Fixing up the Sse41.Insert float HWIntrinsics#18735
Conversation
|
FYI. @fiigii, @CarolEidt, @eerhardt |
| ("SimpleUnOpTest.template", new Dictionary<string, string> { ["Isa"] = "Sse41", ["LoadIsa"] = "Sse", ["Method"] = "Floor", ["RetVectorType"] = "Vector128", ["RetBaseType"] = "Single", ["Op1VectorType"] ="Vector128", ["Op1BaseType"] = "Single", ["LargestVectorSize"] = "16", ["NextValueOp1"] = "(float)(random.NextDouble())", ["ValidateFirstResult"] = "BitConverter.SingleToInt32Bits(result[0]) != BitConverter.SingleToInt32Bits(MathF.Floor(firstOp[0]))", ["ValidateRemainingResults"] = "BitConverter.SingleToInt32Bits(result[i]) != BitConverter.SingleToInt32Bits(MathF.Floor(firstOp[i]))"}), | ||
| ("SimpleBinOpTest.template", new Dictionary<string, string> { ["Isa"] = "Sse41", ["LoadIsa"] = "Sse2", ["Method"] = "FloorScalar", ["RetVectorType"] = "Vector128", ["RetBaseType"] = "Double", ["Op1VectorType"] ="Vector128", ["Op1BaseType"] = "Double", ["Op2VectorType"] = "Vector128", ["Op2BaseType"] = "Double", ["LargestVectorSize"] = "16", ["NextValueOp1"] = "(double)(random.NextDouble())", ["NextValueOp2"] = "(double)(random.NextDouble())", ["ValidateFirstResult"] = "BitConverter.DoubleToInt64Bits(result[0]) != BitConverter.DoubleToInt64Bits(Math.Floor(right[0]))", ["ValidateRemainingResults"] = "BitConverter.DoubleToInt64Bits(result[i]) != BitConverter.DoubleToInt64Bits(left[i])"}), | ||
| ("SimpleBinOpTest.template", new Dictionary<string, string> { ["Isa"] = "Sse41", ["LoadIsa"] = "Sse", ["Method"] = "FloorScalar", ["RetVectorType"] = "Vector128", ["RetBaseType"] = "Single", ["Op1VectorType"] ="Vector128", ["Op1BaseType"] = "Single", ["Op2VectorType"] = "Vector128", ["Op2BaseType"] = "Single", ["LargestVectorSize"] = "16", ["NextValueOp1"] = "(float)(random.NextDouble())", ["NextValueOp2"] = "(float)(random.NextDouble())", ["ValidateFirstResult"] = "BitConverter.SingleToInt32Bits(result[0]) != BitConverter.SingleToInt32Bits(MathF.Floor(right[0]))", ["ValidateRemainingResults"] = "BitConverter.SingleToInt32Bits(result[i]) != BitConverter.SingleToInt32Bits(left[i])"}), | ||
| ("InsertVector128Test.template", new Dictionary<string, string> { ["Isa"] = "Sse41", ["LoadIsa"] = "Sse", ["Method"] = "Insert", ["RetVectorType"] = "Vector128", ["RetBaseType"] = "Single", ["Op1VectorType"] ="Vector128", ["Op1BaseType"] = "Single", ["Op2VectorType"] = "Vector128", ["Op2BaseType"] = "Single", ["Imm"] = "0", ["LargestVectorSize"] = "16", ["NextValueOp1"] = "(float)(random.NextDouble())", ["NextValueOp2"] = "(float)(random.NextDouble())", ["ValidateFirstResult"] = "BitConverter.SingleToInt32Bits(result[0]) != BitConverter.SingleToInt32Bits(right[0])", ["ValidateRemainingResults"] = "BitConverter.SingleToInt32Bits(result[i]) != BitConverter.SingleToInt32Bits(left[i])"}), |
There was a problem hiding this comment.
As usual, tests are auto-generated from this template data, it is worthwhile looking at the base template and the data, but probably not reviewing each test individually.
| ssize_t ival = op3->AsIntCon()->IconValue(); | ||
| assert((ival >= 0) && (ival <= 255)); | ||
|
|
||
| if ((intrinsicId == NI_SSE41_Insert) && (baseType == TYP_FLOAT)) |
There was a problem hiding this comment.
No longer required since the new API shape allows a vector for the second operand, and therefore makes the upper two bits relevant.
|
|
||
| op3 = argList->Current(); | ||
|
|
||
| // The upper two bits of the immediate value are ignored if |
There was a problem hiding this comment.
This is important, as per the architecture manuals:
When the scalar source is a memory operand the Count_S bits are ignored.
As such, we can support containment when the control mask (op3) is <= 0x3F. Otherwise, we need to ensure that op2 is in register and should not support any containment.
|
Some unrelated AVX tests will fail for NoSIMD and NoAVX, these are being handled by: #18734 |
|
@dotnet-bot test Windows_NT x64 Checked jitincompletehwintrinsic @dotnet-bot test Windows_NT x86 Checked jitincompletehwintrinsic @dotnet-bot test Ubuntu x64 Checked jitincompletehwintrinsic |
|
Rebased onto dotnet/master to pick-up the test fixes |
3 participants
Sse41.Insert floattests (temporarily removed due to the API changes made in Improve Intel hardware intrinsic APIs #17637)Sse41.Insert floatimplementation to support immediate values greater than0x3F