forked from MihaZupan/runtime-utils
-
Notifications
You must be signed in to change notification settings - Fork 0
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 our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[X64] MihaZupan/runtime/cheaper-vector-narrow #390
Comments
Top method improvements-36 (-16.00 % of base) - System.Text.Ascii:NarrowUtf16ToAscii_Intrinsified_256(ulong,ulong,ulong):ulong ; Assembly listing for method System.Text.Ascii:NarrowUtf16ToAscii_Intrinsified_256(ulong,ulong,ulong):ulong (FullOpts)
; Emitting BLENDED_CODE for X64 with AVX - Unix
; FullOpts code
; optimized code
; rbp based frame
; fully interruptible
; No PGO data
-; 0 inlinees with PGO data; 0 single block inlinees; 4 inlinees without PGO data
+; 0 inlinees with PGO data; 4 single block inlinees; 8 inlinees without PGO data
; Final local variable assignments
;
; V00 arg0 [V00,T04] ( 3, 3 ) long -> rdi single-def
; V01 arg1 [V01,T03] ( 5, 3.50) long -> rsi single-def
; V02 arg2 [V02,T05] ( 3, 2.50) long -> rdx single-def
; V03 loc0 [V03,T01] ( 5, 10.50) byref -> rdi single-def
-; V04 loc1 [V04,T07] ( 12, 17.50) simd32 -> mm0 <System.Runtime.Intrinsics.Vector256`1[ushort]>
+; V04 loc1 [V04,T07] ( 14, 18.50) simd32 -> mm0 <System.Runtime.Intrinsics.Vector256`1[ushort]>
; V05 loc2 [V05,T02] ( 5, 6 ) byref -> rax single-def
; V06 loc3 [V06,T00] ( 12, 27 ) long -> rcx
; V07 loc4 [V07,T06] ( 2, 4.50) long -> rdx
-; V08 loc5 [V08,T08] ( 3, 12 ) simd32 -> mm3 <System.Runtime.Intrinsics.Vector256`1[ushort]>
+; V08 loc5 [V08,T09] ( 3, 12 ) simd32 -> mm2 <System.Runtime.Intrinsics.Vector256`1[ushort]>
;# V09 OutArgs [V09 ] ( 1, 1 ) struct ( 0) [rsp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
-;* V10 tmp1 [V10 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-;* V11 tmp2 [V11 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
-;* V12 tmp3 [V12 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-;* V13 tmp4 [V13 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V10 tmp1 [V10 ] ( 0, 0 ) simd32 -> zero-ref "spilled call-like call argument"
+;* V11 tmp2 [V11 ] ( 0, 0 ) simd32 -> zero-ref "spilled call-like call argument"
+; V12 tmp3 [V12,T08] ( 2, 16 ) simd32 -> mm0 "Spilling op1 side effects for HWIntrinsic"
+;* V13 tmp4 [V13 ] ( 0, 0 ) simd32 -> zero-ref "spilled call-like call argument"
;* V14 tmp5 [V14 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
;* V15 tmp6 [V15 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
-;* V16 tmp7 [V16 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ushort]>
-;* V17 tmp8 [V17 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-;* V18 tmp9 [V18 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
-; V19 cse0 [V19,T11] ( 3, 1.50) simd32 -> mm0 "CSE #02: moderate"
-; V20 cse1 [V20,T12] ( 3, 1.50) simd32 -> mm0 "CSE #04: moderate"
-; V21 cse2 [V21,T09] ( 7, 10.50) simd32 -> mm2 "CSE #01: moderate"
-; V22 cse3 [V22,T10] ( 5, 7 ) simd32 -> mm1 "CSE #03: moderate"
+;* V16 tmp7 [V16 ] ( 0, 0 ) simd32 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V17 tmp8 [V17 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V18 tmp9 [V18 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+;* V19 tmp10 [V19 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V20 tmp11 [V20 ] ( 0, 0 ) simd32 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V21 tmp12 [V21 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V22 tmp13 [V22 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+;* V23 tmp14 [V23 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V24 tmp15 [V24 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V25 tmp16 [V25 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V26 tmp17 [V26 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+;* V27 tmp18 [V27 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V28 tmp19 [V28 ] ( 0, 0 ) simd32 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V29 tmp20 [V29 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+; V30 cse0 [V30,T10] ( 5, 7 ) simd32 -> mm1 "CSE #01: moderate"
;
; Lcl frame size = 0
G_M60588_IG01:
push rbp
mov rbp, rsp
;; size=4 bbWeight=1 PerfScore 1.25
G_M60588_IG02:
vmovups ymm0, ymmword ptr [rdi]
vmovups ymm1, ymmword ptr [reloc @RWD00]
vptest ymm0, ymm1
jne G_M60588_IG10
;; size=23 bbWeight=1 PerfScore 15.00
G_M60588_IG03:
