168,8 +4160,8 @@ LBB13_79:
MOVQ R12, DI
MOVL BX, SI
MOVL -60(BP), DX
LONG $0xffeb02e8; BYTE $0xff // callq _atof_eisel_lemire64
TESTL AX, AX
LONG $0xffeb34e8; BYTE $0xff // callq _atof_eisel_lemire64
TESTB AX, AX
JE LBB13_83
CMPL -44(BP), $0
JE LBB13_84
@ -4178,8 +4170,8 @@ LBB13_79:
MOVQ R12, DI
MOVL BX, SI
MOVL -60(BP), DX
LONG $0xffeae4e8; BYTE $0xff // callq _atof_eisel_lemire64
TESTL AX, AX
LONG $0xffeb16e8; BYTE $0xff // callq _atof_eisel_lemire64
TESTB AX, AX
JE LBB13_83
LONG $0x4d10fbc5; BYTE $0xb8 // vmovsd $-72(%rbp), %xmm1
LONG $0x4510fbc5; BYTE $0xc8 // vmovsd $-56(%rbp), %xmm0
@ -4194,7 +4186,7 @@ LBB13_83:
ADDQ AX, DI
MOVL 8(CX), SI
SUBL AX, SI
LONG $0xffef40e8; BYTE $0xff // callq _atof_native_decimal
LONG $0xffef53e8; BYTE $0xff // callq _atof_native_decimal
JMP LBB13_85
LBB13_84:
@ -5687,7 +5679,7 @@ _print_mantissa:
LONG $0x64fa6b41 // imull $100, %r10d, %edi
SUBL DI, AX
MOVWLZX AX, R11
LONG $0x9e3d8d48; WORD $0x0059; BYTE $0x00 // leaq $22942(%rip), %rdi /* _Digits(%rip) */
LONG $0xbc3d8d48; WORD $0x0059; BYTE $0x00 // leaq $22972(%rip), %rdi /* _Digits(%rip) */
MOVWLZX 0(DI)(R8*2), AX
MOVW AX, -2(R14)
MOVWLZX 0(DI)(R9*2), AX
@ -5703,7 +5695,7 @@ LBB24_2:
CMPL DI, $10000
JB LBB24_3
MOVL $3518437209, R8
LONG $0x560d8d4c; WORD $0x0059; BYTE $0x00 // leaq $22870(%rip), %r9 /* _Digits(%rip) */
LONG $0x740d8d4c; WORD $0x0059; BYTE $0x00 // leaq $22900(%rip), %r9 /* _Digits(%rip) */
LBB24_5:
MOVL DI, AX
@ -5734,7 +5726,7 @@ LBB24_7:
WORD $0xd16b; BYTE $0x64 // imull $100, %ecx, %edx
SUBL DX, AX
MOVWLZX AX, AX
LONG $0xef158d48; WORD $0x0058; BYTE $0x00 // leaq $22767(%rip), %rdx /* _Digits(%rip) */
LONG $0x0d158d48; WORD $0x0059; BYTE $0x00 // leaq $22797(%rip), %rdx /* _Digits(%rip) */
MOVWLZX 0(DX)(AX*2), AX
MOVW AX, -2(R14)
ADDQ $-2, R14
@ -5744,7 +5736,7 @@ LBB24_8:
CMPL AX, $10
JB LBB24_10
MOVL AX, AX
LONG $0xd20d8d48; WORD $0x0058; BYTE $0x00 // leaq $22738(%rip), %rcx /* _Digits(%rip) */
LONG $0xf00d8d48; WORD $0x0058; BYTE $0x00 // leaq $22768(%rip), %rcx /* _Digits(%rip) */
MOVWLZX 0(CX)(AX*2), AX
MOVW AX, -2(R14)
BYTE $0x5b // popq %rbx
@ -5801,7 +5793,7 @@ LBB25_4:
RET
LBB25_6:
LONG $0xb70d8d48; WORD $0x007a; BYTE $0x00 // leaq $31415(%rip), %rcx /* _U8_LEN_TAB(%rip) */
LONG $0xd50d8d48; WORD $0x007a; BYTE $0x00 // leaq $31445(%rip), %rcx /* _U8_LEN_TAB(%rip) */
MOVBLZX 0(AX)(CX*1), AX
ADDL BX, AX
ADDQ $8, SP
@ -5812,64 +5804,62 @@ LBB25_6:
_left_shift:
BYTE $0x55 // pushq %rbp
WORD $0x8948; BYTE $0xe5 // movq %rsp, %rbp
WORD $0x5741 // pushq %r15
WORD $0x5641 // pushq %r14
BYTE $0x53 // pushq %rbx
MOVL SI, CX
IMUL3Q $804, CX, R10
LONG $0x951d8d4c; WORD $0x007b; BYTE $0x00 // leaq $31637(%rip), %r11 /* _LSHIFT_TAB(%rip) */
MOVL 0(R10)(R11*1), R8
IMUL3Q $804, CX, DX
LONG $0xaf358d48; WORD $0x007b; BYTE $0x00 // leaq $31663(%rip), %rsi /* _LSHIFT_TAB(%rip) */
MOVL 0(DX)(SI*1), R8
MOVLQSX 800(DI), R9
MOVB 4(DX)(SI*1), AX
TESTQ R9, R9
JE LBB26_1
LEAQ 4(R10)(R11*1), SI
XORL AX, AX
JE LBB26_6
LEAQ 5(DX)(SI*1), DX
XORL SI, SI
LBB26_3:
MOVBLZX 0(SI)(AX*1), DX
CMPB 0(DI)(AX*1), DX
JNE LBB26_8
INCQ AX
CMPQ R9, AX
TESTB AX, AX
JE LBB26_8
CMPB 0(DI)(SI*1), AX
JNE LBB26_5
MOVBLZX 0(DX)(SI*1), AX
INCQ SI
CMPQ R9, SI
JNE LBB26_3
LEAL -1(R9), AX
CMPL AX, $798
JA LBB26_10
MOVL R9, AX
JMP LBB26_7
LBB26_1:
XORL AX, AX
LBB26_6:
TESTB AX, AX
JE LBB26_8
LBB26_7:
ADDQ R10, R11
MOVB 4(AX)(R11*1), AX
ADDB $-48, AX
CMPB AX, $10
JAE LBB26_10
LBB26_9:
DECL R8
LBB26_10:
LBB26_8:
TESTL R9, R9
JLE LBB26_23
LEAL -1(R9)(R8*1), R10
JLE LBB26_22
LEAL 0(R8)(R9*1), AX
MOVLQSX AX, BX
MOVQ BX, R10
SHLQ $32, R10
INCQ R9
XORL DX, DX
MOVQ $-3689348814741910323, R11
JMP LBB26_12
MOVQ $-4294967296, R14
JMP LBB26_10
LBB26_13:
LBB26_11:
ADDB $48, AX
MOVLQSX R10, BX
MOVB AX, 0(DI)(BX*1)
DECL R10
MOVB AX, -1(DI)(BX*1)
LBB26_19:
LBB26_18:
DECQ BX
ADDQ R14, R10
DECQ R9
CMPQ R9, $1
JLE LBB26_14
JLE LBB26_12
LBB26_12:
LBB26_10:
MOVBQSX -2(DI)(R9*1), SI
ADDQ $-48, SI
SHLQ CX, SI
@ -5878,62 +5868,48 @@ LBB26_12:
MULQ R11
SHRQ $3, DX
LEAQ 0(DX)(DX*1), AX
LEAQ 0(AX)(AX*4), BX
LEAQ 0(AX)(AX*4), R15
MOVQ SI, AX
SUBQ BX, AX
CMPL R10, $799
JLE LBB26_13
SUBQ R15, AX
CMPQ BX, $800
JLE LBB26_11
TESTQ AX, AX
JE LBB26_19
JE LBB26_18
MOVL $1, 812(DI)
JMP LBB26_19
JMP LBB26_18
LBB26_14:
LBB26_12:
CMPQ SI, $10
JAE LBB26_15
LBB26_23:
MOVL 800(DI), CX
ADDL R8, CX
CMPL CX, $801
MOVL $800, AX
WORD $0x4c0f; BYTE $0xc1 // cmovll %ecx, %eax
MOVL AX, 800(DI)
ADDL R8, 804(DI)
TESTL CX, CX
JLE LBB26_27
MOVL AX, CX
INCQ CX
LBB26_25:
DECL AX
CMPB 0(DI)(AX*1), $48
JNE LBB26_29
MOVL AX, 800(DI)
DECQ CX
CMPQ CX, $1
JG LBB26_25
LBB26_27:
TESTL AX, AX
JE LBB26_28
LBB26_29:
BYTE $0x5b // popq %rbx
BYTE $0x5d // popq %rbp
RET
LBB26_16:
ADDB $48, AX
MOVLQSX R10, SI
MOVB AX, 0(DI)(SI*1)
DECL R10
JAE LBB26_13
LBB26_22:
CMPQ CX, $9
JBE LBB26_23
MOVL 800(DI), AX
ADDL R8, AX
CMPL AX, $801
MOVL $800, CX
WORD $0x4c0f; BYTE $0xc8 // cmovll %eax, %ecx
MOVL CX, 800(DI)
ADDL R8, 804(DI)
BYTE $0x5b // popq %rbx
WORD $0x5e41 // popq %r14
WORD $0x5f41 // popq %r15
