Implement StoreSelectedScalar for Arm64

[SwapnilGaikwad]

Contribute towards #84510.

// AdvSimd// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar64x3(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3)value,byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar64x4(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3,Vector64<byte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3,Vector64<sbyte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3,Vector64<short>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3,Vector64<ushort>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3,Vector64<int>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3,Vector64<uint>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3,Vector64<float>Value4)value,byteindex);// AdvSimd.Arm64// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar128x3(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3),byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar128x4(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3,Vector128<byte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3,Vector128<sbyte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3,Vector128<short>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3,Vector128<ushort>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3,Vector128<int>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3,Vector128<uint>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3,Vector128<long>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3,Vector128<ulong>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3,Vector128<float>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3,Vector128<double>Value4),byteindex);

[ghost]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[ghost]

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

Issue Details

---

Contribute towards #84510.

// AdvSimdpublicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// AdvSimd.Arm64publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);

Author:	SwapnilGaikwad
Assignees:	-
Labels:	area-CodeGen-coreclr, new-api-needs-documentation, community-contribution
Milestone:	-

[SwapnilGaikwad]

This patch would emit the following assembly sequence

1. When the index is a constant value

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),1);

...
stp d8, d16, [fp, #0x10]	// [V18 tmp17], [V19 tmp18]
ldr q16, [fp, #0x10]	// [V04 tmp3]
ldr q17, [fp, #0x18]	// [V04 tmp3+0x08]
st2 {v16.s, v17.s}[1], [x20]
...

2. When index is not a constant

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),elemIndex);

...
mov x0, x21
movz x2, #0xC8A0 // code for System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[int],System.Runtime.Intrinsics.Vector64`1[int]],ubyte)
movk x2, #0x5DB8 LSL #16
movk x2, #0xFFFF LSL #32
ldr x2, [x2]
...
; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[short],System.Runtime.Intrinsics.Vector64`1[short]],ubyte) (FullOpts)
G_M60890_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x20]!
mov fp, sp
stp d0, d1, [fp, #0x10]	// [V01 arg1], [V01 arg1+0x08]
;; size=12 bbWeight=1 PerfScore 2.50
G_M60890_IG02: ;; offset=0x000C
ldr q16, [fp, #0x10]	// [V01 arg1]
ldr q17, [fp, #0x18]	// [V01 arg1+0x08]
uxtb w1, w1
adr x2, [G_M60890_IG03]
add x2, x2, x1, LSL #3
br x2
;; size=24 bbWeight=1 PerfScore 7.00
G_M60890_IG03: ;; offset=0x0024
st2 {v16.h, v17.h}[0], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG04: ;; offset=0x002C
st2 {v16.h, v17.h}[1], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG05: ;; offset=0x0034
st2 {v16.h, v17.h}[2], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG06: ;; offset=0x003C
st2 {v16.h, v17.h}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M60890_IG07: ;; offset=0x0040
ldp fp, lr, [sp], #0x20
ret lr
;; size=8 bbWeight=1 PerfScore 2.00

3. When the index is variable and out of bounds

The current test crashes with an error.

Illegal instruction (core dumped)

Not sure whether it's the desired behaviour. Also, I haven't considered the possibility of creating a ROP gadget with the sequence for Case 2 above. The sequence of instruction where value of an index can be used control the jump target.

[kunalspathak]

As seen in #93197, the API names should be just StoreSelectedScalar because you can deduce the shape and tuple size from the parameters.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

Alright, so if you follow #93197, the immOp is handled in it.

[kunalspathak]

3. When the index is variable and out of bounds

They should throw IndexOutOfBoundException

[kunalspathak]

RunBasicScenario_UnsafeRead segfaults with the following error

Interesting. What is the cause of it?

[SwapnilGaikwad]

Now we get out of range exception for all the invalid scenarios (with constant or variable index) with all the test cases.

System.ArgumentOutOfRangeException: Specified argument was out of the range of valid values.

I never got to ensure from the assembly that it throws InvalidIndexException. 🤔

This mystery is also resolved now. There was a check for array out of bounds missing. Now added it. Can be confirmed from the assembly for StoreSelectedScalar as well.

...
G_M56847_IG04:
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
...

Full assembly

; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; Emitting BLENDED_CODE for generic ARM64 - Unix
; MinOpts code
; debuggable code
; fp based frame
; fully interruptible
; No PGO data
; Final local variable assignments
;
; V00 arg0 [V00 ] ( 1, 1 ) long -> [fp+0x38] do-not-enreg[]
; V01 arg1 [V01 ] ( 1, 1 ) simd16 -> [fp+0x20] HFA(simd16) do-not-enreg[S] <System.Runtime.Intrinsics.Vector128`1[int]>
; V02 arg2 [V02 ] ( 1, 1 ) ubyte -> [fp+0x1C] do-not-enreg[]
;# V03 OutArgs [V03 ] ( 1, 1 ) struct ( 0) [sp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
;
; Lcl frame size = 48
G_M56847_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x40]!
mov fp, sp
str x0, [fp, #0x38]	// [V00 arg0]
str w1, [fp, #0x1C]	// [V02 arg2]
str q0, [fp, #0x20]	// [V01 arg1]
;; size=20 bbWeight=1 PerfScore 4.50
G_M56847_IG02: ;; offset=0x0014
movz x0, #0x45A8
movk x0, #0x4004 LSL #16
movk x0, #0xFFFF LSL #32
ldr w0, [x0]
cbz w0, G_M56847_IG04
;; size=20 bbWeight=1 PerfScore 5.50
G_M56847_IG03: ;; offset=0x0028
bl CORINFO_HELP_DBG_IS_JUST_MY_CODE
;; size=4 bbWeight=0.50 PerfScore 0.50
G_M56847_IG04: ;; offset=0x002C
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
;; size=28 bbWeight=1 PerfScore 9.00
G_M56847_IG05: ;; offset=0x0048
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
adr x2, [G_M56847_IG06]
add x2, x2, x1, LSL #3
br x2
;; size=20 bbWeight=1 PerfScore 5.00
G_M56847_IG06: ;; offset=0x005C
st1 {v16.s}[0], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG07: ;; offset=0x0064
st1 {v16.s}[1], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG08: ;; offset=0x006C
st1 {v16.s}[2], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG09: ;; offset=0x0074
st1 {v16.s}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M56847_IG10: ;; offset=0x0078
nop
nop
;; size=8 bbWeight=1 PerfScore 1.00
G_M56847_IG11: ;; offset=0x0080
ldp fp, lr, [sp], #0x40
ret lr
;; size=8 bbWeight=1 PerfScore 2.00
; Total bytes of code 136, prolog size 20, PerfScore 48.10, instruction count 34, allocated bytes for code 136 (MethodHash=5b1221f0) for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; ============================================================

Interestingly, for some tests I got IndexOutOfRangeException earlier because the test was executing without an error but failing in the validation stage where it was accessing the out of range index. Only the first test was failing with an illegal instruction probably because of the code layout. Invalid index tried to jump to location that did not contain a valid index. For the further tests, the jump target at invalid index contained a valid instruction so it did not throw any exception during the execution.

[kunalspathak]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

[SwapnilGaikwad]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

Had to move the range check after recomputing the simdSize 🤦 . Now the current failures seems while sending the test data to helix system.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

[kunalspathak]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

You're right, it seems to be related to the arg check. Not sure how to debug it though. I'll try to see if I can reproduce the failure. If that doesn't progress, should I put the explicit check as before?

[kunalspathak]

If that doesn't progress, should I put the explicit check as before?

I would say just inspect the code changes and see if the code path for existing SelectedScalar with single vector is unchanged. That should fix the problem.

[kunalspathak]

@SwapnilGaikwad - can you resolve the merge conflicts?

[fanyang-mono]

The change to Mono codebase looks good to me.

[kunalspathak]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

[fanyang-mono]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

Unfortunately, yes, Mono is not super friendly with handling API with 4 and more input arguments. We need to add additional code to handle those cases, rather than automatically handled by the machinery like Vector64/128. There is an issue to track Mono side of the work related to these new set of API's.

[SwapnilGaikwad]

Contribute towards #84510.

// AdvSimd// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar64x3(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3)value,byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar64x4(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3,Vector64<byte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3,Vector64<sbyte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3,Vector64<short>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3,Vector64<ushort>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3,Vector64<int>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3,Vector64<uint>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3,Vector64<float>Value4)value,byteindex);// AdvSimd.Arm64// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar128x3(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3),byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar128x4(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3,Vector128<byte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3,Vector128<sbyte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3,Vector128<short>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3,Vector128<ushort>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3,Vector128<int>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3,Vector128<uint>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3,Vector128<long>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3,Vector128<ulong>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3,Vector128<float>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3,Vector128<double>Value4),byteindex);

[ghost]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[ghost]

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

Issue Details

---

Contribute towards #84510.

// AdvSimdpublicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// AdvSimd.Arm64publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);

Author:	SwapnilGaikwad
Assignees:	-
Labels:	area-CodeGen-coreclr, new-api-needs-documentation, community-contribution
Milestone:	-

[SwapnilGaikwad]

This patch would emit the following assembly sequence

1. When the index is a constant value

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),1);

...
stp d8, d16, [fp, #0x10]	// [V18 tmp17], [V19 tmp18]
ldr q16, [fp, #0x10]	// [V04 tmp3]
ldr q17, [fp, #0x18]	// [V04 tmp3+0x08]
st2 {v16.s, v17.s}[1], [x20]
...

2. When index is not a constant

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),elemIndex);

...
mov x0, x21
movz x2, #0xC8A0 // code for System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[int],System.Runtime.Intrinsics.Vector64`1[int]],ubyte)
movk x2, #0x5DB8 LSL #16
movk x2, #0xFFFF LSL #32
ldr x2, [x2]
...
; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[short],System.Runtime.Intrinsics.Vector64`1[short]],ubyte) (FullOpts)
G_M60890_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x20]!
mov fp, sp
stp d0, d1, [fp, #0x10]	// [V01 arg1], [V01 arg1+0x08]
;; size=12 bbWeight=1 PerfScore 2.50
G_M60890_IG02: ;; offset=0x000C
ldr q16, [fp, #0x10]	// [V01 arg1]
ldr q17, [fp, #0x18]	// [V01 arg1+0x08]
uxtb w1, w1
adr x2, [G_M60890_IG03]
add x2, x2, x1, LSL #3
br x2
;; size=24 bbWeight=1 PerfScore 7.00
G_M60890_IG03: ;; offset=0x0024
st2 {v16.h, v17.h}[0], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG04: ;; offset=0x002C
st2 {v16.h, v17.h}[1], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG05: ;; offset=0x0034
st2 {v16.h, v17.h}[2], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG06: ;; offset=0x003C
st2 {v16.h, v17.h}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M60890_IG07: ;; offset=0x0040
ldp fp, lr, [sp], #0x20
ret lr
;; size=8 bbWeight=1 PerfScore 2.00

3. When the index is variable and out of bounds

The current test crashes with an error.

Illegal instruction (core dumped)

Not sure whether it's the desired behaviour. Also, I haven't considered the possibility of creating a ROP gadget with the sequence for Case 2 above. The sequence of instruction where value of an index can be used control the jump target.

[kunalspathak]

As seen in #93197, the API names should be just StoreSelectedScalar because you can deduce the shape and tuple size from the parameters.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

Alright, so if you follow #93197, the immOp is handled in it.

[kunalspathak]

3. When the index is variable and out of bounds

They should throw IndexOutOfBoundException

[kunalspathak]

RunBasicScenario_UnsafeRead segfaults with the following error

Interesting. What is the cause of it?

[SwapnilGaikwad]

Now we get out of range exception for all the invalid scenarios (with constant or variable index) with all the test cases.

System.ArgumentOutOfRangeException: Specified argument was out of the range of valid values.

I never got to ensure from the assembly that it throws InvalidIndexException. 🤔

This mystery is also resolved now. There was a check for array out of bounds missing. Now added it. Can be confirmed from the assembly for StoreSelectedScalar as well.

...
G_M56847_IG04:
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
...

Full assembly

; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; Emitting BLENDED_CODE for generic ARM64 - Unix
; MinOpts code
; debuggable code
; fp based frame
; fully interruptible
; No PGO data
; Final local variable assignments
;
; V00 arg0 [V00 ] ( 1, 1 ) long -> [fp+0x38] do-not-enreg[]
; V01 arg1 [V01 ] ( 1, 1 ) simd16 -> [fp+0x20] HFA(simd16) do-not-enreg[S] <System.Runtime.Intrinsics.Vector128`1[int]>
; V02 arg2 [V02 ] ( 1, 1 ) ubyte -> [fp+0x1C] do-not-enreg[]
;# V03 OutArgs [V03 ] ( 1, 1 ) struct ( 0) [sp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
;
; Lcl frame size = 48
G_M56847_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x40]!
mov fp, sp
str x0, [fp, #0x38]	// [V00 arg0]
str w1, [fp, #0x1C]	// [V02 arg2]
str q0, [fp, #0x20]	// [V01 arg1]
;; size=20 bbWeight=1 PerfScore 4.50
G_M56847_IG02: ;; offset=0x0014
movz x0, #0x45A8
movk x0, #0x4004 LSL #16
movk x0, #0xFFFF LSL #32
ldr w0, [x0]
cbz w0, G_M56847_IG04
;; size=20 bbWeight=1 PerfScore 5.50
G_M56847_IG03: ;; offset=0x0028
bl CORINFO_HELP_DBG_IS_JUST_MY_CODE
;; size=4 bbWeight=0.50 PerfScore 0.50
G_M56847_IG04: ;; offset=0x002C
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
;; size=28 bbWeight=1 PerfScore 9.00
G_M56847_IG05: ;; offset=0x0048
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
adr x2, [G_M56847_IG06]
add x2, x2, x1, LSL #3
br x2
;; size=20 bbWeight=1 PerfScore 5.00
G_M56847_IG06: ;; offset=0x005C
st1 {v16.s}[0], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG07: ;; offset=0x0064
st1 {v16.s}[1], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG08: ;; offset=0x006C
st1 {v16.s}[2], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG09: ;; offset=0x0074
st1 {v16.s}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M56847_IG10: ;; offset=0x0078
nop
nop
;; size=8 bbWeight=1 PerfScore 1.00
G_M56847_IG11: ;; offset=0x0080
ldp fp, lr, [sp], #0x40
ret lr
;; size=8 bbWeight=1 PerfScore 2.00
; Total bytes of code 136, prolog size 20, PerfScore 48.10, instruction count 34, allocated bytes for code 136 (MethodHash=5b1221f0) for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; ============================================================

Interestingly, for some tests I got IndexOutOfRangeException earlier because the test was executing without an error but failing in the validation stage where it was accessing the out of range index. Only the first test was failing with an illegal instruction probably because of the code layout. Invalid index tried to jump to location that did not contain a valid index. For the further tests, the jump target at invalid index contained a valid instruction so it did not throw any exception during the execution.

[kunalspathak]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

[SwapnilGaikwad]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

Had to move the range check after recomputing the simdSize 🤦 . Now the current failures seems while sending the test data to helix system.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

[kunalspathak]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

You're right, it seems to be related to the arg check. Not sure how to debug it though. I'll try to see if I can reproduce the failure. If that doesn't progress, should I put the explicit check as before?

[kunalspathak]

If that doesn't progress, should I put the explicit check as before?

I would say just inspect the code changes and see if the code path for existing SelectedScalar with single vector is unchanged. That should fix the problem.

[kunalspathak]

@SwapnilGaikwad - can you resolve the merge conflicts?

[fanyang-mono]

The change to Mono codebase looks good to me.

[kunalspathak]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

[fanyang-mono]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

Unfortunately, yes, Mono is not super friendly with handling API with 4 and more input arguments. We need to add additional code to handle those cases, rather than automatically handled by the machinery like Vector64/128. There is an issue to track Mono side of the work related to these new set of API's.

[SwapnilGaikwad]

Contribute towards #84510.

// AdvSimd// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar64x3(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3)value,byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar64x4(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3,Vector64<byte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3,Vector64<sbyte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3,Vector64<short>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3,Vector64<ushort>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3,Vector64<int>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3,Vector64<uint>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3,Vector64<float>Value4)value,byteindex);// AdvSimd.Arm64// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar128x3(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3),byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar128x4(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3,Vector128<byte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3,Vector128<sbyte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3,Vector128<short>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3,Vector128<ushort>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3,Vector128<int>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3,Vector128<uint>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3,Vector128<long>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3,Vector128<ulong>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3,Vector128<float>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3,Vector128<double>Value4),byteindex);

[ghost]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[ghost]

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

Issue Details

---

Contribute towards #84510.

// AdvSimdpublicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// AdvSimd.Arm64publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);

Author:	SwapnilGaikwad
Assignees:	-
Labels:	area-CodeGen-coreclr, new-api-needs-documentation, community-contribution
Milestone:	-

[SwapnilGaikwad]

This patch would emit the following assembly sequence

1. When the index is a constant value

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),1);

...
stp d8, d16, [fp, #0x10]	// [V18 tmp17], [V19 tmp18]
ldr q16, [fp, #0x10]	// [V04 tmp3]
ldr q17, [fp, #0x18]	// [V04 tmp3+0x08]
st2 {v16.s, v17.s}[1], [x20]
...