mov rax, rsi
- vmovups ymm2, ymmword ptr [reloc @RWD32]
- vpand ymm0, ymm2, ymm0
vpackuswb ymm0, ymm0, ymm0
vpermq ymm0, ymm0, -40
vmovups xmmword ptr [rax], xmm0
mov ecx, 16
test sil, 16
jne SHORT G_M60588_IG04
vmovups ymm0, ymmword ptr [rdi+0x20]
vptest ymm0, ymm1
- jne G_M60588_IG08
- vpand ymm0, ymm2, ymm0
+ jne SHORT G_M60588_IG08
vpackuswb ymm0, ymm0, ymm0
vpermq ymm0, ymm0, -40
vmovups xmmword ptr [rax+0x10], xmm0
- ;; size=75 bbWeight=0.50 PerfScore 13.71
+ ;; size=55 bbWeight=0.50 PerfScore 11.38
G_M60588_IG04:
and rsi, 31
mov rcx, rsi
neg rcx
add rcx, 32
add rdx, -32
align [0 bytes for IG05]
;; size=18 bbWeight=0.50 PerfScore 0.62
G_M60588_IG05:
vmovups ymm0, ymmword ptr [rdi+2*rcx]
- vmovups ymm3, ymmword ptr [rdi+2*rcx+0x20]
- vpor ymm4, ymm0, ymm3
- vptest ymm4, ymm1
+ vmovups ymm2, ymmword ptr [rdi+2*rcx+0x20]
+ vpor ymm3, ymm0, ymm2
+ vptest ymm3, ymm1
je SHORT G_M60588_IG07
;; size=22 bbWeight=4 PerfScore 65.33
G_M60588_IG06:
vptest ymm0, ymm1
jne SHORT G_M60588_IG08
- vpand ymm3, ymm2, ymm0
- vpand ymm0, ymm2, ymm0
- vpackuswb ymm2, ymm3, ymm0
- vpermq ymm1, ymm2, -40
- vmovups xmmword ptr [rax+rcx], xmm1
+ vpackuswb ymm0, ymm0, ymm0
+ vpermq ymm2, ymm0, -40
+ vmovups xmmword ptr [rax+rcx], xmm2
add rcx, 16
jmp SHORT G_M60588_IG08
- ;; size=36 bbWeight=0.50 PerfScore 6.96
+ ;; size=28 bbWeight=0.50 PerfScore 6.62
G_M60588_IG07:
- vpand ymm0, ymm2, ymm0
- vpand ymm3, ymm2, ymm3
- vpackuswb ymm0, ymm0, ymm3
+ vpackuswb ymm0, ymm0, ymm2
vpermq ymm0, ymm0, -40
vmovups ymmword ptr [rax+rcx], ymm0
add rcx, 32
cmp rcx, rdx
jbe SHORT G_M60588_IG05
- ;; size=32 bbWeight=4 PerfScore 28.67
+ ;; size=24 bbWeight=4 PerfScore 26.00
G_M60588_IG08:
mov rax, rcx
;; size=3 bbWeight=0.50 PerfScore 0.12
G_M60588_IG09:
vzeroupper
pop rbp
ret
;; size=5 bbWeight=0.50 PerfScore 1.25
G_M60588_IG10:
xor eax, eax
;; size=2 bbWeight=0.50 PerfScore 0.12
G_M60588_IG11:
vzeroupper
pop rbp
ret
;; size=5 bbWeight=0.50 PerfScore 1.25
RWD00 dq FF80FF80FF80FF80h, FF80FF80FF80FF80h, FF80FF80FF80FF80h, FF80FF80FF80FF80h
-RWD32 dq 00FF00FF00FF00FFh, 00FF00FF00FF00FFh, 00FF00FF00FF00FFh, 00FF00FF00FF00FFh
-; Total bytes of code 225, prolog size 4, PerfScore 134.29, instruction count 58, allocated bytes for code 225 (MethodHash=910c1353) for method System.Text.Ascii:NarrowUtf16ToAscii_Intrinsified_256(ulong,ulong,ulong):ulong (FullOpts)
+; Total bytes of code 189, prolog size 4, PerfScore 128.96, instruction count 51, allocated bytes for code 189 (MethodHash=910c1353) for method System.Text.Ascii:NarrowUtf16ToAscii_Intrinsified_256(ulong,ulong,ulong):ulong (FullOpts) -36 (-5.99 % of base) - System.Text.Ascii:NarrowUtf16ToAscii(ulong,ulong,ulong):ulong ; Assembly listing for method System.Text.Ascii:NarrowUtf16ToAscii(ulong,ulong,ulong):ulong (FullOpts)
; Emitting BLENDED_CODE for X64 with AVX - Unix
; FullOpts code
; optimized code
; rbp based frame
; fully interruptible
; No PGO data
-; 0 inlinees with PGO data; 9 single block inlinees; 16 inlinees without PGO data
+; 0 inlinees with PGO data; 13 single block inlinees; 20 inlinees without PGO data
; Final local variable assignments
;
; V00 arg0 [V00,T05] ( 8, 8.50) long -> rdi single-def
; V01 arg1 [V01,T04] ( 12, 10.50) long -> rsi single-def
; V02 arg2 [V02,T09] ( 7, 5 ) long -> rdx single-def
; V03 loc0 [V03,T00] ( 22, 29.50) long -> rax
; V04 loc1 [V04,T10] ( 13, 6.50) int -> rcx
;* V05 loc2 [V05 ] ( 0, 0 ) int -> zero-ref
; V06 loc3 [V06,T03] ( 7, 14 ) long -> registers
; V07 loc4 [V07,T18] ( 5, 2.50) long -> rdx
; V08 loc5 [V08,T13] ( 2, 4.50) long -> rcx
;# V09 OutArgs [V09 ] ( 1, 1 ) struct ( 0) [rsp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
; V10 tmp1 [V10,T19] ( 3, 1.50) long -> rax "Inline return value spill temp"
; V11 tmp2 [V11,T06] ( 5, 9.50) byref -> rax single-def "Inline stloc first use temp"
-; V12 tmp3 [V12,T25] ( 12, 16.50) simd32 -> mm0 "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+; V12 tmp3 [V12,T24] ( 14, 17.50) simd32 -> mm0 "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
; V13 tmp4 [V13,T11] ( 5, 6 ) byref -> rcx single-def "Inline stloc first use temp"
-; V14 tmp5 [V14,T01] ( 12, 27 ) long -> r8 "Inline stloc first use temp"
-; V15 tmp6 [V15,T14] ( 2, 4.50) long -> r9 "Inline stloc first use temp"
-; V16 tmp7 [V16,T27] ( 3, 12 ) simd32 -> mm3 "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
-;* V17 tmp8 [V17 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-;* V18 tmp9 [V18 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
-;* V19 tmp10 [V19 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-;* V20 tmp11 [V20 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V14 tmp5 [V14 ] ( 0, 0 ) simd32 -> zero-ref "spilled call-like call argument"
+; V15 tmp6 [V15,T01] ( 12, 27 ) long -> r8 "Inline stloc first use temp"
+; V16 tmp7 [V16,T14] ( 2, 4.50) long -> r9 "Inline stloc first use temp"
+; V17 tmp8 [V17,T28] ( 3, 12 ) simd32 -> mm2 "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V18 tmp9 [V18 ] ( 0, 0 ) simd32 -> zero-ref "spilled call-like call argument"
+; V19 tmp10 [V19,T26] ( 2, 16 ) simd32 -> mm0 "Spilling op1 side effects for HWIntrinsic"
+;* V20 tmp11 [V20 ] ( 0, 0 ) simd32 -> zero-ref "spilled call-like call argument"
;* V21 tmp12 [V21 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
;* V22 tmp13 [V22 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
-;* V23 tmp14 [V23 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ushort]>
-;* V24 tmp15 [V24 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-;* V25 tmp16 [V25 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
-; V26 tmp17 [V26,T20] ( 3, 1.50) long -> rax "Inline return value spill temp"
-;* V27 tmp18 [V27,T22] ( 0, 0 ) int -> zero-ref "Inline stloc first use temp"
-;* V28 tmp19 [V28 ] ( 0, 0 ) long -> zero-ref "Inline stloc first use temp"
-; V29 tmp20 [V29,T07] ( 5, 9.50) byref -> rax single-def "Inline stloc first use temp"
-; V30 tmp21 [V30,T24] ( 14, 17.50) simd16 -> mm0 "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
-; V31 tmp22 [V31,T12] ( 5, 6 ) byref -> rcx single-def "Inline stloc first use temp"
-;* V32 tmp23 [V32 ] ( 0, 0 ) simd16 -> zero-ref "spilled call-like call argument"
-; V33 tmp24 [V33,T02] ( 11, 26.50) long -> r8 "Inline stloc first use temp"
-; V34 tmp25 [V34,T15] ( 2, 4.50) long -> r9 "Inline stloc first use temp"
-; V35 tmp26 [V35,T28] ( 3, 12 ) simd16 -> mm2 "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
-;* V36 tmp27 [V36 ] ( 0, 0 ) simd16 -> zero-ref "spilled call-like call argument"
-; V37 tmp28 [V37,T26] ( 2, 16 ) simd16 -> mm0 "Spilling op1 side effects for HWIntrinsic"
-;* V38 tmp29 [V38 ] ( 0, 0 ) simd16 -> zero-ref "spilled call-like call argument"
-;* V39 tmp30 [V39 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-;* V40 tmp31 [V40 ] ( 0, 0 ) simd16 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
-;* V41 tmp32 [V41 ] ( 0, 0 ) simd16 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
-;* V42 tmp33 [V42 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-;* V43 tmp34 [V43 ] ( 0, 0 ) simd16 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
-;* V44 tmp35 [V44 ] ( 0, 0 ) simd16 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
-;* V45 tmp36 [V45 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-;* V46 tmp37 [V46 ] ( 0, 0 ) simd16 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
-;* V47 tmp38 [V47 ] ( 0, 0 ) simd16 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector128`1[ushort]>
-;* V48 tmp39 [V48 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-;* V49 tmp40 [V49 ] ( 0, 0 ) simd16 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
-;* V50 tmp41 [V50 ] ( 0, 0 ) simd16 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
-;* V51 tmp42 [V51 ] ( 0, 0 ) long -> zero-ref "Inlining Arg"
-; V52 tmp43 [V52,T23] ( 3, 24 ) simd16 -> mm0 "dup spill"
-;* V53 tmp44 [V53 ] ( 0, 0 ) simd16 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[uint]>
-;* V54 tmp45 [V54 ] ( 0, 0 ) byref -> zero-ref "Inlining Arg"
-; V55 tmp46 [V55,T16] ( 3, 3 ) byref -> r8 single-def "Inlining Arg"
-; V56 tmp47 [V56,T17] ( 3, 3 ) byref -> rdx "Inlining Arg"
-;* V57 tmp48 [V57,T21] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
-; V58 cse0 [V58,T08] ( 3, 8.50) long -> r10 "CSE #07: moderate"
-; V59 cse1 [V59,T32] ( 3, 1.50) simd32 -> mm0 "CSE #02: moderate"
-; V60 cse2 [V60,T33] ( 3, 1.50) simd32 -> mm0 "CSE #04: moderate"
-; V61 cse3 [V61,T29] ( 7, 10.50) simd32 -> mm2 "CSE #01: aggressive"
-; V62 cse4 [V62,T30] ( 5, 6 ) simd32 -> mm1 "CSE #03: moderate"
-; V63 cse5 [V63,T31] ( 5, 6 ) simd16 -> mm1 "CSE #06: moderate"