BYTE $0x5d // popq %rbp
JMP _trim
LBB26_13:
SARQ $32, R10
JMP LBB26_14
LBB26_15:
ADDB $48, AX
MOVB AX, -1(DI)(R10*1)
LBB26_21:
DECQ R10
CMPQ CX, $9
JBE LBB26_22
LBB26_14:
MOVQ DX, CX
MOVQ DX, AX
MULQ R11
@ -5942,22 +5918,16 @@ LBB26_15:
LEAQ 0(AX)(AX*4), SI
MOVQ CX, AX
SUBQ SI, AX
CMPL R10, $799
JLE LBB26_16
CMPQ R10, $800
JLE LBB26_15
TESTQ AX, AX
JE LBB26_22
JE LBB26_21
MOVL $1, 812(DI)
JMP LBB26_22
JMP LBB26_21
LBB26_8:
JLE LBB26_9
JMP LBB26_10
LBB26_28:
MOVL $0, 804(DI)
BYTE $0x5b // popq %rbx
BYTE $0x5d // popq %rbp
RET
LBB26_5:
JL LBB26_7
JMP LBB26_8
_right_shift:
BYTE $0x55 // pushq %rbp
@ -6093,6 +6063,36 @@ LBB27_23:
BYTE $0x5d // popq %rbp
RET
_trim:
BYTE $0x55 // pushq %rbp
WORD $0x8948; BYTE $0xe5 // movq %rsp, %rbp
MOVL 800(DI), AX
TESTL AX, AX
JLE LBB28_4
LEAQ 1(AX), CX
LBB28_2:
DECL AX
CMPB 0(DI)(AX*1), $48
JNE LBB28_6
MOVL AX, 800(DI)
DECQ CX
CMPQ CX, $1
JG LBB28_2
LBB28_4:
TESTL AX, AX
JE LBB28_5
LBB28_6:
BYTE $0x5d // popq %rbp
RET
LBB28_5:
MOVL $0, 804(DI)
BYTE $0x5d // popq %rbp
RET
_POW10_M128_TAB:
QUAD $0x1732c869cd60e453 // .quad 1671618768450675795
QUAD $0xfa8fd5a0081c0288 // .quad -391859759250406776
@ -12666,7 +12666,7 @@ TEXT ·__skip_array(SB), NOSPLIT, $0 - 32
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ m+16(FP), DX
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+15938(SB) // _skip_array
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+15888(SB) // _skip_array
MOVQ AX, ret+24(FP)
RET
@ -12674,7 +12674,7 @@ TEXT ·__skip_object(SB), NOSPLIT, $0 - 32
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ m+16(FP), DX
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+15973(SB) // _skip_object
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+15923(SB) // _skip_object
MOVQ AX, ret+24(FP)
RET
@ -12682,7 +12682,7 @@ TEXT ·__skip_one(SB), NOSPLIT, $0 - 32
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ m+16(FP), DX
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+13616(SB) // _skip_one
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+13566(SB) // _skip_one
MOVQ AX, ret+24(FP)
RET
@ -12709,7 +12709,7 @@ TEXT ·__value(SB), NOSPLIT, $0 - 48
MOVQ p+16(FP), DX
MOVQ v+24(FP), CX
MOVQ allow_control+32(FP), R8
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+9329(SB) // _value
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+9279(SB) // _value
MOVQ AX, ret+40(FP)
RET
@ -12717,26 +12717,26 @@ TEXT ·__vnumber(SB), NOSPLIT, $0 - 24
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ v+16(FP), DX
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+11904(SB), AX // _vnumber
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+11854(SB), AX // _vnumber
JMP AX
TEXT ·__vsigned(SB), NOSPLIT, $0 - 24
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ v+16(FP), DX
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+13066(SB), AX // _vsigned
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+13016(SB), AX // _vsigned
JMP AX
TEXT ·__vstring(SB), NOSPLIT, $0 - 24
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ v+16(FP), DX
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+10935(SB), AX // _vstring
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+10885(SB), AX // _vstring
JMP AX
TEXT ·__vunsigned(SB), NOSPLIT, $0 - 24
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ v+16(FP), DX
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+13343(SB), AX // _vunsigned
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+13293(SB), AX // _vunsigned
JMP AX

16
internal/native/avx/native_subr_amd64.go
View file

@ -19,16 +19,16 @@ var (
_subr__lspace = **(**uintptr)(unsafe.Pointer(&_func__base)) + 238
_subr__lzero = **(**uintptr)(unsafe.Pointer(&_func__base)) + 0
_subr__quote = **(**uintptr)(unsafe.Pointer(&_func__base)) + 4854
_subr__skip_array = **(**uintptr)(unsafe.Pointer(&_func__base)) + 15938
_subr__skip_object = **(**uintptr)(unsafe.Pointer(&_func__base)) + 15973
_subr__skip_one = **(**uintptr)(unsafe.Pointer(&_func__base)) + 13616
_subr__skip_array = **(**uintptr)(unsafe.Pointer(&_func__base)) + 15888
_subr__skip_object = **(**uintptr)(unsafe.Pointer(&_func__base)) + 15923
_subr__skip_one = **(**uintptr)(unsafe.Pointer(&_func__base)) + 13566
_subr__u64toa = **(**uintptr)(unsafe.Pointer(&_func__base)) + 3634
_subr__unquote = **(**uintptr)(unsafe.Pointer(&_func__base)) + 5875
_subr__value = **(**uintptr)(unsafe.Pointer(&_func__base)) + 9329
_subr__vnumber = **(**uintptr)(unsafe.Pointer(&_func__base)) + 11904
_subr__vsigned = **(**uintptr)(unsafe.Pointer(&_func__base)) + 13066
_subr__vstring = **(**uintptr)(unsafe.Pointer(&_func__base)) + 10935
_subr__vunsigned = **(**uintptr)(unsafe.Pointer(&_func__base)) + 13343
_subr__value = **(**uintptr)(unsafe.Pointer(&_func__base)) + 9279
_subr__vnumber = **(**uintptr)(unsafe.Pointer(&_func__base)) + 11854
_subr__vsigned = **(**uintptr)(unsafe.Pointer(&_func__base)) + 13016
_subr__vstring = **(**uintptr)(unsafe.Pointer(&_func__base)) + 10885
_subr__vunsigned = **(**uintptr)(unsafe.Pointer(&_func__base)) + 13293
)
var (

518
internal/native/avx2/native_amd64.s
View file

@ -335,7 +335,7 @@ LBB2_8:
LONG $0x4fdd6941; WORD $0x1293; BYTE $0x00 // imull $1217359, %r13d, %ebx
MOVQ R13, AX
SHLQ $4, AX
LONG $0x450d8d48; WORD $0x007b; BYTE $0x00 // leaq $31557(%rip), %rcx /* _DOUBLE_POW5_INV_SPLIT(%rip) */
LONG $0x310d8d48; WORD $0x007b; BYTE $0x00 // leaq $31537(%rip), %rcx /* _DOUBLE_POW5_INV_SPLIT(%rip) */
MOVQ R8, DI
ORQ $2, DI
MOVQ 0(AX)(CX*1), R10
@ -423,7 +423,7 @@ LBB2_15:
SHRL $19, DI
MOVLQSX AX, SI
SHLQ $4, SI
LONG $0x6d0d8d4c; WORD $0x008f; BYTE $0x00 // leaq $36717(%rip), %r9 /* _DOUBLE_POW5_SPLIT(%rip) */
LONG $0x590d8d4c; WORD $0x008f; BYTE $0x00 // leaq $36697(%rip), %r9 /* _DOUBLE_POW5_SPLIT(%rip) */
MOVQ R8, BX
ORQ $2, BX
MOVQ 0(SI)(R9*1), R15
@ -826,7 +826,7 @@ LBB2_65:
LEAQ 1(R13), BX
MOVQ BX, SI
MOVL R15, DX
LONG $0x004480e8; BYTE $0x00 // callq _print_mantissa
LONG $0x00444ee8; BYTE $0x00 // callq _print_mantissa
MOVB 1(R13), AX
MOVB AX, 0(R13)
MOVL $1, CX
@ -855,7 +855,7 @@ LBB2_70:
LEAL 0(DX)(DX*1), CX
LEAL 0(CX)(CX*4), CX
SUBL CX, AX
LONG $0x630d8d48; WORD $0x009e; BYTE $0x00 // leaq $40547(%rip), %rcx /* _Digits(%rip) */
LONG $0x4f0d8d48; WORD $0x009e; BYTE $0x00 // leaq $40527(%rip), %rcx /* _Digits(%rip) */
MOVWLZX 0(CX)(DX*2), CX
MOVL BX, DX
MOVW CX, 0(R13)(DX*1)
@ -879,7 +879,7 @@ LBB2_73:
MOVL BX, SI
ADDQ R13, SI
MOVL R15, DX
LONG $0x0043dce8; BYTE $0x00 // callq _print_mantissa
LONG $0x0043aae8; BYTE $0x00 // callq _print_mantissa
TESTL R12, R12
JE LBB2_84
LEAL 0(R12)(BX*1), AX
@ -910,7 +910,7 @@ LBB2_78:
CMPL AX, $10
JL LBB2_86
WORD $0x9848 // cltq
LONG $0xba0d8d48; WORD $0x009d; BYTE $0x00 // leaq $40378(%rip), %rcx /* _Digits(%rip) */
LONG $0xa60d8d48; WORD $0x009d; BYTE $0x00 // leaq $40358(%rip), %rcx /* _Digits(%rip) */
MOVWLZX 0(CX)(AX*2), AX
MOVL BX, CX
MOVW AX, 0(R13)(CX*1)
@ -1112,7 +1112,7 @@ LBB2_109:
ADDQ R13, SI
MOVL R15, DX
WORD $0xf8c5; BYTE $0x77 // vzeroupper
LONG $0x004062e8; BYTE $0x00 // callq _print_mantissa
LONG $0x004030e8; BYTE $0x00 // callq _print_mantissa
ADDL BX, R15
MOVL R15, BX
@ -1194,7 +1194,7 @@ _u64toa:
ADDQ AX, AX
CMPL SI, $1000
JB LBB4_3
LONG $0xdd0d8d48; WORD $0x0099; BYTE $0x00 // leaq $39389(%rip), %rcx /* _Digits(%rip) */
LONG $0xc90d8d48; WORD $0x0099; BYTE $0x00 // leaq $39369(%rip), %rcx /* _Digits(%rip) */
MOVB 0(DX)(CX*1), CX
MOVB CX, 0(DI)
MOVL $1, CX
@ -1208,14 +1208,14 @@ LBB4_3:
LBB4_4:
MOVWLZX DX, DX
ORQ $1, DX
LONG $0xbc358d48; WORD $0x0099; BYTE $0x00 // leaq $39356(%rip), %rsi /* _Digits(%rip) */
LONG $0xa8358d48; WORD $0x0099; BYTE $0x00 // leaq $39336(%rip), %rsi /* _Digits(%rip) */
MOVB 0(DX)(SI*1), DX
MOVL CX, SI
INCL CX
MOVB DX, 0(DI)(SI*1)
LBB4_6:
LONG $0xab158d48; WORD $0x0099; BYTE $0x00 // leaq $39339(%rip), %rdx /* _Digits(%rip) */
LONG $0x97158d48; WORD $0x0099; BYTE $0x00 // leaq $39319(%rip), %rdx /* _Digits(%rip) */
MOVB 0(AX)(DX*1), DX
MOVL CX, SI
INCL CX
@ -1224,7 +1224,7 @@ LBB4_6:
LBB4_7:
MOVWLZX AX, AX
ORQ $1, AX
LONG $0x93158d48; WORD $0x0099; BYTE $0x00 // leaq $39315(%rip), %rdx /* _Digits(%rip) */
LONG $0x7f158d48; WORD $0x0099; BYTE $0x00 // leaq $39295(%rip), %rdx /* _Digits(%rip) */
MOVB 0(AX)(DX*1), AX
MOVL CX, DX
INCL CX
@ -1271,7 +1271,7 @@ LBB4_8:
ADDQ R11, R11
CMPL SI, $10000000
JB LBB4_11
LONG $0xfc058d48; WORD $0x0098; BYTE $0x00 // leaq $39164(%rip), %rax /* _Digits(%rip) */
LONG $0xe8058d48; WORD $0x0098; BYTE $0x00 // leaq $39144(%rip), %rax /* _Digits(%rip) */
MOVB 0(R10)(AX*1), AX
MOVB AX, 0(DI)
MOVL $1, CX
@ -1285,14 +1285,14 @@ LBB4_11:
LBB4_12:
MOVL R10, AX
ORQ $1, AX
LONG $0xd7358d48; WORD $0x0098; BYTE $0x00 // leaq $39127(%rip), %rsi /* _Digits(%rip) */
LONG $0xc3358d48; WORD $0x0098; BYTE $0x00 // leaq $39107(%rip), %rsi /* _Digits(%rip) */
MOVB 0(AX)(SI*1), AX
MOVL CX, SI
INCL CX
MOVB AX, 0(DI)(SI*1)
LBB4_14:
LONG $0xc6058d48; WORD $0x0098; BYTE $0x00 // leaq $39110(%rip), %rax /* _Digits(%rip) */
LONG $0xb2058d48; WORD $0x0098; BYTE $0x00 // leaq $39090(%rip), %rax /* _Digits(%rip) */
MOVB 0(R9)(AX*1), AX
MOVL CX, SI
INCL CX
@ -1301,7 +1301,7 @@ LBB4_14:
LBB4_15:
MOVWLZX R9, AX
ORQ $1, AX
LONG $0xac358d48; WORD $0x0098; BYTE $0x00 // leaq $39084(%rip), %rsi /* _Digits(%rip) */
LONG $0x98358d48; WORD $0x0098; BYTE $0x00 // leaq $39064(%rip), %rsi /* _Digits(%rip) */
MOVB 0(AX)(SI*1), AX
MOVL CX, DX
MOVB AX, 0(DI)(DX*1)
@ -1383,7 +1383,7 @@ LBB4_16:
MOVL $16, CX
SUBL AX, CX
SHLQ $4, AX
LONG $0x1f158d48; WORD $0x0098; BYTE $0x00 // leaq $38943(%rip), %rdx /* _VecShiftShuffles(%rip) */
LONG $0x0b158d48; WORD $0x0098; BYTE $0x00 // leaq $38923(%rip), %rdx /* _VecShiftShuffles(%rip) */
LONG $0x0071e2c4; WORD $0x1004 // vpshufb (%rax,%rdx), %xmm1, %xmm0
LONG $0x077ffac5 // vmovdqu %xmm0, (%rdi)
MOVL CX, AX
@ -1409,7 +1409,7 @@ LBB4_20:
CMPL DX, $99
JA LBB4_22
MOVL DX, AX
LONG $0x020d8d48; WORD $0x0097; BYTE $0x00 // leaq $38658(%rip), %rcx /* _Digits(%rip) */
LONG $0xee0d8d48; WORD $0x0096; BYTE $0x00 // leaq $38638(%rip), %rcx /* _Digits(%rip) */
MOVB 0(CX)(AX*2), DX
MOVB 1(CX)(AX*2), AX
MOVB DX, 0(DI)
@ -1434,7 +1434,7 @@ LBB4_22:
WORD $0xc96b; BYTE $0x64 // imull $100, %ecx, %ecx
SUBL CX, AX
MOVWLZX AX, AX
LONG $0xb10d8d48; WORD $0x0096; BYTE $0x00 // leaq $38577(%rip), %rcx /* _Digits(%rip) */
LONG $0x9d0d8d48; WORD $0x0096; BYTE $0x00 // leaq $38557(%rip), %rcx /* _Digits(%rip) */
MOVB 0(CX)(AX*2), DX
MOVB 1(CX)(AX*2), AX
MOVB DX, 1(DI)
@ -1446,7 +1446,7 @@ LBB4_24:
WORD $0xc86b; BYTE $0x64 // imull $100, %eax, %ecx
SUBL CX, DX
MOVWLZX AX, AX
LONG $0x8e058d4c; WORD $0x0096; BYTE $0x00 // leaq $38542(%rip), %r8 /* _Digits(%rip) */
LONG $0x7a058d4c; WORD $0x0096; BYTE $0x00 // leaq $38522(%rip), %r8 /* _Digits(%rip) */
MOVB 0(R8)(AX*2), CX
MOVB 1(R8)(AX*2), AX
MOVB CX, 0(DI)
@ -1544,8 +1544,8 @@ _quote:
MOVQ CX, R14
MOVQ DI, R12
TESTB $1, R8
LONG $0x2b058d48; WORD $0x0096; BYTE $0x00 // leaq $38443(%rip), %rax /* __SingleQuoteTab(%rip) */
LONG $0x24158d4c; WORD $0x00a6; BYTE $0x00 // leaq $42532(%rip), %r10 /* __DoubleQuoteTab(%rip) */
LONG $0x17058d48; WORD $0x0096; BYTE $0x00 // leaq $38423(%rip), %rax /* __SingleQuoteTab(%rip) */
LONG $0x10158d4c; WORD $0x00a6; BYTE $0x00 // leaq $42512(%rip), %r10 /* __DoubleQuoteTab(%rip) */
LONG $0xd0440f4c // cmoveq %rax, %r10
MOVQ DX, R8
MOVQ DI, AX
@ -1770,7 +1770,7 @@ LBB5_26:
LBB5_27:
TESTQ CX, CX
MOVQ BX, R14
LONG $0xca0d8d4c; WORD $0x0092; BYTE $0x00 // leaq $37578(%rip), %r9 /* __SingleQuoteTab(%rip) */
LONG $0xb60d8d4c; WORD $0x0092; BYTE $0x00 // leaq $37558(%rip), %r9 /* __SingleQuoteTab(%rip) */
JLE LBB5_60
MOVQ SI, DX
TESTQ SI, SI
@ -2282,7 +2282,7 @@ LBB6_20:
LBB6_34:
ADDQ BX, AX
MOVBLZX -1(R9), CX
LONG $0x7c1d8d48; WORD $0x00ad; BYTE $0x00 // leaq $44412(%rip), %rbx /* __UnquoteTab(%rip) */
LONG $0x681d8d48; WORD $0x00ad; BYTE $0x00 // leaq $44392(%rip), %rbx /* __UnquoteTab(%rip) */
MOVB 0(CX)(BX*1), CX
CMPB CX, $-1
JE LBB6_38
@ -2693,64 +2693,61 @@ _atof_eisel_lemire64:
WORD $0x5441 // pushq %r12
BYTE $0x53 // pushq %rbx
BYTE $0x50 // pushq %rax
MOVL SI, R13
LEAL 348(R13), R14
XORL R12, R12
CMPL R14, $695
JA LBB7_10
MOVL SI, R12
LEAL 348(R12), BX
CMPL BX, $695
JA LBB7_1
MOVQ CX, R14
MOVL DX, R15
MOVQ DI, BX
MOVQ CX, -48(BP)
LONG $0x002e05e8; BYTE $0x00 // callq _count_len_u64
MOVQ DI, R13
LONG $0x002dd7e8; BYTE $0x00 // callq _count_len_u64
MOVL AX, R8
MOVL AX, CX
NEGL CX
SHLQ CX, BX
MOVL R14, CX
SHLQ CX, R13
MOVL BX, CX
SHLQ $4, CX
LONG $0x2b0d8d4c; WORD $0x0031; BYTE $0x00 // leaq $12587(%rip), %r9 /* _POW10_M128_TAB(%rip) */
MOVQ BX, AX
MULQ 8(CX)(R9*1)
MOVQ AX, R10
LONG $0x1c158d4c; WORD $0x0031; BYTE $0x00 // leaq $12572(%rip), %r10 /* _POW10_M128_TAB(%rip) */
MOVQ R13, AX
MULQ 8(CX)(R10*1)
MOVQ AX, R9
MOVQ DX, DI
ANDL $511, DX
MOVQ BX, CX
NOTQ CX
CMPQ AX, CX
JBE LBB7_6
MOVQ R13, SI
NOTQ SI
CMPQ AX, SI
JBE LBB7_8
CMPL DX, $511
JNE LBB7_6
MOVLQSX R13, SI
SHLQ $4, SI
MOVQ BX, AX
MULQ 5552(SI)(R9*1)
ADDQ DX, R10
JNE LBB7_8
MOVQ R13, AX
MULQ 0(CX)(R10*1)
ADDQ DX, R9
ADCQ $0, DI
MOVL DI, DX
ANDL $511, DX
CMPQ AX, CX
JBE LBB7_6
CMPQ R10, $-1
JNE LBB7_6
CMPQ AX, SI
JBE LBB7_8
CMPQ R9, $-1
JNE LBB7_8
CMPL DX, $511
JE LBB7_10
JE LBB7_1
LBB7_6:
LBB7_8:
MOVQ DI, AX
SHRQ $63, AX
LEAL 9(AX), CX
SHRQ CX, DI
ORQ R10, DX
JNE LBB7_8
ORQ R9, DX
JNE LBB7_11
MOVL DI, CX
ANDL $3, CX
CMPL CX, $1
JE LBB7_10
JE LBB7_1
LBB7_8:
LBB7_11:
MOVL $64, CX
SUBL R8, CX
LONG $0x6af56941; WORD $0x0352; BYTE $0x00 // imull $217706, %r13d, %esi
LONG $0x6af46941; WORD $0x0352; BYTE $0x00 // imull $217706, %r12d, %esi
SARL $16, SI
MOVLQSX CX, BX
MOVL DI, DX
@ -2766,7 +2763,22 @@ LBB7_8:
SBBQ $-1, AX
LEAQ -1(AX), SI
CMPQ SI, $2045
JA LBB7_10
JBE LBB7_13
LBB7_1:
XORL AX, AX
LBB7_14:
ADDQ $8, SP
BYTE $0x5b // popq %rbx
WORD $0x5c41 // popq %r12
WORD $0x5d41 // popq %r13
WORD $0x5e41 // popq %r14
WORD $0x5f41 // popq %r15
BYTE $0x5d // popq %rbp
RET
LBB7_13:
CMPQ CX, $1
MOVB $2, CX
SBBB $0, CX
@ -2779,20 +2791,9 @@ LBB7_8:
ORQ CX, AX
CMPL R15, $-1
LONG $0xc1450f48 // cmovneq %rcx, %rax
MOVQ -48(BP), CX
MOVQ AX, 0(CX)
MOVL $1, R12
LBB7_10:
MOVL R12, AX
ADDQ $8, SP
BYTE $0x5b // popq %rbx
WORD $0x5c41 // popq %r12
WORD $0x5d41 // popq %r13
WORD $0x5e41 // popq %r14
WORD $0x5f41 // popq %r15
BYTE $0x5d // popq %rbp
RET
MOVQ AX, 0(R14)
MOVB $1, AX
JMP LBB7_14
LCPI8_0:
QUAD $0x0000000000000000 // .space 8, '\x00\x00\x00\x00\x00\x00\x00\x00'
@ -2843,7 +2844,7 @@ LBB8_5:
MOVQ BX, -48(BP)
JLE LBB8_12
XORL R12, R12
LONG $0xee358d4c; WORD $0x005a; BYTE $0x00 // leaq $23278(%rip), %r14 /* _POW_TAB(%rip) */
LONG $0xf9358d4c; WORD $0x005a; BYTE $0x00 // leaq $23289(%rip), %r14 /* _POW_TAB(%rip) */
JMP LBB8_8
LBB8_10:
@ -2855,7 +2856,7 @@ LBB8_10:
LBB8_11:
MOVQ R15, DI
MOVL BX, SI
LONG $0x002de8e8; BYTE $0x00 // callq _right_shift
LONG $0x002db8e8; BYTE $0x00 // callq _right_shift
LBB8_7:
ADDL BX, R12
@ -2872,7 +2873,7 @@ LBB8_8:
JMP LBB8_7
LBB8_12:
LONG $0xa7358d4c; WORD $0x005a; BYTE $0x00 // leaq $23207(%rip), %r14 /* _POW_TAB(%rip) */
LONG $0xb2358d4c; WORD $0x005a; BYTE $0x00 // leaq $23218(%rip), %r14 /* _POW_TAB(%rip) */
JMP LBB8_14
LBB8_18:
@ -2883,7 +2884,7 @@ LBB8_18:
LBB8_20:
MOVQ R15, DI
MOVL BX, SI
LONG $0x002c1ee8; BYTE $0x00 // callq _left_shift
LONG $0x002c0be8; BYTE $0x00 // callq _left_shift
MOVL 804(R15), AX
LBB8_13:
@ -2923,7 +2924,7 @@ LBB8_21:
LBB8_25:
MOVQ R15, DI
MOVL $60, SI
LONG $0x002d3ae8; BYTE $0x00 // callq _right_shift
LONG $0x002d0ae8; BYTE $0x00 // callq _right_shift
ADDL $60, R12
CMPL R12, $-120
JL LBB8_25
@ -2949,7 +2950,7 @@ LBB8_31:
NEGL R12
MOVQ R15, DI
MOVL R12, SI
LONG $0x002cf4e8; BYTE $0x00 // callq _right_shift
LONG $0x002cc4e8; BYTE $0x00 // callq _right_shift
MOVL $-1022, R14
LBB8_32:
@ -2957,7 +2958,7 @@ LBB8_32:
JE LBB8_34
MOVQ R15, DI
MOVL $53, SI
LONG $0x002b53e8; BYTE $0x00 // callq _left_shift
LONG $0x002b40e8; BYTE $0x00 // callq _left_shift
LBB8_34:
MOVL 804(R15), AX
@ -3249,19 +3250,15 @@ _atof_native_decimal:
LONG $0x4511fcc5; BYTE $0xb8 // vmovups %ymm0, $-72(%rbp)
LONG $0xc057f8c5 // vxorps %xmm0, %xmm0, %xmm0
LONG $0x4511f8c5; BYTE $0xd8 // vmovups %xmm0, $-40(%rbp)
MOVB 0(DI), CX
MOVL $1, AX
CMPB CX, $43
JE LBB9_4
CMPB CX, $45
JNE LBB9_3
XORL AX, AX
CMPB 0(DI), $45
JNE LBB9_2
MOVL $1, -32(BP)
MOVL $1, AX
LBB9_4:
LBB9_2:
CMPL AX, SI
JGE LBB9_5
LBB9_6:
JGE LBB9_3
MOVL SI, R10
MOVL AX, AX
MOVLQSX SI, R11
@ -3269,9 +3266,9 @@ LBB9_6:
XORL R9, R9
XORL R14, R14
XORL R8, R8
JMP LBB9_7
JMP LBB9_5
LBB9_13:
LBB9_11:
MOVLQSX R14, CX
MOVB BX, -840(BP)(CX*1)
MOVL -40(BP), R9
@ -3279,129 +3276,124 @@ LBB9_13:
MOVL R9, -40(BP)
MOVL R9, R14
LBB9_17:
LBB9_15:
INCQ AX
CMPQ AX, R11
SETLT CX
CMPQ R10, AX
JE LBB9_18
JE LBB9_16
LBB9_7:
LBB9_5:
MOVBLZX 0(DI)(AX*1), BX
LEAL -48(BX), DX
CMPB DX, $9
JA LBB9_15
JA LBB9_13
CMPB BX, $48
JNE LBB9_12
JNE LBB9_10
TESTL R14, R14
JE LBB9_10
JE LBB9_8
CMPL R14, $800
JL LBB9_13
JMP LBB9_17
JL LBB9_11
JMP LBB9_15
LBB9_15:
LBB9_13:
CMPB BX, $46
JNE LBB9_19
JNE LBB9_17
MOVL R14, -36(BP)
MOVL $1, R8
JMP LBB9_17
LBB9_12:
CMPL R14, $799
JLE LBB9_13
MOVL $1, -28(BP)
JMP LBB9_17
JMP LBB9_15
LBB9_10:
CMPL R14, $799
JLE LBB9_11
MOVL $1, -28(BP)
JMP LBB9_15
LBB9_8:
DECL -36(BP)
XORL R14, R14
JMP LBB9_17
JMP LBB9_15
LBB9_3:
XORL AX, AX
CMPL AX, SI
JL LBB9_6
LBB9_5:
MOVL $0, -36(BP)
JMP LBB9_35
JMP LBB9_33
LBB9_18:
LBB9_16:
MOVL R10, AX
LBB9_19:
LBB9_17:
TESTL R8, R8
JE LBB9_20
JE LBB9_18
TESTB $1, CX
JNE LBB9_22
JMP LBB9_35
JNE LBB9_20
JMP LBB9_33
LBB9_20:
LBB9_18:
MOVL R9, -36(BP)
TESTB $1, CX
JE LBB9_35
JE LBB9_33
LBB9_22:
LBB9_20:
MOVL AX, DX
MOVB 0(DI)(DX*1), CX
ORB $32, CX
CMPB CX, $101
JNE LBB9_35
JNE LBB9_33
MOVB 1(DI)(DX*1), BX
CMPB BX, $45
JE LBB9_27
JE LBB9_25
MOVL $1, R8
CMPB BX, $43
JNE LBB9_25
JNE LBB9_23
ADDL $2, AX
JMP LBB9_28
JMP LBB9_26
LBB9_27:
LBB9_25:
ADDL $2, AX
MOVL $-1, R8
LBB9_28:
LBB9_26:
MOVL AX, DX
XORL AX, AX
CMPL DX, SI
JL LBB9_30
JMP LBB9_34
JL LBB9_28
JMP LBB9_32
LBB9_25:
LBB9_23:
INCQ DX
XORL AX, AX
CMPL DX, SI
JGE LBB9_34
JGE LBB9_32
LBB9_30:
LBB9_28:
MOVLQSX DX, AX
SUBL DX, SI
ADDQ AX, DI
XORL DX, DX
XORL AX, AX
LBB9_31:
LBB9_29:
CMPL AX, $9999
JG LBB9_34
JG LBB9_32
MOVBLZX 0(DI)(DX*1), BX
LEAL -48(BX), CX
CMPB CX, $9
JA LBB9_34
JA LBB9_32
LEAL 0(AX)(AX*4), AX
LEAL -48(BX)(AX*2), AX
INCQ DX
CMPL SI, DX
JNE LBB9_31
JNE LBB9_29
LBB9_34:
LBB9_32:
IMULL R8, AX
ADDL AX, -36(BP)
LBB9_35:
LBB9_33:
LEAQ -840(BP), DI
LEAQ -24(BP), SI
WORD $0xf8c5; BYTE $0x77 // vzeroupper
LONG $0xfff8c2e8; BYTE $0xff // callq _decimal_to_f64
LONG $0xfff8d5e8; BYTE $0xff // callq _decimal_to_f64
LONG $0x4510fbc5; BYTE $0xe8 // vmovsd $-24(%rbp), %xmm0
ADDQ $832, SP
BYTE $0x5b // popq %rbx
@ -4584,7 +4576,7 @@ LBB13_66:
CMPL BX, $23
JL LBB13_75
LEAL -22(BX), AX
LONG $0xee0d8d48; WORD $0x0152; BYTE $0x00 // leaq $86766(%rip), %rcx /* _P10_TAB(%rip) */
LONG $0x0c0d8d48; WORD $0x0153; BYTE $0x00 // leaq $86796(%rip), %rcx /* _P10_TAB(%rip) */
LONG $0x0459fbc5; BYTE $0xc1 // vmulsd (%rcx,%rax,8), %xmm0, %xmm0
MOVL $22, AX
JMP LBB13_76
@ -4603,7 +4595,7 @@ LBB13_72:
CMPL BX, $-22
JB LBB13_79
NEGL BX
LONG $0xb5058d48; WORD $0x0152; BYTE $0x00 // leaq $86709(%rip), %rax /* _P10_TAB(%rip) */
LONG $0xd3058d48; WORD $0x0152; BYTE $0x00 // leaq $86739(%rip), %rax /* _P10_TAB(%rip) */
LONG $0x045efbc5; BYTE $0xd8 // vdivsd (%rax,%rbx,8), %xmm0, %xmm0
LONG $0x4511fbc5; BYTE $0xc8 // vmovsd %xmm0, $-56(%rbp)
JMP LBB13_85
@ -4624,7 +4616,7 @@ LBB13_76:
LONG $0xc82ef9c5 // vucomisd %xmm0, %xmm1
JA LBB13_79
MOVL AX, AX
LONG $0x710d8d48; WORD $0x0152; BYTE $0x00 // leaq $86641(%rip), %rcx /* _P10_TAB(%rip) */
LONG $0x8f0d8d48; WORD $0x0152; BYTE $0x00 // leaq $86671(%rip), %rcx /* _P10_TAB(%rip) */
LONG $0x0459fbc5; BYTE $0xc1 // vmulsd (%rcx,%rax,8), %xmm0, %xmm0
LONG $0x4511fbc5; BYTE $0xc8 // vmovsd %xmm0, $-56(%rbp)
JMP LBB13_85
@ -4635,8 +4627,8 @@ LBB13_79:
MOVQ R12, DI
MOVL BX, SI
MOVL -60(BP), DX
LONG $0xffe876e8; BYTE $0xff // callq _atof_eisel_lemire64
TESTL AX, AX
LONG $0xffe8a8e8; BYTE $0xff // callq _atof_eisel_lemire64
TESTB AX, AX
JE LBB13_83
CMPL -44(BP), $0
JE LBB13_84
@ -4645,8 +4637,8 @@ LBB13_79:
MOVQ R12, DI
MOVL BX, SI
MOVL -60(BP), DX
LONG $0xffe858e8; BYTE $0xff // callq _atof_eisel_lemire64
TESTL AX, AX
LONG $0xffe88ae8; BYTE $0xff // callq _atof_eisel_lemire64
TESTB AX, AX
JE LBB13_83
LONG $0x4d10fbc5; BYTE $0xb8 // vmovsd $-72(%rbp), %xmm1
LONG $0x4510fbc5; BYTE $0xc8 // vmovsd $-56(%rbp), %xmm0
@ -4661,7 +4653,7 @@ LBB13_83:
ADDQ AX, DI
MOVL 8(CX), SI
SUBL AX, SI
LONG $0xffeecbe8; BYTE $0xff // callq _atof_native_decimal
LONG $0xffeedee8; BYTE $0xff // callq _atof_native_decimal
JMP LBB13_85
LBB13_84:
@ -6388,7 +6380,7 @@ _print_mantissa:
LONG $0x64fa6b41 // imull $100, %r10d, %edi
SUBL DI, AX
MOVWLZX AX, R11
LONG $0x9e3d8d48; WORD $0x0059; BYTE $0x00 // leaq $22942(%rip), %rdi /* _Digits(%rip) */
LONG $0xbc3d8d48; WORD $0x0059; BYTE $0x00 // leaq $22972(%rip), %rdi /* _Digits(%rip) */
MOVWLZX 0(DI)(R8*2), AX
MOVW AX, -2(R14)
MOVWLZX 0(DI)(R9*2), AX
@ -6404,7 +6396,7 @@ LBB24_2:
CMPL DI, $10000
JB LBB24_3
MOVL $3518437209, R8
LONG $0x560d8d4c; WORD $0x0059; BYTE $0x00 // leaq $22870(%rip), %r9 /* _Digits(%rip) */
LONG $0x740d8d4c; WORD $0x0059; BYTE $0x00 // leaq $22900(%rip), %r9 /* _Digits(%rip) */
LBB24_5:
MOVL DI, AX
@ -6435,7 +6427,7 @@ LBB24_7:
WORD $0xd16b; BYTE $0x64 // imull $100, %ecx, %edx
SUBL DX, AX
MOVWLZX AX, AX
LONG $0xef158d48; WORD $0x0058; BYTE $0x00 // leaq $22767(%rip), %rdx /* _Digits(%rip) */
LONG $0x0d158d48; WORD $0x0059; BYTE $0x00 // leaq $22797(%rip), %rdx /* _Digits(%rip) */
MOVWLZX 0(DX)(AX*2), AX
MOVW AX, -2(R14)
ADDQ $-2, R14
@ -6445,7 +6437,7 @@ LBB24_8:
CMPL AX, $10
JB LBB24_10
MOVL AX, AX
LONG $0xd20d8d48; WORD $0x0058; BYTE $0x00 // leaq $22738(%rip), %rcx /* _Digits(%rip) */
LONG $0xf00d8d48; WORD $0x0058; BYTE $0x00 // leaq $22768(%rip), %rcx /* _Digits(%rip) */
MOVWLZX 0(CX)(AX*2), AX
MOVW AX, -2(R14)
BYTE $0x5b // popq %rbx
@ -6502,7 +6494,7 @@ LBB25_4:
RET
LBB25_6:
LONG $0xb70d8d48; WORD $0x007a; BYTE $0x00 // leaq $31415(%rip), %rcx /* _U8_LEN_TAB(%rip) */
LONG $0xd50d8d48; WORD $0x007a; BYTE $0x00 // leaq $31445(%rip), %rcx /* _U8_LEN_TAB(%rip) */
MOVBLZX 0(AX)(CX*1), AX
ADDL BX, AX
ADDQ $8, SP
@ -6513,64 +6505,62 @@ LBB25_6:
_left_shift:
BYTE $0x55 // pushq %rbp
WORD $0x8948; BYTE $0xe5 // movq %rsp, %rbp
WORD $0x5741 // pushq %r15
WORD $0x5641 // pushq %r14
BYTE $0x53 // pushq %rbx
MOVL SI, CX
IMUL3Q $804, CX, R10
LONG $0x951d8d4c; WORD $0x007b; BYTE $0x00 // leaq $31637(%rip), %r11 /* _LSHIFT_TAB(%rip) */
MOVL 0(R10)(R11*1), R8
IMUL3Q $804, CX, DX
LONG $0xaf358d48; WORD $0x007b; BYTE $0x00 // leaq $31663(%rip), %rsi /* _LSHIFT_TAB(%rip) */
MOVL 0(DX)(SI*1), R8
MOVLQSX 800(DI), R9
MOVB 4(DX)(SI*1), AX
TESTQ R9, R9
JE LBB26_1
LEAQ 4(R10)(R11*1), SI
XORL AX, AX
JE LBB26_6
LEAQ 5(DX)(SI*1), DX
XORL SI, SI
LBB26_3:
MOVBLZX 0(SI)(AX*1), DX
CMPB 0(DI)(AX*1), DX
JNE LBB26_8
INCQ AX
CMPQ R9, AX
TESTB AX, AX
JE LBB26_8
CMPB 0(DI)(SI*1), AX
JNE LBB26_5
MOVBLZX 0(DX)(SI*1), AX
INCQ SI
CMPQ R9, SI
JNE LBB26_3
LEAL -1(R9), AX
CMPL AX, $798
JA LBB26_10
MOVL R9, AX
JMP LBB26_7
LBB26_1:
XORL AX, AX
LBB26_6:
TESTB AX, AX
JE LBB26_8
LBB26_7:
ADDQ R10, R11
MOVB 4(AX)(R11*1), AX
ADDB $-48, AX
CMPB AX, $10
JAE LBB26_10
LBB26_9:
DECL R8
LBB26_10:
LBB26_8:
TESTL R9, R9
JLE LBB26_23
LEAL -1(R9)(R8*1), R10
JLE LBB26_22
LEAL 0(R8)(R9*1), AX
MOVLQSX AX, BX
MOVQ BX, R10
SHLQ $32, R10
INCQ R9
XORL DX, DX
MOVQ $-3689348814741910323, R11
JMP LBB26_12
MOVQ $-4294967296, R14
JMP LBB26_10
LBB26_13:
LBB26_11:
ADDB $48, AX
MOVLQSX R10, BX
MOVB AX, 0(DI)(BX*1)
DECL R10
MOVB AX, -1(DI)(BX*1)
LBB26_19:
LBB26_18:
DECQ BX
ADDQ R14, R10
DECQ R9
CMPQ R9, $1
JLE LBB26_14
JLE LBB26_12
LBB26_12:
LBB26_10:
MOVBQSX -2(DI)(R9*1), SI
ADDQ $-48, SI
SHLQ CX, SI
@ -6579,62 +6569,48 @@ LBB26_12:
MULQ R11
SHRQ $3, DX
LEAQ 0(DX)(DX*1), AX
LEAQ 0(AX)(AX*4), BX
LEAQ 0(AX)(AX*4), R15
MOVQ SI, AX
SUBQ BX, AX
CMPL R10, $799
JLE LBB26_13
SUBQ R15, AX
CMPQ BX, $800
JLE LBB26_11
TESTQ AX, AX
JE LBB26_19
JE LBB26_18
MOVL $1, 812(DI)
JMP LBB26_19
JMP LBB26_18
LBB26_14:
LBB26_12:
CMPQ SI, $10
JAE LBB26_15
LBB26_23:
MOVL 800(DI), CX
ADDL R8, CX
CMPL CX, $801
MOVL $800, AX
WORD $0x4c0f; BYTE $0xc1 // cmovll %ecx, %eax
MOVL AX, 800(DI)
ADDL R8, 804(DI)
TESTL CX, CX
JLE LBB26_27
MOVL AX, CX
INCQ CX
LBB26_25:
DECL AX
CMPB 0(DI)(AX*1), $48
JNE LBB26_29
MOVL AX, 800(DI)
DECQ CX
CMPQ CX, $1
JG LBB26_25
LBB26_27:
TESTL AX, AX
JE LBB26_28
LBB26_29:
BYTE $0x5b // popq %rbx
BYTE $0x5d // popq %rbp
RET
LBB26_16:
ADDB $48, AX
MOVLQSX R10, SI
MOVB AX, 0(DI)(SI*1)
DECL R10
JAE LBB26_13
LBB26_22:
CMPQ CX, $9
JBE LBB26_23
MOVL 800(DI), AX
ADDL R8, AX
CMPL AX, $801
MOVL $800, CX
WORD $0x4c0f; BYTE $0xc8 // cmovll %eax, %ecx
MOVL CX, 800(DI)
ADDL R8, 804(DI)
BYTE $0x5b // popq %rbx
WORD $0x5e41 // popq %r14
WORD $0x5f41 // popq %r15
BYTE $0x5d // popq %rbp
JMP _trim
LBB26_13:
SARQ $32, R10
JMP LBB26_14
LBB26_15:
ADDB $48, AX
MOVB AX, -1(DI)(R10*1)
LBB26_21:
DECQ R10
CMPQ CX, $9
JBE LBB26_22
LBB26_14:
MOVQ DX, CX
MOVQ DX, AX
MULQ R11
@ -6643,22 +6619,16 @@ LBB26_15:
LEAQ 0(AX)(AX*4), SI
MOVQ CX, AX
SUBQ SI, AX
CMPL R10, $799
JLE LBB26_16
CMPQ R10, $800
JLE LBB26_15
TESTQ AX, AX
JE LBB26_22
JE LBB26_21
MOVL $1, 812(DI)
JMP LBB26_22
JMP LBB26_21
LBB26_8:
JLE LBB26_9
JMP LBB26_10
LBB26_28:
MOVL $0, 804(DI)
BYTE $0x5b // popq %rbx
BYTE $0x5d // popq %rbp
RET
LBB26_5:
JL LBB26_7
JMP LBB26_8
_right_shift:
BYTE $0x55 // pushq %rbp
@ -6794,6 +6764,36 @@ LBB27_23:
BYTE $0x5d // popq %rbp
RET
_trim:
BYTE $0x55 // pushq %rbp
WORD $0x8948; BYTE $0xe5 // movq %rsp, %rbp
MOVL 800(DI), AX
TESTL AX, AX
JLE LBB28_4
LEAQ 1(AX), CX
LBB28_2:
DECL AX
CMPB 0(DI)(AX*1), $48
JNE LBB28_6
MOVL AX, 800(DI)
DECQ CX
CMPQ CX, $1
JG LBB28_2
LBB28_4:
TESTL AX, AX
JE LBB28_5
LBB28_6:
BYTE $0x5d // popq %rbp
RET
LBB28_5:
MOVL $0, 804(DI)
BYTE $0x5d // popq %rbp
RET
_POW10_M128_TAB:
QUAD $0x1732c869cd60e453 // .quad 1671618768450675795
QUAD $0xfa8fd5a0081c0288 // .quad -391859759250406776
@ -13367,7 +13367,7 @@ TEXT ·__skip_array(SB), NOSPLIT, $0 - 32
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ m+16(FP), DX
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+18358(SB) // _skip_array
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+18308(SB) // _skip_array
MOVQ AX, ret+24(FP)
RET
@ -13375,7 +13375,7 @@ TEXT ·__skip_object(SB), NOSPLIT, $0 - 32
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ m+16(FP), DX
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+18393(SB) // _skip_object
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+18343(SB) // _skip_object
MOVQ AX, ret+24(FP)
RET
@ -13383,7 +13383,7 @@ TEXT ·__skip_one(SB), NOSPLIT, $0 - 32
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ m+16(FP), DX
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+15558(SB) // _skip_one
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+15508(SB) // _skip_one
MOVQ AX, ret+24(FP)
RET
@ -13410,7 +13410,7 @@ TEXT ·__value(SB), NOSPLIT, $0 - 48
MOVQ p+16(FP), DX
MOVQ v+24(FP), CX
MOVQ allow_control+32(FP), R8
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+11282(SB) // _value
CALL ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+11232(SB) // _value
MOVQ AX, ret+40(FP)
RET
@ -13418,26 +13418,26 @@ TEXT ·__vnumber(SB), NOSPLIT, $0 - 24
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ v+16(FP), DX
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+13846(SB), AX // _vnumber
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+13796(SB), AX // _vnumber
JMP AX
TEXT ·__vsigned(SB), NOSPLIT, $0 - 24
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ v+16(FP), DX
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+15008(SB), AX // _vsigned
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+14958(SB), AX // _vsigned
JMP AX
TEXT ·__vstring(SB), NOSPLIT, $0 - 24
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ v+16(FP), DX
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+12993(SB), AX // _vstring
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+12943(SB), AX // _vstring
JMP AX
TEXT ·__vunsigned(SB), NOSPLIT, $0 - 24
MOVQ s+0(FP), DI
MOVQ p+8(FP), SI
MOVQ v+16(FP), DX
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+15285(SB), AX // _vunsigned
LEAQ ·___asm2asm_compiled_code__DO_NOT_CALL_THIS_SYMBOL___+15235(SB), AX // _vunsigned
JMP AX

16
internal/native/avx2/native_subr_amd64.go
View file

@ -19,16 +19,16 @@ var (
_subr__lspace = **(**uintptr)(unsafe.Pointer(&_func__base)) + 366
_subr__lzero = **(**uintptr)(unsafe.Pointer(&_func__base)) + 0
_subr__quote = **(**uintptr)(unsafe.Pointer(&_func__base)) + 5202
_subr__skip_array = **(**uintptr)(unsafe.Pointer(&_func__base)) + 18358
_subr__skip_object = **(**uintptr)(unsafe.Pointer(&_func__base)) + 18393
_subr__skip_one = **(**uintptr)(unsafe.Pointer(&_func__base)) + 15558
_subr__skip_array = **(**uintptr)(unsafe.Pointer(&_func__base)) + 18308
_subr__skip_object = **(**uintptr)(unsafe.Pointer(&_func__base)) + 18343
_subr__skip_one = **(**uintptr)(unsafe.Pointer(&_func__base)) + 15508
_subr__u64toa = **(**uintptr)(unsafe.Pointer(&_func__base)) + 3882
_subr__unquote = **(**uintptr)(unsafe.Pointer(&_func__base)) + 7039
_subr__value = **(**uintptr)(unsafe.Pointer(&_func__base)) + 11282
_subr__vnumber = **(**uintptr)(unsafe.Pointer(&_func__base)) + 13846
_subr__vsigned = **(**uintptr)(unsafe.Pointer(&_func__base)) + 15008
_subr__vstring = **(**uintptr)(unsafe.Pointer(&_func__base)) + 12993
_subr__vunsigned = **(**uintptr)(unsafe.Pointer(&_func__base)) + 15285
_subr__value = **(**uintptr)(unsafe.Pointer(&_func__base)) + 11232
_subr__vnumber = **(**uintptr)(unsafe.Pointer(&_func__base)) + 13796
_subr__vsigned = **(**uintptr)(unsafe.Pointer(&_func__base)) + 14958
_subr__vstring = **(**uintptr)(unsafe.Pointer(&_func__base)) + 12943
_subr__vunsigned = **(**uintptr)(unsafe.Pointer(&_func__base)) + 15235
)
var (

16
native/atof_eisel_lemire.c
View file

@ -54,9 +54,9 @@ static inline u128_output mul_u64(uint64_t x, uint64_t y) {
/* This is Eisel Lemire ParseFloat algorithm implemented in C.
* reference: https://nigeltao.github.io/blog/2020/eisel-lemire.html
*/
int atof_eisel_lemire64(uint64_t mant, int exp10, int sgn, double *val) {
bool atof_eisel_lemire64(uint64_t mant, int exp10, int sgn, double *val) {
if (exp10 < -348 || exp10 > 347) {
return 0;
return false;
}
/* Calculate the 2-base exponent of float */
@ -78,7 +78,7 @@ int atof_eisel_lemire64(uint64_t mant, int exp10, int sgn, double *val) {
*/
if ((x_hi & 0x1FF) == 0x1FF && (x_lo + mant) < mant) {
/* lower 64 bits of POW10 mantissa */
u128_output y = mul_u64(mant, POW10_M128_TAB[exp10 + 347][0]);
u128_output y = mul_u64(mant, POW10_M128_TAB[exp10 + 348][0]);
uint64_t merged_hi = x_hi;
uint64_t merged_lo = x_lo + y.hi;
@ -89,7 +89,7 @@ int atof_eisel_lemire64(uint64_t mant, int exp10, int sgn, double *val) {
/* Still ambigunous */
if ((merged_hi & 0x1FF) == 0x1FF && (merged_lo + 1) == 0 && (y.lo + mant) < mant) {
return 0;
return false;
}
x_hi = merged_hi;
@ -103,7 +103,7 @@ int atof_eisel_lemire64(uint64_t mant, int exp10, int sgn, double *val) {
/* Half-way Ambiguity */
if ((x_lo == 0) && ((x_hi & 0x1FF) == 0) && ((ret_man & 3) == 1)) {
return 0;
return false;
}
/* If not half-way, then it's rounding to-nearest
@ -121,17 +121,17 @@ int atof_eisel_lemire64(uint64_t mant, int exp10, int sgn, double *val) {
* ret_exp2 >= 0x7FF means INF/inf.
*/
if ((ret_exp2 - 1) >= (0x7FF - 1)) {
return 0;
return false;
}
/* Get the lower 52 bits as finnal mantissa */
uint64_t bits = (ret_exp2 << 52) | (ret_man & 0x000FFFFFFFFFFFFF);
if (sgn == -1) {
bits |= ((uint64_t)1) << 63;
bits |= 1ull << 63;
}
*(int64_t*)val = bits;
return 1;
return true;
}

44
native/atof_native.c
View file

@ -70,10 +70,7 @@ static inline void decimal_set(Decimal *d, const char *s, int len) {
int i = 0;
decimal_init(d);
if (s[i] == '+') {
i++;
}
else if (s[i] == '-') {
if (s[i] == '-') {
i++;
d->neg = 1;
}
@ -161,7 +158,7 @@ static inline void right_shift(Decimal *d, uint32_t k) {
}
d->dp -= r - 1; // point shift left
uint64_t mask = (((uint64_t)1) << k) - 1;
uint64_t mask = (1ull << k) - 1;
uint64_t dig = 0;
/* Pick up a digit, put down a digit */
@ -191,44 +188,41 @@ static inline void right_shift(Decimal *d, uint32_t k) {
}
/* Compare the leading prefix, if b is lexicographically less, return 0 */
static inline int prefix_cmp(const char *b, const char *s, uint64_t len) {
static inline bool prefix_is_less(const char *b, const char *s, uint64_t bn) {
int i = 0;
for (; i < len; i++) {
for (; i < bn; i++) {
if (s[i] == '\0') {
return false;
}
if (b[i] != s[i]) {
return b[i] > s[i];
return b[i] < s[i];
}
}
/* when b is substr of s */
if (i < DECIMAL_MAX_DNUM && s[i] >= '0' && s[i] <= '9') {
return 0;
}
return 1;
return s[i] != '\0';
}
/* Binary shift left (* 2) by k bits. k <= maxShift to avoid overflow */
static inline void left_shift(Decimal *d, uint32_t k) {
int delta = LSHIFT_TAB[k].delta;
if (prefix_cmp(d->d, LSHIFT_TAB[k].cutoff, d->nd) == 0){
if (prefix_is_less(d->d, LSHIFT_TAB[k].cutoff, d->nd)){
delta--;
}
int r = d->nd - 1; // read index
int w = d->nd + delta - 1; // write index
int r = d->nd; // read index
int w = d->nd + delta; // write index
uint64_t n = 0;
uint64_t quo = 0;
uint64_t rem = 0;
/* Pick up a digit, put down a digit */
for (; r >= 0; r--) {
for (r--; r >= 0; r--) {
n += (uint64_t)(d->d[r] - '0') << k;
quo = n / 10;
rem = n - 10 * quo;
w--;
if (w < DECIMAL_MAX_DNUM) {
d->d[w] = (char)(rem + '0');
w--;
} else if (rem != 0) {
/* truncated */
d->trunc = 1;
@ -240,9 +234,9 @@ static inline void left_shift(Decimal *d, uint32_t k) {
while (n > 0) {
quo = n / 10;
rem = n - 10 * quo;
w--;
if (w < DECIMAL_MAX_DNUM) {
d->d[w] = (char)(rem + '0');
w--;
} else if (rem != 0) {
/* truncated */
d->trunc = 1;
@ -356,7 +350,7 @@ int decimal_to_f64(Decimal *d, double *val) {
decimal_shift(d, -n); // shift right
exp2 += n;
}
while ((d->dp < 0) || (d->dp == 0) && (d->d[0] < '5')) { // d < 0.5
while ((d->dp < 0) || ((d->dp == 0) && (d->d[0] < '5'))) { // d < 0.5
if (-d->dp >= 9) {
n = 27;
} else {
@ -389,7 +383,7 @@ int decimal_to_f64(Decimal *d, double *val) {
mant = rounded_integer(d);
/* Rounding might have added a bit; shift down. */
if (mant == (((uint64_t)2) << 52)) { // mant has 54 bits
if (mant == (2ull << 52)) { // mant has 54 bits
mant >>= 1;
exp2 ++;
if ((exp2 + 1023) >= 0x7FF) {
@ -398,7 +392,7 @@ int decimal_to_f64(Decimal *d, double *val) {
}
/* Denormalized? */
if ((mant & (((uint64_t)1) << 52)) == 0) {
if ((mant & (1ull << 52)) == 0) {
exp2 = -1023;
}
goto out;
@ -413,7 +407,7 @@ out:
bits = mant & 0x000FFFFFFFFFFFFF;
bits |= (uint64_t)((exp2 + 1023) & 0x7FF) << 52;
if (d->neg) {
bits |= ((uint64_t)1) << 63;
bits |= 1ull << 63;
}
*(uint64_t*)val = bits;
return 0;

2
native/native.h
View file

@ -114,7 +114,7 @@ long skip_string(const GoString *src, long *p);
long skip_negative(const GoString *src, long *p);
long skip_positive(const GoString *src, long *p);
int atof_eisel_lemire64(uint64_t mant, int exp10, int sgn, double *val);
bool atof_eisel_lemire64(uint64_t mant, int exp10, int sgn, double *val);
double atof_native_decimal(const char *buf, int len);
#endif

14
native/scanning.c
View file

@ -483,11 +483,11 @@ void vstring(const GoString *src, long *p, JsonState *ret) {
ret->iv = val;
/** check whether float can represent the val exactly **/
static inline int is_atof_exact(uint64_t man, int exp, int sgn, double *val) {
static inline bool is_atof_exact(uint64_t man, int exp, int sgn, double *val) {
double f = (double)man;
if (man >> 52 != 0) {
return 0;
return false;
}
if (sgn == -1) {
@ -497,7 +497,7 @@ static inline int is_atof_exact(uint64_t man, int exp, int sgn, double *val) {
if (exp == 0 || man == 0) {
*val = f;
return 1;
return true;
} else if (exp > 0 && exp <= 15+22) {
/* uint64 integers: accurate range <= 10^15 *
* Powers of 10: accurate range <= 10^22, as P10_TAB *
@ -509,17 +509,17 @@ static inline int is_atof_exact(uint64_t man, int exp, int sgn, double *val) {
/* f is not accurate when too larger */
if (f > 1e15 || f < -1e15) {
return 0;
return false;
}
*val = f * P10_TAB[exp];
return 1;
return true;
} else if (exp < 0 && exp >= -22) {
*val = f / P10_TAB[-exp];
return 1;
return true;
}
return 0;
return false;
}
static inline double parse_float64(uint64_t man, int exp, int sgn, int trunc, const GoString *src, long idx) {