2. When index is not a constant

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),elemIndex);

...
mov x0, x21
movz x2, #0xC8A0 // code for System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[int],System.Runtime.Intrinsics.Vector64`1[int]],ubyte)
movk x2, #0x5DB8 LSL #16
movk x2, #0xFFFF LSL #32
ldr x2, [x2]
...
; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[short],System.Runtime.Intrinsics.Vector64`1[short]],ubyte) (FullOpts)
G_M60890_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x20]!
mov fp, sp
stp d0, d1, [fp, #0x10]	// [V01 arg1], [V01 arg1+0x08]
;; size=12 bbWeight=1 PerfScore 2.50
G_M60890_IG02: ;; offset=0x000C
ldr q16, [fp, #0x10]	// [V01 arg1]
ldr q17, [fp, #0x18]	// [V01 arg1+0x08]
uxtb w1, w1
adr x2, [G_M60890_IG03]
add x2, x2, x1, LSL #3
br x2
;; size=24 bbWeight=1 PerfScore 7.00
G_M60890_IG03: ;; offset=0x0024
st2 {v16.h, v17.h}[0], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG04: ;; offset=0x002C
st2 {v16.h, v17.h}[1], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG05: ;; offset=0x0034
st2 {v16.h, v17.h}[2], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG06: ;; offset=0x003C
st2 {v16.h, v17.h}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M60890_IG07: ;; offset=0x0040
ldp fp, lr, [sp], #0x20
ret lr
;; size=8 bbWeight=1 PerfScore 2.00

3. When the index is variable and out of bounds

The current test crashes with an error.

Illegal instruction (core dumped)

Not sure whether it's the desired behaviour. Also, I haven't considered the possibility of creating a ROP gadget with the sequence for Case 2 above. The sequence of instruction where value of an index can be used control the jump target.

[kunalspathak]

As seen in #93197, the API names should be just StoreSelectedScalar because you can deduce the shape and tuple size from the parameters.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

Alright, so if you follow #93197, the immOp is handled in it.

[kunalspathak]

3. When the index is variable and out of bounds

They should throw IndexOutOfBoundException

[kunalspathak]

RunBasicScenario_UnsafeRead segfaults with the following error

Interesting. What is the cause of it?

[SwapnilGaikwad]

Now we get out of range exception for all the invalid scenarios (with constant or variable index) with all the test cases.

System.ArgumentOutOfRangeException: Specified argument was out of the range of valid values.

I never got to ensure from the assembly that it throws InvalidIndexException. 🤔

This mystery is also resolved now. There was a check for array out of bounds missing. Now added it. Can be confirmed from the assembly for StoreSelectedScalar as well.

...
G_M56847_IG04:
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
...

Full assembly

; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; Emitting BLENDED_CODE for generic ARM64 - Unix
; MinOpts code
; debuggable code
; fp based frame
; fully interruptible
; No PGO data
; Final local variable assignments
;
; V00 arg0 [V00 ] ( 1, 1 ) long -> [fp+0x38] do-not-enreg[]
; V01 arg1 [V01 ] ( 1, 1 ) simd16 -> [fp+0x20] HFA(simd16) do-not-enreg[S] <System.Runtime.Intrinsics.Vector128`1[int]>
; V02 arg2 [V02 ] ( 1, 1 ) ubyte -> [fp+0x1C] do-not-enreg[]
;# V03 OutArgs [V03 ] ( 1, 1 ) struct ( 0) [sp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
;
; Lcl frame size = 48
G_M56847_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x40]!
mov fp, sp
str x0, [fp, #0x38]	// [V00 arg0]
str w1, [fp, #0x1C]	// [V02 arg2]
str q0, [fp, #0x20]	// [V01 arg1]
;; size=20 bbWeight=1 PerfScore 4.50
G_M56847_IG02: ;; offset=0x0014
movz x0, #0x45A8
movk x0, #0x4004 LSL #16
movk x0, #0xFFFF LSL #32
ldr w0, [x0]
cbz w0, G_M56847_IG04
;; size=20 bbWeight=1 PerfScore 5.50
G_M56847_IG03: ;; offset=0x0028
bl CORINFO_HELP_DBG_IS_JUST_MY_CODE
;; size=4 bbWeight=0.50 PerfScore 0.50
G_M56847_IG04: ;; offset=0x002C
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
;; size=28 bbWeight=1 PerfScore 9.00
G_M56847_IG05: ;; offset=0x0048
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
adr x2, [G_M56847_IG06]
add x2, x2, x1, LSL #3
br x2
;; size=20 bbWeight=1 PerfScore 5.00
G_M56847_IG06: ;; offset=0x005C
st1 {v16.s}[0], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG07: ;; offset=0x0064
st1 {v16.s}[1], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG08: ;; offset=0x006C
st1 {v16.s}[2], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG09: ;; offset=0x0074
st1 {v16.s}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M56847_IG10: ;; offset=0x0078
nop
nop
;; size=8 bbWeight=1 PerfScore 1.00
G_M56847_IG11: ;; offset=0x0080
ldp fp, lr, [sp], #0x40
ret lr
;; size=8 bbWeight=1 PerfScore 2.00
; Total bytes of code 136, prolog size 20, PerfScore 48.10, instruction count 34, allocated bytes for code 136 (MethodHash=5b1221f0) for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; ============================================================

Interestingly, for some tests I got IndexOutOfRangeException earlier because the test was executing without an error but failing in the validation stage where it was accessing the out of range index. Only the first test was failing with an illegal instruction probably because of the code layout. Invalid index tried to jump to location that did not contain a valid index. For the further tests, the jump target at invalid index contained a valid instruction so it did not throw any exception during the execution.

[kunalspathak]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

[SwapnilGaikwad]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

Had to move the range check after recomputing the simdSize 🤦 . Now the current failures seems while sending the test data to helix system.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

[kunalspathak]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

You're right, it seems to be related to the arg check. Not sure how to debug it though. I'll try to see if I can reproduce the failure. If that doesn't progress, should I put the explicit check as before?

[kunalspathak]

If that doesn't progress, should I put the explicit check as before?

I would say just inspect the code changes and see if the code path for existing SelectedScalar with single vector is unchanged. That should fix the problem.

[kunalspathak]

@SwapnilGaikwad - can you resolve the merge conflicts?

[fanyang-mono]

The change to Mono codebase looks good to me.

[kunalspathak]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

[fanyang-mono]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

Unfortunately, yes, Mono is not super friendly with handling API with 4 and more input arguments. We need to add additional code to handle those cases, rather than automatically handled by the machinery like Vector64/128. There is an issue to track Mono side of the work related to these new set of API's.

[SwapnilGaikwad]

Contribute towards #84510.

// AdvSimd// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar64x3(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3)value,byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar64x4(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3,Vector64<byte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3,Vector64<sbyte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3,Vector64<short>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3,Vector64<ushort>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3,Vector64<int>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3,Vector64<uint>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3,Vector64<float>Value4)value,byteindex);// AdvSimd.Arm64// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar128x3(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3),byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar128x4(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3,Vector128<byte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3,Vector128<sbyte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3,Vector128<short>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3,Vector128<ushort>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3,Vector128<int>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3,Vector128<uint>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3,Vector128<long>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3,Vector128<ulong>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3,Vector128<float>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3,Vector128<double>Value4),byteindex);

[ghost]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[ghost]

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

Issue Details

---

Contribute towards #84510.

// AdvSimdpublicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// AdvSimd.Arm64publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);

Author:	SwapnilGaikwad
Assignees:	-
Labels:	area-CodeGen-coreclr, new-api-needs-documentation, community-contribution
Milestone:	-

[SwapnilGaikwad]

This patch would emit the following assembly sequence

1. When the index is a constant value

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),1);

...
stp d8, d16, [fp, #0x10]	// [V18 tmp17], [V19 tmp18]
ldr q16, [fp, #0x10]	// [V04 tmp3]
ldr q17, [fp, #0x18]	// [V04 tmp3+0x08]
st2 {v16.s, v17.s}[1], [x20]
...

2. When index is not a constant

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),elemIndex);

...
mov x0, x21
movz x2, #0xC8A0 // code for System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[int],System.Runtime.Intrinsics.Vector64`1[int]],ubyte)
movk x2, #0x5DB8 LSL #16
movk x2, #0xFFFF LSL #32
ldr x2, [x2]
...
; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[short],System.Runtime.Intrinsics.Vector64`1[short]],ubyte) (FullOpts)
G_M60890_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x20]!
mov fp, sp
stp d0, d1, [fp, #0x10]	// [V01 arg1], [V01 arg1+0x08]
;; size=12 bbWeight=1 PerfScore 2.50
G_M60890_IG02: ;; offset=0x000C
ldr q16, [fp, #0x10]	// [V01 arg1]
ldr q17, [fp, #0x18]	// [V01 arg1+0x08]
uxtb w1, w1
adr x2, [G_M60890_IG03]
add x2, x2, x1, LSL #3
br x2
;; size=24 bbWeight=1 PerfScore 7.00
G_M60890_IG03: ;; offset=0x0024
st2 {v16.h, v17.h}[0], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG04: ;; offset=0x002C
st2 {v16.h, v17.h}[1], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG05: ;; offset=0x0034
st2 {v16.h, v17.h}[2], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG06: ;; offset=0x003C
st2 {v16.h, v17.h}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M60890_IG07: ;; offset=0x0040
ldp fp, lr, [sp], #0x20
ret lr
;; size=8 bbWeight=1 PerfScore 2.00

3. When the index is variable and out of bounds

The current test crashes with an error.

Illegal instruction (core dumped)

Not sure whether it's the desired behaviour. Also, I haven't considered the possibility of creating a ROP gadget with the sequence for Case 2 above. The sequence of instruction where value of an index can be used control the jump target.

[kunalspathak]

As seen in #93197, the API names should be just StoreSelectedScalar because you can deduce the shape and tuple size from the parameters.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

Alright, so if you follow #93197, the immOp is handled in it.

[kunalspathak]

3. When the index is variable and out of bounds

They should throw IndexOutOfBoundException

[kunalspathak]

RunBasicScenario_UnsafeRead segfaults with the following error

Interesting. What is the cause of it?

[SwapnilGaikwad]

Now we get out of range exception for all the invalid scenarios (with constant or variable index) with all the test cases.

System.ArgumentOutOfRangeException: Specified argument was out of the range of valid values.

I never got to ensure from the assembly that it throws InvalidIndexException. 🤔

This mystery is also resolved now. There was a check for array out of bounds missing. Now added it. Can be confirmed from the assembly for StoreSelectedScalar as well.

...
G_M56847_IG04:
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
...

Full assembly

; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; Emitting BLENDED_CODE for generic ARM64 - Unix
; MinOpts code
; debuggable code
; fp based frame
; fully interruptible
; No PGO data
; Final local variable assignments
;
; V00 arg0 [V00 ] ( 1, 1 ) long -> [fp+0x38] do-not-enreg[]
; V01 arg1 [V01 ] ( 1, 1 ) simd16 -> [fp+0x20] HFA(simd16) do-not-enreg[S] <System.Runtime.Intrinsics.Vector128`1[int]>
; V02 arg2 [V02 ] ( 1, 1 ) ubyte -> [fp+0x1C] do-not-enreg[]
;# V03 OutArgs [V03 ] ( 1, 1 ) struct ( 0) [sp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
;
; Lcl frame size = 48
G_M56847_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x40]!
mov fp, sp
str x0, [fp, #0x38]	// [V00 arg0]
str w1, [fp, #0x1C]	// [V02 arg2]
str q0, [fp, #0x20]	// [V01 arg1]
;; size=20 bbWeight=1 PerfScore 4.50
G_M56847_IG02: ;; offset=0x0014
movz x0, #0x45A8
movk x0, #0x4004 LSL #16
movk x0, #0xFFFF LSL #32
ldr w0, [x0]
cbz w0, G_M56847_IG04
;; size=20 bbWeight=1 PerfScore 5.50
G_M56847_IG03: ;; offset=0x0028
bl CORINFO_HELP_DBG_IS_JUST_MY_CODE
;; size=4 bbWeight=0.50 PerfScore 0.50
G_M56847_IG04: ;; offset=0x002C
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
;; size=28 bbWeight=1 PerfScore 9.00
G_M56847_IG05: ;; offset=0x0048
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
adr x2, [G_M56847_IG06]
add x2, x2, x1, LSL #3
br x2
;; size=20 bbWeight=1 PerfScore 5.00
G_M56847_IG06: ;; offset=0x005C
st1 {v16.s}[0], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG07: ;; offset=0x0064
st1 {v16.s}[1], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG08: ;; offset=0x006C
st1 {v16.s}[2], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG09: ;; offset=0x0074
st1 {v16.s}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M56847_IG10: ;; offset=0x0078
nop
nop
;; size=8 bbWeight=1 PerfScore 1.00
G_M56847_IG11: ;; offset=0x0080
ldp fp, lr, [sp], #0x40
ret lr
;; size=8 bbWeight=1 PerfScore 2.00
; Total bytes of code 136, prolog size 20, PerfScore 48.10, instruction count 34, allocated bytes for code 136 (MethodHash=5b1221f0) for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; ============================================================

Interestingly, for some tests I got IndexOutOfRangeException earlier because the test was executing without an error but failing in the validation stage where it was accessing the out of range index. Only the first test was failing with an illegal instruction probably because of the code layout. Invalid index tried to jump to location that did not contain a valid index. For the further tests, the jump target at invalid index contained a valid instruction so it did not throw any exception during the execution.

[kunalspathak]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

[SwapnilGaikwad]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

Had to move the range check after recomputing the simdSize 🤦 . Now the current failures seems while sending the test data to helix system.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

[kunalspathak]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

You're right, it seems to be related to the arg check. Not sure how to debug it though. I'll try to see if I can reproduce the failure. If that doesn't progress, should I put the explicit check as before?

[kunalspathak]

If that doesn't progress, should I put the explicit check as before?

I would say just inspect the code changes and see if the code path for existing SelectedScalar with single vector is unchanged. That should fix the problem.

[kunalspathak]

@SwapnilGaikwad - can you resolve the merge conflicts?

[fanyang-mono]

The change to Mono codebase looks good to me.

[kunalspathak]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

[fanyang-mono]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

Unfortunately, yes, Mono is not super friendly with handling API with 4 and more input arguments. We need to add additional code to handle those cases, rather than automatically handled by the machinery like Vector64/128. There is an issue to track Mono side of the work related to these new set of API's.

[SwapnilGaikwad]

Contribute towards #84510.

// AdvSimd// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar64x3(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3)value,byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar64x4(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3,Vector64<byte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3,Vector64<sbyte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3,Vector64<short>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3,Vector64<ushort>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3,Vector64<int>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3,Vector64<uint>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3,Vector64<float>Value4)value,byteindex);// AdvSimd.Arm64// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar128x3(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3),byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar128x4(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3,Vector128<byte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3,Vector128<sbyte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3,Vector128<short>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3,Vector128<ushort>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3,Vector128<int>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3,Vector128<uint>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3,Vector128<long>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3,Vector128<ulong>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3,Vector128<float>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3,Vector128<double>Value4),byteindex);

[ghost]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[ghost]

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

Issue Details

---

Contribute towards #84510.

// AdvSimdpublicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// AdvSimd.Arm64publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);

Author:	SwapnilGaikwad
Assignees:	-
Labels:	area-CodeGen-coreclr, new-api-needs-documentation, community-contribution
Milestone:	-

[SwapnilGaikwad]

This patch would emit the following assembly sequence

1. When the index is a constant value

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),1);

...
stp d8, d16, [fp, #0x10]	// [V18 tmp17], [V19 tmp18]
ldr q16, [fp, #0x10]	// [V04 tmp3]
ldr q17, [fp, #0x18]	// [V04 tmp3+0x08]
st2 {v16.s, v17.s}[1], [x20]
...

2. When index is not a constant

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),elemIndex);

...
mov x0, x21
movz x2, #0xC8A0 // code for System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[int],System.Runtime.Intrinsics.Vector64`1[int]],ubyte)
movk x2, #0x5DB8 LSL #16
movk x2, #0xFFFF LSL #32
ldr x2, [x2]
...
; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[short],System.Runtime.Intrinsics.Vector64`1[short]],ubyte) (FullOpts)
G_M60890_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x20]!
mov fp, sp
stp d0, d1, [fp, #0x10]	// [V01 arg1], [V01 arg1+0x08]
;; size=12 bbWeight=1 PerfScore 2.50
G_M60890_IG02: ;; offset=0x000C
ldr q16, [fp, #0x10]	// [V01 arg1]
ldr q17, [fp, #0x18]	// [V01 arg1+0x08]
uxtb w1, w1
adr x2, [G_M60890_IG03]
add x2, x2, x1, LSL #3
br x2
;; size=24 bbWeight=1 PerfScore 7.00
G_M60890_IG03: ;; offset=0x0024
st2 {v16.h, v17.h}[0], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG04: ;; offset=0x002C
st2 {v16.h, v17.h}[1], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG05: ;; offset=0x0034
st2 {v16.h, v17.h}[2], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG06: ;; offset=0x003C
st2 {v16.h, v17.h}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M60890_IG07: ;; offset=0x0040
ldp fp, lr, [sp], #0x20
ret lr
;; size=8 bbWeight=1 PerfScore 2.00

3. When the index is variable and out of bounds

The current test crashes with an error.

