[MichalPetryka] Intrinsify Array GetArrayDataReference for SZ arrays #76
Comments
Top method improvements-6 (-30.00 % of base) - System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[double,long]:GetPinnableReference(ulong[]):byref ; Assembly listing for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[double,long]:GetPinnableReference(ulong[]):byref (FullOpts)
; Emitting BLENDED_CODE for X64 with AVX - Unix
; FullOpts code
; optimized code
; rsp based frame
; partially interruptible
; No PGO data
-; 0 inlinees with PGO data; 2 single block inlinees; 0 inlinees without PGO data
; Final local variable assignments
;
-; V00 arg0 [V00,T00] ( 5, 4 ) ref -> rdi class-hnd single-def
+; V00 arg0 [V00,T00] ( 4, 4 ) ref -> rdi class-hnd single-def
;# V01 OutArgs [V01 ] ( 1, 1 ) struct ( 0) [rsp+00H] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
-;* V02 tmp1 [V02 ] ( 0, 0 ) byref -> zero-ref "impAppendStmt"
;
; Lcl frame size = 0
G_M54100_IG01:
;; size=0 bbWeight=1 PerfScore 0.00
G_M54100_IG02:
+ lea rax, bword ptr [rdi+10H]
+ xor rcx, rcx
test rdi, rdi
- jne SHORT G_M54100_IG05
- ;; size=5 bbWeight=1 PerfScore 1.25
+ cmove rax, rcx
+ ;; size=13 bbWeight=1 PerfScore 1.25
G_M54100_IG03:
- xor rax, rax
- ;; size=2 bbWeight=0.50 PerfScore 0.12
-G_M54100_IG04:
ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
-G_M54100_IG05:
- mov rax, qword ptr [rdi]
- mov eax, dword ptr [rax+04H]
- lea rax, bword ptr [rdi+rax-08H]
- ;; size=11 bbWeight=0.50 PerfScore 2.50
-G_M54100_IG06:
- ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
+ ;; size=1 bbWeight=1 PerfScore 1.00
-; Total bytes of code 20, prolog size 0, PerfScore 6.88, instruction count 8, allocated bytes for code 20 (MethodHash=33252cab) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[double,long]:GetPinnableReference(ulong[]):byref (FullOpts)
+; Total bytes of code 14, prolog size 0, PerfScore 3.65, instruction count 5, allocated bytes for code 14 (MethodHash=33252cab) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[double,long]:GetPinnableReference(ulong[]):byref (FullOpts)-6 (-30.00 % of base) - System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[int,long]:GetPinnableReference(ulong[]):byref ; Assembly listing for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[int,long]:GetPinnableReference(ulong[]):byref (FullOpts)
; Emitting BLENDED_CODE for X64 with AVX - Unix
; FullOpts code
; optimized code
; rsp based frame
; partially interruptible
; No PGO data
-; 0 inlinees with PGO data; 2 single block inlinees; 0 inlinees without PGO data
; Final local variable assignments
;
-; V00 arg0 [V00,T00] ( 5, 4 ) ref -> rdi class-hnd single-def
+; V00 arg0 [V00,T00] ( 4, 4 ) ref -> rdi class-hnd single-def
;# V01 OutArgs [V01 ] ( 1, 1 ) struct ( 0) [rsp+00H] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
-;* V02 tmp1 [V02 ] ( 0, 0 ) byref -> zero-ref "impAppendStmt"
;
; Lcl frame size = 0
G_M44274_IG01:
;; size=0 bbWeight=1 PerfScore 0.00
G_M44274_IG02:
+ lea rax, bword ptr [rdi+10H]
+ xor rcx, rcx
test rdi, rdi
- jne SHORT G_M44274_IG05
- ;; size=5 bbWeight=1 PerfScore 1.25
+ cmove rax, rcx
+ ;; size=13 bbWeight=1 PerfScore 1.25
G_M44274_IG03:
- xor rax, rax
- ;; size=2 bbWeight=0.50 PerfScore 0.12
-G_M44274_IG04:
ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
-G_M44274_IG05:
- mov rax, qword ptr [rdi]
- mov eax, dword ptr [rax+04H]
- lea rax, bword ptr [rdi+rax-08H]
- ;; size=11 bbWeight=0.50 PerfScore 2.50
-G_M44274_IG06:
- ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
+ ;; size=1 bbWeight=1 PerfScore 1.00
-; Total bytes of code 20, prolog size 0, PerfScore 6.88, instruction count 8, allocated bytes for code 20 (MethodHash=78b3530d) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[int,long]:GetPinnableReference(ulong[]):byref (FullOpts)
+; Total bytes of code 14, prolog size 0, PerfScore 3.65, instruction count 5, allocated bytes for code 14 (MethodHash=78b3530d) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[int,long]:GetPinnableReference(ulong[]):byref (FullOpts)-6 (-30.00 % of base) - System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[long,long]:GetPinnableReference(ulong[]):byref ; Assembly listing for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[long,long]:GetPinnableReference(ulong[]):byref (FullOpts)
; Emitting BLENDED_CODE for X64 with AVX - Unix
; FullOpts code
; optimized code
; rsp based frame
; partially interruptible
; No PGO data
-; 0 inlinees with PGO data; 2 single block inlinees; 0 inlinees without PGO data
; Final local variable assignments
;
-; V00 arg0 [V00,T00] ( 5, 4 ) ref -> rdi class-hnd single-def
+; V00 arg0 [V00,T00] ( 4, 4 ) ref -> rdi class-hnd single-def
;# V01 OutArgs [V01 ] ( 1, 1 ) struct ( 0) [rsp+00H] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
-;* V02 tmp1 [V02 ] ( 0, 0 ) byref -> zero-ref "impAppendStmt"
;
; Lcl frame size = 0
G_M27755_IG01:
;; size=0 bbWeight=1 PerfScore 0.00
G_M27755_IG02:
+ lea rax, bword ptr [rdi+10H]
+ xor rcx, rcx
test rdi, rdi
- jne SHORT G_M27755_IG05
- ;; size=5 bbWeight=1 PerfScore 1.25
+ cmove rax, rcx
+ ;; size=13 bbWeight=1 PerfScore 1.25
G_M27755_IG03:
- xor rax, rax
- ;; size=2 bbWeight=0.50 PerfScore 0.12
-G_M27755_IG04:
ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
-G_M27755_IG05:
- mov rax, qword ptr [rdi]
- mov eax, dword ptr [rax+04H]
- lea rax, bword ptr [rdi+rax-08H]
- ;; size=11 bbWeight=0.50 PerfScore 2.50
-G_M27755_IG06:
- ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
+ ;; size=1 bbWeight=1 PerfScore 1.00
-; Total bytes of code 20, prolog size 0, PerfScore 6.88, instruction count 8, allocated bytes for code 20 (MethodHash=1eeb9394) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[long,long]:GetPinnableReference(ulong[]):byref (FullOpts)
+; Total bytes of code 14, prolog size 0, PerfScore 3.65, instruction count 5, allocated bytes for code 14 (MethodHash=1eeb9394) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[long,long]:GetPinnableReference(ulong[]):byref (FullOpts)-6 (-30.00 % of base) - System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[short,long]:GetPinnableReference(ulong[]):byref ; Assembly listing for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[short,long]:GetPinnableReference(ulong[]):byref (FullOpts)
; Emitting BLENDED_CODE for X64 with AVX - Unix
; FullOpts code
; optimized code
; rsp based frame
; partially interruptible
; No PGO data
-; 0 inlinees with PGO data; 2 single block inlinees; 0 inlinees without PGO data
; Final local variable assignments
;
-; V00 arg0 [V00,T00] ( 5, 4 ) ref -> rdi class-hnd single-def
+; V00 arg0 [V00,T00] ( 4, 4 ) ref -> rdi class-hnd single-def
;# V01 OutArgs [V01 ] ( 1, 1 ) struct ( 0) [rsp+00H] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
-;* V02 tmp1 [V02 ] ( 0, 0 ) byref -> zero-ref "impAppendStmt"
;
; Lcl frame size = 0
G_M3059_IG01:
;; size=0 bbWeight=1 PerfScore 0.00
G_M3059_IG02:
+ lea rax, bword ptr [rdi+10H]
+ xor rcx, rcx
test rdi, rdi
- jne SHORT G_M3059_IG05
- ;; size=5 bbWeight=1 PerfScore 1.25
+ cmove rax, rcx
+ ;; size=13 bbWeight=1 PerfScore 1.25
G_M3059_IG03:
- xor rax, rax
- ;; size=2 bbWeight=0.50 PerfScore 0.12
-G_M3059_IG04:
ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
-G_M3059_IG05:
- mov rax, qword ptr [rdi]
- mov eax, dword ptr [rax+04H]
- lea rax, bword ptr [rdi+rax-08H]
- ;; size=11 bbWeight=0.50 PerfScore 2.50
-G_M3059_IG06:
- ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
+ ;; size=1 bbWeight=1 PerfScore 1.00
-; Total bytes of code 20, prolog size 0, PerfScore 6.88, instruction count 8, allocated bytes for code 20 (MethodHash=4012f40c) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[short,long]:GetPinnableReference(ulong[]):byref (FullOpts)
+; Total bytes of code 14, prolog size 0, PerfScore 3.65, instruction count 5, allocated bytes for code 14 (MethodHash=4012f40c) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[short,long]:GetPinnableReference(ulong[]):byref (FullOpts)-6 (-30.00 % of base) - System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[System.Numerics.Vector`1[float],long]:GetPinnableReference(ulong[]):byref ; Assembly listing for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[System.Numerics.Vector`1[float],long]:GetPinnableReference(ulong[]):byref (FullOpts)
