Test failure System.Numerics.Tests.Matrix3x2Tests.Matrix3x2CreateRotationCenterTest #70124

Activity

1. VincentBu

   added

   area-System.Numerics

   os-windows

   JitStressCLR JIT issues involving JIT internal stress modesCLR JIT issues involving JIT internal stress modes

   arch-x64

   blocking-clean-ci-optionalBlocking optional rolling runsBlocking optional rolling runs

   on Jun 2, 2022
2. ghost added

   untriagedNew issue has not been triaged by the area ownerNew issue has not been triaged by the area owner

   on Jun 2, 2022
4. SingleAccretion

   self-assigned this

   on Jun 5, 2022
5. SingleAccretion

   added this to the 7.0.0 milestone on Jun 5, 2022
6. ghost removed

   untriagedNew issue has not been triaged by the area ownerNew issue has not been triaged by the area owner

   on Jun 5, 2022
7. SingleAccretion commented on Jun 5, 2022

   

   SingleAccretion

   on Jun 5, 2022

   Last edited by SingleAccretion

   Contributor

   More actions

   So, as one would suspect, this has been exposed (but not caused) by #68874.

   In the test we have retyping performed by fgMorphArgs generate the following tree:

   N004 ( 23, 19) [000014] S-CXG------                         *  CALL      void   System.Numerics.Matrix3x2.CreateRotation $VN.Void
   N001 (  3,  3) [000016] ----------- retbuf in rcx           +--*  LCL_VAR_ADDR long   V03 loc2          $200
   N002 (  3,  4) [000012] ----------- arg1 in mm1             +--*  CNS_DBL   float  0.52359879016876221 $41
   N003 (  3,  4) [000013] ----------- arg2 in r8              \--*  CNS_VEC   long  <0x00000000, 0x00000000> $80

   Then CSE comes along and CSEs [000013]:

   optValnumCSE morphed tree:
   N008 ( 22, 18)              [000014] SACXG------                         *  CALL      void   System.Numerics.Matrix3x2.CreateRotation $VN.Void
   N001 (  3,  3)              [000016] ----------- retbuf in rcx           +--*  LCL_VAR_ADDR long   V03 loc2          $200
   N002 (  1,  2)              [000430] ----------- arg1 in mm1             +--*  LCL_VAR   float  V32 cse0         u:1 $41
   N007 (  4,  5)              [000435] -A--------- arg2 in r8              \--*  COMMA     long   $80
   N005 (  3,  4) CSE #02 (def)[000433] -A------R--                            +--*  ASG       long   $VN.Void
   N004 (  1,  1)              [000432] D------N---                            |  +--*  LCL_VAR   long   V33 cse1         d:1 $80
   N003 (  3,  4)              [000013] -----------                            |  \--*  CNS_VEC   long  <0x00000000, 0x00000000> $80
   N006 (  1,  1)              [000434] -----------                            \--*  LCL_VAR   long   V33 cse1         u:1 $80
   
   
   Working on the replacement of the CSE #02 use at [000202] in BB08
   optValnumCSE morphed tree:
   N005 (  3,  3)              [000045] -A------R--                         *  ASG       simd8  (copy) $VN.Void
   N004 (  1,  1)              [000043] D------N---                         +--*  LCL_VAR   simd8 <System.Numerics.Vector2> V08 tmp1         d:2 $VN.Void
   N003 (  3,  3)              [000204] --------R--                         \--*  HWINTRINSIC simd8  float Subtract $1c1
   N002 (  1,  1)              [000436] -----------                            +--*  LCL_VAR   long   V33 cse1         u:1 $80
   N001 (  1,  1)              [000037] -----------                            \--*  LCL_VAR   simd8 <System.Numerics.Vector2> V02 loc1         u:2 $1c0

   It is the latter use that (presumably) causes trouble later on, we emit this for it:

   Generating: N281 (  1,  1) [000037] -----------                   t37 =    LCL_VAR   simd8 <System.Numerics.Vector2> V02 loc1         u:2 mm6 REG mm6 $1c0
   Generating: N283 (  1,  1) [000436] -----------                  t436 =    LCL_VAR   long   V33 cse1         u:1 rsi (last use) REG rsi $80
                                                                           /--*  t436   long   
                                                                           +--*  t37    simd8  
   Generating: N285 (  3,  3) [000204] -----------                  t204 = *  HWINTRINSIC simd8  float Subtract REG mm7 $1c1
   							V33 in reg rsi is becoming dead  [000436]
   							Live regs: 00000040 {rsi xmm6} => 00000000 {xmm6}
   							Live vars: {V02 V29 V33} => {V02 V29}
   IN003b:        vsubps   xmm7, rsi, xmm6
8. SingleAccretion

   added

   area-CodeGen-coreclrCLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMICLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI

   and removed

   area-System.Numerics

   on Jun 5, 2022
10. ghost added

    in-prThere is an active PR which will close this issue when it is mergedThere is an active PR which will close this issue when it is merged

    on Jun 5, 2022
11. BruceForstall

    closed this as completedin #70265on Jun 7, 2022
12. ghost removed

    in-prThere is an active PR which will close this issue when it is mergedThere is an active PR which will close this issue when it is merged

    on Jun 7, 2022
13. ghost locked as resolved and limited conversation to collaborators on Jul 7, 2022