Illegal instruction (core dumped)

Not sure whether it's the desired behaviour. Also, I haven't considered the possibility of creating a ROP gadget with the sequence for Case 2 above. The sequence of instruction where value of an index can be used control the jump target.

[kunalspathak]

As seen in #93197, the API names should be just StoreSelectedScalar because you can deduce the shape and tuple size from the parameters.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

Alright, so if you follow #93197, the immOp is handled in it.

[kunalspathak]

3. When the index is variable and out of bounds

They should throw IndexOutOfBoundException

[kunalspathak]

RunBasicScenario_UnsafeRead segfaults with the following error

Interesting. What is the cause of it?

[SwapnilGaikwad]

Now we get out of range exception for all the invalid scenarios (with constant or variable index) with all the test cases.

System.ArgumentOutOfRangeException: Specified argument was out of the range of valid values.

I never got to ensure from the assembly that it throws InvalidIndexException. 🤔

This mystery is also resolved now. There was a check for array out of bounds missing. Now added it. Can be confirmed from the assembly for StoreSelectedScalar as well.

...
G_M56847_IG04:
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
...

Full assembly

; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; Emitting BLENDED_CODE for generic ARM64 - Unix
; MinOpts code
; debuggable code
; fp based frame
; fully interruptible
; No PGO data
; Final local variable assignments
;
; V00 arg0 [V00 ] ( 1, 1 ) long -> [fp+0x38] do-not-enreg[]
; V01 arg1 [V01 ] ( 1, 1 ) simd16 -> [fp+0x20] HFA(simd16) do-not-enreg[S] <System.Runtime.Intrinsics.Vector128`1[int]>
; V02 arg2 [V02 ] ( 1, 1 ) ubyte -> [fp+0x1C] do-not-enreg[]
;# V03 OutArgs [V03 ] ( 1, 1 ) struct ( 0) [sp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
;
; Lcl frame size = 48
G_M56847_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x40]!
mov fp, sp
str x0, [fp, #0x38]	// [V00 arg0]
str w1, [fp, #0x1C]	// [V02 arg2]
str q0, [fp, #0x20]	// [V01 arg1]
;; size=20 bbWeight=1 PerfScore 4.50
G_M56847_IG02: ;; offset=0x0014
movz x0, #0x45A8
movk x0, #0x4004 LSL #16
movk x0, #0xFFFF LSL #32
ldr w0, [x0]
cbz w0, G_M56847_IG04
;; size=20 bbWeight=1 PerfScore 5.50
G_M56847_IG03: ;; offset=0x0028
bl CORINFO_HELP_DBG_IS_JUST_MY_CODE
;; size=4 bbWeight=0.50 PerfScore 0.50
G_M56847_IG04: ;; offset=0x002C
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
;; size=28 bbWeight=1 PerfScore 9.00
G_M56847_IG05: ;; offset=0x0048
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
adr x2, [G_M56847_IG06]
add x2, x2, x1, LSL #3
br x2
;; size=20 bbWeight=1 PerfScore 5.00
G_M56847_IG06: ;; offset=0x005C
st1 {v16.s}[0], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG07: ;; offset=0x0064
st1 {v16.s}[1], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG08: ;; offset=0x006C
st1 {v16.s}[2], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG09: ;; offset=0x0074
st1 {v16.s}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M56847_IG10: ;; offset=0x0078
nop
nop
;; size=8 bbWeight=1 PerfScore 1.00
G_M56847_IG11: ;; offset=0x0080
ldp fp, lr, [sp], #0x40
ret lr
;; size=8 bbWeight=1 PerfScore 2.00
; Total bytes of code 136, prolog size 20, PerfScore 48.10, instruction count 34, allocated bytes for code 136 (MethodHash=5b1221f0) for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; ============================================================

Interestingly, for some tests I got IndexOutOfRangeException earlier because the test was executing without an error but failing in the validation stage where it was accessing the out of range index. Only the first test was failing with an illegal instruction probably because of the code layout. Invalid index tried to jump to location that did not contain a valid index. For the further tests, the jump target at invalid index contained a valid instruction so it did not throw any exception during the execution.

[kunalspathak]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

[SwapnilGaikwad]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

Had to move the range check after recomputing the simdSize 🤦 . Now the current failures seems while sending the test data to helix system.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

[kunalspathak]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

You're right, it seems to be related to the arg check. Not sure how to debug it though. I'll try to see if I can reproduce the failure. If that doesn't progress, should I put the explicit check as before?

[kunalspathak]

If that doesn't progress, should I put the explicit check as before?

I would say just inspect the code changes and see if the code path for existing SelectedScalar with single vector is unchanged. That should fix the problem.

[kunalspathak]

@SwapnilGaikwad - can you resolve the merge conflicts?

[fanyang-mono]

The change to Mono codebase looks good to me.

[kunalspathak]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

[fanyang-mono]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

Unfortunately, yes, Mono is not super friendly with handling API with 4 and more input arguments. We need to add additional code to handle those cases, rather than automatically handled by the machinery like Vector64/128. There is an issue to track Mono side of the work related to these new set of API's.

[SwapnilGaikwad]

Contribute towards #84510.

// AdvSimd// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar64x3(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3)value,byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar64x4(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3,Vector64<byte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3,Vector64<sbyte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3,Vector64<short>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3,Vector64<ushort>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3,Vector64<int>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3,Vector64<uint>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3,Vector64<float>Value4)value,byteindex);// AdvSimd.Arm64// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar128x3(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3),byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar128x4(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3,Vector128<byte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3,Vector128<sbyte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3,Vector128<short>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3,Vector128<ushort>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3,Vector128<int>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3,Vector128<uint>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3,Vector128<long>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3,Vector128<ulong>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3,Vector128<float>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3,Vector128<double>Value4),byteindex);

[ghost]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[ghost]

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

Issue Details

---

Contribute towards #84510.

// AdvSimdpublicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// AdvSimd.Arm64publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);

Author:	SwapnilGaikwad
Assignees:	-
Labels:	area-CodeGen-coreclr, new-api-needs-documentation, community-contribution
Milestone:	-

[SwapnilGaikwad]

This patch would emit the following assembly sequence

1. When the index is a constant value

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),1);

...
stp d8, d16, [fp, #0x10]	// [V18 tmp17], [V19 tmp18]
ldr q16, [fp, #0x10]	// [V04 tmp3]
ldr q17, [fp, #0x18]	// [V04 tmp3+0x08]
st2 {v16.s, v17.s}[1], [x20]
...

2. When index is not a constant

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),elemIndex);

...
mov x0, x21
movz x2, #0xC8A0 // code for System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[int],System.Runtime.Intrinsics.Vector64`1[int]],ubyte)
movk x2, #0x5DB8 LSL #16
movk x2, #0xFFFF LSL #32
ldr x2, [x2]
...
; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[short],System.Runtime.Intrinsics.Vector64`1[short]],ubyte) (FullOpts)
G_M60890_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x20]!
mov fp, sp
stp d0, d1, [fp, #0x10]	// [V01 arg1], [V01 arg1+0x08]
;; size=12 bbWeight=1 PerfScore 2.50
G_M60890_IG02: ;; offset=0x000C
ldr q16, [fp, #0x10]	// [V01 arg1]
ldr q17, [fp, #0x18]	// [V01 arg1+0x08]
uxtb w1, w1
adr x2, [G_M60890_IG03]
add x2, x2, x1, LSL #3
br x2
;; size=24 bbWeight=1 PerfScore 7.00
G_M60890_IG03: ;; offset=0x0024
st2 {v16.h, v17.h}[0], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG04: ;; offset=0x002C
st2 {v16.h, v17.h}[1], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG05: ;; offset=0x0034
st2 {v16.h, v17.h}[2], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG06: ;; offset=0x003C
st2 {v16.h, v17.h}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M60890_IG07: ;; offset=0x0040
ldp fp, lr, [sp], #0x20
ret lr
;; size=8 bbWeight=1 PerfScore 2.00

3. When the index is variable and out of bounds

The current test crashes with an error.

Illegal instruction (core dumped)

Not sure whether it's the desired behaviour. Also, I haven't considered the possibility of creating a ROP gadget with the sequence for Case 2 above. The sequence of instruction where value of an index can be used control the jump target.

[kunalspathak]

As seen in #93197, the API names should be just StoreSelectedScalar because you can deduce the shape and tuple size from the parameters.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

Alright, so if you follow #93197, the immOp is handled in it.

[kunalspathak]

3. When the index is variable and out of bounds

They should throw IndexOutOfBoundException

[kunalspathak]

RunBasicScenario_UnsafeRead segfaults with the following error

Interesting. What is the cause of it?

[SwapnilGaikwad]

Now we get out of range exception for all the invalid scenarios (with constant or variable index) with all the test cases.

System.ArgumentOutOfRangeException: Specified argument was out of the range of valid values.

I never got to ensure from the assembly that it throws InvalidIndexException. 🤔

This mystery is also resolved now. There was a check for array out of bounds missing. Now added it. Can be confirmed from the assembly for StoreSelectedScalar as well.

...
G_M56847_IG04:
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
...

Full assembly

; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; Emitting BLENDED_CODE for generic ARM64 - Unix
; MinOpts code
; debuggable code
; fp based frame
; fully interruptible
; No PGO data
; Final local variable assignments
;
; V00 arg0 [V00 ] ( 1, 1 ) long -> [fp+0x38] do-not-enreg[]
; V01 arg1 [V01 ] ( 1, 1 ) simd16 -> [fp+0x20] HFA(simd16) do-not-enreg[S] <System.Runtime.Intrinsics.Vector128`1[int]>
; V02 arg2 [V02 ] ( 1, 1 ) ubyte -> [fp+0x1C] do-not-enreg[]
;# V03 OutArgs [V03 ] ( 1, 1 ) struct ( 0) [sp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
;
; Lcl frame size = 48
G_M56847_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x40]!
mov fp, sp
str x0, [fp, #0x38]	// [V00 arg0]
str w1, [fp, #0x1C]	// [V02 arg2]
str q0, [fp, #0x20]	// [V01 arg1]
;; size=20 bbWeight=1 PerfScore 4.50
G_M56847_IG02: ;; offset=0x0014
movz x0, #0x45A8
movk x0, #0x4004 LSL #16
movk x0, #0xFFFF LSL #32
ldr w0, [x0]
cbz w0, G_M56847_IG04
;; size=20 bbWeight=1 PerfScore 5.50
G_M56847_IG03: ;; offset=0x0028
bl CORINFO_HELP_DBG_IS_JUST_MY_CODE
;; size=4 bbWeight=0.50 PerfScore 0.50
G_M56847_IG04: ;; offset=0x002C
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
;; size=28 bbWeight=1 PerfScore 9.00
G_M56847_IG05: ;; offset=0x0048
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
adr x2, [G_M56847_IG06]
add x2, x2, x1, LSL #3
br x2
;; size=20 bbWeight=1 PerfScore 5.00
G_M56847_IG06: ;; offset=0x005C
st1 {v16.s}[0], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG07: ;; offset=0x0064
st1 {v16.s}[1], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG08: ;; offset=0x006C
st1 {v16.s}[2], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG09: ;; offset=0x0074
st1 {v16.s}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M56847_IG10: ;; offset=0x0078
nop
nop
;; size=8 bbWeight=1 PerfScore 1.00
G_M56847_IG11: ;; offset=0x0080
ldp fp, lr, [sp], #0x40
ret lr
;; size=8 bbWeight=1 PerfScore 2.00
; Total bytes of code 136, prolog size 20, PerfScore 48.10, instruction count 34, allocated bytes for code 136 (MethodHash=5b1221f0) for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; ============================================================

Interestingly, for some tests I got IndexOutOfRangeException earlier because the test was executing without an error but failing in the validation stage where it was accessing the out of range index. Only the first test was failing with an illegal instruction probably because of the code layout. Invalid index tried to jump to location that did not contain a valid index. For the further tests, the jump target at invalid index contained a valid instruction so it did not throw any exception during the execution.

[kunalspathak]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

[SwapnilGaikwad]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

Had to move the range check after recomputing the simdSize 🤦 . Now the current failures seems while sending the test data to helix system.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

[kunalspathak]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

You're right, it seems to be related to the arg check. Not sure how to debug it though. I'll try to see if I can reproduce the failure. If that doesn't progress, should I put the explicit check as before?

[kunalspathak]

If that doesn't progress, should I put the explicit check as before?

I would say just inspect the code changes and see if the code path for existing SelectedScalar with single vector is unchanged. That should fix the problem.

[kunalspathak]

@SwapnilGaikwad - can you resolve the merge conflicts?

[fanyang-mono]

The change to Mono codebase looks good to me.

[kunalspathak]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

[fanyang-mono]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

Unfortunately, yes, Mono is not super friendly with handling API with 4 and more input arguments. We need to add additional code to handle those cases, rather than automatically handled by the machinery like Vector64/128. There is an issue to track Mono side of the work related to these new set of API's.

[SwapnilGaikwad]

Contribute towards #84510.

// AdvSimd// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar64x3(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3)value,byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar64x4(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3,Vector64<byte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3,Vector64<sbyte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3,Vector64<short>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3,Vector64<ushort>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3,Vector64<int>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3,Vector64<uint>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3,Vector64<float>Value4)value,byteindex);// AdvSimd.Arm64// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar128x3(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3),byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar128x4(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3,Vector128<byte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3,Vector128<sbyte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3,Vector128<short>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3,Vector128<ushort>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3,Vector128<int>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3,Vector128<uint>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3,Vector128<long>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3,Vector128<ulong>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3,Vector128<float>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3,Vector128<double>Value4),byteindex);

[ghost]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[ghost]

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

Issue Details

---

Contribute towards #84510.

// AdvSimdpublicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// AdvSimd.Arm64publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);

Author:	SwapnilGaikwad
Assignees:	-
Labels:	area-CodeGen-coreclr, new-api-needs-documentation, community-contribution
Milestone:	-

[SwapnilGaikwad]

This patch would emit the following assembly sequence

1. When the index is a constant value

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),1);

...
stp d8, d16, [fp, #0x10]	// [V18 tmp17], [V19 tmp18]
ldr q16, [fp, #0x10]	// [V04 tmp3]
ldr q17, [fp, #0x18]	// [V04 tmp3+0x08]
st2 {v16.s, v17.s}[1], [x20]
...

2. When index is not a constant

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),elemIndex);

...
mov x0, x21
movz x2, #0xC8A0 // code for System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[int],System.Runtime.Intrinsics.Vector64`1[int]],ubyte)
movk x2, #0x5DB8 LSL #16
movk x2, #0xFFFF LSL #32
ldr x2, [x2]
...
; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[short],System.Runtime.Intrinsics.Vector64`1[short]],ubyte) (FullOpts)
G_M60890_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x20]!
mov fp, sp
stp d0, d1, [fp, #0x10]	// [V01 arg1], [V01 arg1+0x08]
;; size=12 bbWeight=1 PerfScore 2.50
G_M60890_IG02: ;; offset=0x000C
ldr q16, [fp, #0x10]	// [V01 arg1]
ldr q17, [fp, #0x18]	// [V01 arg1+0x08]
uxtb w1, w1
adr x2, [G_M60890_IG03]
add x2, x2, x1, LSL #3
br x2
;; size=24 bbWeight=1 PerfScore 7.00
G_M60890_IG03: ;; offset=0x0024
st2 {v16.h, v17.h}[0], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG04: ;; offset=0x002C
st2 {v16.h, v17.h}[1], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG05: ;; offset=0x0034
st2 {v16.h, v17.h}[2], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG06: ;; offset=0x003C
st2 {v16.h, v17.h}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M60890_IG07: ;; offset=0x0040
ldp fp, lr, [sp], #0x20
ret lr
;; size=8 bbWeight=1 PerfScore 2.00

3. When the index is variable and out of bounds

The current test crashes with an error.

Illegal instruction (core dumped)

Not sure whether it's the desired behaviour. Also, I haven't considered the possibility of creating a ROP gadget with the sequence for Case 2 above. The sequence of instruction where value of an index can be used control the jump target.

[kunalspathak]

As seen in #93197, the API names should be just StoreSelectedScalar because you can deduce the shape and tuple size from the parameters.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

Alright, so if you follow #93197, the immOp is handled in it.

[kunalspathak]

3. When the index is variable and out of bounds

They should throw IndexOutOfBoundException

[kunalspathak]

RunBasicScenario_UnsafeRead segfaults with the following error

Interesting. What is the cause of it?

[SwapnilGaikwad]

Now we get out of range exception for all the invalid scenarios (with constant or variable index) with all the test cases.

System.ArgumentOutOfRangeException: Specified argument was out of the range of valid values.

I never got to ensure from the assembly that it throws InvalidIndexException. 🤔

This mystery is also resolved now. There was a check for array out of bounds missing. Now added it. Can be confirmed from the assembly for StoreSelectedScalar as well.

...
G_M56847_IG04:
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
...

Full assembly

; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; Emitting BLENDED_CODE for generic ARM64 - Unix
; MinOpts code
; debuggable code
; fp based frame
; fully interruptible
; No PGO data
; Final local variable assignments
;
; V00 arg0 [V00 ] ( 1, 1 ) long -> [fp+0x38] do-not-enreg[]
; V01 arg1 [V01 ] ( 1, 1 ) simd16 -> [fp+0x20] HFA(simd16) do-not-enreg[S] <System.Runtime.Intrinsics.Vector128`1[int]>
; V02 arg2 [V02 ] ( 1, 1 ) ubyte -> [fp+0x1C] do-not-enreg[]
;# V03 OutArgs [V03 ] ( 1, 1 ) struct ( 0) [sp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
;
; Lcl frame size = 48
G_M56847_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x40]!
mov fp, sp
str x0, [fp, #0x38]	// [V00 arg0]
str w1, [fp, #0x1C]	// [V02 arg2]
str q0, [fp, #0x20]	// [V01 arg1]
;; size=20 bbWeight=1 PerfScore 4.50
G_M56847_IG02: ;; offset=0x0014
movz x0, #0x45A8
movk x0, #0x4004 LSL #16
movk x0, #0xFFFF LSL #32
ldr w0, [x0]
cbz w0, G_M56847_IG04
;; size=20 bbWeight=1 PerfScore 5.50
G_M56847_IG03: ;; offset=0x0028
bl CORINFO_HELP_DBG_IS_JUST_MY_CODE
;; size=4 bbWeight=0.50 PerfScore 0.50
G_M56847_IG04: ;; offset=0x002C
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
;; size=28 bbWeight=1 PerfScore 9.00
G_M56847_IG05: ;; offset=0x0048
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
adr x2, [G_M56847_IG06]
add x2, x2, x1, LSL #3
br x2
;; size=20 bbWeight=1 PerfScore 5.00
G_M56847_IG06: ;; offset=0x005C
st1 {v16.s}[0], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG07: ;; offset=0x0064
st1 {v16.s}[1], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG08: ;; offset=0x006C
st1 {v16.s}[2], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG09: ;; offset=0x0074
st1 {v16.s}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M56847_IG10: ;; offset=0x0078
nop
nop
;; size=8 bbWeight=1 PerfScore 1.00
G_M56847_IG11: ;; offset=0x0080
ldp fp, lr, [sp], #0x40
ret lr
;; size=8 bbWeight=1 PerfScore 2.00
; Total bytes of code 136, prolog size 20, PerfScore 48.10, instruction count 34, allocated bytes for code 136 (MethodHash=5b1221f0) for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; ============================================================

Interestingly, for some tests I got IndexOutOfRangeException earlier because the test was executing without an error but failing in the validation stage where it was accessing the out of range index. Only the first test was failing with an illegal instruction probably because of the code layout. Invalid index tried to jump to location that did not contain a valid index. For the further tests, the jump target at invalid index contained a valid instruction so it did not throw any exception during the execution.

[kunalspathak]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

[SwapnilGaikwad]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

Had to move the range check after recomputing the simdSize 🤦 . Now the current failures seems while sending the test data to helix system.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

[kunalspathak]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

You're right, it seems to be related to the arg check. Not sure how to debug it though. I'll try to see if I can reproduce the failure. If that doesn't progress, should I put the explicit check as before?

[kunalspathak]

If that doesn't progress, should I put the explicit check as before?

I would say just inspect the code changes and see if the code path for existing SelectedScalar with single vector is unchanged. That should fix the problem.

[kunalspathak]

@SwapnilGaikwad - can you resolve the merge conflicts?

[fanyang-mono]

The change to Mono codebase looks good to me.

[kunalspathak]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

[fanyang-mono]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

Unfortunately, yes, Mono is not super friendly with handling API with 4 and more input arguments. We need to add additional code to handle those cases, rather than automatically handled by the machinery like Vector64/128. There is an issue to track Mono side of the work related to these new set of API's.

[SwapnilGaikwad]

Contribute towards #84510.

// AdvSimd// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar64x3(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x3(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3)value,byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar64x4(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2,Vector64<byte>Value3,Vector64<byte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2,Vector64<sbyte>Value3,Vector64<sbyte>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(short*address,(Vector64<short>Value1,Vector64<short>Value2,Vector64<short>Value3,Vector64<short>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2,Vector64<ushort>Value3,Vector64<ushort>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(int*address,(Vector64<int>Value1,Vector64<int>Value2,Vector64<int>Value3,Vector64<int>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2,Vector64<uint>Value3,Vector64<uint>Value4)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x4(float*address,(Vector64<float>Value1,Vector64<float>Value2,Vector64<float>Value3,Vector64<float>Value4)value,byteindex);// AdvSimd.Arm64// ST2 (single structure)publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);// ST3 (single structure)publicstaticunsafevoidStoreSelectedScalar128x3(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3),byteindex);publicstaticunsafevoidStoreSelectedScalar128x3(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3),byteindex);// ST4 (single structure)publicstaticunsafevoidStoreSelectedScalar128x4(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2,Vector128<byte>Value3,Vector128<byte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2,Vector128<sbyte>Value3,Vector128<sbyte>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(short*address,(Vector128<short>Value1,Vector128<short>Value2,Vector128<short>Value3,Vector128<short>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2,Vector128<ushort>Value3,Vector128<ushort>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(int*address,(Vector128<int>Value1,Vector128<int>Value2,Vector128<int>Value3,Vector128<int>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2,Vector128<uint>Value3,Vector128<uint>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(long*address,(Vector128<long>Value1,Vector128<long>Value2,Vector128<long>Value3,Vector128<long>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2,Vector128<ulong>Value3,Vector128<ulong>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(float*address,(Vector128<float>Value1,Vector128<float>Value2,Vector128<float>Value3,Vector128<float>Value4),byteindex);publicstaticunsafevoidStoreSelectedScalar128x4(double*address,(Vector128<double>Value1,Vector128<double>Value2,Vector128<double>Value3,Vector128<double>Value4),byteindex);

[ghost]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[ghost]

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

Issue Details

---

Contribute towards #84510.

// AdvSimdpublicstaticunsafevoidStoreSelectedScalar64x2(byte*address,(Vector64<byte>Value1,Vector64<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(sbyte*address,(Vector64<sbyte>Value1,Vector64<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(short*address,(Vector64<short>Value1,Vector64<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(ushort*address,(Vector64<ushort>Value1,Vector64<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(int*address,(Vector64<int>Value1,Vector64<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(uint*address,(Vector64<uint>Value1,Vector64<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar64x2(float*address,(Vector64<float>Value1,Vector64<float>Value2)value,byteindex);// AdvSimd.Arm64publicstaticunsafevoidStoreSelectedScalar128x2(byte*address,(Vector128<byte>Value1,Vector128<byte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(sbyte*address,(Vector128<sbyte>Value1,Vector128<sbyte>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(short*address,(Vector128<short>Value1,Vector128<short>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ushort*address,(Vector128<ushort>Value1,Vector128<ushort>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(int*address,(Vector128<int>Value1,Vector128<int>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(uint*address,(Vector128<uint>Value1,Vector128<uint>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(long*address,(Vector128<long>Value1,Vector128<long>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(ulong*address,(Vector128<ulong>Value1,Vector128<ulong>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(float*address,(Vector128<float>Value1,Vector128<float>Value2)value,byteindex);publicstaticunsafevoidStoreSelectedScalar128x2(double*address,(Vector128<double>Value1,Vector128<double>Value2)value,byteindex);

Author:	SwapnilGaikwad
Assignees:	-
Labels:	area-CodeGen-coreclr, new-api-needs-documentation, community-contribution
Milestone:	-

[SwapnilGaikwad]

This patch would emit the following assembly sequence

1. When the index is a constant value

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),1);

...
stp d8, d16, [fp, #0x10]	// [V18 tmp17], [V19 tmp18]
ldr q16, [fp, #0x10]	// [V04 tmp3]
ldr q17, [fp, #0x18]	// [V04 tmp3+0x08]
st2 {v16.s, v17.s}[1], [x20]
...

2. When index is not a constant

AdvSimd.StoreSelectedScalar64x2((Int32*)_dataTable.outArrayPtr,(AdvSimd.LoadVector64((Int32*)(_dataTable.inArray1Ptr)),AdvSimd.LoadVector64((Int32*)(_dataTable.inArray2Ptr))),elemIndex);

...
mov x0, x21
movz x2, #0xC8A0 // code for System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[int],System.Runtime.Intrinsics.Vector64`1[int]],ubyte)
movk x2, #0x5DB8 LSL #16
movk x2, #0xFFFF LSL #32
ldr x2, [x2]
...
; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar64x2(ulong,System.ValueTuple`2[System.Runtime.Intrinsics.Vector64`1[short],System.Runtime.Intrinsics.Vector64`1[short]],ubyte) (FullOpts)
G_M60890_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x20]!
mov fp, sp
stp d0, d1, [fp, #0x10]	// [V01 arg1], [V01 arg1+0x08]
;; size=12 bbWeight=1 PerfScore 2.50
G_M60890_IG02: ;; offset=0x000C
ldr q16, [fp, #0x10]	// [V01 arg1]
ldr q17, [fp, #0x18]	// [V01 arg1+0x08]
uxtb w1, w1
adr x2, [G_M60890_IG03]
add x2, x2, x1, LSL #3
br x2
;; size=24 bbWeight=1 PerfScore 7.00
G_M60890_IG03: ;; offset=0x0024
st2 {v16.h, v17.h}[0], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG04: ;; offset=0x002C
st2 {v16.h, v17.h}[1], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG05: ;; offset=0x0034
st2 {v16.h, v17.h}[2], [x0]
b G_M60890_IG07
;; size=8 bbWeight=1 PerfScore 2.00
G_M60890_IG06: ;; offset=0x003C
st2 {v16.h, v17.h}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M60890_IG07: ;; offset=0x0040
ldp fp, lr, [sp], #0x20
ret lr
;; size=8 bbWeight=1 PerfScore 2.00

3. When the index is variable and out of bounds

The current test crashes with an error.

Illegal instruction (core dumped)

Not sure whether it's the desired behaviour. Also, I haven't considered the possibility of creating a ROP gadget with the sequence for Case 2 above. The sequence of instruction where value of an index can be used control the jump target.

[kunalspathak]

As seen in #93197, the API names should be just StoreSelectedScalar because you can deduce the shape and tuple size from the parameters.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

[kunalspathak]

I will double check about the immediate operand out of range scenario.

Alright, so if you follow #93197, the immOp is handled in it.

[kunalspathak]

3. When the index is variable and out of bounds

They should throw IndexOutOfBoundException

[kunalspathak]

RunBasicScenario_UnsafeRead segfaults with the following error

Interesting. What is the cause of it?

[SwapnilGaikwad]

Now we get out of range exception for all the invalid scenarios (with constant or variable index) with all the test cases.

System.ArgumentOutOfRangeException: Specified argument was out of the range of valid values.

I never got to ensure from the assembly that it throws InvalidIndexException. 🤔

This mystery is also resolved now. There was a check for array out of bounds missing. Now added it. Can be confirmed from the assembly for StoreSelectedScalar as well.

...
G_M56847_IG04:
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
...

Full assembly

; Assembly listing for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; Emitting BLENDED_CODE for generic ARM64 - Unix
; MinOpts code
; debuggable code
; fp based frame
; fully interruptible
; No PGO data
; Final local variable assignments
;
; V00 arg0 [V00 ] ( 1, 1 ) long -> [fp+0x38] do-not-enreg[]
; V01 arg1 [V01 ] ( 1, 1 ) simd16 -> [fp+0x20] HFA(simd16) do-not-enreg[S] <System.Runtime.Intrinsics.Vector128`1[int]>
; V02 arg2 [V02 ] ( 1, 1 ) ubyte -> [fp+0x1C] do-not-enreg[]
;# V03 OutArgs [V03 ] ( 1, 1 ) struct ( 0) [sp+0x00] do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
;
; Lcl frame size = 48
G_M56847_IG01: ;; offset=0x0000
stp fp, lr, [sp, #-0x40]!
mov fp, sp
str x0, [fp, #0x38]	// [V00 arg0]
str w1, [fp, #0x1C]	// [V02 arg2]
str q0, [fp, #0x20]	// [V01 arg1]
;; size=20 bbWeight=1 PerfScore 4.50
G_M56847_IG02: ;; offset=0x0014
movz x0, #0x45A8
movk x0, #0x4004 LSL #16
movk x0, #0xFFFF LSL #32
ldr w0, [x0]
cbz w0, G_M56847_IG04
;; size=20 bbWeight=1 PerfScore 5.50
G_M56847_IG03: ;; offset=0x0028
bl CORINFO_HELP_DBG_IS_JUST_MY_CODE
;; size=4 bbWeight=0.50 PerfScore 0.50
G_M56847_IG04: ;; offset=0x002C
ldr x0, [fp, #0x38]	// [V00 arg0]
ldr q16, [fp, #0x20]	// [V01 arg1]
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
cmp w1, #4
blo G_M56847_IG05
bl CORINFO_HELP_THROW_ARGUMENTOUTOFRANGEEXCEPTION
;; size=28 bbWeight=1 PerfScore 9.00
G_M56847_IG05: ;; offset=0x0048
ldr w1, [fp, #0x1C]	// [V02 arg2]
uxtb w1, w1
adr x2, [G_M56847_IG06]
add x2, x2, x1, LSL #3
br x2
;; size=20 bbWeight=1 PerfScore 5.00
G_M56847_IG06: ;; offset=0x005C
st1 {v16.s}[0], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG07: ;; offset=0x0064
st1 {v16.s}[1], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG08: ;; offset=0x006C
st1 {v16.s}[2], [x0]
b G_M56847_IG10
;; size=8 bbWeight=1 PerfScore 2.00
G_M56847_IG09: ;; offset=0x0074
st1 {v16.s}[3], [x0]
;; size=4 bbWeight=1 PerfScore 1.00
G_M56847_IG10: ;; offset=0x0078
nop
nop
;; size=8 bbWeight=1 PerfScore 1.00
G_M56847_IG11: ;; offset=0x0080
ldp fp, lr, [sp], #0x40
ret lr
;; size=8 bbWeight=1 PerfScore 2.00
; Total bytes of code 136, prolog size 20, PerfScore 48.10, instruction count 34, allocated bytes for code 136 (MethodHash=5b1221f0) for method System.Runtime.Intrinsics.Arm.AdvSimd:StoreSelectedScalar(ulong,System.Runtime.Intrinsics.Vector128`1[int],ubyte) (MinOpts)
; ============================================================

Interestingly, for some tests I got IndexOutOfRangeException earlier because the test was executing without an error but failing in the validation stage where it was accessing the out of range index. Only the first test was failing with an illegal instruction probably because of the code layout. Invalid index tried to jump to location that did not contain a valid index. For the further tests, the jump target at invalid index contained a valid instruction so it did not throw any exception during the execution.

[kunalspathak]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

[SwapnilGaikwad]

Seems the indices of few test cases needs adjustment for them to not throw Exception.

Had to move the range check after recomputing the simdSize 🤦 . Now the current failures seems while sending the test data to helix system.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

[kunalspathak]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

[SwapnilGaikwad]

can you double check?

Sure! It's tricky to debug, and it isn't obvious why compilation of System.Runtime.Serialization.SerializationGuard would fail. I'll try rebasing/merging main, if the error persists then can debug further.

From what I remember, these tests were passing earlier and failing with the recent commit and are in arm64, so most likly related. I could be wrong.

You're right, it seems to be related to the arg check. Not sure how to debug it though. I'll try to see if I can reproduce the failure. If that doesn't progress, should I put the explicit check as before?

[kunalspathak]

If that doesn't progress, should I put the explicit check as before?

I would say just inspect the code changes and see if the code path for existing SelectedScalar with single vector is unchanged. That should fix the problem.

[kunalspathak]

@SwapnilGaikwad - can you resolve the merge conflicts?

[fanyang-mono]

The change to Mono codebase looks good to me.

[kunalspathak]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

[fanyang-mono]

The change to Mono codebase looks good to me.

@fanyang-mono - was the change essentially to make sure that mono continue to ignore newer APIs whose 2nd parameter is tuple and only support the existing StoreSelectedScalar API?

Unfortunately, yes, Mono is not super friendly with handling API with 4 and more input arguments. We need to add additional code to handle those cases, rather than automatically handled by the machinery like Vector64/128. There is an issue to track Mono side of the work related to these new set of API's.