Skip to content

Commit f81f0a8

Browse files
kunalspathakmichaelgsharp
authored andcommitted
Arm64/Sve: Implement SVE Math Min*/Max* APIs (dotnet#101859)
* Add the missing else * Max, MaxAcross, MaxNumber, MaxNumberAcross, Min, MinAcross, MinNumber, MinNumberAcross * Map APIs to instruction * Add test cases * Remove the space * fix the test case
1 parent 71d264c commit f81f0a8

File tree

6 files changed

+896
-11
lines changed

6 files changed

+896
-11
lines changed

src/coreclr/jit/hwintrinsiccodegenarm64.cpp

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -550,6 +550,12 @@ void CodeGen::genHWIntrinsic(GenTreeHWIntrinsic* node)
550550
// and `embMaskOp2Reg` is the second operand.
551551
GetEmitter()->emitIns_R_R_R(insEmbMask, emitSize, targetReg, maskReg, embMaskOp2Reg, opt);
552552
}
553+
else
554+
{
555+
// Just perform the actual "predicated" operation so that `targetReg` is the first operand
556+
// and `embMaskOp2Reg` is the second operand.
557+
GetEmitter()->emitIns_R_R_R(insEmbMask, emitSize, targetReg, maskReg, embMaskOp2Reg, opt);
558+
}
553559

554560
break;
555561

src/coreclr/jit/hwintrinsiclistarm64sve.h

Lines changed: 8 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -71,6 +71,14 @@ HARDWARE_INTRINSIC(Sve, LoadVectorUInt16ZeroExtendToUInt32,
7171
HARDWARE_INTRINSIC(Sve, LoadVectorUInt16ZeroExtendToUInt64, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ld1h, INS_invalid, INS_invalid}, HW_Category_MemoryLoad, HW_Flag_Scalable|HW_Flag_ExplicitMaskedOperation|HW_Flag_LowMaskedOperation)
7272
HARDWARE_INTRINSIC(Sve, LoadVectorUInt32ZeroExtendToInt64, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ld1w, INS_invalid, INS_invalid, INS_invalid}, HW_Category_MemoryLoad, HW_Flag_Scalable|HW_Flag_ExplicitMaskedOperation|HW_Flag_LowMaskedOperation)
7373
HARDWARE_INTRINSIC(Sve, LoadVectorUInt32ZeroExtendToUInt64, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ld1w, INS_invalid, INS_invalid}, HW_Category_MemoryLoad, HW_Flag_Scalable|HW_Flag_ExplicitMaskedOperation|HW_Flag_LowMaskedOperation)
74+
HARDWARE_INTRINSIC(Sve, Max, -1, -1, false, {INS_sve_smax, INS_sve_umax, INS_sve_smax, INS_sve_umax, INS_sve_smax, INS_sve_umax, INS_sve_smax, INS_sve_umax, INS_sve_fmax, INS_sve_fmax}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation|HW_Flag_HasRMWSemantics)
75+
HARDWARE_INTRINSIC(Sve, MaxAcross, -1, -1, false, {INS_sve_smaxv, INS_sve_umaxv, INS_sve_smaxv, INS_sve_umaxv, INS_sve_smaxv, INS_sve_umaxv, INS_sve_smaxv, INS_sve_umaxv, INS_sve_fmaxv, INS_sve_fmaxv}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation)
76+
HARDWARE_INTRINSIC(Sve, MaxNumber, -1, -1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_fmaxnm, INS_sve_fmaxnm}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation|HW_Flag_HasRMWSemantics)
77+
HARDWARE_INTRINSIC(Sve, MaxNumberAcross, -1, -1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_fmaxnmv, INS_sve_fmaxnmv}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation)
78+
HARDWARE_INTRINSIC(Sve, Min, -1, -1, false, {INS_sve_smin, INS_sve_umin, INS_sve_smin, INS_sve_umin, INS_sve_smin, INS_sve_umin, INS_sve_smin, INS_sve_umin, INS_sve_fmin, INS_sve_fmin}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation|HW_Flag_HasRMWSemantics)
79+
HARDWARE_INTRINSIC(Sve, MinAcross, -1, -1, false, {INS_sve_sminv, INS_sve_uminv, INS_sve_sminv, INS_sve_uminv, INS_sve_sminv, INS_sve_uminv, INS_sve_sminv, INS_sve_uminv, INS_sve_fminv, INS_sve_fminv}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation)
80+
HARDWARE_INTRINSIC(Sve, MinNumber, -1, -1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_fminnm, INS_sve_fminnm}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation|HW_Flag_HasRMWSemantics)
81+
HARDWARE_INTRINSIC(Sve, MinNumberAcross, -1, -1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_fminnmv, INS_sve_fminnmv}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation)
7482
HARDWARE_INTRINSIC(Sve, Multiply, -1, 2, true, {INS_sve_mul, INS_sve_mul, INS_sve_mul, INS_sve_mul, INS_sve_mul, INS_sve_mul, INS_sve_mul, INS_sve_mul, INS_sve_fmul, INS_sve_fmul}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation|HW_Flag_HasRMWSemantics)
7583
HARDWARE_INTRINSIC(Sve, Or, -1, -1, false, {INS_sve_orr, INS_sve_orr, INS_sve_orr, INS_sve_orr, INS_sve_orr, INS_sve_orr, INS_sve_orr, INS_sve_orr, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_OptionalEmbeddedMaskedOperation|HW_Flag_HasRMWSemantics|HW_Flag_LowMaskedOperation)
7684
HARDWARE_INTRINSIC(Sve, OrAcross, -1, -1, false, {INS_sve_orv, INS_sve_orv, INS_sve_orv, INS_sve_orv, INS_sve_orv, INS_sve_orv, INS_sve_orv, INS_sve_orv, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation|HW_Flag_LowMaskedOperation)

0 commit comments

Comments
 (0)