Test failure System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02 #83316

v-wenyuxu · 2023-03-13T01:06:29Z

Failed in Run: runtime-coreclr libraries-jitstress2-jitstressregs 20230311.1

Failed tests:

net8.0-windows-Release-x64-CoreCLR_checked-jitstress2_jitstressregs3-Windows.10.Amd64.Open
    - System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02

Error message:

Matrix4x4.CreateShadow did not provide expected value.
Expected: True
Actual:   False

Stack trace:

  at System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02() in /_/src/libraries/System.Numerics.Vectors/tests/Matrix4x4Tests.cs:line 762
  at System.RuntimeMethodHandle.InvokeMethod(Object target, Void** arguments, Signature sig, Boolean isConstructor)
  at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr) in /_/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodInvoker.cs:line 59

The text was updated successfully, but these errors were encountered:

ghost · 2023-03-13T01:06:40Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch, @kunalspathak
See info in area-owners.md if you want to be subscribed.

Issue Details

Failed in Run: runtime-coreclr libraries-jitstress2-jitstressregs 20230311.1

Failed tests:

net8.0-windows-Release-x64-CoreCLR_checked-jitstress2_jitstressregs3-Windows.10.Amd64.Open
    - System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02

Error message:

Matrix4x4.CreateShadow did not provide expected value.
Expected: True
Actual:   False

Stack trace:

  at System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02() in /_/src/libraries/System.Numerics.Vectors/tests/Matrix4x4Tests.cs:line 762
  at System.RuntimeMethodHandle.InvokeMethod(Object target, Void** arguments, Signature sig, Boolean isConstructor)
  at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr) in /_/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodInvoker.cs:line 59

Author:	v-wenyuxu
Assignees:	-
Labels:	`os-windows`, `JitStress`, `arch-x64`, `area-CodeGen-coreclr`
Milestone:	-

JulieLeeMSFT · 2023-03-14T11:28:48Z

@tannergooding PTAL.

v-wenyuxu · 2023-05-15T01:21:17Z

Failed again in: runtime-coreclr libraries-jitstress2-jitstressregs 20230513.1

Failed tests:

net8.0-windows-Release-x64-CoreCLR_checked-jitstress2_jitstressregs3-Windows.10.Amd64.Open
    - System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02

�Error message:

 Matrix4x4.CreateShadow did not provide expected value.
Expected: True
Actual:   False

�Stack trace:

    at System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02() in /_/src/libraries/System.Numerics.Vectors/tests/Matrix4x4Tests.cs:line 762
   at System.RuntimeMethodHandle.InvokeMethod(Object target, Void** arguments, Signature sig, Boolean isConstructor)
   at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr) in /_/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodInvoker.cs:line 59

�

BruceForstall · 2023-05-17T19:12:54Z

Still failing

https://dev.azure.com/dnceng-public/public/_build/results?buildId=277068&view=ms.vss-test-web.build-test-results-tab&runId=5495120&resultId=133658&paneView=debug

@kunalspathak This apparently only fails in JitStressRegs=3; maybe you should take a look?

v-wenyuxu · 2023-05-29T01:29:02Z

Failed again in: runtime-coreclr libraries-jitstress2-jitstressregs 20230527.1

Failed tests:

net8.0-windows-Release-x64-CoreCLR_checked-jitstress2_jitstressregs3-Windows.10.Amd64.Open
    - System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02

Error message:

 Matrix4x4.CreateShadow did not provide expected value.
Expected: True
Actual:   False

Stack trace:

   at System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02() in /_/src/libraries/System.Numerics.Vectors/tests/Matrix4x4Tests.cs:line 762
   at System.RuntimeMethodHandle.InvokeMethod(Object target, Void** arguments, Signature sig, Boolean isConstructor)
   at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr) in /_/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodInvoker.cs:line 59

v-wenyuxu · 2023-06-05T01:56:41Z

Failed again in: runtime-coreclr libraries-jitstress2-jitstressregs 20230603.1

Failed test:

net8.0-windows-Release-x64-CoreCLR_checked-jitstress2_jitstressregs3-Windows.10.Amd64.Open

- System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02

Error message:

Matrix4x4.CreateShadow did not provide expected value.
Expected: True
Actual:   False



Stack trace
   at System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02() in /_/src/libraries/System.Numerics.Vectors/tests/Matrix4x4Tests.cs:line 762
   at System.RuntimeMethodHandle.InvokeMethod(Object target, Void** arguments, Signature sig, Boolean isConstructor)
   at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr) in /_/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodInvoker.cs:line 59

v-wenyuxu · 2023-06-12T01:20:21Z

Failed again in: runtime-coreclr libraries-jitstress2-jitstressregs 20230610.1

Failed tests:

net8.0-windows-Release-x64-CoreCLR_checked-jitstress2_jitstressregs3-Windows.10.Amd64.Open
    - System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02

�Error message:

 Matrix4x4.CreateShadow did not provide expected value.
Expected: True
Actual:   False