; Emitting BLENDED_CODE for X64 with AVX - Unix
; FullOpts code
; optimized code
; rsp based frame
; partially interruptible
; No PGO data
-; 0 inlinees with PGO data; 2 single block inlinees; 0 inlinees without PGO data
; Final local variable assignments
;
-; V00 arg0 [V00,T00] ( 5, 4 ) ref -> rdi class-hnd single-def
+; V00 arg0 [V00,T00] ( 4, 4 ) ref -> rdi class-hnd single-def
;# V01 OutArgs [V01 ] ( 1, 1 ) struct ( 0) [rsp+00H] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
-;* V02 tmp1 [V02 ] ( 0, 0 ) byref -> zero-ref "impAppendStmt"
;
; Lcl frame size = 0
G_M13410_IG01:
;; size=0 bbWeight=1 PerfScore 0.00
G_M13410_IG02:
+ lea rax, bword ptr [rdi+10H]
+ xor rcx, rcx
test rdi, rdi
- jne SHORT G_M13410_IG05
- ;; size=5 bbWeight=1 PerfScore 1.25
+ cmove rax, rcx
+ ;; size=13 bbWeight=1 PerfScore 1.25
G_M13410_IG03:
- xor rax, rax
- ;; size=2 bbWeight=0.50 PerfScore 0.12
-G_M13410_IG04:
ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
-G_M13410_IG05:
- mov rax, qword ptr [rdi]
- mov eax, dword ptr [rax+04H]
- lea rax, bword ptr [rdi+rax-08H]
- ;; size=11 bbWeight=0.50 PerfScore 2.50
-G_M13410_IG06:
- ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
+ ;; size=1 bbWeight=1 PerfScore 1.00
-; Total bytes of code 20, prolog size 0, PerfScore 6.88, instruction count 8, allocated bytes for code 20 (MethodHash=84f6cb9d) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[System.Numerics.Vector`1[float],long]:GetPinnableReference(ulong[]):byref (FullOpts)
+; Total bytes of code 14, prolog size 0, PerfScore 3.65, instruction count 5, allocated bytes for code 14 (MethodHash=84f6cb9d) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[System.Numerics.Vector`1[float],long]:GetPinnableReference(ulong[]):byref (FullOpts)-6 (-30.00 % of base) - System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[ubyte,long]:GetPinnableReference(ulong[]):byref ; Assembly listing for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[ubyte,long]:GetPinnableReference(ulong[]):byref (FullOpts)
; Emitting BLENDED_CODE for X64 with AVX - Unix
; FullOpts code
; optimized code
; rsp based frame
; partially interruptible
; No PGO data
-; 0 inlinees with PGO data; 2 single block inlinees; 0 inlinees without PGO data
; Final local variable assignments
;
-; V00 arg0 [V00,T00] ( 5, 4 ) ref -> rdi class-hnd single-def
+; V00 arg0 [V00,T00] ( 4, 4 ) ref -> rdi class-hnd single-def
;# V01 OutArgs [V01 ] ( 1, 1 ) struct ( 0) [rsp+00H] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
-;* V02 tmp1 [V02 ] ( 0, 0 ) byref -> zero-ref "impAppendStmt"
;
; Lcl frame size = 0
G_M26878_IG01:
;; size=0 bbWeight=1 PerfScore 0.00
G_M26878_IG02:
+ lea rax, bword ptr [rdi+10H]
+ xor rcx, rcx
test rdi, rdi
- jne SHORT G_M26878_IG05
- ;; size=5 bbWeight=1 PerfScore 1.25
+ cmove rax, rcx
+ ;; size=13 bbWeight=1 PerfScore 1.25
G_M26878_IG03:
- xor rax, rax
- ;; size=2 bbWeight=0.50 PerfScore 0.12
-G_M26878_IG04:
ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
-G_M26878_IG05:
- mov rax, qword ptr [rdi]
- mov eax, dword ptr [rax+04H]
- lea rax, bword ptr [rdi+rax-08H]
- ;; size=11 bbWeight=0.50 PerfScore 2.50
-G_M26878_IG06:
- ret
- ;; size=1 bbWeight=0.50 PerfScore 0.50
+ ;; size=1 bbWeight=1 PerfScore 1.00
-; Total bytes of code 20, prolog size 0, PerfScore 6.88, instruction count 8, allocated bytes for code 20 (MethodHash=31f69701) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[ubyte,long]:GetPinnableReference(ulong[]):byref (FullOpts)
+; Total bytes of code 14, prolog size 0, PerfScore 3.65, instruction count 5, allocated bytes for code 14 (MethodHash=31f69701) for method System.Runtime.InteropServices.Marshalling.PointerArrayMarshaller`2+ManagedToUnmanagedIn[ubyte,long]:GetPinnableReference(ulong[]):byref (FullOpts) |
|
cc: @MichalPetryka |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Build completed in 1 hour 57 minutes.
dotnet/runtime#87374
CoreLib diffs
Frameworks diffs
Diffs
Artifacts:
The text was updated successfully, but these errors were encountered: