Skip to content
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.

Sign up for GitHub

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

Already on GitHub? Sign in to your account

Jump to bottom

Simplify and optimize Math(F).Round #98186

Merged
merged 2 commits into from
Feb 9, 2024
Merged

Simplify and optimize Math(F).Round #98186

merged 2 commits into from
Feb 9, 2024

Conversation

MichalPetryka
Copy link
Contributor

@MichalPetryka MichalPetryka commented Feb 8, 2024 •
edited
Loading

Simplifies and optimizes parameterized Round overloads.

Code:

internal static class RoundDisasm
{
	public static double RoundAway(double d) => Math.Round(d, MidpointRounding.AwayFromZero);
	public static float RoundAway(float f) => MathF.Round(f, MidpointRounding.AwayFromZero);
	public static double RoundTruncate(double d) => Math.Round(d, MidpointRounding.ToZero);
	public static float RoundTruncate(float f) => MathF.Round(f, MidpointRounding.ToZero);
	public static double Round(double d, MidpointRounding m) => Math.Round(d, m);
	public static float Round(float f, MidpointRounding m) => MathF.Round(f, m);
	public static double Round2(double d) => Math.Round(d, 2);
	public static float Round2(float f) => MathF.Round(f, 2);
	public static double Round4(double d) => Math.Round(d, 4, MidpointRounding.ToZero);
	public static float Round4(float f) => 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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       C5F828C8             vmovaps  xmm1, xmm0
       C5F8101521000000     vmovups  xmm2, xmmword ptr [reloc @RWD00]
       C5F154CA             vandpd   xmm1, xmm1, xmm2
       62F1F518560D23000000 vorpd    xmm1, xmm1, qword ptr [reloc @RWD16] {1to2}
       C5F358C0             vaddsd   xmm0, xmm1, xmm0
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
 
G_M000_IG03:                ;; offset=0x0027
       C3                   ret      
 
RWD00  	dq	8000000000000000h, 8000000000000000h
RWD16  	dq	3FDFFFFFFFFFFFFFh	;          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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       C5F828C8             vmovaps  xmm1, xmm0
       C5F8101521000000     vmovups  xmm2, xmmword ptr [reloc @RWD00]
       C5F054CA             vandps   xmm1, xmm1, xmm2
       62F17418560D23000000 vorps    xmm1, xmm1, dword ptr [reloc @RWD16] {1to4}
       C5F258C0             vaddss   xmm0, xmm1, xmm0
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
 
G_M000_IG03:                ;; offset=0x0027
       C3                   ret      
 
RWD00  	dq	8000000080000000h, 8000000080000000h
RWD16  	dd	3EFFFFFFh		;       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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       33D2                 xor      edx, edx
       41B802000000         mov      r8d, 2
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       33D2                 xor      edx, edx
       41B802000000         mov      r8d, 2
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       448BC2               mov      r8d, edx
       33D2                 xor      edx, edx
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       448BC2               mov      r8d, edx
       33D2                 xor      edx, edx
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       BA02000000           mov      edx, 2
       4533C0               xor      r8d, r8d
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       BA02000000           mov      edx, 2
       4533C0               xor      r8d, r8d
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       BA04000000           mov      edx, 4
       41B802000000         mov      r8d, 2
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       BA04000000           mov      edx, 4
       41B802000000         mov      r8d, 2
 
G_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 = 0

G_M25588_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M25588_IG02:  ;; offset=0x0003
       C5F8100D25000000     vmovups  xmm1, xmmword ptr [reloc @RWD00]
       C5F828D0             vmovaps  xmm2, xmm0
       62F3ED08250D26000000CA vpternlogq xmm1, xmm2, xmmword ptr [reloc @RWD16], -54
       C5F358C0             vaddsd   xmm0, xmm1, xmm0
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
						;; size=33 bbWeight=1 PerfScore 15.25
G_M25588_IG03:  ;; offset=0x0024
       C3                   ret      
						;; size=1 bbWeight=1 PerfScore 1.00
RWD00  	dq	8000000000000000h, 8000000000000000h
RWD16  	dq	3FDFFFFFFFFFFFFFh, 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 = 0

G_M10964_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M10964_IG02:  ;; offset=0x0003
       C5F8100D25000000     vmovups  xmm1, xmmword ptr [reloc @RWD00]
       C5F828D0             vmovaps  xmm2, xmm0
       62F36D08250D26000000CA vpternlogd xmm1, xmm2, xmmword ptr [reloc @RWD16], -54
       C5F258C0             vaddss   xmm0, xmm1, xmm0
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
						;; size=33 bbWeight=1 PerfScore 15.25
G_M10964_IG03:  ;; offset=0x0024
       C3                   ret      
						;; size=1 bbWeight=1 PerfScore 1.00
RWD00  	dq	8000000080000000h, 8000000080000000h
RWD16  	dq	3EFFFFFF3EFFFFFFh, 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 = 0

G_M39156_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M39156_IG02:  ;; offset=0x0003
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
						;; size=6 bbWeight=1 PerfScore 7.00
G_M39156_IG03:  ;; offset=0x0009
       C3                   ret      
						;; 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 = 0

G_M8148_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M8148_IG02:  ;; offset=0x0003
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
						;; size=6 bbWeight=1 PerfScore 7.00
G_M8148_IG03:  ;; offset=0x0009
       C3                   ret      
						;; 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 = 32

G_M6149_IG01:  ;; offset=0x0000
       53                   push     rbx
       4883EC20             sub      rsp, 32
       C5F877               vzeroupper 
       8BDA                 mov      ebx, edx
						;; size=10 bbWeight=1 PerfScore 2.50
G_M6149_IG02:  ;; offset=0x000A
       83FB04               cmp      ebx, 4
       775F                 ja       SHORT G_M6149_IG09
       8BCB                 mov      ecx, ebx
       488D1578000000       lea      rdx, [reloc @RWD00]
       8B148A               mov      edx, dword ptr [rdx+4*rcx]
       488D05E8FFFFFF       lea      rax, G_M6149_IG02
       4803D0               add      rdx, rax
       FFE2                 jmp      rdx
						;; size=29 bbWeight=1 PerfScore 7.25
G_M6149_IG03:  ;; offset=0x0027
       C5F8100D81000000     vmovups  xmm1, xmmword ptr [reloc @RWD32]
       C5F828D0             vmovaps  xmm2, xmm0
       62F3ED08250D82000000CA vpternlogq xmm1, xmm2, xmmword ptr [reloc @RWD48], -54
       C5F358C0             vaddsd   xmm0, xmm1, xmm0
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
       EB1E                 jmp      SHORT G_M6149_IG08
						;; size=35 bbWeight=0.50 PerfScore 8.62
G_M6149_IG04:  ;; offset=0x004A
       C4E3790BC004         vroundsd xmm0, xmm0, xmm0, 4
       EB16                 jmp      SHORT G_M6149_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M6149_IG05:  ;; offset=0x0052
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
       EB0E                 jmp      SHORT G_M6149_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M6149_IG06:  ;; offset=0x005A
       C4E3790BC009         vroundsd xmm0, xmm0, xmm0, 9
       EB06                 jmp      SHORT G_M6149_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M6149_IG07:  ;; offset=0x0062
       C4E3790BC00A         vroundsd xmm0, xmm0, xmm0, 10
						;; size=6 bbWeight=0.50 PerfScore 3.50
G_M6149_IG08:  ;; offset=0x0068
       4883C420             add      rsp, 32
       5B                   pop      rbx
       C3                   ret      
						;; size=6 bbWeight=1 PerfScore 1.75
G_M6149_IG09:  ;; offset=0x006E
       B9840B0000           mov      ecx, 0xB84
       48BA00400F6CFD7F0000 mov      rdx, 0x7FFD6C0F4000
       E83EE64D5F           call     CORINFO_HELP_STRCNS
       488BD0               mov      rdx, rax
       8BCB                 mov      ecx, ebx
       FF15E3957C00         call     [System.ThrowHelper:ThrowArgumentException_InvalidEnumValue[int](int,System.String)]
       CC                   int3     
						;; size=32 bbWeight=0 PerfScore 0.00
RWD00  	dd	00000040h ; case G_M6149_IG04
       	dd	0000001Dh ; case G_M6149_IG03
       	dd	00000048h ; case G_M6149_IG05
       	dd	00000050h ; case G_M6149_IG06
       	dd	00000058h ; case G_M6149_IG07
RWD20  	dd	00000000h, 00000000h, 00000000h
RWD32  	dq	8000000000000000h, 8000000000000000h
RWD48  	dq	3FDFFFFFFFFFFFFFh, 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 = 32

G_M10373_IG01:  ;; offset=0x0000
       53                   push     rbx
       4883EC20             sub      rsp, 32
       C5F877               vzeroupper 
       8BDA                 mov      ebx, edx
						;; size=10 bbWeight=1 PerfScore 2.50
G_M10373_IG02:  ;; offset=0x000A
       83FB04               cmp      ebx, 4
       775F                 ja       SHORT G_M10373_IG09
       8BCB                 mov      ecx, ebx
       488D1578000000       lea      rdx, [reloc @RWD00]
       8B148A               mov      edx, dword ptr [rdx+4*rcx]
       488D05E8FFFFFF       lea      rax, G_M10373_IG02
       4803D0               add      rdx, rax
       FFE2                 jmp      rdx
						;; size=29 bbWeight=1 PerfScore 7.25
G_M10373_IG03:  ;; offset=0x0027
       C5F8100D81000000     vmovups  xmm1, xmmword ptr [reloc @RWD32]
       C5F828D0             vmovaps  xmm2, xmm0
       62F36D08250D82000000CA vpternlogd xmm1, xmm2, xmmword ptr [reloc @RWD48], -54
       C5F258C0             vaddss   xmm0, xmm1, xmm0
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
       EB1E                 jmp      SHORT G_M10373_IG08
						;; size=35 bbWeight=0.50 PerfScore 8.62
G_M10373_IG04:  ;; offset=0x004A
       C4E3790AC004         vroundss xmm0, xmm0, xmm0, 4
       EB16                 jmp      SHORT G_M10373_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M10373_IG05:  ;; offset=0x0052
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
       EB0E                 jmp      SHORT G_M10373_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M10373_IG06:  ;; offset=0x005A
       C4E3790AC009         vroundss xmm0, xmm0, xmm0, 9
       EB06                 jmp      SHORT G_M10373_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M10373_IG07:  ;; offset=0x0062
       C4E3790AC00A         vroundss xmm0, xmm0, xmm0, 10
						;; size=6 bbWeight=0.50 PerfScore 3.50
G_M10373_IG08:  ;; offset=0x0068
       4883C420             add      rsp, 32
       5B                   pop      rbx
       C3                   ret      
						;; size=6 bbWeight=1 PerfScore 1.75
G_M10373_IG09:  ;; offset=0x006E
       B9840B0000           mov      ecx, 0xB84
       48BA00400F6CFD7F0000 mov      rdx, 0x7FFD6C0F4000
       E84EE54D5F           call     CORINFO_HELP_STRCNS
       488BD0               mov      rdx, rax
       8BCB                 mov      ecx, ebx
       FF15F3947C00         call     [System.ThrowHelper:ThrowArgumentException_InvalidEnumValue[int](int,System.String)]
       CC                   int3     
						;; size=32 bbWeight=0 PerfScore 0.00
RWD00  	dd	00000040h ; case G_M10373_IG04
       	dd	0000001Dh ; case G_M10373_IG03
       	dd	00000048h ; case G_M10373_IG05
       	dd	00000050h ; case G_M10373_IG06
       	dd	00000058h ; case G_M10373_IG07
RWD20  	dd	00000000h, 00000000h, 00000000h
RWD32  	dq	8000000080000000h, 8000000080000000h
RWD48  	dq	3EFFFFFF3EFFFFFFh, 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 = 0

G_M8040_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M8040_IG02:  ;; offset=0x0003
       C5F828C8             vmovaps  xmm1, xmm0
       C5F0541531000000     vandps   xmm2, xmm1, xmmword ptr [reloc @RWD00]
       C5FB101D39000000     vmovsd   xmm3, qword ptr [reloc @RWD16]
       C5F92EDA             vucomisd xmm3, xmm2
       7616                 jbe      SHORT G_M8040_IG04
						;; size=26 bbWeight=1 PerfScore 8.25
G_M8040_IG03:  ;; offset=0x001D
       C5FB590D33000000     vmulsd   xmm1, xmm0, qword ptr [reloc @RWD24]
       C4E3710BC904         vroundsd xmm1, xmm1, xmm1, 4
       C5F35E0D25000000     vdivsd   xmm1, xmm1, qword ptr [reloc @RWD24]
						;; size=22 bbWeight=0.50 PerfScore 13.00
G_M8040_IG04:  ;; offset=0x0033
       C5F828C1             vmovaps  xmm0, xmm1
						;; size=4 bbWeight=1 PerfScore 0.25
G_M8040_IG05:  ;; offset=0x0037
       C3                   ret      
						;; size=1 bbWeight=1 PerfScore 1.00
RWD00  	dq	7FFFFFFFFFFFFFFFh, 7FFFFFFFFFFFFFFFh
RWD16  	dq	4341C37937E08000h	;        1e+16
RWD24  	dq	4059000000000000h	;          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 = 0

G_M12104_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M12104_IG02:  ;; offset=0x0003
       C5F828C8             vmovaps  xmm1, xmm0
       C5F0541531000000     vandps   xmm2, xmm1, xmmword ptr [reloc @RWD00]
       C5FA101D39000000     vmovss   xmm3, dword ptr [reloc @RWD16]
       C5F82EDA             vucomiss xmm3, xmm2
       7616                 jbe      SHORT G_M12104_IG04
						;; size=26 bbWeight=1 PerfScore 8.25
G_M12104_IG03:  ;; offset=0x001D
       C5FA590D2F000000     vmulss   xmm1, xmm0, dword ptr [reloc @RWD20]
       C4E3710AC904         vroundss xmm1, xmm1, xmm1, 4
       C5F25E0D21000000     vdivss   xmm1, xmm1, dword ptr [reloc @RWD20]
						;; size=22 bbWeight=0.50 PerfScore 12.00
G_M12104_IG04:  ;; offset=0x0033
       C5F828C1             vmovaps  xmm0, xmm1
						;; size=4 bbWeight=1 PerfScore 0.25
G_M12104_IG05:  ;; offset=0x0037
       C3                   ret      
						;; size=1 bbWeight=1 PerfScore 1.00
RWD00  	dq	7FFFFFFF7FFFFFFFh, 7FFFFFFF7FFFFFFFh
RWD16  	dd	4CBEBC20h		;     1e+08
RWD20  	dd	42C80000h		;       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 = 0

G_M22062_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M22062_IG02:  ;; offset=0x0003
       C5F828C8             vmovaps  xmm1, xmm0
       C5F0541531000000     vandps   xmm2, xmm1, xmmword ptr [reloc @RWD00]
       C5FB101D39000000     vmovsd   xmm3, qword ptr [reloc @RWD16]
       C5F92EDA             vucomisd xmm3, xmm2
       7616                 jbe      SHORT G_M22062_IG04
						;; size=26 bbWeight=1 PerfScore 8.25
G_M22062_IG03:  ;; offset=0x001D
       C5FB590D33000000     vmulsd   xmm1, xmm0, qword ptr [reloc @RWD24]
       C4E3710BC90B         vroundsd xmm1, xmm1, xmm1, 11
       C5F35E0D25000000     vdivsd   xmm1, xmm1, qword ptr [reloc @RWD24]
						;; size=22 bbWeight=0.50 PerfScore 13.00
G_M22062_IG04:  ;; offset=0x0033
       C5F828C1             vmovaps  xmm0, xmm1
						;; size=4 bbWeight=1 PerfScore 0.25
G_M22062_IG05:  ;; offset=0x0037
       C3                   ret      
						;; size=1 bbWeight=1 PerfScore 1.00
RWD00  	dq	7FFFFFFFFFFFFFFFh, 7FFFFFFFFFFFFFFFh
RWD16  	dq	4341C37937E08000h	;        1e+16
RWD24  	dq	40C3880000000000h	;        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 = 0

G_M65422_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M65422_IG02:  ;; offset=0x0003
       C5F828C8             vmovaps  xmm1, xmm0
       C5F0541531000000     vandps   xmm2, xmm1, xmmword ptr [reloc @RWD00]
       C5FA101D39000000     vmovss   xmm3, dword ptr [reloc @RWD16]
       C5F82EDA             vucomiss xmm3, xmm2
       7616                 jbe      SHORT G_M65422_IG04
						;; size=26 bbWeight=1 PerfScore 8.25
G_M65422_IG03:  ;; offset=0x001D
       C5FA590D2F000000     vmulss   xmm1, xmm0, dword ptr [reloc @RWD20]
       C4E3710AC90B         vroundss xmm1, xmm1, xmm1, 11
       C5F25E0D21000000     vdivss   xmm1, xmm1, dword ptr [reloc @RWD20]
						;; size=22 bbWeight=0.50 PerfScore 12.00
G_M65422_IG04:  ;; offset=0x0033
       C5F828C1             vmovaps  xmm0, xmm1
						;; size=4 bbWeight=1 PerfScore 0.25
G_M65422_IG05:  ;; offset=0x0037
       C3                   ret      
						;; size=1 bbWeight=1 PerfScore 1.00
RWD00  	dq	7FFFFFFF7FFFFFFFh, 7FFFFFFF7FFFFFFFh
RWD16  	dd	4CBEBC20h		;     1e+08
RWD20  	dd	461C4000h		;     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.

@ghost ghost added the community-contribution Indicates that the PR has been added by a community member label Feb 8, 2024
@ghost
Copy link

ghost commented Feb 8, 2024

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

Issue Details

Simplifies and optimizes parameterized Round overloads.

Code:

internal static class RoundDisasm
{
	public static double RoundAway(double d) => Math.Round(d, MidpointRounding.AwayFromZero);
	public static double RoundAway(float f) => MathF.Round(f, MidpointRounding.AwayFromZero);
	public static double RoundTruncate(double d) => Math.Round(d, MidpointRounding.ToZero);
	public static double RoundTruncate(float f) => MathF.Round(f, MidpointRounding.ToZero);
	public static double Round(double d, MidpointRounding m) => Math.Round(d, m);
	public static double Round(float f, MidpointRounding m) => MathF.Round(f, m);
	public static double Round2(double d) => Math.Round(d, 2);
	public static double Round2(float f) => MathF.Round(f, 2);
	public static double Round4(double d) => Math.Round(d, 4, MidpointRounding.ToZero);
	public static double Round4(float f) => 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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       C5F828C8             vmovaps  xmm1, xmm0
       C5F8101521000000     vmovups  xmm2, xmmword ptr [reloc @RWD00]
       C5F154CA             vandpd   xmm1, xmm1, xmm2
       62F1F518560D23000000 vorpd    xmm1, xmm1, qword ptr [reloc @RWD16] {1to2}
       C5F358C0             vaddsd   xmm0, xmm1, xmm0
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
 
G_M000_IG03:                ;; offset=0x0027
       C3                   ret      
 
RWD00  	dq	8000000000000000h, 8000000000000000h
RWD16  	dq	3FDFFFFFFFFFFFFFh	;          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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       C5F828C8             vmovaps  xmm1, xmm0
       C5F8101521000000     vmovups  xmm2, xmmword ptr [reloc @RWD00]
       C5F054CA             vandps   xmm1, xmm1, xmm2
       62F17418560D23000000 vorps    xmm1, xmm1, dword ptr [reloc @RWD16] {1to4}
       C5F258C0             vaddss   xmm0, xmm1, xmm0
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
       C5FA5AC0             vcvtss2sd xmm0, xmm0, xmm0
 
G_M000_IG03:                ;; offset=0x002B
       C3                   ret      
 
RWD00  	dq	8000000080000000h, 8000000080000000h
RWD16  	dd	3EFFFFFFh		;       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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       33D2                 xor      edx, edx
       41B802000000         mov      r8d, 2
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       4883EC28             sub      rsp, 40
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0007
       33D2                 xor      edx, edx
       41B802000000         mov      r8d, 2
       FF15F38F1E00         call     [System.MathF:Round(float,int,int):float]
       C5FA5AC0             vcvtss2sd xmm0, xmm0, xmm0
 
G_M000_IG03:                ;; offset=0x0019
       4883C428             add      rsp, 40
       C3                   ret      
 
; 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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       448BC2               mov      r8d, edx
       33D2                 xor      edx, edx
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       4883EC28             sub      rsp, 40
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0007
       448BC2               mov      r8d, edx
       33D2                 xor      edx, edx
       FF15968F1E00         call     [System.MathF:Round(float,int,int):float]
       C5FA5AC0             vcvtss2sd xmm0, xmm0, xmm0
 
G_M000_IG03:                ;; offset=0x0016
       4883C428             add      rsp, 40
       C3                   ret      
 
; 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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       BA02000000           mov      edx, 2
       4533C0               xor      r8d, r8d
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       4883EC28             sub      rsp, 40
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0007
       BA02000000           mov      edx, 2
       4533C0               xor      r8d, r8d
       FF15338F1E00         call     [System.MathF:Round(float,int,int):float]
       C5FA5AC0             vcvtss2sd xmm0, xmm0, xmm0
 
G_M000_IG03:                ;; offset=0x0019
       4883C428             add      rsp, 40
       C3                   ret      
 
; 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 data

G_M000_IG01:                ;; offset=0x0000
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0003
       BA04000000           mov      edx, 4
       41B802000000         mov      r8d, 2
 
G_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 data

G_M000_IG01:                ;; offset=0x0000
       4883EC28             sub      rsp, 40
       C5F877               vzeroupper 
 
G_M000_IG02:                ;; offset=0x0007
       BA04000000           mov      edx, 4
       41B802000000         mov      r8d, 2
       FF15C08E1E00         call     [System.MathF:Round(float,int,int):float]
       C5FA5AC0             vcvtss2sd xmm0, xmm0, xmm0
 
G_M000_IG03:                ;; offset=0x001C
       4883C428             add      rsp, 40
       C3                   ret      
 
; 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 = 0

G_M25588_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M25588_IG02:  ;; offset=0x0003
       C5F8100D25000000     vmovups  xmm1, xmmword ptr [reloc @RWD00]
       C5F828D0             vmovaps  xmm2, xmm0
       62F3ED08250D26000000CA vpternlogq xmm1, xmm2, xmmword ptr [reloc @RWD16], -54
       C5F358C0             vaddsd   xmm0, xmm1, xmm0
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
						;; size=33 bbWeight=1 PerfScore 15.25
G_M25588_IG03:  ;; offset=0x0024
       C3                   ret      
						;; size=1 bbWeight=1 PerfScore 1.00
RWD00  	dq	8000000000000000h, 8000000000000000h
RWD16  	dq	3FDFFFFFFFFFFFFFh, 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 = 0

G_M34641_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M34641_IG02:  ;; offset=0x0003
       C5F8100D25000000     vmovups  xmm1, xmmword ptr [reloc @RWD00]
       C5F828D0             vmovaps  xmm2, xmm0
       62F36D08250D26000000CA vpternlogd xmm1, xmm2, xmmword ptr [reloc @RWD16], -54
       C5F258C0             vaddss   xmm0, xmm1, xmm0
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
       C5FA5AC0             vcvtss2sd xmm0, xmm0, xmm0
						;; size=37 bbWeight=1 PerfScore 19.25
G_M34641_IG03:  ;; offset=0x0028
       C3                   ret      
						;; size=1 bbWeight=1 PerfScore 1.00
RWD00  	dq	8000000080000000h, 8000000080000000h
RWD16  	dq	3EFFFFFF3EFFFFFFh, 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 = 0

G_M39156_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M39156_IG02:  ;; offset=0x0003
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
						;; size=6 bbWeight=1 PerfScore 7.00
G_M39156_IG03:  ;; offset=0x0009
       C3                   ret      
						;; 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 = 0

G_M7249_IG01:  ;; offset=0x0000
       C5F877               vzeroupper 
						;; size=3 bbWeight=1 PerfScore 1.00
G_M7249_IG02:  ;; offset=0x0003
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
       C5FA5AC0             vcvtss2sd xmm0, xmm0, xmm0
						;; size=10 bbWeight=1 PerfScore 11.00
G_M7249_IG03:  ;; offset=0x000D
       C3                   ret      
						;; 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 = 32

G_M6149_IG01:  ;; offset=0x0000
       53                   push     rbx
       4883EC20             sub      rsp, 32
       C5F877               vzeroupper 
       8BDA                 mov      ebx, edx
						;; size=10 bbWeight=1 PerfScore 2.50
G_M6149_IG02:  ;; offset=0x000A
       83FB04               cmp      ebx, 4
       775F                 ja       SHORT G_M6149_IG09
       8BCB                 mov      ecx, ebx
       488D1578000000       lea      rdx, [reloc @RWD00]
       8B148A               mov      edx, dword ptr [rdx+4*rcx]
       488D05E8FFFFFF       lea      rax, G_M6149_IG02
       4803D0               add      rdx, rax
       FFE2                 jmp      rdx
						;; size=29 bbWeight=1 PerfScore 7.25
G_M6149_IG03:  ;; offset=0x0027
       C5F8100D81000000     vmovups  xmm1, xmmword ptr [reloc @RWD32]
       C5F828D0             vmovaps  xmm2, xmm0
       62F3ED08250D82000000CA vpternlogq xmm1, xmm2, xmmword ptr [reloc @RWD48], -54
       C5F358C0             vaddsd   xmm0, xmm1, xmm0
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
       EB1E                 jmp      SHORT G_M6149_IG08
						;; size=35 bbWeight=0.50 PerfScore 8.62
G_M6149_IG04:  ;; offset=0x004A
       C4E3790BC004         vroundsd xmm0, xmm0, xmm0, 4
       EB16                 jmp      SHORT G_M6149_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M6149_IG05:  ;; offset=0x0052
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
       EB0E                 jmp      SHORT G_M6149_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M6149_IG06:  ;; offset=0x005A
       C4E3790BC009         vroundsd xmm0, xmm0, xmm0, 9
       EB06                 jmp      SHORT G_M6149_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M6149_IG07:  ;; offset=0x0062
       C4E3790BC00A         vroundsd xmm0, xmm0, xmm0, 10
						;; size=6 bbWeight=0.50 PerfScore 3.50
G_M6149_IG08:  ;; offset=0x0068
       4883C420             add      rsp, 32
       5B                   pop      rbx
       C3                   ret      
						;; size=6 bbWeight=1 PerfScore 1.75
G_M6149_IG09:  ;; offset=0x006E
       B9840B0000           mov      ecx, 0xB84
       48BA004022BEFA7F0000 mov      rdx, 0x7FFABE224000
       E88EE44E5F           call     CORINFO_HELP_STRCNS
       488BD0               mov      rdx, rax
       8BCB                 mov      ecx, ebx
       FF1533947C00         call     [System.ThrowHelper:ThrowArgumentException_InvalidEnumValue[int](int,System.String)]
       CC                   int3     
						;; size=32 bbWeight=0 PerfScore 0.00
RWD00  	dd	00000040h ; case G_M6149_IG04
       	dd	0000001Dh ; case G_M6149_IG03
       	dd	00000048h ; case G_M6149_IG05
       	dd	00000050h ; case G_M6149_IG06
       	dd	00000058h ; case G_M6149_IG07
RWD20  	dd	00000000h, 00000000h, 00000000h
RWD32  	dq	8000000000000000h, 8000000000000000h
RWD48  	dq	3FDFFFFFFFFFFFFFh, 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 = 32

G_M39904_IG01:  ;; offset=0x0000
       53                   push     rbx
       4883EC20             sub      rsp, 32
       C5F877               vzeroupper 
       8BDA                 mov      ebx, edx
						;; size=10 bbWeight=1 PerfScore 2.50
G_M39904_IG02:  ;; offset=0x000A
       83FB04               cmp      ebx, 4
       7763                 ja       SHORT G_M39904_IG10
       8BCB                 mov      ecx, ebx
       488D1588000000       lea      rdx, [reloc @RWD00]
       8B148A               mov      edx, dword ptr [rdx+4*rcx]
       488D05E8FFFFFF       lea      rax, G_M39904_IG02
       4803D0               add      rdx, rax
       FFE2                 jmp      rdx
						;; size=29 bbWeight=1 PerfScore 7.25
G_M39904_IG03:  ;; offset=0x0027
       C5F8100D91000000     vmovups  xmm1, xmmword ptr [reloc @RWD32]
       C5F828D0             vmovaps  xmm2, xmm0
       62F36D08250D92000000CA vpternlogd xmm1, xmm2, xmmword ptr [reloc @RWD48], -54
       C5F258C0             vaddss   xmm0, xmm1, xmm0
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
       EB1E                 jmp      SHORT G_M39904_IG08
						;; size=35 bbWeight=0.50 PerfScore 8.62
G_M39904_IG04:  ;; offset=0x004A
       C4E3790AC004         vroundss xmm0, xmm0, xmm0, 4
       EB16                 jmp      SHORT G_M39904_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M39904_IG05:  ;; offset=0x0052
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
       EB0E                 jmp      SHORT G_M39904_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M39904_IG06:  ;; offset=0x005A
       C4E3790AC009         vroundss xmm0, xmm0, xmm0, 9
       EB06                 jmp      SHORT G_M39904_IG08
						;; size=8 bbWeight=0.50 PerfScore 4.50
G_M39904_IG07:  ;; offset=0x0062
       C4E3790AC00A         vroundss xmm0, xmm0, xmm0, 10
						;; size=6 bbWeight=0.50 PerfScore 3.50
G_M39904_IG08:  ;; offset=0x0068
       C5FA5AC0             vcvtss2sd xmm0, xmm0, xmm0
						;; size=4 bbWeight=1 PerfScore 4.00
G_M39904_IG09:  ;; offset=0x006C
       4883C420             add      rsp, 32
       5B                   pop      rbx
       C3                   ret      
						;; size=6 bbWeight=1 PerfScore 1.75
G_M39904_IG10:  ;; offset=0x0072
       B9840B0000           mov      ecx, 0xB84
       48BA004022BEFA7F0000 mov      rdx, 0x7FFABE224000
       E89AE34E5F           call     CORINFO_HELP_STRCNS
       488BD0               mov      rdx, rax
       8BCB                 mov      ecx, ebx
       FF153F937C00         call     [System.ThrowHelper:ThrowArgumentException_InvalidEnumValue[int](int,System.String)]
       CC                   int3     
						;; size=32 bbWeight=0 PerfScore 0.00
RWD00  	dd	00000040h ; case G_M39904_IG04
       	dd	0000001Dh ; case G_M39904_IG03
       	dd	00000048h ; case G_M39904_IG05
       	dd	00000050h ; case G_M39904_IG06
       	dd	00000058h ; case G_M39904_IG07
RWD20  	dd	00000000h, 00000000h, 00000000h
RWD32  	dq	8000000080000000h, 8000000080000000h
RWD48  	dq	3EFFFFFF3EFFFFFFh, 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 = 56

G_M8040_IG01:  ;; offset=0x0000
       4883EC38             sub      rsp, 56
       C5F877               vzeroupper 
       33C0                 xor      eax, eax
       4889442428           mov      qword ptr [rsp+0x28], rax
						;; size=14 bbWeight=1 PerfScore 2.50
G_M8040_IG02:  ;; offset=0x000E
       C5F828C8             vmovaps  xmm1, xmm0
       C5F0541546000000     vandps   xmm2, xmm1, xmmword ptr [reloc @RWD00]
       C5FB101D4E000000     vmovsd   xmm3, qword ptr [reloc @RWD16]
       C5F92EDA             vucomisd xmm3, xmm2
       762F                 jbe      SHORT G_M8040_IG04
						;; size=26 bbWeight=1 PerfScore 8.25
G_M8040_IG03:  ;; offset=0x0028
       C744243010000000     mov      dword ptr [rsp+0x30], 16
       48B8206AB71AFB7F0000 mov      rax, 0x7FFB1AB76A20      ; static handle
       4889442428           mov      bword ptr [rsp+0x28], rax
       488B442428           mov      rax, bword ptr [rsp+0x28]
       C5FB104810           vmovsd   xmm1, qword ptr [rax+0x10]
       C5FB59C1             vmulsd   xmm0, xmm0, xmm1
       C4E3790BC004         vroundsd xmm0, xmm0, xmm0, 4
       C5FB5EC9             vdivsd   xmm1, xmm0, xmm1
						;; size=47 bbWeight=0.50 PerfScore 14.62
G_M8040_IG04:  ;; offset=0x0057
       C5F828C1             vmovaps  xmm0, xmm1
						;; size=4 bbWeight=1 PerfScore 0.25
G_M8040_IG05:  ;; offset=0x005B
       4883C438             add      rsp, 56
       C3                   ret      
						;; size=5 bbWeight=1 PerfScore 1.25
RWD00  	dq	7FFFFFFFFFFFFFFFh, 7FFFFFFFFFFFFFFFh
RWD16  	dq	4341C37937E08000h	;        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 = 56

G_M52813_IG01:  ;; offset=0x0000
       4883EC38             sub      rsp, 56
       C5F877               vzeroupper 
       33C0                 xor      eax, eax
       4889442428           mov      qword ptr [rsp+0x28], rax
						;; size=14 bbWeight=1 PerfScore 2.50
G_M52813_IG02:  ;; offset=0x000E
       C5F828C8             vmovaps  xmm1, xmm0
       C5F0541546000000     vandps   xmm2, xmm1, xmmword ptr [reloc @RWD00]
       C5FA101D4E000000     vmovss   xmm3, dword ptr [reloc @RWD16]
       C5F82EDA             vucomiss xmm3, xmm2
       762F                 jbe      SHORT G_M52813_IG04
						;; size=26 bbWeight=1 PerfScore 8.25
G_M52813_IG03:  ;; offset=0x0028
       C744243007000000     mov      dword ptr [rsp+0x30], 7
       48B868DCB71AFB7F0000 mov      rax, 0x7FFB1AB7DC68      ; static handle
       4889442428           mov      bword ptr [rsp+0x28], rax
       488B442428           mov      rax, bword ptr [rsp+0x28]
       C5FA104808           vmovss   xmm1, dword ptr [rax+0x08]
       C5FA59C1             vmulss   xmm0, xmm0, xmm1
       C4E3790AC004         vroundss xmm0, xmm0, xmm0, 4
       C5FA5EC9             vdivss   xmm1, xmm0, xmm1
						;; size=47 bbWeight=0.50 PerfScore 13.62
G_M52813_IG04:  ;; offset=0x0057
       C5FA5AC1             vcvtss2sd xmm0, xmm0, xmm1
						;; size=4 bbWeight=1 PerfScore 4.00
G_M52813_IG05:  ;; offset=0x005B
       4883C438             add      rsp, 56
       C3                   ret      
						;; size=5 bbWeight=1 PerfScore 1.25
RWD00  	dq	7FFFFFFF7FFFFFFFh, 7FFFFFFF7FFFFFFFh
RWD16  	dd	4CBEBC20h		;     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 = 56

G_M22062_IG01:  ;; offset=0x0000
       4883EC38             sub      rsp, 56
       C5F877               vzeroupper 
       33C0                 xor      eax, eax
       4889442428           mov      qword ptr [rsp+0x28], rax
						;; size=14 bbWeight=1 PerfScore 2.50
G_M22062_IG02:  ;; offset=0x000E
       C5F828C8             vmovaps  xmm1, xmm0
       C5F0541546000000     vandps   xmm2, xmm1, xmmword ptr [reloc @RWD00]
       C5FB101D4E000000     vmovsd   xmm3, qword ptr [reloc @RWD16]
       C5F92EDA             vucomisd xmm3, xmm2
       762F                 jbe      SHORT G_M22062_IG04
						;; size=26 bbWeight=1 PerfScore 8.25
G_M22062_IG03:  ;; offset=0x0028
       C744243010000000     mov      dword ptr [rsp+0x30], 16
       48B8206AB71AFB7F0000 mov      rax, 0x7FFB1AB76A20      ; static handle
       4889442428           mov      bword ptr [rsp+0x28], rax
       488B442428           mov      rax, bword ptr [rsp+0x28]
       C5FB104820           vmovsd   xmm1, qword ptr [rax+0x20]
       C5FB59C1             vmulsd   xmm0, xmm0, xmm1
       C4E3790BC00B         vroundsd xmm0, xmm0, xmm0, 11
       C5FB5EC9             vdivsd   xmm1, xmm0, xmm1
						;; size=47 bbWeight=0.50 PerfScore 14.62
G_M22062_IG04:  ;; offset=0x0057
       C5F828C1             vmovaps  xmm0, xmm1
						;; size=4 bbWeight=1 PerfScore 0.25
G_M22062_IG05:  ;; offset=0x005B
       4883C438             add      rsp, 56
       C3                   ret      
						;; size=5 bbWeight=1 PerfScore 1.25
RWD00  	dq	7FFFFFFFFFFFFFFFh, 7FFFFFFFFFFFFFFFh
RWD16  	dq	4341C37937E08000h	;        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 = 56

G_M34635_IG01:  ;; offset=0x0000
       4883EC38             sub      rsp, 56
       C5F877               vzeroupper 
       33C0                 xor      eax, eax
       4889442428           mov      qword ptr [rsp+0x28], rax
						;; size=14 bbWeight=1 PerfScore 2.50
G_M34635_IG02:  ;; offset=0x000E
       C5F828C8             vmovaps  xmm1, xmm0
       C5F0541546000000     vandps   xmm2, xmm1, xmmword ptr [reloc @RWD00]
       C5FA101D4E000000     vmovss   xmm3, dword ptr [reloc @RWD16]
       C5F82EDA             vucomiss xmm3, xmm2
       762F                 jbe      SHORT G_M34635_IG04
						;; size=26 bbWeight=1 PerfScore 8.25
G_M34635_IG03:  ;; offset=0x0028
       C744243007000000     mov      dword ptr [rsp+0x30], 7
       48B868DCB71AFB7F0000 mov      rax, 0x7FFB1AB7DC68      ; static handle
       4889442428           mov      bword ptr [rsp+0x28], rax
       488B442428           mov      rax, bword ptr [rsp+0x28]
       C5FA104810           vmovss   xmm1, dword ptr [rax+0x10]
       C5FA59C1             vmulss   xmm0, xmm0, xmm1
       C4E3790AC00B         vroundss xmm0, xmm0, xmm0, 11
       C5FA5EC9             vdivss   xmm1, xmm0, xmm1
						;; size=47 bbWeight=0.50 PerfScore 13.62
G_M34635_IG04:  ;; offset=0x0057
       C5FA5AC1             vcvtss2sd xmm0, xmm0, xmm1
						;; size=4 bbWeight=1 PerfScore 4.00
G_M34635_IG05:  ;; offset=0x005B
       4883C438             add      rsp, 56
       C3                   ret      
						;; size=5 bbWeight=1 PerfScore 1.25
RWD00  	dq	7FFFFFFF7FFFFFFFh, 7FFFFFFF7FFFFFFFh
RWD16  	dd	4CBEBC20h		;     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)
; ============================================================

Fixes #98164.

Author: MichalPetryka
Assignees: -
Labels:

area-System.Numerics
Milestone: -

@MichalPetryka
Copy link
Contributor Author

@MihuBot

@MihuBot MihuBot mentioned this pull request Feb 8, 2024
Open
This was referenced Feb 9, 2024
Closed
Closed
Closed
@jkotas jkotas mentioned this pull request Feb 9, 2024
Closed
@tannergooding tannergooding merged commit e7f8b66 into dotnet:main Feb 9, 2024
174 of 178 checks passed
@build-analysis build-analysis bot mentioned this pull request Feb 9, 2024
Closed
@tannergooding tannergooding mentioned this pull request Feb 10, 2024
Merged
@github-actions github-actions bot locked and limited conversation to collaborators Mar 12, 2024
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Reviewers

@tannergooding tannergooding tannergooding approved these changes

Assignees
No one assigned
Labels
area-System.Numerics community-contribution Indicates that the PR has been added by a community member
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

Math.Round[F](x, MidpointRounding) doesn't use VROUNDS[SD] outside of "to even"
2 participants
@MichalPetryka @tannergooding