�Stack trace:

    at System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02() in /_/src/libraries/System.Numerics.Vectors/tests/Matrix4x4Tests.cs:line 762
   at System.RuntimeMethodHandle.InvokeMethod(Object target, Void** arguments, Signature sig, Boolean isConstructor)
   at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr) in /_/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodInvoker.cs:line 59

�

BruceForstall · 2023-06-12T20:43:22Z

@tannergooding @kunalspathak Since this is only failing under JitStressRegs=3 I assigned it to Kunal.

kunalspathak · 2023-06-21T17:39:14Z

Taking a look now.

kunalspathak · 2023-06-22T02:14:51Z

The problematic code is the following:

public void Matrix4x4CreateShadowTest02()
{
  if (lightDirInfo.Length() < 0.1f)
       continue;
  Vector3 lightDir = Vector3.Normalize(lightDirInfo);
 ...
 ...
}

[Intrinsic]
[MethodImpl(MethodImplOptions.AggressiveInlining)]
public readonly float Length()
{
    float lengthSquared = LengthSquared();
    return MathF.Sqrt(lengthSquared);
}


[Intrinsic]
[MethodImpl(MethodImplOptions.AggressiveInlining)]
public static Vector3 Normalize(Vector3 value)
{
    return value / value.Length();
}

Here, we CSE lightDirInfo.Length() and store its value as float. In Normalize() when we reload the value, we reload the 64 bits float instead of 128 bits in xmm. That leads to incomplete vector getting loaded with few fields 0. This makes the division returning Nan.

G_M15576_IG58:  ;; offset=0187H
       vmovss   dword ptr [rsp+58H], xmm6
						;; size=6 bbWeight=16 PerfScore 16.00
G_M15576_IG59:  ;; offset=018DH
       vsqrtss  xmm7, xmm7, xmm6
						;; size=4 bbWeight=16 PerfScore 176.00
G_M15576_IG60:  ;; offset=0191H
       vmovss   xmm6, dword ptr [rsp+5CH]
						;; size=6 bbWeight=16 PerfScore 48.00
G_M15576_IG61:  ;; offset=0197H
       vucomiss xmm6, xmm7
						;; size=4 bbWeight=16 PerfScore 32.00
G_M15576_IG62:  ;; offset=019BH
       ja       G_M15576_IG166
						;; size=6 bbWeight=16 PerfScore 16.00
G_M15576_IG63:  ;; offset=01A1H
       vmovss   xmm7, dword ptr [rsp+58H]
						;; size=6 bbWeight=8 PerfScore 24.00
G_M15576_IG64:  ;; offset=01A7H
       vsqrtps  xmm7, xmm7
						;; size=4 bbWeight=8 PerfScore 88.00
G_M15576_IG65:  ;; offset=01ABH
       vdivps   xmm7, xmm2, xmm7

                                                                        ┌──▌  t719   simd16 
Generating: N269 (  3,  3) [000094] -----------                   t94 = ▌  HWINTRINSIC float  float ToScalar REG mm6 <l:$688, c:$689>
                                                                        ┌──▌  t94    float  
Generating: N271 (  3,  3) [000635] DA--------z                         ▌  STORE_LCL_VAR float  V62 cse4         d:1 mm6 REG mm6 $VN.Void
Saved:
      G_M15576_IG57:        ; offs=00012CH, size=0006H, bbWeight=16, BB09 [0002], extend
Created:
      G_M15576_IG58:        ; offs=000132H, size=0000H, bbWeight=16, gcrefRegs=0000 {}
Mapped BB09 to G_M15576_IG58
IN0039:        vmovss   dword ptr [V62 rsp+58H], xmm6

...
...
Generating: N287 (  1,  2) [000638] ----------z                  t638 =    LCL_VAR   float  V62 cse4         u:1 mm7 (last use) REG mm7 <l:$688, c:$689>
                                                                        ┌──▌  t638   float  
Generating: N289 (  2,  3) [000107] -----------                  t107 = ▌  HWINTRINSIC simd16 float Sqrt REG mm7 <l:$207, c:$208>
Mapped BB10 to G_M15576_IG63
IN003e:        vmovss   xmm7, dword ptr [V62 rsp+58H]

The / operator that takes float is:

       public static Vector3 operator /(Vector3 value1, float value2)
       {
            return value1 / new Vector3(value2);
       }

        [Intrinsic]
        public Vector3(float value) : this(value, value, value)
        {
        }

which seems that there are one or both the problems:

the nodes that does the conversion of value2 -> Vector3(value2)
we should not be CSEing the value directly.

Still investigating.

kunalspathak · 2023-06-22T02:28:05Z

Mostly caused by #81335. JitStressRegs seems to just highlight the problem where we end up spilling the value.

tannergooding · 2023-06-22T04:38:31Z

Noting this comes down to a CSE of HWINTRINSIC float float Dot and HWINTRINSIC TYP_SIMD12 float Dot.

Need to determine why VN decides these can be CSE'd even though they have differing return types.

v-wenyuxu · 2023-06-25T03:09:09Z

Failed again in: runtime-coreclr libraries-jitstress2-jitstressregs 20230624.1

Failed tests:

net8.0-windows-Release-x64-CoreCLR_checked-jitstress2_jitstressregs3-Windows.10.Amd64.Open
    - System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02

Error message:

 Matrix4x4.CreateShadow did not provide expected value.
Expected: True
Actual:   False

Stack trace:

    at System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02() in /_/src/libraries/System.Numerics.Vectors/tests/Matrix4x4Tests.cs:line 762
   at System.RuntimeMethodHandle.InvokeMethod(Object target, Void** arguments, Signature sig, Boolean isConstructor)
   at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr) in /_/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodInvoker.cs:line 59

v-wenyuxu · 2023-07-03T01:24:45Z

Failed again in: runtime-coreclr libraries-jitstress2-jitstressregs 20230701.1

Failed tests:

net8.0-windows-Release-x64-CoreCLR_checked-jitstress2_jitstressregs3-Windows.10.Amd64.Open
    - System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02

Error message:

Matrix4x4.CreateShadow did not provide expected value.
Expected: True
Actual:   False

Stack trace:

   at System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02() in /_/src/libraries/System.Numerics.Vectors/tests/Matrix4x4Tests.cs:line 762
   at System.RuntimeMethodHandle.InvokeMethod(Object target, Void** arguments, Signature sig, Boolean isConstructor)
   at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr) in /_/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodInvoker.cs:line 59

kunalspathak · 2023-07-05T17:05:55Z

Assigning to @tannergooding to resolve other problems that does not necessarily show up in this issue.

v-wenyuxu added os-windows JitStress CLR JIT issues involving JIT internal stress modes arch-x64 labels Mar 13, 2023

dotnet-issue-labeler bot added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label Mar 13, 2023

ghost added the untriaged New issue has not been triaged by the area owner label Mar 13, 2023

JulieLeeMSFT assigned tannergooding Mar 14, 2023

JulieLeeMSFT removed the untriaged New issue has not been triaged by the area owner label Mar 14, 2023

JulieLeeMSFT added this to the 8.0.0 milestone Mar 14, 2023

kunalspathak mentioned this issue Jun 7, 2023

Combine two table lookups into one while formatting Guid on Arm64 #87126

Merged

BruceForstall assigned kunalspathak and unassigned tannergooding Jun 12, 2023

BruceForstall added the blocking-clean-ci-optional Blocking optional rolling runs label Jun 12, 2023

kunalspathak mentioned this issue Jul 4, 2023

Fix SIMD node DOT retType #88363

Merged

kunalspathak assigned tannergooding and unassigned kunalspathak Jul 5, 2023

tannergooding mentioned this issue Jul 5, 2023

Ensure that NI_Vector_Dot is always handled as TYP_SIMD #88447

Merged

kunalspathak closed this as completed in #88447 Jul 10, 2023

ghost locked as resolved and limited conversation to collaborators Aug 9, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Test failure System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02 #83316

Test failure System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02 #83316

v-wenyuxu commented Mar 13, 2023

ghost commented Mar 13, 2023

JulieLeeMSFT commented Mar 14, 2023

v-wenyuxu commented May 15, 2023

BruceForstall commented May 17, 2023

v-wenyuxu commented May 29, 2023

v-wenyuxu commented Jun 5, 2023

v-wenyuxu commented Jun 12, 2023

BruceForstall commented Jun 12, 2023

kunalspathak commented Jun 21, 2023

kunalspathak commented Jun 22, 2023 •

edited

Loading

kunalspathak commented Jun 22, 2023 •

edited

Loading

tannergooding commented Jun 22, 2023

v-wenyuxu commented Jun 25, 2023 •

edited

Loading

v-wenyuxu commented Jul 3, 2023

kunalspathak commented Jul 5, 2023

Test failure System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02 #83316

Test failure System.Numerics.Tests.Matrix4x4Tests.Matrix4x4CreateShadowTest02 #83316

Comments

v-wenyuxu commented Mar 13, 2023

ghost commented Mar 13, 2023

JulieLeeMSFT commented Mar 14, 2023

v-wenyuxu commented May 15, 2023

BruceForstall commented May 17, 2023

v-wenyuxu commented May 29, 2023

v-wenyuxu commented Jun 5, 2023

v-wenyuxu commented Jun 12, 2023

BruceForstall commented Jun 12, 2023

kunalspathak commented Jun 21, 2023

kunalspathak commented Jun 22, 2023 • edited Loading

kunalspathak commented Jun 22, 2023 • edited Loading

tannergooding commented Jun 22, 2023

v-wenyuxu commented Jun 25, 2023 • edited Loading

v-wenyuxu commented Jul 3, 2023

kunalspathak commented Jul 5, 2023

kunalspathak commented Jun 22, 2023 •

edited

Loading

kunalspathak commented Jun 22, 2023 •

edited

Loading

v-wenyuxu commented Jun 25, 2023 •

edited

Loading