NotificationsYou must be signed in to change notification settings
Fork5.2k
Star17.2k

Simplify and optimize Math(F).Round#98186

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to ourterms of service andprivacy statement. We’ll occasionally send you account related emails.

Already on GitHub?Sign in to your account

Jump to bottom

Merged

tannergooding merged 2 commits intodotnet:mainfromMichalPetryka:round-fast

Feb 9, 2024

Merged

Simplify and optimize Math(F).Round#98186

tannergooding merged 2 commits intodotnet:mainfromMichalPetryka:round-fast

Feb 9, 2024

Conversation

Copy link

Contributor

MichalPetryka commentedFeb 8, 2024•
edited
Loading

Simplifies and optimizes parameterized Round overloads.

Code:

internalstaticclassRoundDisasm{publicstaticdoubleRoundAway(doubled)=>Math.Round(d,MidpointRounding.AwayFromZero);publicstaticfloatRoundAway(floatf)=>MathF.Round(f,MidpointRounding.AwayFromZero);publicstaticdoubleRoundTruncate(doubled)=>Math.Round(d,MidpointRounding.ToZero);publicstaticfloatRoundTruncate(floatf)=>MathF.Round(f,MidpointRounding.ToZero);publicstaticdoubleRound(doubled,MidpointRoundingm)=>Math.Round(d,m);publicstaticfloatRound(floatf,MidpointRoundingm)=>MathF.Round(f,m);publicstaticdoubleRound2(doubled)=>Math.Round(d,2);publicstaticfloatRound2(floatf)=>MathF.Round(f,2);publicstaticdoubleRound4(doubled)=>Math.Round(d,4,MidpointRounding.ToZero);publicstaticfloatRound4(floatf)=>MathF.Round(f,4,MidpointRounding.ToZero);}

.NET 8

; Assembly listing for method RoundDisasm:RoundAway(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 3 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       C5F828C8             vmovapsxmm1,xmm0       C5F8101521000000vmovupsxmm2, xmmword ptr[reloc @RWD00]       C5F154CA             vandpdxmm1,xmm1,xmm2       62F1F518560D23000000vorpdxmm1,xmm1, qword ptr[reloc @RWD16] {1to2}       C5F358C0             vaddsdxmm0,xmm1,xmm0       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11G_M000_IG03:                ;; offset=0x0027       C3retRWD00  dq8000000000000000h,8000000000000000hRWD16  dq3FDFFFFFFFFFFFFFh;          0.5; Total bytes of code 40; Assembly listing for method RoundDisasm:RoundAway(float):float (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 3 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       C5F828C8             vmovapsxmm1,xmm0       C5F8101521000000vmovupsxmm2, xmmword ptr[reloc @RWD00]       C5F054CAvandpsxmm1,xmm1,xmm2       62F17418560D23000000vorpsxmm1,xmm1, dword ptr[reloc @RWD16] {1to4}       C5F258C0             vaddssxmm0,xmm1,xmm0       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11G_M000_IG03:                ;; offset=0x0027       C3retRWD00  dq8000000080000000h,8000000080000000hRWD16  dd3EFFFFFFh;       0.5; Total bytes of code 40; Assembly listing for method RoundDisasm:RoundTruncate(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 1 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       33D2xoredx,edx       41B802000000movr8d,2G_M000_IG03:                ;; offset=0x000B       FF259F041E00         tail.jmp[System.Math:Round(double,int,int):double]; Total bytes of code 17; Assembly listing for method RoundDisasm:RoundTruncate(float):float (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 1 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       33D2xoredx,edx       41B802000000movr8d,2G_M000_IG03:                ;; offset=0x000B       FF25F78F1E00         tail.jmp[System.MathF:Round(float,int,int):float]; Total bytes of code 17; Assembly listing for method RoundDisasm:Round(double,int):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 3 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       448BC2movr8d,edx       33D2xoredx,edxG_M000_IG03:                ;; offset=0x0008       FF2542041E00         tail.jmp[System.Math:Round(double,int,int):double]; Total bytes of code 14; Assembly listing for method RoundDisasm:Round(float,int):float (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 3 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       448BC2movr8d,edx       33D2xoredx,edxG_M000_IG03:                ;; offset=0x0008       FF25AA8F1E00         tail.jmp[System.MathF:Round(float,int,int):float]; Total bytes of code 14; Assembly listing for method RoundDisasm:Round2(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO data; 0 inlinees with PGO data; 1 single block inlinees; 0 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       BA02000000movedx,2       4533C0xorr8d,r8dG_M000_IG03:                ;; offset=0x000B       FF25FF031E00         tail.jmp[System.Math:Round(double,int,int):double]; Total bytes of code 17; Assembly listing for method RoundDisasm:Round2(float):float (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO data; 0 inlinees with PGO data; 1 single block inlinees; 0 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       BA02000000movedx,2       4533C0xorr8d,r8dG_M000_IG03:                ;; offset=0x000B       FF25578F1E00         tail.jmp[System.MathF:Round(float,int,int):float]; Total bytes of code 17; Assembly listing for method RoundDisasm:Round4(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       BA04000000movedx,4       41B802000000movr8d,2G_M000_IG03:                ;; offset=0x000E       FF259C031E00         tail.jmp[System.Math:Round(double,int,int):double]; Total bytes of code 20; Assembly listing for method RoundDisasm:Round4(float):float (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       BA04000000movedx,4       41B802000000movr8d,2G_M000_IG03:                ;; offset=0x000E       FF25F48E1E00         tail.jmp[System.MathF:Round(float,int,int):float]; Total bytes of code 20

This branch

; Assembly listing for method RoundDisasm:RoundAway(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T00] (  4,  4   )  double  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  2,  2   )  double  ->  mm0         "Inline return value spill temp";* V03 tmp2         [V03    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V04 tmp3         [V04    ] (  0,  0   )  double  ->  zero-ref    "Inline return value spill temp";* V05 tmp4         [V05    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M25588_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M25588_IG02:  ;; offset=0x0003       C5F8100D25000000vmovupsxmm1, xmmword ptr[reloc @RWD00]       C5F828D0             vmovapsxmm2,xmm0       62F3ED08250D26000000CAvpternlogqxmm1,xmm2, xmmword ptr[reloc @RWD16],-54       C5F358C0             vaddsdxmm0,xmm1,xmm0       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11;; size=33 bbWeight=1 PerfScore 15.25G_M25588_IG03:  ;; offset=0x0024       C3ret;; size=1 bbWeight=1 PerfScore 1.00RWD00  dq8000000000000000h,8000000000000000hRWD16  dq3FDFFFFFFFFFFFFFh,3FDFFFFFFFFFFFFFh; Total bytes of code 37, prolog size 3, PerfScore 17.25, instruction count 7, allocated bytes for code 37 (MethodHash=318b9c0b) for method RoundDisasm:RoundAway(double):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:RoundAway(float):float (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T00] (  4,  4   )   float  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  2,  2   )   float  ->  mm0         "Inline return value spill temp";* V03 tmp2         [V03    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V04 tmp3         [V04    ] (  0,  0   )   float  ->  zero-ref    "Inline return value spill temp";* V05 tmp4         [V05    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M10964_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M10964_IG02:  ;; offset=0x0003       C5F8100D25000000vmovupsxmm1, xmmword ptr[reloc @RWD00]       C5F828D0             vmovapsxmm2,xmm0       62F36D08250D26000000CAvpternlogdxmm1,xmm2, xmmword ptr[reloc @RWD16],-54       C5F258C0             vaddssxmm0,xmm1,xmm0       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11;; size=33 bbWeight=1 PerfScore 15.25G_M10964_IG03:  ;; offset=0x0024       C3ret;; size=1 bbWeight=1 PerfScore 1.00RWD00  dq8000000080000000h,8000000080000000hRWD16  dq3EFFFFFF3EFFFFFFh,3EFFFFFF3EFFFFFFh; Total bytes of code 37, prolog size 3, PerfScore 17.25, instruction count 7, allocated bytes for code 37 (MethodHash=b997d52b) for method RoundDisasm:RoundAway(float):float (FullOpts); ============================================================; Assembly listing for method RoundDisasm:RoundTruncate(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T00] (  3,  3   )  double  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  2,  2   )  double  ->  mm0         "Inline return value spill temp";* V03 tmp2         [V03    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V04 tmp3         [V04    ] (  0,  0   )  double  ->  zero-ref    "Inline return value spill temp";* V05 tmp4         [V05    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M39156_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M39156_IG02:  ;; offset=0x0003       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11;; size=6 bbWeight=1 PerfScore 7.00G_M39156_IG03:  ;; offset=0x0009       C3ret;; size=1 bbWeight=1 PerfScore 1.00; Total bytes of code 10, prolog size 3, PerfScore 9.00, instruction count 3, allocated bytes for code 10 (MethodHash=6fe1670b) for method RoundDisasm:RoundTruncate(double):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:RoundTruncate(float):float (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T00] (  3,  3   )   float  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  2,  2   )   float  ->  mm0         "Inline return value spill temp";* V03 tmp2         [V03    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V04 tmp3         [V04    ] (  0,  0   )   float  ->  zero-ref    "Inline return value spill temp";* V05 tmp4         [V05    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M8148_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M8148_IG02:  ;; offset=0x0003       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11;; size=6 bbWeight=1 PerfScore 7.00G_M8148_IG03:  ;; offset=0x0009       C3ret;; size=1 bbWeight=1 PerfScore 1.00; Total bytes of code 10, prolog size 3, PerfScore 9.00, instruction count 3, allocated bytes for code 10 (MethodHash=41fee02b) for method RoundDisasm:RoundTruncate(float):float (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round(double,int):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T01] (  8,  5   )  double  ->  mm0         single-def;  V01 arg1         [V01,T00] (  5,  4   )     int  ->  rbx         single-def;  V02 OutArgs      [V02    ] (  1,  1   )  struct (32) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V03 tmp1         [V03,T02] (  6,  3.50)  double  ->  mm0         "Inline return value spill temp";* V04 tmp2         [V04    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V05 tmp3         [V05    ] (  0,  0   )  double  ->  zero-ref    "Inline return value spill temp";  V06 tmp4         [V06,T03] (  2,  0   )     ref  ->  rdx         single-def "argument with side effect";; Lcl frame size = 32G_M6149_IG01:  ;; offset=0x000053pushrbx       4883EC20subrsp,32       C5F877vzeroupper       8BDAmovebx,edx;; size=10 bbWeight=1 PerfScore 2.50G_M6149_IG02:  ;; offset=0x000A       83FB04cmpebx,4       775Fja       SHORT G_M6149_IG09       8BCBmovecx,ebx       488D1578000000leardx,[reloc @RWD00]       8B148Amovedx, dword ptr[rdx+4*rcx]       488D05E8FFFFFFlearax, G_M6149_IG02       4803D0addrdx,rax       FFE2jmprdx;; size=29 bbWeight=1 PerfScore 7.25G_M6149_IG03:  ;; offset=0x0027       C5F8100D81000000vmovupsxmm1, xmmword ptr[reloc @RWD32]       C5F828D0             vmovapsxmm2,xmm0       62F3ED08250D82000000CAvpternlogqxmm1,xmm2, xmmword ptr[reloc @RWD48],-54       C5F358C0             vaddsdxmm0,xmm1,xmm0       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11       EB1Ejmp      SHORT G_M6149_IG08;; size=35 bbWeight=0.50 PerfScore 8.62G_M6149_IG04:  ;; offset=0x004A       C4E3790BC004         vroundsdxmm0,xmm0,xmm0,4       EB16jmp      SHORT G_M6149_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M6149_IG05:  ;; offset=0x0052       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11       EB0Ejmp      SHORT G_M6149_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M6149_IG06:  ;; offset=0x005A       C4E3790BC009         vroundsdxmm0,xmm0,xmm0,9       EB06jmp      SHORT G_M6149_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M6149_IG07:  ;; offset=0x0062       C4E3790BC00A         vroundsdxmm0,xmm0,xmm0,10;; size=6 bbWeight=0.50 PerfScore 3.50G_M6149_IG08:  ;; offset=0x0068       4883C420addrsp,32       5Bpoprbx       C3ret;; size=6 bbWeight=1 PerfScore 1.75G_M6149_IG09:  ;; offset=0x006E       B9840B0000movecx,0xB84       48BA00400F6CFD7F0000movrdx,0x7FFD6C0F4000       E83EE64D5Fcall     CORINFO_HELP_STRCNS       488BD0movrdx,rax       8BCBmovecx,ebx       FF15E3957C00call[System.ThrowHelper:ThrowArgumentException_InvalidEnumValue[int](int,System.String)]       CCint3;; size=32 bbWeight=0 PerfScore 0.00RWD00  dd00000040h ; case G_M6149_IG04dd0000001Dh ; case G_M6149_IG03dd00000048h ; case G_M6149_IG05dd00000050h ; case G_M6149_IG06dd00000058h ; case G_M6149_IG07RWD20  dd00000000h,00000000h,00000000hRWD32  dq8000000000000000h,8000000000000000hRWD48  dq3FDFFFFFFFFFFFFFh,3FDFFFFFFFFFFFFFh; Total bytes of code 142, prolog size 8, PerfScore 37.12, instruction count 35, allocated bytes for code 142 (MethodHash=d24ae7fa) for method RoundDisasm:Round(double,int):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round(float,int):float (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T01] (  8,  5   )   float  ->  mm0         single-def;  V01 arg1         [V01,T00] (  5,  4   )     int  ->  rbx         single-def;  V02 OutArgs      [V02    ] (  1,  1   )  struct (32) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V03 tmp1         [V03,T02] (  6,  3.50)   float  ->  mm0         "Inline return value spill temp";* V04 tmp2         [V04    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V05 tmp3         [V05    ] (  0,  0   )   float  ->  zero-ref    "Inline return value spill temp";  V06 tmp4         [V06,T03] (  2,  0   )     ref  ->  rdx         single-def "argument with side effect";; Lcl frame size = 32G_M10373_IG01:  ;; offset=0x000053pushrbx       4883EC20subrsp,32       C5F877vzeroupper       8BDAmovebx,edx;; size=10 bbWeight=1 PerfScore 2.50G_M10373_IG02:  ;; offset=0x000A       83FB04cmpebx,4       775Fja       SHORT G_M10373_IG09       8BCBmovecx,ebx       488D1578000000leardx,[reloc @RWD00]       8B148Amovedx, dword ptr[rdx+4*rcx]       488D05E8FFFFFFlearax, G_M10373_IG02       4803D0addrdx,rax       FFE2jmprdx;; size=29 bbWeight=1 PerfScore 7.25G_M10373_IG03:  ;; offset=0x0027       C5F8100D81000000vmovupsxmm1, xmmword ptr[reloc @RWD32]       C5F828D0             vmovapsxmm2,xmm0       62F36D08250D82000000CAvpternlogdxmm1,xmm2, xmmword ptr[reloc @RWD48],-54       C5F258C0             vaddssxmm0,xmm1,xmm0       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11       EB1Ejmp      SHORT G_M10373_IG08;; size=35 bbWeight=0.50 PerfScore 8.62G_M10373_IG04:  ;; offset=0x004A       C4E3790AC004         vroundssxmm0,xmm0,xmm0,4       EB16jmp      SHORT G_M10373_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M10373_IG05:  ;; offset=0x0052       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11       EB0Ejmp      SHORT G_M10373_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M10373_IG06:  ;; offset=0x005A       C4E3790AC009         vroundssxmm0,xmm0,xmm0,9       EB06jmp      SHORT G_M10373_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M10373_IG07:  ;; offset=0x0062       C4E3790AC00A         vroundssxmm0,xmm0,xmm0,10;; size=6 bbWeight=0.50 PerfScore 3.50G_M10373_IG08:  ;; offset=0x0068       4883C420addrsp,32       5Bpoprbx       C3ret;; size=6 bbWeight=1 PerfScore 1.75G_M10373_IG09:  ;; offset=0x006E       B9840B0000movecx,0xB84       48BA00400F6CFD7F0000movrdx,0x7FFD6C0F4000       E84EE54D5Fcall     CORINFO_HELP_STRCNS       488BD0movrdx,rax       8BCBmovecx,ebx       FF15F3947C00call[System.ThrowHelper:ThrowArgumentException_InvalidEnumValue[int](int,System.String)]       CCint3;; size=32 bbWeight=0 PerfScore 0.00RWD00  dd00000040h ; case G_M10373_IG04dd0000001Dh ; case G_M10373_IG03dd00000048h ; case G_M10373_IG05dd00000050h ; case G_M10373_IG06dd00000058h ; case G_M10373_IG07RWD20  dd00000000h,00000000h,00000000hRWD32  dq8000000080000000h,8000000080000000hRWD48  dq3EFFFFFF3EFFFFFFh,3EFFFFFF3EFFFFFFh; Total bytes of code 142, prolog size 8, PerfScore 37.12, instruction count 35, allocated bytes for code 142 (MethodHash=e4b5d77a) for method RoundDisasm:Round(float,int):float (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round2(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 1 inlinees with PGO data; 2 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T01] (  4,  3.50)  double  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T00] (  4,  7   )  double  ->  mm1         "Inlining Arg";* V03 tmp2         [V03    ] (  0,  0   )  struct (16) zero-ref    ld-addr-op "Inline stloc first use temp" <System.ReadOnlySpan`1[double]>;* V04 tmp3         [V04,T04] (  0,  0   )  double  ->  zero-ref    "Inline stloc first use temp";* V05 tmp4         [V05    ] (  0,  0   )  struct (16) zero-ref    "ReadOnlySpan<T> for CreateSpan<T>" <System.ReadOnlySpan`1[double]>;  V06 tmp5         [V06,T03] (  2,  1   )  double  ->  mm1         "Inline return value spill temp";* V07 tmp6         [V07    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";  V08 tmp7         [V08,T02] (  2,  2   )  double  ->  mm1         "Inlining Arg";* V09 tmp8         [V09    ] (  0,  0   )  double  ->  zero-ref    "Inline return value spill temp";* V10 tmp9         [V10    ] (  0,  0   )   byref  ->  zero-ref    single-def "field V03._reference (fldOffset=0x0)" P-INDEP;* V11 tmp10        [V11    ] (  0,  0   )     int  ->  zero-ref    "field V03._length (fldOffset=0x8)" P-INDEP;* V12 tmp11        [V12    ] (  0,  0   )   byref  ->  zero-ref    single-def "field V05._reference (fldOffset=0x0)" P-INDEP;* V13 tmp12        [V13    ] (  0,  0   )     int  ->  zero-ref    "field V05._length (fldOffset=0x8)" P-INDEP;* V14 tmp13        [V14    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M8040_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M8040_IG02:  ;; offset=0x0003       C5F828C8             vmovapsxmm1,xmm0       C5F0541531000000vandpsxmm2,xmm1, xmmword ptr[reloc @RWD00]       C5FB101D39000000     vmovsdxmm3, qword ptr[reloc @RWD16]       C5F92EDA             vucomisdxmm3,xmm27616jbe      SHORT G_M8040_IG04;; size=26 bbWeight=1 PerfScore 8.25G_M8040_IG03:  ;; offset=0x001D       C5FB590D33000000     vmulsdxmm1,xmm0, qword ptr[reloc @RWD24]       C4E3710BC904         vroundsdxmm1,xmm1,xmm1,4       C5F35E0D25000000     vdivsdxmm1,xmm1, qword ptr[reloc @RWD24];; size=22 bbWeight=0.50 PerfScore 13.00G_M8040_IG04:  ;; offset=0x0033       C5F828C1             vmovapsxmm0,xmm1;; size=4 bbWeight=1 PerfScore 0.25G_M8040_IG05:  ;; offset=0x0037       C3ret;; size=1 bbWeight=1 PerfScore 1.00RWD00  dq7FFFFFFFFFFFFFFFh,7FFFFFFFFFFFFFFFhRWD16  dq4341C37937E08000h;        1e+16RWD24  dq4059000000000000h;          100; Total bytes of code 56, prolog size 3, PerfScore 23.50, instruction count 11, allocated bytes for code 56 (MethodHash=166ce097) for method RoundDisasm:Round2(double):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round2(float):float (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 2 single block inlinees; 3 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T01] (  4,  3.50)   float  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T00] (  4,  7   )   float  ->  mm1         "Inlining Arg";* V03 tmp2         [V03    ] (  0,  0   )  struct (16) zero-ref    ld-addr-op "Inline stloc first use temp" <System.ReadOnlySpan`1[float]>;* V04 tmp3         [V04,T04] (  0,  0   )   float  ->  zero-ref    "Inline stloc first use temp";* V05 tmp4         [V05    ] (  0,  0   )  struct (16) zero-ref    "ReadOnlySpan<T> for CreateSpan<T>" <System.ReadOnlySpan`1[float]>;  V06 tmp5         [V06,T03] (  2,  1   )   float  ->  mm1         "Inline return value spill temp";* V07 tmp6         [V07    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";  V08 tmp7         [V08,T02] (  2,  2   )   float  ->  mm1         "Inlining Arg";* V09 tmp8         [V09    ] (  0,  0   )   float  ->  zero-ref    "Inline return value spill temp";* V10 tmp9         [V10    ] (  0,  0   )   byref  ->  zero-ref    single-def "field V03._reference (fldOffset=0x0)" P-INDEP;* V11 tmp10        [V11    ] (  0,  0   )     int  ->  zero-ref    "field V03._length (fldOffset=0x8)" P-INDEP;* V12 tmp11        [V12    ] (  0,  0   )   byref  ->  zero-ref    single-def "field V05._reference (fldOffset=0x0)" P-INDEP;* V13 tmp12        [V13    ] (  0,  0   )     int  ->  zero-ref    "field V05._length (fldOffset=0x8)" P-INDEP;* V14 tmp13        [V14    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M12104_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M12104_IG02:  ;; offset=0x0003       C5F828C8             vmovapsxmm1,xmm0       C5F0541531000000vandpsxmm2,xmm1, xmmword ptr[reloc @RWD00]       C5FA101D39000000vmovssxmm3, dword ptr[reloc @RWD16]       C5F82EDA             vucomissxmm3,xmm27616jbe      SHORT G_M12104_IG04;; size=26 bbWeight=1 PerfScore 8.25G_M12104_IG03:  ;; offset=0x001D       C5FA590D2F000000     vmulssxmm1,xmm0, dword ptr[reloc @RWD20]       C4E3710AC904         vroundssxmm1,xmm1,xmm1,4       C5F25E0D21000000     vdivssxmm1,xmm1, dword ptr[reloc @RWD20];; size=22 bbWeight=0.50 PerfScore 12.00G_M12104_IG04:  ;; offset=0x0033       C5F828C1             vmovapsxmm0,xmm1;; size=4 bbWeight=1 PerfScore 0.25G_M12104_IG05:  ;; offset=0x0037       C3ret;; size=1 bbWeight=1 PerfScore 1.00RWD00  dq7FFFFFFF7FFFFFFFh,7FFFFFFF7FFFFFFFhRWD16  dd4CBEBC20h;     1e+08RWD20  dd42C80000h;       100; Total bytes of code 56, prolog size 3, PerfScore 22.50, instruction count 11, allocated bytes for code 56 (MethodHash=a66ed0b7) for method RoundDisasm:Round2(float):float (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round4(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 1 inlinees with PGO data; 1 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T01] (  4,  3.50)  double  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T00] (  4,  7   )  double  ->  mm1         "Inlining Arg";* V03 tmp2         [V03    ] (  0,  0   )  struct (16) zero-ref    ld-addr-op "Inline stloc first use temp" <System.ReadOnlySpan`1[double]>;* V04 tmp3         [V04,T04] (  0,  0   )  double  ->  zero-ref    "Inline stloc first use temp";* V05 tmp4         [V05    ] (  0,  0   )  struct (16) zero-ref    "ReadOnlySpan<T> for CreateSpan<T>" <System.ReadOnlySpan`1[double]>;  V06 tmp5         [V06,T03] (  2,  1   )  double  ->  mm1         "Inline return value spill temp";* V07 tmp6         [V07    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";  V08 tmp7         [V08,T02] (  2,  2   )  double  ->  mm1         "Inlining Arg";* V09 tmp8         [V09    ] (  0,  0   )  double  ->  zero-ref    "Inline return value spill temp";* V10 tmp9         [V10    ] (  0,  0   )   byref  ->  zero-ref    single-def "field V03._reference (fldOffset=0x0)" P-INDEP;* V11 tmp10        [V11    ] (  0,  0   )     int  ->  zero-ref    "field V03._length (fldOffset=0x8)" P-INDEP;* V12 tmp11        [V12    ] (  0,  0   )   byref  ->  zero-ref    single-def "field V05._reference (fldOffset=0x0)" P-INDEP;* V13 tmp12        [V13    ] (  0,  0   )     int  ->  zero-ref    "field V05._length (fldOffset=0x8)" P-INDEP;* V14 tmp13        [V14    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M22062_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M22062_IG02:  ;; offset=0x0003       C5F828C8             vmovapsxmm1,xmm0       C5F0541531000000vandpsxmm2,xmm1, xmmword ptr[reloc @RWD00]       C5FB101D39000000     vmovsdxmm3, qword ptr[reloc @RWD16]       C5F92EDA             vucomisdxmm3,xmm27616jbe      SHORT G_M22062_IG04;; size=26 bbWeight=1 PerfScore 8.25G_M22062_IG03:  ;; offset=0x001D       C5FB590D33000000     vmulsdxmm1,xmm0, qword ptr[reloc @RWD24]       C4E3710BC90B         vroundsdxmm1,xmm1,xmm1,11       C5F35E0D25000000     vdivsdxmm1,xmm1, qword ptr[reloc @RWD24];; size=22 bbWeight=0.50 PerfScore 13.00G_M22062_IG04:  ;; offset=0x0033       C5F828C1             vmovapsxmm0,xmm1;; size=4 bbWeight=1 PerfScore 0.25G_M22062_IG05:  ;; offset=0x0037       C3ret;; size=1 bbWeight=1 PerfScore 1.00RWD00  dq7FFFFFFFFFFFFFFFh,7FFFFFFFFFFFFFFFhRWD16  dq4341C37937E08000h;        1e+16RWD24  dq40C3880000000000h;        10000; Total bytes of code 56, prolog size 3, PerfScore 23.50, instruction count 11, allocated bytes for code 56 (MethodHash=b456a9d1) for method RoundDisasm:Round4(double):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round4(float):float (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 1 single block inlinees; 3 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T01] (  4,  3.50)   float  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T00] (  4,  7   )   float  ->  mm1         "Inlining Arg";* V03 tmp2         [V03    ] (  0,  0   )  struct (16) zero-ref    ld-addr-op "Inline stloc first use temp" <System.ReadOnlySpan`1[float]>;* V04 tmp3         [V04,T04] (  0,  0   )   float  ->  zero-ref    "Inline stloc first use temp";* V05 tmp4         [V05    ] (  0,  0   )  struct (16) zero-ref    "ReadOnlySpan<T> for CreateSpan<T>" <System.ReadOnlySpan`1[float]>;  V06 tmp5         [V06,T03] (  2,  1   )   float  ->  mm1         "Inline return value spill temp";* V07 tmp6         [V07    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";  V08 tmp7         [V08,T02] (  2,  2   )   float  ->  mm1         "Inlining Arg";* V09 tmp8         [V09    ] (  0,  0   )   float  ->  zero-ref    "Inline return value spill temp";* V10 tmp9         [V10    ] (  0,  0   )   byref  ->  zero-ref    single-def "field V03._reference (fldOffset=0x0)" P-INDEP;* V11 tmp10        [V11    ] (  0,  0   )     int  ->  zero-ref    "field V03._length (fldOffset=0x8)" P-INDEP;* V12 tmp11        [V12    ] (  0,  0   )   byref  ->  zero-ref    single-def "field V05._reference (fldOffset=0x0)" P-INDEP;* V13 tmp12        [V13    ] (  0,  0   )     int  ->  zero-ref    "field V05._length (fldOffset=0x8)" P-INDEP;* V14 tmp13        [V14    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M65422_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M65422_IG02:  ;; offset=0x0003       C5F828C8             vmovapsxmm1,xmm0       C5F0541531000000vandpsxmm2,xmm1, xmmword ptr[reloc @RWD00]       C5FA101D39000000vmovssxmm3, dword ptr[reloc @RWD16]       C5F82EDA             vucomissxmm3,xmm27616jbe      SHORT G_M65422_IG04;; size=26 bbWeight=1 PerfScore 8.25G_M65422_IG03:  ;; offset=0x001D       C5FA590D2F000000     vmulssxmm1,xmm0, dword ptr[reloc @RWD20]       C4E3710AC90B         vroundssxmm1,xmm1,xmm1,11       C5F25E0D21000000     vdivssxmm1,xmm1, dword ptr[reloc @RWD20];; size=22 bbWeight=0.50 PerfScore 12.00G_M65422_IG04:  ;; offset=0x0033       C5F828C1             vmovapsxmm0,xmm1;; size=4 bbWeight=1 PerfScore 0.25G_M65422_IG05:  ;; offset=0x0037       C3ret;; size=1 bbWeight=1 PerfScore 1.00RWD00  dq7FFFFFFF7FFFFFFFh,7FFFFFFF7FFFFFFFhRWD16  dd4CBEBC20h;     1e+08RWD20  dd461C4000h;     10000; Total bytes of code 56, prolog size 3, PerfScore 22.50, instruction count 11, allocated bytes for code 56 (MethodHash=15960071) for method RoundDisasm:Round4(float):float (FullOpts); ============================================================

Fixes#98164.

Simplify and optimize Math(F).Round

159ec83

ghost added area-System.Numerics community-contributionIndicates that the PR has been added by a community member labels

Feb 8, 2024

Copy link

ghost commentedFeb 8, 2024

Tagging subscribers to this area: @dotnet/area-system-numerics
See info inarea-owners.md if you want to be subscribed.

Issue Details

Simplifies and optimizes parameterized Round overloads.

Code:

internalstaticclassRoundDisasm{publicstaticdoubleRoundAway(doubled)=>Math.Round(d,MidpointRounding.AwayFromZero);publicstaticdoubleRoundAway(floatf)=>MathF.Round(f,MidpointRounding.AwayFromZero);publicstaticdoubleRoundTruncate(doubled)=>Math.Round(d,MidpointRounding.ToZero);publicstaticdoubleRoundTruncate(floatf)=>MathF.Round(f,MidpointRounding.ToZero);publicstaticdoubleRound(doubled,MidpointRoundingm)=>Math.Round(d,m);publicstaticdoubleRound(floatf,MidpointRoundingm)=>MathF.Round(f,m);publicstaticdoubleRound2(doubled)=>Math.Round(d,2);publicstaticdoubleRound2(floatf)=>MathF.Round(f,2);publicstaticdoubleRound4(doubled)=>Math.Round(d,4,MidpointRounding.ToZero);publicstaticdoubleRound4(floatf)=>MathF.Round(f,4,MidpointRounding.ToZero);}

.NET 8

; Assembly listing for method RoundDisasm:RoundAway(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 3 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       C5F828C8             vmovapsxmm1,xmm0       C5F8101521000000vmovupsxmm2, xmmword ptr[reloc @RWD00]       C5F154CA             vandpdxmm1,xmm1,xmm2       62F1F518560D23000000vorpdxmm1,xmm1, qword ptr[reloc @RWD16] {1to2}       C5F358C0             vaddsdxmm0,xmm1,xmm0       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11G_M000_IG03:                ;; offset=0x0027       C3retRWD00  dq8000000000000000h,8000000000000000hRWD16  dq3FDFFFFFFFFFFFFFh;          0.5; Total bytes of code 40; Assembly listing for method RoundDisasm:RoundAway(float):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 3 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       C5F828C8             vmovapsxmm1,xmm0       C5F8101521000000vmovupsxmm2, xmmword ptr[reloc @RWD00]       C5F054CAvandpsxmm1,xmm1,xmm2       62F17418560D23000000vorpsxmm1,xmm1, dword ptr[reloc @RWD16] {1to4}       C5F258C0             vaddssxmm0,xmm1,xmm0       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11       C5FA5AC0             vcvtss2sdxmm0,xmm0,xmm0G_M000_IG03:                ;; offset=0x002B       C3retRWD00  dq8000000080000000h,8000000080000000hRWD16  dd3EFFFFFFh;       0.5; Total bytes of code 44; Assembly listing for method RoundDisasm:RoundTruncate(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 1 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       33D2xoredx,edx       41B802000000movr8d,2G_M000_IG03:                ;; offset=0x000B       FF259F041E00         tail.jmp[System.Math:Round(double,int,int):double]; Total bytes of code 17; Assembly listing for method RoundDisasm:RoundTruncate(float):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 1 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       4883EC28subrsp,40       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0007       33D2xoredx,edx       41B802000000movr8d,2       FF15F38F1E00call[System.MathF:Round(float,int,int):float]       C5FA5AC0             vcvtss2sdxmm0,xmm0,xmm0G_M000_IG03:                ;; offset=0x0019       4883C428addrsp,40       C3ret; Total bytes of code 30; Assembly listing for method RoundDisasm:Round(double,int):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 3 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       448BC2movr8d,edx       33D2xoredx,edxG_M000_IG03:                ;; offset=0x0008       FF2532041E00         tail.jmp[System.Math:Round(double,int,int):double]; Total bytes of code 14; Assembly listing for method RoundDisasm:Round(float,int):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 3 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       4883EC28subrsp,40       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0007       448BC2movr8d,edx       33D2xoredx,edx       FF15968F1E00call[System.MathF:Round(float,int,int):float]       C5FA5AC0             vcvtss2sdxmm0,xmm0,xmm0G_M000_IG03:                ;; offset=0x0016       4883C428addrsp,40       C3ret; Total bytes of code 27; Assembly listing for method RoundDisasm:Round2(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO data; 0 inlinees with PGO data; 1 single block inlinees; 0 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       BA02000000movedx,2       4533C0xorr8d,r8dG_M000_IG03:                ;; offset=0x000B       FF25DF031E00         tail.jmp[System.Math:Round(double,int,int):double]; Total bytes of code 17; Assembly listing for method RoundDisasm:Round2(float):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 1 single block inlinees; 0 inlinees without PGO dataG_M000_IG01:                ;; offset=0x0000       4883EC28subrsp,40       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0007       BA02000000movedx,2       4533C0xorr8d,r8d       FF15338F1E00call[System.MathF:Round(float,int,int):float]       C5FA5AC0             vcvtss2sdxmm0,xmm0,xmm0G_M000_IG03:                ;; offset=0x0019       4883C428addrsp,40       C3ret; Total bytes of code 30; Assembly listing for method RoundDisasm:Round4(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; fully interruptible; No PGO dataG_M000_IG01:                ;; offset=0x0000       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0003       BA04000000movedx,4       41B802000000movr8d,2G_M000_IG03:                ;; offset=0x000E       FF256C031E00         tail.jmp[System.Math:Round(double,int,int):double]; Total bytes of code 20; Assembly listing for method RoundDisasm:Round4(float):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO dataG_M000_IG01:                ;; offset=0x0000       4883EC28subrsp,40       C5F877vzeroupperG_M000_IG02:                ;; offset=0x0007       BA04000000movedx,4       41B802000000movr8d,2       FF15C08E1E00call[System.MathF:Round(float,int,int):float]       C5FA5AC0             vcvtss2sdxmm0,xmm0,xmm0G_M000_IG03:                ;; offset=0x001C       4883C428addrsp,40       C3ret; Total bytes of code 33

This branch

; Assembly listing for method RoundDisasm:RoundAway(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T00] (  4,  4   )  double  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  2,  2   )  double  ->  mm0         "Inline return value spill temp";* V03 tmp2         [V03    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V04 tmp3         [V04    ] (  0,  0   )  double  ->  zero-ref    "Inline return value spill temp";* V05 tmp4         [V05    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M25588_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M25588_IG02:  ;; offset=0x0003       C5F8100D25000000vmovupsxmm1, xmmword ptr[reloc @RWD00]       C5F828D0             vmovapsxmm2,xmm0       62F3ED08250D26000000CAvpternlogqxmm1,xmm2, xmmword ptr[reloc @RWD16],-54       C5F358C0             vaddsdxmm0,xmm1,xmm0       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11;; size=33 bbWeight=1 PerfScore 15.25G_M25588_IG03:  ;; offset=0x0024       C3ret;; size=1 bbWeight=1 PerfScore 1.00RWD00  dq8000000000000000h,8000000000000000hRWD16  dq3FDFFFFFFFFFFFFFh,3FDFFFFFFFFFFFFFh; Total bytes of code 37, prolog size 3, PerfScore 17.25, instruction count 7, allocated bytes for code 37 (MethodHash=318b9c0b) for method RoundDisasm:RoundAway(double):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:RoundAway(float):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T00] (  4,  4   )   float  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  2,  2   )   float  ->  mm0         "Inline return value spill temp";* V03 tmp2         [V03    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V04 tmp3         [V04    ] (  0,  0   )   float  ->  zero-ref    "Inline return value spill temp";* V05 tmp4         [V05    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M34641_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M34641_IG02:  ;; offset=0x0003       C5F8100D25000000vmovupsxmm1, xmmword ptr[reloc @RWD00]       C5F828D0             vmovapsxmm2,xmm0       62F36D08250D26000000CAvpternlogdxmm1,xmm2, xmmword ptr[reloc @RWD16],-54       C5F258C0             vaddssxmm0,xmm1,xmm0       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11       C5FA5AC0             vcvtss2sdxmm0,xmm0,xmm0;; size=37 bbWeight=1 PerfScore 19.25G_M34641_IG03:  ;; offset=0x0028       C3ret;; size=1 bbWeight=1 PerfScore 1.00RWD00  dq8000000080000000h,8000000080000000hRWD16  dq3EFFFFFF3EFFFFFFh,3EFFFFFF3EFFFFFFh; Total bytes of code 41, prolog size 3, PerfScore 21.25, instruction count 8, allocated bytes for code 41 (MethodHash=f9e078ae) for method RoundDisasm:RoundAway(float):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:RoundTruncate(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T00] (  3,  3   )  double  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  2,  2   )  double  ->  mm0         "Inline return value spill temp";* V03 tmp2         [V03    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V04 tmp3         [V04    ] (  0,  0   )  double  ->  zero-ref    "Inline return value spill temp";* V05 tmp4         [V05    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M39156_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M39156_IG02:  ;; offset=0x0003       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11;; size=6 bbWeight=1 PerfScore 7.00G_M39156_IG03:  ;; offset=0x0009       C3ret;; size=1 bbWeight=1 PerfScore 1.00; Total bytes of code 10, prolog size 3, PerfScore 9.00, instruction count 3, allocated bytes for code 10 (MethodHash=6fe1670b) for method RoundDisasm:RoundTruncate(double):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:RoundTruncate(float):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T00] (  3,  3   )   float  ->  mm0         single-def;# V01 OutArgs      [V01    ] (  1,  1   )  struct ( 0) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  2,  2   )   float  ->  mm0         "Inline return value spill temp";* V03 tmp2         [V03    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V04 tmp3         [V04    ] (  0,  0   )   float  ->  zero-ref    "Inline return value spill temp";* V05 tmp4         [V05    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 0G_M7249_IG01:  ;; offset=0x0000       C5F877vzeroupper;; size=3 bbWeight=1 PerfScore 1.00G_M7249_IG02:  ;; offset=0x0003       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11       C5FA5AC0             vcvtss2sdxmm0,xmm0,xmm0;; size=10 bbWeight=1 PerfScore 11.00G_M7249_IG03:  ;; offset=0x000D       C3ret;; size=1 bbWeight=1 PerfScore 1.00; Total bytes of code 14, prolog size 3, PerfScore 13.00, instruction count 4, allocated bytes for code 14 (MethodHash=8f28e3ae) for method RoundDisasm:RoundTruncate(float):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round(double,int):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T01] (  8,  5   )  double  ->  mm0         single-def;  V01 arg1         [V01,T00] (  5,  4   )     int  ->  rbx         single-def;  V02 OutArgs      [V02    ] (  1,  1   )  struct (32) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V03 tmp1         [V03,T02] (  6,  3.50)  double  ->  mm0         "Inline return value spill temp";* V04 tmp2         [V04    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V05 tmp3         [V05    ] (  0,  0   )  double  ->  zero-ref    "Inline return value spill temp";  V06 tmp4         [V06,T03] (  2,  0   )     ref  ->  rdx         single-def "argument with side effect";; Lcl frame size = 32G_M6149_IG01:  ;; offset=0x000053pushrbx       4883EC20subrsp,32       C5F877vzeroupper       8BDAmovebx,edx;; size=10 bbWeight=1 PerfScore 2.50G_M6149_IG02:  ;; offset=0x000A       83FB04cmpebx,4       775Fja       SHORT G_M6149_IG09       8BCBmovecx,ebx       488D1578000000leardx,[reloc @RWD00]       8B148Amovedx, dword ptr[rdx+4*rcx]       488D05E8FFFFFFlearax, G_M6149_IG02       4803D0addrdx,rax       FFE2jmprdx;; size=29 bbWeight=1 PerfScore 7.25G_M6149_IG03:  ;; offset=0x0027       C5F8100D81000000vmovupsxmm1, xmmword ptr[reloc @RWD32]       C5F828D0             vmovapsxmm2,xmm0       62F3ED08250D82000000CAvpternlogqxmm1,xmm2, xmmword ptr[reloc @RWD48],-54       C5F358C0             vaddsdxmm0,xmm1,xmm0       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11       EB1Ejmp      SHORT G_M6149_IG08;; size=35 bbWeight=0.50 PerfScore 8.62G_M6149_IG04:  ;; offset=0x004A       C4E3790BC004         vroundsdxmm0,xmm0,xmm0,4       EB16jmp      SHORT G_M6149_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M6149_IG05:  ;; offset=0x0052       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11       EB0Ejmp      SHORT G_M6149_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M6149_IG06:  ;; offset=0x005A       C4E3790BC009         vroundsdxmm0,xmm0,xmm0,9       EB06jmp      SHORT G_M6149_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M6149_IG07:  ;; offset=0x0062       C4E3790BC00A         vroundsdxmm0,xmm0,xmm0,10;; size=6 bbWeight=0.50 PerfScore 3.50G_M6149_IG08:  ;; offset=0x0068       4883C420addrsp,32       5Bpoprbx       C3ret;; size=6 bbWeight=1 PerfScore 1.75G_M6149_IG09:  ;; offset=0x006E       B9840B0000movecx,0xB84       48BA004022BEFA7F0000movrdx,0x7FFABE224000       E88EE44E5Fcall     CORINFO_HELP_STRCNS       488BD0movrdx,rax       8BCBmovecx,ebx       FF1533947C00call[System.ThrowHelper:ThrowArgumentException_InvalidEnumValue[int](int,System.String)]       CCint3;; size=32 bbWeight=0 PerfScore 0.00RWD00  dd00000040h ; case G_M6149_IG04dd0000001Dh ; case G_M6149_IG03dd00000048h ; case G_M6149_IG05dd00000050h ; case G_M6149_IG06dd00000058h ; case G_M6149_IG07RWD20  dd00000000h,00000000h,00000000hRWD32  dq8000000000000000h,8000000000000000hRWD48  dq3FDFFFFFFFFFFFFFh,3FDFFFFFFFFFFFFFh; Total bytes of code 142, prolog size 8, PerfScore 37.12, instruction count 35, allocated bytes for code 142 (MethodHash=d24ae7fa) for method RoundDisasm:Round(double,int):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round(float,int):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 0 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T01] (  8,  5   )   float  ->  mm0         single-def;  V01 arg1         [V01,T00] (  5,  4   )     int  ->  rbx         single-def;  V02 OutArgs      [V02    ] (  1,  1   )  struct (32) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V03 tmp1         [V03,T02] (  6,  3.50)   float  ->  mm0         "Inline return value spill temp";* V04 tmp2         [V04    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";* V05 tmp3         [V05    ] (  0,  0   )   float  ->  zero-ref    "Inline return value spill temp";  V06 tmp4         [V06,T03] (  2,  0   )     ref  ->  rdx         single-def "argument with side effect";; Lcl frame size = 32G_M39904_IG01:  ;; offset=0x000053pushrbx       4883EC20subrsp,32       C5F877vzeroupper       8BDAmovebx,edx;; size=10 bbWeight=1 PerfScore 2.50G_M39904_IG02:  ;; offset=0x000A       83FB04cmpebx,47763ja       SHORT G_M39904_IG10       8BCBmovecx,ebx       488D1588000000leardx,[reloc @RWD00]       8B148Amovedx, dword ptr[rdx+4*rcx]       488D05E8FFFFFFlearax, G_M39904_IG02       4803D0addrdx,rax       FFE2jmprdx;; size=29 bbWeight=1 PerfScore 7.25G_M39904_IG03:  ;; offset=0x0027       C5F8100D91000000vmovupsxmm1, xmmword ptr[reloc @RWD32]       C5F828D0             vmovapsxmm2,xmm0       62F36D08250D92000000CAvpternlogdxmm1,xmm2, xmmword ptr[reloc @RWD48],-54       C5F258C0             vaddssxmm0,xmm1,xmm0       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11       EB1Ejmp      SHORT G_M39904_IG08;; size=35 bbWeight=0.50 PerfScore 8.62G_M39904_IG04:  ;; offset=0x004A       C4E3790AC004         vroundssxmm0,xmm0,xmm0,4       EB16jmp      SHORT G_M39904_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M39904_IG05:  ;; offset=0x0052       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11       EB0Ejmp      SHORT G_M39904_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M39904_IG06:  ;; offset=0x005A       C4E3790AC009         vroundssxmm0,xmm0,xmm0,9       EB06jmp      SHORT G_M39904_IG08;; size=8 bbWeight=0.50 PerfScore 4.50G_M39904_IG07:  ;; offset=0x0062       C4E3790AC00A         vroundssxmm0,xmm0,xmm0,10;; size=6 bbWeight=0.50 PerfScore 3.50G_M39904_IG08:  ;; offset=0x0068       C5FA5AC0             vcvtss2sdxmm0,xmm0,xmm0;; size=4 bbWeight=1 PerfScore 4.00G_M39904_IG09:  ;; offset=0x006C       4883C420addrsp,32       5Bpoprbx       C3ret;; size=6 bbWeight=1 PerfScore 1.75G_M39904_IG10:  ;; offset=0x0072       B9840B0000movecx,0xB84       48BA004022BEFA7F0000movrdx,0x7FFABE224000       E89AE34E5Fcall     CORINFO_HELP_STRCNS       488BD0movrdx,rax       8BCBmovecx,ebx       FF153F937C00call[System.ThrowHelper:ThrowArgumentException_InvalidEnumValue[int](int,System.String)]       CCint3;; size=32 bbWeight=0 PerfScore 0.00RWD00  dd00000040h ; case G_M39904_IG04dd0000001Dh ; case G_M39904_IG03dd00000048h ; case G_M39904_IG05dd00000050h ; case G_M39904_IG06dd00000058h ; case G_M39904_IG07RWD20  dd00000000h,00000000h,00000000hRWD32  dq8000000080000000h,8000000080000000hRWD48  dq3EFFFFFF3EFFFFFFh,3EFFFFFF3EFFFFFFh; Total bytes of code 146, prolog size 8, PerfScore 41.12, instruction count 36, allocated bytes for code 146 (MethodHash=8938641f) for method RoundDisasm:Round(float,int):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round2(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 1 inlinees with PGO data; 2 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T02] (  4,  3.50)  double  ->  mm0         single-def;  V01 OutArgs      [V01    ] (  1,  1   )  struct (32) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  4,  7   )  double  ->  mm1         "Inlining Arg";* V03 tmp2         [V03    ] (  0,  0   )  struct (16) zero-ref    do-not-enreg[SF] ld-addr-op "Inline stloc first use temp" <System.ReadOnlySpan`1[double]>;  V04 tmp3         [V04,T04] (  3,  1.50)  double  ->  mm1         "Inline stloc first use temp";  V05 tmp4         [V05,T00] (  3,  3   )  struct (16) [rsp+0x28]  do-not-enreg[SF] must-init "ReadOnlySpan<T> for CreateSpan<T>" <System.ReadOnlySpan`1[double]>;  V06 tmp5         [V06,T05] (  2,  1   )  double  ->  mm0         "Inline return value spill temp";* V07 tmp6         [V07    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";  V08 tmp7         [V08,T03] (  2,  2   )  double  ->  mm0         "Inlining Arg";* V09 tmp8         [V09    ] (  0,  0   )  double  ->  zero-ref    "Inline return value spill temp";* V10 tmp9         [V10    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 56G_M8040_IG01:  ;; offset=0x0000       4883EC38subrsp,56       C5F877vzeroupper       33C0xoreax,eax4889442428mov      qword ptr[rsp+0x28],rax;; size=14 bbWeight=1 PerfScore 2.50G_M8040_IG02:  ;; offset=0x000E       C5F828C8             vmovapsxmm1,xmm0       C5F0541546000000vandpsxmm2,xmm1, xmmword ptr[reloc @RWD00]       C5FB101D4E000000     vmovsdxmm3, qword ptr[reloc @RWD16]       C5F92EDA             vucomisdxmm3,xmm2       762Fjbe      SHORT G_M8040_IG04;; size=26 bbWeight=1 PerfScore 8.25G_M8040_IG03:  ;; offset=0x0028       C744243010000000mov      dword ptr[rsp+0x30],16       48B8206AB71AFB7F0000movrax,0x7FFB1AB76A20      ; static handle4889442428mov      bword ptr[rsp+0x28],rax       488B442428movrax, bword ptr[rsp+0x28]       C5FB104810           vmovsdxmm1, qword ptr[rax+0x10]       C5FB59C1             vmulsdxmm0,xmm0,xmm1       C4E3790BC004         vroundsdxmm0,xmm0,xmm0,4       C5FB5EC9             vdivsdxmm1,xmm0,xmm1;; size=47 bbWeight=0.50 PerfScore 14.62G_M8040_IG04:  ;; offset=0x0057       C5F828C1             vmovapsxmm0,xmm1;; size=4 bbWeight=1 PerfScore 0.25G_M8040_IG05:  ;; offset=0x005B       4883C438addrsp,56       C3ret;; size=5 bbWeight=1 PerfScore 1.25RWD00  dq7FFFFFFFFFFFFFFFh,7FFFFFFFFFFFFFFFhRWD16  dq4341C37937E08000h;        1e+16; Total bytes of code 96, prolog size 14, PerfScore 26.88, instruction count 20, allocated bytes for code 96 (MethodHash=166ce097) for method RoundDisasm:Round2(double):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round2(float):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 2 single block inlinees; 3 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T02] (  4,  3.50)   float  ->  mm0         single-def;  V01 OutArgs      [V01    ] (  1,  1   )  struct (32) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  4,  7   )   float  ->  mm1         "Inlining Arg";* V03 tmp2         [V03    ] (  0,  0   )  struct (16) zero-ref    do-not-enreg[SF] ld-addr-op "Inline stloc first use temp" <System.ReadOnlySpan`1[float]>;  V04 tmp3         [V04,T04] (  3,  1.50)   float  ->  mm1         "Inline stloc first use temp";  V05 tmp4         [V05,T00] (  3,  3   )  struct (16) [rsp+0x28]  do-not-enreg[SF] must-init "ReadOnlySpan<T> for CreateSpan<T>" <System.ReadOnlySpan`1[float]>;  V06 tmp5         [V06,T05] (  2,  1   )   float  ->  mm0         "Inline return value spill temp";* V07 tmp6         [V07    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";  V08 tmp7         [V08,T03] (  2,  2   )   float  ->  mm0         "Inlining Arg";* V09 tmp8         [V09    ] (  0,  0   )   float  ->  zero-ref    "Inline return value spill temp";* V10 tmp9         [V10    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 56G_M52813_IG01:  ;; offset=0x0000       4883EC38subrsp,56       C5F877vzeroupper       33C0xoreax,eax4889442428mov      qword ptr[rsp+0x28],rax;; size=14 bbWeight=1 PerfScore 2.50G_M52813_IG02:  ;; offset=0x000E       C5F828C8             vmovapsxmm1,xmm0       C5F0541546000000vandpsxmm2,xmm1, xmmword ptr[reloc @RWD00]       C5FA101D4E000000vmovssxmm3, dword ptr[reloc @RWD16]       C5F82EDA             vucomissxmm3,xmm2       762Fjbe      SHORT G_M52813_IG04;; size=26 bbWeight=1 PerfScore 8.25G_M52813_IG03:  ;; offset=0x0028       C744243007000000mov      dword ptr[rsp+0x30],7       48B868DCB71AFB7F0000movrax,0x7FFB1AB7DC68      ; static handle4889442428mov      bword ptr[rsp+0x28],rax       488B442428movrax, bword ptr[rsp+0x28]       C5FA104808vmovssxmm1, dword ptr[rax+0x08]       C5FA59C1             vmulssxmm0,xmm0,xmm1       C4E3790AC004         vroundssxmm0,xmm0,xmm0,4       C5FA5EC9             vdivssxmm1,xmm0,xmm1;; size=47 bbWeight=0.50 PerfScore 13.62G_M52813_IG04:  ;; offset=0x0057       C5FA5AC1             vcvtss2sdxmm0,xmm0,xmm1;; size=4 bbWeight=1 PerfScore 4.00G_M52813_IG05:  ;; offset=0x005B       4883C438addrsp,56       C3ret;; size=5 bbWeight=1 PerfScore 1.25RWD00  dq7FFFFFFF7FFFFFFFh,7FFFFFFF7FFFFFFFhRWD16  dd4CBEBC20h;     1e+08; Total bytes of code 96, prolog size 14, PerfScore 29.62, instruction count 20, allocated bytes for code 96 (MethodHash=81b131b2) for method RoundDisasm:Round2(float):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round4(double):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 1 inlinees with PGO data; 1 single block inlinees; 2 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T02] (  4,  3.50)  double  ->  mm0         single-def;  V01 OutArgs      [V01    ] (  1,  1   )  struct (32) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  4,  7   )  double  ->  mm1         "Inlining Arg";* V03 tmp2         [V03    ] (  0,  0   )  struct (16) zero-ref    do-not-enreg[SF] ld-addr-op "Inline stloc first use temp" <System.ReadOnlySpan`1[double]>;  V04 tmp3         [V04,T04] (  3,  1.50)  double  ->  mm1         "Inline stloc first use temp";  V05 tmp4         [V05,T00] (  3,  3   )  struct (16) [rsp+0x28]  do-not-enreg[SF] must-init "ReadOnlySpan<T> for CreateSpan<T>" <System.ReadOnlySpan`1[double]>;  V06 tmp5         [V06,T05] (  2,  1   )  double  ->  mm0         "Inline return value spill temp";* V07 tmp6         [V07    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";  V08 tmp7         [V08,T03] (  2,  2   )  double  ->  mm0         "Inlining Arg";* V09 tmp8         [V09    ] (  0,  0   )  double  ->  zero-ref    "Inline return value spill temp";* V10 tmp9         [V10    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 56G_M22062_IG01:  ;; offset=0x0000       4883EC38subrsp,56       C5F877vzeroupper       33C0xoreax,eax4889442428mov      qword ptr[rsp+0x28],rax;; size=14 bbWeight=1 PerfScore 2.50G_M22062_IG02:  ;; offset=0x000E       C5F828C8             vmovapsxmm1,xmm0       C5F0541546000000vandpsxmm2,xmm1, xmmword ptr[reloc @RWD00]       C5FB101D4E000000     vmovsdxmm3, qword ptr[reloc @RWD16]       C5F92EDA             vucomisdxmm3,xmm2       762Fjbe      SHORT G_M22062_IG04;; size=26 bbWeight=1 PerfScore 8.25G_M22062_IG03:  ;; offset=0x0028       C744243010000000mov      dword ptr[rsp+0x30],16       48B8206AB71AFB7F0000movrax,0x7FFB1AB76A20      ; static handle4889442428mov      bword ptr[rsp+0x28],rax       488B442428movrax, bword ptr[rsp+0x28]       C5FB104820           vmovsdxmm1, qword ptr[rax+0x20]       C5FB59C1             vmulsdxmm0,xmm0,xmm1       C4E3790BC00B         vroundsdxmm0,xmm0,xmm0,11       C5FB5EC9             vdivsdxmm1,xmm0,xmm1;; size=47 bbWeight=0.50 PerfScore 14.62G_M22062_IG04:  ;; offset=0x0057       C5F828C1             vmovapsxmm0,xmm1;; size=4 bbWeight=1 PerfScore 0.25G_M22062_IG05:  ;; offset=0x005B       4883C438addrsp,56       C3ret;; size=5 bbWeight=1 PerfScore 1.25RWD00  dq7FFFFFFFFFFFFFFFh,7FFFFFFFFFFFFFFFhRWD16  dq4341C37937E08000h;        1e+16; Total bytes of code 96, prolog size 14, PerfScore 26.88, instruction count 20, allocated bytes for code 96 (MethodHash=b456a9d1) for method RoundDisasm:Round4(double):double (FullOpts); ============================================================; Assembly listing for method RoundDisasm:Round4(float):double (FullOpts); Emitting BLENDED_CODE for X64 with AVX512 - Windows; FullOpts code; optimized code; rsp based frame; partially interruptible; No PGO data; 0 inlinees with PGO data; 1 single block inlinees; 3 inlinees without PGO data; Final local variable assignments;;  V00 arg0         [V00,T02] (  4,  3.50)   float  ->  mm0         single-def;  V01 OutArgs      [V01    ] (  1,  1   )  struct (32) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace";  V02 tmp1         [V02,T01] (  4,  7   )   float  ->  mm1         "Inlining Arg";* V03 tmp2         [V03    ] (  0,  0   )  struct (16) zero-ref    do-not-enreg[SF] ld-addr-op "Inline stloc first use temp" <System.ReadOnlySpan`1[float]>;  V04 tmp3         [V04,T04] (  3,  1.50)   float  ->  mm1         "Inline stloc first use temp";  V05 tmp4         [V05,T00] (  3,  3   )  struct (16) [rsp+0x28]  do-not-enreg[SF] must-init "ReadOnlySpan<T> for CreateSpan<T>" <System.ReadOnlySpan`1[float]>;  V06 tmp5         [V06,T05] (  2,  1   )   float  ->  mm0         "Inline return value spill temp";* V07 tmp6         [V07    ] (  0,  0   )     int  ->  zero-ref    "Inline stloc first use temp";  V08 tmp7         [V08,T03] (  2,  2   )   float  ->  mm0         "Inlining Arg";* V09 tmp8         [V09    ] (  0,  0   )   float  ->  zero-ref    "Inline return value spill temp";* V10 tmp9         [V10    ] (  0,  0   )     ref  ->  zero-ref    "argument with side effect";; Lcl frame size = 56G_M34635_IG01:  ;; offset=0x0000       4883EC38subrsp,56       C5F877vzeroupper       33C0xoreax,eax4889442428mov      qword ptr[rsp+0x28],rax;; size=14 bbWeight=1 PerfScore 2.50G_M34635_IG02:  ;; offset=0x000E       C5F828C8             vmovapsxmm1,xmm0       C5F0541546000000vandpsxmm2,xmm1, xmmword ptr[reloc @RWD00]       C5FA101D4E000000vmovssxmm3, dword ptr[reloc @RWD16]       C5F82EDA             vucomissxmm3,xmm2       762Fjbe      SHORT G_M34635_IG04;; size=26 bbWeight=1 PerfScore 8.25G_M34635_IG03:  ;; offset=0x0028       C744243007000000mov      dword ptr[rsp+0x30],7       48B868DCB71AFB7F0000movrax,0x7FFB1AB7DC68      ; static handle4889442428mov      bword ptr[rsp+0x28],rax       488B442428movrax, bword ptr[rsp+0x28]       C5FA104810vmovssxmm1, dword ptr[rax+0x10]       C5FA59C1             vmulssxmm0,xmm0,xmm1       C4E3790AC00B         vroundssxmm0,xmm0,xmm0,11       C5FA5EC9             vdivssxmm1,xmm0,xmm1;; size=47 bbWeight=0.50 PerfScore 13.62G_M34635_IG04:  ;; offset=0x0057       C5FA5AC1             vcvtss2sdxmm0,xmm0,xmm1;; size=4 bbWeight=1 PerfScore 4.00G_M34635_IG05:  ;; offset=0x005B       4883C438addrsp,56       C3ret;; size=5 bbWeight=1 PerfScore 1.25RWD00  dq7FFFFFFF7FFFFFFFh,7FFFFFFF7FFFFFFFhRWD16  dd4CBEBC20h;     1e+08; Total bytes of code 96, prolog size 14, PerfScore 29.62, instruction count 20, allocated bytes for code 96 (MethodHash=cd4078b4) for method RoundDisasm:Round4(float):double (FullOpts); ============================================================