+;* V23 tmp14 [V23 ] ( 0, 0 ) simd32 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V24 tmp15 [V24 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V25 tmp16 [V25 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+;* V26 tmp17 [V26 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V27 tmp18 [V27 ] ( 0, 0 ) simd32 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V28 tmp19 [V28 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V29 tmp20 [V29 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+;* V30 tmp21 [V30 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V31 tmp22 [V31 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V32 tmp23 [V32 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V33 tmp24 [V33 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+;* V34 tmp25 [V34 ] ( 0, 0 ) simd32 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector256`1[ushort]>
+;* V35 tmp26 [V35 ] ( 0, 0 ) simd32 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+;* V36 tmp27 [V36 ] ( 0, 0 ) simd32 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector256`1[ubyte]>
+; V37 tmp28 [V37,T20] ( 3, 1.50) long -> rax "Inline return value spill temp"
+;* V38 tmp29 [V38,T22] ( 0, 0 ) int -> zero-ref "Inline stloc first use temp"
+;* V39 tmp30 [V39 ] ( 0, 0 ) long -> zero-ref "Inline stloc first use temp"
+; V40 tmp31 [V40,T07] ( 5, 9.50) byref -> rax single-def "Inline stloc first use temp"
+; V41 tmp32 [V41,T25] ( 14, 17.50) simd16 -> mm0 "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
+; V42 tmp33 [V42,T12] ( 5, 6 ) byref -> rcx single-def "Inline stloc first use temp"
+;* V43 tmp34 [V43 ] ( 0, 0 ) simd16 -> zero-ref "spilled call-like call argument"
+; V44 tmp35 [V44,T02] ( 11, 26.50) long -> r8 "Inline stloc first use temp"
+; V45 tmp36 [V45,T15] ( 2, 4.50) long -> r9 "Inline stloc first use temp"
+; V46 tmp37 [V46,T29] ( 3, 12 ) simd16 -> mm2 "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
+;* V47 tmp38 [V47 ] ( 0, 0 ) simd16 -> zero-ref "spilled call-like call argument"
+; V48 tmp39 [V48,T27] ( 2, 16 ) simd16 -> mm0 "Spilling op1 side effects for HWIntrinsic"
+;* V49 tmp40 [V49 ] ( 0, 0 ) simd16 -> zero-ref "spilled call-like call argument"
+;* V50 tmp41 [V50 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+;* V51 tmp42 [V51 ] ( 0, 0 ) simd16 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
+;* V52 tmp43 [V52 ] ( 0, 0 ) simd16 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
+;* V53 tmp44 [V53 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+;* V54 tmp45 [V54 ] ( 0, 0 ) simd16 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
+;* V55 tmp46 [V55 ] ( 0, 0 ) simd16 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
+;* V56 tmp47 [V56 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+;* V57 tmp48 [V57 ] ( 0, 0 ) simd16 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
+;* V58 tmp49 [V58 ] ( 0, 0 ) simd16 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector128`1[ushort]>
+;* V59 tmp50 [V59 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+;* V60 tmp51 [V60 ] ( 0, 0 ) simd16 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
+;* V61 tmp52 [V61 ] ( 0, 0 ) simd16 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
+;* V62 tmp53 [V62 ] ( 0, 0 ) long -> zero-ref "Inlining Arg"
+; V63 tmp54 [V63,T23] ( 3, 24 ) simd16 -> mm0 "dup spill"
+;* V64 tmp55 [V64 ] ( 0, 0 ) simd16 -> zero-ref "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[uint]>
+;* V65 tmp56 [V65 ] ( 0, 0 ) byref -> zero-ref "Inlining Arg"
+; V66 tmp57 [V66,T16] ( 3, 3 ) byref -> r8 single-def "Inlining Arg"
+; V67 tmp58 [V67,T17] ( 3, 3 ) byref -> rdx "Inlining Arg"
+;* V68 tmp59 [V68,T21] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
+; V69 cse0 [V69,T08] ( 3, 8.50) long -> r10 "CSE #03: moderate"
+; V70 cse1 [V70,T30] ( 5, 6 ) simd32 -> mm1 "CSE #01: moderate"
+; V71 cse2 [V71,T31] ( 5, 6 ) simd16 -> mm1 "CSE #02: moderate"
;
; Lcl frame size = 0
G_M6063_IG01:
push rbp
mov rbp, rsp
;; size=4 bbWeight=1 PerfScore 1.25
G_M6063_IG02:
xor eax, eax
cmp rdx, 32
jb G_M6063_IG18
;; size=12 bbWeight=1 PerfScore 1.50
G_M6063_IG03:
mov rcx, qword ptr [rdi]
mov r8, 0xD1FFAB1E
test rcx, r8
mov r8, rcx
jne G_M6063_IG20
cmp rdx, 64
jae G_M6063_IG11
mov rax, rdi
vmovups xmm0, xmmword ptr [rax]
vmovups xmm1, xmmword ptr [reloc @RWD00]
vptest xmm0, xmm1
jne G_M6063_IG09
mov rcx, rsi
vpackuswb xmm0, xmm0, xmm0
vmovsd qword ptr [rcx], xmm0
mov r8d, 8
test sil, 8
jne SHORT G_M6063_IG04
vmovups xmm0, xmmword ptr [rax+0x10]
vptest xmm0, xmm1
jne SHORT G_M6063_IG08
vpackuswb xmm0, xmm0, xmm0
vmovsd qword ptr [rcx+0x08], xmm0
;; size=105 bbWeight=0.50 PerfScore 16.00
G_M6063_IG04:
mov r8, rsi
and r8, 15
neg r8
add r8, 16
lea r9, [rdx-0x10]
align [0 bytes for IG05]
;; size=18 bbWeight=0.50 PerfScore 0.75
G_M6063_IG05:
vmovups xmm0, xmmword ptr [rax+2*r8]
lea r10, [r8+0x08]
vmovups xmm2, xmmword ptr [rax+2*r10]
vpor xmm3, xmm0, xmm2
vptest xmm3, xmm1
je SHORT G_M6063_IG07
;; size=27 bbWeight=4 PerfScore 51.33
G_M6063_IG06:
vptest xmm0, xmm1
jne SHORT G_M6063_IG08
vpackuswb xmm0, xmm0, xmm0
vmovsd qword ptr [rcx+r8], xmm0
mov r8, r10
jmp SHORT G_M6063_IG08
align [0 bytes for IG13]
;; size=22 bbWeight=0.50 PerfScore 4.62
G_M6063_IG07:
vpackuswb xmm0, xmm0, xmm2
vmovups xmmword ptr [rcx+r8], xmm0
add r8, 16
cmp r8, r9
jbe SHORT G_M6063_IG05
;; size=19 bbWeight=4 PerfScore 18.00
G_M6063_IG08:
mov rax, r8
jmp SHORT G_M6063_IG10
;; size=5 bbWeight=0.50 PerfScore 1.12
G_M6063_IG09:
xor eax, eax
;; size=2 bbWeight=0.50 PerfScore 0.12
G_M6063_IG10:
jmp G_M6063_IG18
;; size=5 bbWeight=0.50 PerfScore 1.00
G_M6063_IG11:
mov rax, rdi
vmovups ymm0, ymmword ptr [rax]
vmovups ymm1, ymmword ptr [reloc @RWD32]
vptest ymm0, ymm1
jne G_M6063_IG17
mov rcx, rsi
- vmovups ymm2, ymmword ptr [reloc @RWD64]
- vpand ymm0, ymm2, ymm0
vpackuswb ymm0, ymm0, ymm0
vpermq ymm0, ymm0, -40
vmovups xmmword ptr [rcx], xmm0
mov r8d, 16
test sil, 16
jne SHORT G_M6063_IG12
vmovups ymm0, ymmword ptr [rax+0x20]
vptest ymm0, ymm1
- jne G_M6063_IG16
- vpand ymm0, ymm2, ymm0
+ jne SHORT G_M6063_IG16
vpackuswb ymm0, ymm0, ymm0
vpermq ymm0, ymm0, -40
vmovups xmmword ptr [rcx+0x10], xmm0
- ;; size=102 bbWeight=0.50 PerfScore 21.33
+ ;; size=82 bbWeight=0.50 PerfScore 19.00
G_M6063_IG12:
mov r8, rsi
and r8, 31
neg r8
add r8, 32
lea r9, [rdx-0x20]
;; size=18 bbWeight=0.50 PerfScore 0.75
G_M6063_IG13:
vmovups ymm0, ymmword ptr [rax+2*r8]
- vmovups ymm3, ymmword ptr [rax+2*r8+0x20]
- vpor ymm4, ymm0, ymm3
- vptest ymm4, ymm1
+ vmovups ymm2, ymmword ptr [rax+2*r8+0x20]
+ vpor ymm3, ymm0, ymm2
+ vptest ymm3, ymm1
je SHORT G_M6063_IG15
;; size=24 bbWeight=4 PerfScore 65.33
G_M6063_IG14:
vptest ymm0, ymm1
jne SHORT G_M6063_IG16
- vpand ymm3, ymm2, ymm0
- vpand ymm0, ymm2, ymm0
- vpackuswb ymm2, ymm3, ymm0
- vpermq ymm1, ymm2, -40
- vmovups xmmword ptr [rcx+r8], xmm1
+ vpackuswb ymm0, ymm0, ymm0
+ vpermq ymm2, ymm0, -40
+ vmovups xmmword ptr [rcx+r8], xmm2
add r8, 16
jmp SHORT G_M6063_IG16
align [0 bytes for IG19]
- ;; size=37 bbWeight=0.50 PerfScore 6.96
+ ;; size=29 bbWeight=0.50 PerfScore 6.62
G_M6063_IG15:
- vpand ymm0, ymm2, ymm0
- vpand ymm3, ymm2, ymm3
- vpackuswb ymm0, ymm0, ymm3
+ vpackuswb ymm0, ymm0, ymm2
vpermq ymm0, ymm0, -40
vmovups ymmword ptr [rcx+r8], ymm0
add r8, 32
cmp r8, r9
jbe SHORT G_M6063_IG13
- ;; size=33 bbWeight=4 PerfScore 28.67
+ ;; size=25 bbWeight=4 PerfScore 26.00
G_M6063_IG16:
mov rax, r8
jmp SHORT G_M6063_IG18
;; size=5 bbWeight=0.50 PerfScore 1.12
G_M6063_IG17:
xor eax, eax
;; size=2 bbWeight=0.50 PerfScore 0.12
G_M6063_IG18:
sub rdx, rax
cmp rdx, 4
jb SHORT G_M6063_IG22
lea rcx, [rax+rdx-0x04]
;; size=14 bbWeight=0.50 PerfScore 1.25
G_M6063_IG19:
mov r8, qword ptr [rdi+2*rax]
mov r9, 0xD1FFAB1E
test r8, r9
je SHORT G_M6063_IG21
;; size=19 bbWeight=4 PerfScore 14.00
G_M6063_IG20:
mov ecx, r8d
test ecx, 0xD1FFAB1E
jne SHORT G_M6063_IG23
lea rdx, [rsi+rax]
mov byte ptr [rdx], cl
shr ecx, 16
mov byte ptr [rdx+0x01], cl
shr r8, 32
mov ecx, r8d
add rax, 2
jmp SHORT G_M6063_IG23
;; size=36 bbWeight=0.50 PerfScore 3.75
G_M6063_IG21:
vmovd xmm0, r8
vpackuswb xmm0, xmm0, xmm0
vmovd dword ptr [rsi+rax], xmm0
add rax, 4
cmp rax, rcx
jbe SHORT G_M6063_IG19
;; size=23 bbWeight=4 PerfScore 26.00
G_M6063_IG22:
test dl, 2
je SHORT G_M6063_IG25
mov ecx, dword ptr [rdi+2*rax]
test ecx, 0xD1FFAB1E
je SHORT G_M6063_IG24
;; size=16 bbWeight=0.50 PerfScore 2.25
G_M6063_IG23:
test ecx, 0xFF80
je SHORT G_M6063_IG26
jmp SHORT G_M6063_IG27
;; size=10 bbWeight=0.50 PerfScore 1.62
G_M6063_IG24:
lea r8, [rsi+rax]
mov byte ptr [r8], cl
shr ecx, 16
mov byte ptr [r8+0x01], cl
add rax, 2
;; size=18 bbWeight=0.50 PerfScore 1.62
G_M6063_IG25:
test dl, 1
je SHORT G_M6063_IG27
movzx rcx, word ptr [rdi+2*rax]
cmp ecx, 127
ja SHORT G_M6063_IG27
;; size=14 bbWeight=0.50 PerfScore 2.25
G_M6063_IG26:
mov byte ptr [rsi+rax], cl
inc rax
;; size=6 bbWeight=0.50 PerfScore 0.62
G_M6063_IG27:
vzeroupper
pop rbp
ret
;; size=5 bbWeight=1 PerfScore 2.50
RWD00 dq FF80FF80FF80FF80h, FF80FF80FF80FF80h
RWD16 dd 00000000h, 00000000h, 00000000h, 00000000h
RWD32 dq FF80FF80FF80FF80h, FF80FF80FF80FF80h, FF80FF80FF80FF80h, FF80FF80FF80FF80h
-RWD64 dq 00FF00FF00FF00FFh, 00FF00FF00FF00FFh, 00FF00FF00FF00FFh, 00FF00FF00FF00FFh
-; Total bytes of code 601, prolog size 4, PerfScore 275.88, instruction count 156, allocated bytes for code 605 (MethodHash=53fae850) for method System.Text.Ascii:NarrowUtf16ToAscii(ulong,ulong,ulong):ulong (FullOpts)
+; Total bytes of code 565, prolog size 4, PerfScore 270.54, instruction count 149, allocated bytes for code 573 (MethodHash=53fae850) for method System.Text.Ascii:NarrowUtf16ToAscii(ulong,ulong,ulong):ulong (FullOpts) -19 (-5.40 % of base) - System.HexConverter:TryDecodeFromUtf16_Vector128(System.ReadOnlySpan`1[ushort],System.Span`1[ubyte],byref):ubyte ; Assembly listing for method System.HexConverter:TryDecodeFromUtf16_Vector128(System.ReadOnlySpan`1[ushort],System.Span`1[ubyte],byref):ubyte (FullOpts)
; Emitting BLENDED_CODE for X64 with AVX - Unix
; FullOpts code
; optimized code
; rbp based frame
; fully interruptible
; No PGO data
; 0 inlinees with PGO data; 6 single block inlinees; 6 inlinees without PGO data
; Final local variable assignments
;
;* V00 arg0 [V00 ] ( 0, 0 ) struct (16) zero-ref multireg-arg ld-addr-op single-def <System.ReadOnlySpan`1[ushort]>
;* V01 arg1 [V01 ] ( 0, 0 ) struct (16) zero-ref multireg-arg ld-addr-op single-def <System.Span`1[ubyte]>
; V02 arg2 [V02,T06] ( 4, 3 ) byref -> rbx single-def
; V03 loc0 [V03,T00] ( 12, 42.50) long -> r15
; V04 loc1 [V04,T02] ( 3, 9 ) long -> r13
;* V05 loc2 [V05,T19] ( 0, 0 ) byref -> zero-ref single-def
;* V06 loc3 [V06,T20] ( 0, 0 ) byref -> zero-ref single-def
; V07 loc4 [V07 ] ( 2, 1 ) int -> [rbp-0x28] do-not-enreg[X] addr-exposed ld-addr-op
-; V08 loc5 [V08,T22] ( 3, 24 ) simd16 -> mm8 <System.Runtime.Intrinsics.Vector128`1[ushort]>
-; V09 loc6 [V09,T23] ( 3, 24 ) simd16 -> mm9 <System.Runtime.Intrinsics.Vector128`1[ushort]>
+; V08 loc5 [V08,T21] ( 3, 24 ) simd16 -> mm7 <System.Runtime.Intrinsics.Vector128`1[ushort]>
+; V09 loc6 [V09,T22] ( 3, 24 ) simd16 -> mm8 <System.Runtime.Intrinsics.Vector128`1[ushort]>
;* V10 loc7 [V10 ] ( 0, 0 ) simd16 -> zero-ref <System.Runtime.Intrinsics.Vector128`1[ubyte]>
;* V11 loc8 [V11 ] ( 0, 0 ) simd16 -> zero-ref <System.Runtime.Intrinsics.Vector128`1[ubyte]>
-; V12 loc9 [V12,T25] ( 3, 16 ) simd16 -> mm10 <System.Runtime.Intrinsics.Vector128`1[ubyte]>
+; V12 loc9 [V12,T24] ( 3, 16 ) simd16 -> mm9 <System.Runtime.Intrinsics.Vector128`1[ubyte]>
;* V13 loc10 [V13 ] ( 0, 0 ) simd16 -> zero-ref <System.Runtime.Intrinsics.Vector128`1[ubyte]>
;* V14 loc11 [V14 ] ( 0, 0 ) simd16 -> zero-ref <System.Runtime.Intrinsics.Vector128`1[short]>
;* V15 loc12 [V15 ] ( 0, 0 ) simd16 -> zero-ref <System.Runtime.Intrinsics.Vector128`1[short]>
;# V16 OutArgs [V16 ] ( 1, 1 ) struct ( 0) [rsp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
-; V17 tmp1 [V17,T21] ( 3, 48 ) simd16 -> mm10 "dup spill"
+; V17 tmp1 [V17,T23] ( 3, 24 ) simd16 -> mm9
;* V18 tmp2 [V18 ] ( 0, 0 ) struct (16) zero-ref "impAppendStmt" <System.ReadOnlySpan`1[ushort]>
;* V19 tmp3 [V19 ] ( 0, 0 ) struct (16) zero-ref "spilled call-like call argument" <System.Span`1[ubyte]>
; V20 tmp4 [V20,T12] ( 2, 2 ) int -> rax "impAppendStmt"
;* V21 tmp5 [V21 ] ( 0, 0 ) simd16 -> zero-ref "spilled call-like call argument"
;* V22 tmp6 [V22 ] ( 0, 0 ) struct (16) zero-ref ld-addr-op "Inlining Arg" <System.ReadOnlySpan`1[ushort]>
;* V23 tmp7 [V23 ] ( 0, 0 ) struct (16) zero-ref ld-addr-op "Inlining Arg" <System.Span`1[ubyte]>
;* V24 tmp8 [V24 ] ( 0, 0 ) simd16 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
;* V25 tmp9 [V25 ] ( 0, 0 ) simd16 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
;* V26 tmp10 [V26 ] ( 0, 0 ) simd16 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
;* V27 tmp11 [V27 ] ( 0, 0 ) simd16 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
;* V28 tmp12 [V28 ] ( 0, 0 ) simd16 -> zero-ref "Inlining Arg" <System.Runtime.Intrinsics.Vector128`1[ushort]>
;* V29 tmp13 [V29 ] ( 0, 0 ) simd16 -> zero-ref "spilled call-like call argument"
;* V30 tmp14 [V30 ] ( 0, 0 ) simd16 -> zero-ref ld-addr-op "Inline stloc first use temp" <System.Runtime.Intrinsics.Vector128`1[ushort]>
;* V31 tmp15 [V31 ] ( 0, 0 ) ubyte -> zero-ref "Inline return value spill temp"
;* V32 tmp16 [V32 ] ( 0, 0 ) simd16 -> zero-ref "Inline return value spill temp" <System.Runtime.Intrinsics.Vector128`1[ubyte]>
; V33 tmp17 [V33,T07] ( 4, 4 ) int -> r8 "Inlining Arg"
;* V34 tmp18 [V34 ] ( 0, 0 ) struct (16) zero-ref multireg-arg ld-addr-op "NewObj constructor temp" <System.ReadOnlySpan`1[ushort]>
; V35 tmp19 [V35,T10] ( 2, 2 ) byref -> rdi single-def "Inlining Arg"
; V36 tmp20 [V36,T13] ( 2, 2 ) int -> rsi "Inlining Arg"
; V37 tmp21 [V37,T08] ( 4, 4 ) int -> r8 "Inlining Arg"
;* V38 tmp22 [V38 ] ( 0, 0 ) struct (16) zero-ref multireg-arg ld-addr-op "NewObj constructor temp" <System.Span`1[ubyte]>
; V39 tmp23 [V39,T11] ( 2, 2 ) byref -> rdx single-def "Inlining Arg"
; V40 tmp24 [V40,T14] ( 2, 2 ) int -> rcx "Inlining Arg"
; V41 tmp25 [V41,T01] ( 4, 17.50) byref -> rdi single-def "field V00._reference (fldOffset=0x0)" P-INDEP
; V42 tmp26 [V42,T05] ( 5, 3.50) int -> rsi single-def "field V00._length (fldOffset=0x8)" P-INDEP
; V43 tmp27 [V43,T03] ( 3, 5.50) byref -> rdx single-def "field V01._reference (fldOffset=0x0)" P-INDEP
; V44 tmp28 [V44,T09] ( 3, 2 ) int -> rcx single-def "field V01._length (fldOffset=0x8)" P-INDEP
;* V45 tmp29 [V45 ] ( 0, 0 ) byref -> zero-ref single-def "field V18._reference (fldOffset=0x0)" P-INDEP
;* V46 tmp30 [V46 ] ( 0, 0 ) int -> zero-ref "field V18._length (fldOffset=0x8)" P-INDEP
;* V47 tmp31 [V47 ] ( 0, 0 ) byref -> zero-ref "field V19._reference (fldOffset=0x0)" P-INDEP
;* V48 tmp32 [V48 ] ( 0, 0 ) int -> zero-ref "field V19._length (fldOffset=0x8)" P-INDEP
;* V49 tmp33 [V49 ] ( 0, 0 ) byref -> zero-ref single-def "field V22._reference (fldOffset=0x0)" P-INDEP
;* V50 tmp34 [V50 ] ( 0, 0 ) int -> zero-ref "field V22._length (fldOffset=0x8)" P-INDEP
;* V51 tmp35 [V51 ] ( 0, 0 ) byref -> zero-ref single-def "field V23._reference (fldOffset=0x0)" P-INDEP
;* V52 tmp36 [V52 ] ( 0, 0 ) int -> zero-ref "field V23._length (fldOffset=0x8)" P-INDEP
; V53 tmp37 [V53,T15] ( 2, 1 ) byref -> rdi single-def "field V34._reference (fldOffset=0x0)" P-INDEP
; V54 tmp38 [V54,T17] ( 2, 1 ) int -> rsi "field V34._length (fldOffset=0x8)" P-INDEP
; V55 tmp39 [V55,T16] ( 2, 1 ) byref -> rdx single-def "field V38._reference (fldOffset=0x0)" P-INDEP
; V56 tmp40 [V56,T18] ( 2, 1 ) int -> rcx "field V38._length (fldOffset=0x8)" P-INDEP
-; V57 cse0 [V57,T24] ( 3, 17 ) simd16 -> mm0 hoist "CSE #01: aggressive"
+; V57 cse0 [V57,T25] ( 2, 9 ) simd16 -> mm0 hoist "CSE #01: aggressive"
; V58 cse1 [V58,T26] ( 2, 9 ) simd16 -> mm1 hoist "CSE #02: aggressive"
; V59 cse2 [V59,T27] ( 2, 9 ) simd16 -> mm2 hoist "CSE #03: aggressive"
; V60 cse3 [V60,T28] ( 2, 9 ) simd16 -> mm3 hoist "CSE #04: aggressive"
; V61 cse4 [V61,T29] ( 2, 9 ) simd16 -> mm4 hoist "CSE #05: aggressive"
; V62 cse5 [V62,T30] ( 2, 9 ) simd16 -> mm5 hoist "CSE #06: aggressive"
; V63 cse6 [V63,T31] ( 2, 9 ) simd16 -> mm6 hoist "CSE #07: aggressive"
-; V64 cse7 [V64,T32] ( 2, 9 ) simd16 -> mm7 hoist "CSE #08: aggressive"
-; V65 cse8 [V65,T04] ( 3, 6 ) long -> r14 "CSE #09: aggressive"
+; V64 cse7 [V64,T04] ( 3, 6 ) long -> r14 "CSE #08: aggressive"
;
; Lcl frame size = 16
G_M6966_IG01:
push rbp
push r15
push r14
push r13
push rbx
sub rsp, 16
lea rbp, [rsp+0x30]
mov rbx, r8
;; size=20 bbWeight=1 PerfScore 6.00
G_M6966_IG02:
xor r15d, r15d
mov r14d, esi
lea r13, [r14-0x10]
vmovups xmm0, xmmword ptr [reloc @RWD00]
vmovups xmm1, xmmword ptr [reloc @RWD16]
vmovups xmm2, xmmword ptr [reloc @RWD32]
vmovups xmm3, xmmword ptr [reloc @RWD48]
vmovups xmm4, xmmword ptr [reloc @RWD64]
vmovups xmm5, xmmword ptr [reloc @RWD80]
vmovups xmm6, xmmword ptr [reloc @RWD96]
- vmovups xmm7, xmmword ptr [reloc @RWD112]
jmp SHORT G_M6966_IG04
align [0 bytes for IG03]
- ;; size=76 bbWeight=1 PerfScore 27.00
+ ;; size=68 bbWeight=1 PerfScore 24.00
G_M6966_IG03:
mov r15, r13
;; size=3 bbWeight=4 PerfScore 1.00
G_M6966_IG04:
- vmovups xmm8, xmmword ptr [rdi+2*r15]
- vmovups xmm9, xmmword ptr [rdi+2*r15+0x10]
- vpand xmm10, xmm0, xmm8
- vpand xmm11, xmm0, xmm9
- vpackuswb xmm10, xmm10, xmm11
- vpaddb xmm11, xmm1, xmm10
- vpsubusb xmm11, xmm11, xmm2
- vpsubb xmm11, xmm11, xmm3
- vpand xmm10, xmm4, xmm10
- vpsubb xmm10, xmm10, xmm5
- vpaddusb xmm10, xmm10, xmm6
- vpminub xmm10, xmm11, xmm10
- vpor xmm8, xmm8, xmm9
- vptest xmm8, xmm7
+ vmovups xmm7, xmmword ptr [rdi+2*r15]
+ vmovups xmm8, xmmword ptr [rdi+2*r15+0x10]
+ vpackuswb xmm9, xmm7, xmm8
+ vpaddb xmm10, xmm0, xmm9
+ vpsubusb xmm10, xmm10, xmm1
+ vpsubb xmm10, xmm10, xmm2
+ vpand xmm9, xmm3, xmm9
+ vpsubb xmm9, xmm9, xmm4
+ vpaddusb xmm9, xmm9, xmm5
+ vpminub xmm9, xmm10, xmm9
+ vpor xmm7, xmm7, xmm8
+ vptest xmm7, xmm6
jne SHORT G_M6966_IG06
- ;; size=71 bbWeight=8 PerfScore 132.00
+ ;; size=61 bbWeight=8 PerfScore 126.67
G_M6966_IG05:
- vpaddusb xmm8, xmm10, xmmword ptr [reloc @RWD128]
- vpmovmskb r8d, xmm8
+ vpaddusb xmm7, xmm9, xmmword ptr [reloc @RWD112]
+ vpmovmskb r8d, xmm7
test r8d, r8d
je SHORT G_M6966_IG08
- ;; size=18 bbWeight=4 PerfScore 21.00
+ ;; size=17 bbWeight=4 PerfScore 21.00
G_M6966_IG06:
mov r8d, r15d
cmp r8d, esi
ja G_M6966_IG11
mov eax, r8d
lea rdi, bword ptr [rdi+2*rax]
sub esi, r8d
mov r8, r15
shr r8, 1
cmp r8d, ecx
ja SHORT G_M6966_IG11
mov eax, r8d
add rdx, rax
sub ecx, r8d
lea r8, [rbp-0x28]
mov rax, 0xD1FFAB1E ; code for System.HexConverter:TryDecodeFromUtf16_Scalar(System.ReadOnlySpan`1[ushort],System.Span`1[ubyte],byref):ubyte
call [rax]System.HexConverter:TryDecodeFromUtf16_Scalar(System.ReadOnlySpan`1[ushort],System.Span`1[ubyte],byref):ubyte
add r15d, dword ptr [rbp-0x28]
mov dword ptr [rbx], r15d
;; size=65 bbWeight=0.50 PerfScore 6.00
G_M6966_IG07:
add rsp, 16
pop rbx
pop r13
pop r14
pop r15
pop rbp
ret
;; size=13 bbWeight=0.50 PerfScore 1.88
G_M6966_IG08:
- vpmaddubsw xmm8, xmm10, xmmword ptr [reloc @RWD144]
- vpshufb xmm8, xmm8, xmmword ptr [reloc @RWD160]
+ vpmaddubsw xmm7, xmm9, xmmword ptr [reloc @RWD128]
+ vpshufb xmm7, xmm7, xmmword ptr [reloc @RWD144]
mov rax, r15
shr rax, 1
- vmovd qword ptr [rdx+rax], xmm8
+ vmovd qword ptr [rdx+rax], xmm7
add r15, 16
cmp r15, r14
je SHORT G_M6966_IG09
cmp r15, r13
jbe G_M6966_IG04
jmp G_M6966_IG03
;; size=53 bbWeight=4 PerfScore 62.00
G_M6966_IG09:
mov dword ptr [rbx], esi
mov eax, 1
;; size=7 bbWeight=0.50 PerfScore 0.62
G_M6966_IG10:
add rsp, 16
pop rbx
pop r13
pop r14
pop r15
pop rbp
ret
;; size=13 bbWeight=0.50 PerfScore 1.88
G_M6966_IG11:
mov rax, 0xD1FFAB1E ; code for System.ThrowHelper:ThrowArgumentOutOfRangeException()
call [rax]System.ThrowHelper:ThrowArgumentOutOfRangeException()
int3
;; size=13 bbWeight=0 PerfScore 0.00
-RWD00 dq 00FF00FF00FF00FFh, 00FF00FF00FF00FFh
-RWD16 dq C6C6C6C6C6C6C6C6h, C6C6C6C6C6C6C6C6h
-RWD32 dq 0606060606060606h, 0606060606060606h
-RWD48 dq F0F0F0F0F0F0F0F0h, F0F0F0F0F0F0F0F0h
-RWD64 dq DFDFDFDFDFDFDFDFh, DFDFDFDFDFDFDFDFh
-RWD80 dq 4141414141414141h, 4141414141414141h
-RWD96 dq 0A0A0A0A0A0A0A0Ah, 0A0A0A0A0A0A0A0Ah
-RWD112 dq FF80FF80FF80FF80h, FF80FF80FF80FF80h
-RWD128 dq 7070707070707070h, 7070707070707070h
-RWD144 dq 0110011001100110h, 0110011001100110h
-RWD160 dq 0E0C0A0806040200h, 0000000000000000h
+RWD00 dq C6C6C6C6C6C6C6C6h, C6C6C6C6C6C6C6C6h
+RWD16 dq 0606060606060606h, 0606060606060606h
+RWD32 dq F0F0F0F0F0F0F0F0h, F0F0F0F0F0F0F0F0h
+RWD48 dq DFDFDFDFDFDFDFDFh, DFDFDFDFDFDFDFDFh
+RWD64 dq 4141414141414141h, 4141414141414141h
+RWD80 dq 0A0A0A0A0A0A0A0Ah, 0A0A0A0A0A0A0A0Ah
+RWD96 dq FF80FF80FF80FF80h, FF80FF80FF80FF80h
+RWD112 dq 7070707070707070h, 7070707070707070h
+RWD128 dq 0110011001100110h, 0110011001100110h
+RWD144 dq 0E0C0A0806040200h, 0000000000000000h
-; Total bytes of code 352, prolog size 20, PerfScore 259.38, instruction count 89, allocated bytes for code 352 (MethodHash=bb7ae4c9) for method System.HexConverter:TryDecodeFromUtf16_Vector128(System.ReadOnlySpan`1[ushort],System.Span`1[ubyte],byref):ubyte (FullOpts)
+; Total bytes of code 333, prolog size 20, PerfScore 251.04, instruction count 86, allocated bytes for code 333 (MethodHash=bb7ae4c9) for method System.HexConverter:TryDecodeFromUtf16_Vector128(System.ReadOnlySpan`1[ushort],System.Span`1[ubyte],byref):ubyte (FullOpts) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Job completed in 22 minutes.
Diffs
Artifacts:
The text was updated successfully, but these errors were encountered: