internal/simdgen: update documentations

This CL is generated by Gemini

diff --git a/internal/simdgen/categories.yaml b/internal/simdgen/categories.yaml
index 1ffe16d..05ae129 100644
--- a/internal/simdgen/categories.yaml
+++ b/internal/simdgen/categories.yaml
@@ -2,75 +2,131 @@
 - go: Add
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Add adds corresponding elements of two vectors.
 - go: SaturatedAdd
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    SaturatedAdd adds corresponding elements of two vectors with saturation.
 - go: MaskedAdd
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedAdd adds corresponding elements of two vectors, masked.
+  docUnmasked: !string |-
+    Add adds corresponding elements of two vectors.
 - go: MaskedSaturatedAdd
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedSaturatedAdd adds corresponding elements of two vectors with saturation, masked.
+  docUnmasked: !string |-
+    SaturatedAdd adds corresponding elements of two vectors with saturation.
 - go: Sub
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    Sub subtracts corresponding elements of two vectors.
 - go: SaturatedSub
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    SaturatedSub subtracts corresponding elements of two vectors with saturation.
 - go: MaskedSub
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedSub subtracts corresponding elements of two vectors, masked.
+  docUnmasked: !string |-
+    Sub subtracts corresponding elements of two vectors.
 - go: MaskedSaturatedSub
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedSaturatedSub subtracts corresponding elements of two vectors with saturation, masked.
+  docUnmasked: !string |-
+    SaturatedSub subtracts corresponding elements of two vectors with saturation.
 - go: PairwiseAdd
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Add pairs of elements in vector x and store them in higher half of the target; Add pairs of elements in vector y and store them in lower half of the target"
+  documentation: !string |-
+    PairwiseAdd horizontally adds adjacent pairs of elements.
+    For x = [x0, x1, x2, x3] and y = [y0, y1, y2, y3], the result is [y0+y1, y2+y3, x0+x1, x2+x3].
 - go: PairwiseSub
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Sub pairs of elements in vector x and store them in higher half of the target; Sub pairs of elements in vector y and store them in lower half of the target"
+  documentation: !string |-
+    PairwiseSub horizontally subtracts adjacent pairs of elements.
+    For x = [x0, x1, x2, x3] and y = [y0, y1, y2, y3], the result is [y0-y1, y2-y3, x0-x1, x2-x3].
 - go: SaturatedPairwiseAdd
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Add pairs of elements in vector x and store them in higher half of the target; Add pairs of elements in vector y and store them in lower half of the target; With saturation"
+  documentation: !string |-
+    SaturatedPairwiseAdd horizontally adds adjacent pairs of elements with saturation.
+    For x = [x0, x1, x2, x3] and y = [y0, y1, y2, y3], the result is [y0+y1, y2+y3, x0+x1, x2+x3].
 - go: SaturatedPairwiseSub
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Sub pairs of elements in vector x and store them in higher half of the target; Sub pairs of elements in vector y and store them in lower half of the target; With saturation"
+  documentation: !string |-
+    SaturatedPairwiseSub horizontally subtracts adjacent pairs of elements with saturation.
+    For x = [x0, x1, x2, x3] and y = [y0, y1, y2, y3], the result is [y0-y1, y2-y3, x0-x1, x2-x3].
 - go: And
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    And performs a bitwise AND operation between two vectors.
 - go: MaskedAnd
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedAnd performs a masked bitwise AND operation between two vectors.
+  docUnmasked: !string |-
+    And performs a bitwise AND operation between two vectors.
 - go: Or
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Or performs a bitwise OR operation between two vectors.
 - go: MaskedOr
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedOr performs a masked bitwise OR operation between two vectors.
+  docUnmasked: !string |-
+    Or performs a bitwise OR operation between two vectors.
 - go: AndNot
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    AndNot performs a bitwise AND NOT operation between two vectors.
 - go: MaskedAndNot
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedAndNot performs a masked bitwise AND NOT operation between two vectors.
+  docUnmasked: !string |-
+    AndNot performs a bitwise AND NOT operation between two vectors.
 - go: Xor
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Xor performs a bitwise XOR operation between two vectors.
 - go: MaskedXor
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedXor performs a masked bitwise XOR operation between two vectors.
+  docUnmasked: !string |-
+    Xor performs a bitwise XOR operation between two vectors.
 # We also have PTEST and VPTERNLOG, those should be hidden from the users
 # and only appear in rewrite rules.
 # const imm predicate(holds for both float and int|uint):
@@ -84,311 +140,560 @@
   constImm: 0
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 0 if it has;"
+  documentation: !string |-
+    Equal compares for equality.
+    Const Immediate = 0.
 - go: Less
   constImm: 1
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 1 if it has;"
+  documentation: !string |-
+    Less compares for less than.
+    Const Immediate = 1.
 - go: LessEqual
   constImm: 2
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 2 if it has;"
+  documentation: !string |-
+    LessEqual compares for less than or equal.
+    Const Immediate = 2.
 - go: IsNan # For float only.
   constImm: 3
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 3 if it has; Returns mask element True if either one of the input\\'s element is Nan; Please use this method as x\\.IsNan\\(x\\) to check x only;"
+  documentation: !string |-
+    IsNan checks if elements are NaN. Use as x.IsNan(x).
+    Const Immediate = 3.
 - go: NotEqual
   constImm: 4
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 4 if it has;"
+  documentation: !string |-
+    NotEqual compares for inequality.
+    Const Immediate = 4.
 - go: GreaterEqual
   constImm: 5
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 5 if it has;"
+  documentation: !string |-
+    GreaterEqual compares for greater than or equal.
+    Const Immediate = 5.
 - go: Greater
   constImm: 6
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 6 if it has;"
+  documentation: !string |-
+    Greater compares for greater than.
+    Const Immediate = 6.
 
 - go: MaskedEqual
   constImm: 0
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 0 if it has;"
+  documentation: !string |-
+    MaskedEqual compares for equality, masked.
+    Const Immediate = 0.
+  docUnmasked: !string |-
+    Equal compares for equality.
+    Const Immediate = 0.
 - go: MaskedLess
   constImm: 1
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 1 if it has;"
+  documentation: !string |-
+    MaskedLess compares for less than, masked.
+    Const Immediate = 1.
+  docUnmasked: !string |-
+    Less compares for less than.
+    Const Immediate = 1.
 - go: MaskedLessEqual
   constImm: 2
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 2 if it has;"
+  documentation: !string |-
+    MaskedLessEqual compares for less than or equal, masked.
+    Const Immediate = 2.
+  docUnmasked: !string |-
+    LessEqual compares for less than or equal.
+    Const Immediate = 2.
 - go: MaskedIsNan # For float only.
   constImm: 3
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 3 if it has; Returns mask element True if either one of the input\\'s element is Nan; Please use this method as x\\.IsNan\\(x\\) to check x only;"
+  documentation: !string |-
+    MaskedIsNan checks if elements are NaN, masked. Use as x.IsNan(x).
+    Const Immediate = 3.
+  docUnmasked: !string |-
+    IsNan checks if elements are NaN. Use as x.IsNan(x).
+    Const Immediate = 3.
 - go: MaskedNotEqual
   constImm: 4
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 4 if it has;"
+  documentation: !string |-
+    MaskedNotEqual compares for inequality, masked.
+    Const Immediate = 4.
+  docUnmasked: !string |-
+    NotEqual compares for inequality.
+    Const Immediate = 4.
 - go: MaskedGreaterEqual
   constImm: 5
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 5 if it has;"
+  documentation: !string |-
+    MaskedGreaterEqual compares for greater than or equal, masked.
+    Const Immediate = 5.
+  docUnmasked: !string |-
+    GreaterEqual compares for greater than or equal.
+    Const Immediate = 5.
 - go: MaskedGreater
   constImm: 6
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 6 if it has;"
+  documentation: !string |-
+    MaskedGreater compares for greater than, masked.
+    Const Immediate = 6.
+  docUnmasked: !string |-
+    Greater compares for greater than.
+    Const Immediate = 6.
 - go: Div
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    Div divides elements of two vectors.
 - go: MaskedDiv
   commutative: "false"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedDiv divides elements of two vectors, masked.
+  docUnmasked: !string |-
+    Div divides elements of two vectors.
 - go: Sqrt
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    Sqrt computes the square root of each element.
 - go: MaskedSqrt
   commutative: "false"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedSqrt computes the square root of each element, masked.
+  docUnmasked: !string |-
+    Sqrt computes the square root of each element.
 - go: ApproximateReciprocal
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    ApproximateReciprocal computes an approximate reciprocal of each element.
 - go: MaskedApproximateReciprocal
   commutative: "false"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedApproximateReciprocal computes an approximate reciprocal of each element, masked.
+  docUnmasked: !string |-
+    ApproximateReciprocal computes an approximate reciprocal of each element.
 - go: ApproximateReciprocalOfSqrt
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    ApproximateReciprocalOfSqrt computes an approximate reciprocal of the square root of each element.
 - go: MaskedApproximateReciprocalOfSqrt
   commutative: "false"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedApproximateReciprocalOfSqrt computes an approximate reciprocal of the square root of each element, masked.
+  docUnmasked: !string |-
+    ApproximateReciprocalOfSqrt computes an approximate reciprocal of the square root of each element.
 - go: MaskedMulByPowOf2 # This operation is all after AVX512, the unmasked version will be generated.
   commutative: "false"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedMulByPowOf2 multiplies elements by a power of 2, masked.
+  docUnmasked: !string |-
+    MulByPowOf2 multiplies elements by a power of 2.
 
 - go: Round
   commutative: "false"
   extension: "AVX.*"
   constImm: 0
+  documentation: !string |-
+    Round rounds elements to the nearest integer.
+    Const Immediate = 0.
 - go: MaskedRoundWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 0
   masked: "true"
+  documentation: !string |-
+    MaskedRoundWithPrecision rounds elements with specified precision, masked.
+    Const Immediate = 0.
+  docUnmasked: !string |-
+    RoundWithPrecision rounds elements with specified precision.
+    Const Immediate = 0.
 - go: MaskedRoundSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 8
   masked: "true"
+  documentation: !string |-
+    MaskedRoundSuppressExceptionWithPrecision rounds elements with specified precision, suppressing exceptions, masked.
+    Const Immediate = 8.
+  docUnmasked: !string |-
+    RoundSuppressExceptionWithPrecision rounds elements with specified precision, suppressing exceptions.
+    Const Immediate = 8.
 - go: MaskedDiffWithRoundWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 0
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithRoundWithPrecision computes the difference after rounding with specified precision, masked.
+    Const Immediate = 0.
+  docUnmasked: !string |-
+    DiffWithRoundWithPrecision computes the difference after rounding with specified precision.
+    Const Immediate = 0.
 - go: MaskedDiffWithRoundSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 8
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithRoundSuppressExceptionWithPrecision computes the difference after rounding with specified precision, suppressing exceptions, masked.
+    Const Immediate = 8.
+  docUnmasked: !string |-
+    DiffWithRoundSuppressExceptionWithPrecision computes the difference after rounding with specified precision, suppressing exceptions.
+    Const Immediate = 8.
 
 - go: Floor
   commutative: "false"
   extension: "AVX.*"
   constImm: 1
+  documentation: !string |-
+    Floor rounds elements down to the nearest integer.
+    Const Immediate = 1.
 - go: MaskedFloorWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 1
   masked: "true"
+  documentation: !string |-
+    MaskedFloorWithPrecision rounds elements down with specified precision, masked.
+    Const Immediate = 1.
+  docUnmasked: !string |-
+    FloorWithPrecision rounds elements down with specified precision.
+    Const Immediate = 1.
 - go: MaskedFloorSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 9
   masked: "true"
+  documentation: !string |-
+    MaskedFloorSuppressExceptionWithPrecision rounds elements down with specified precision, suppressing exceptions, masked.
+    Const Immediate = 9.
+  docUnmasked: !string |-
+    FloorSuppressExceptionWithPrecision rounds elements down with specified precision, suppressing exceptions.
+    Const Immediate = 9.
 - go: MaskedDiffWithFloorWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 1
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithFloorWithPrecision computes the difference after flooring with specified precision, masked.
+    Const Immediate = 1.
+  docUnmasked: !string |-
+    DiffWithFloorWithPrecision computes the difference after flooring with specified precision.
+    Const Immediate = 1.
 - go: MaskedDiffWithFloorSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 9
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithFloorSuppressExceptionWithPrecision computes the difference after flooring with specified precision, suppressing exceptions, masked.
+    Const Immediate = 9.
+  docUnmasked: !string |-
+    DiffWithFloorSuppressExceptionWithPrecision computes the difference after flooring with specified precision, suppressing exceptions.
+    Const Immediate = 9.
 
 - go: Ceil
   commutative: "false"
   extension: "AVX.*"
   constImm: 2
+  documentation: !string |-
+    Ceil rounds elements up to the nearest integer.
+    Const Immediate = 2.
 - go: MaskedCeilWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 2
   masked: "true"
+  documentation: !string |-
+    MaskedCeilWithPrecision rounds elements up with specified precision, masked.
+    Const Immediate = 2.
+  docUnmasked: !string |-
+    CeilWithPrecision rounds elements up with specified precision.
+    Const Immediate = 2.
 - go: MaskedCeilSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 10
   masked: "true"
+  documentation: !string |-
+    MaskedCeilSuppressExceptionWithPrecision rounds elements up with specified precision, suppressing exceptions, masked.
+    Const Immediate = 10.
+  docUnmasked: !string |-
+    CeilSuppressExceptionWithPrecision rounds elements up with specified precision, suppressing exceptions.
+    Const Immediate = 10.
 - go: MaskedDiffWithCeilWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 2
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithCeilWithPrecision computes the difference after ceiling with specified precision, masked.
+    Const Immediate = 2.
+  docUnmasked: !string |-
+    DiffWithCeilWithPrecision computes the difference after ceiling with specified precision.
+    Const Immediate = 2.
 - go: MaskedDiffWithCeilSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 10
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithCeilSuppressExceptionWithPrecision computes the difference after ceiling with specified precision, suppressing exceptions, masked.
+    Const Immediate = 10.
+  docUnmasked: !string |-
+    DiffWithCeilSuppressExceptionWithPrecision computes the difference after ceiling with specified precision, suppressing exceptions.
+    Const Immediate = 10.
 
 - go: Trunc
   commutative: "false"
   extension: "AVX.*"
   constImm: 3
+  documentation: !string |-
+    Trunc truncates elements towards zero.
+    Const Immediate = 3.
 - go: MaskedTruncWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 3
   masked: "true"
+  documentation: !string |-
+    MaskedTruncWithPrecision truncates elements with specified precision, masked.
+    Const Immediate = 3.
+  docUnmasked: !string |-
+    TruncWithPrecision truncates elements with specified precision.
+    Const Immediate = 3.
 - go: MaskedTruncSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 11
   masked: "true"
+  documentation: !string |-
+    MaskedTruncSuppressExceptionWithPrecision truncates elements with specified precision, suppressing exceptions, masked.
+    Const Immediate = 11.
+  docUnmasked: !string |-
+    TruncSuppressExceptionWithPrecision truncates elements with specified precision, suppressing exceptions.
+    Const Immediate = 11.
 - go: MaskedDiffWithTruncWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 3
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithTruncWithPrecision computes the difference after truncating with specified precision, masked.
+    Const Immediate = 3.
+  docUnmasked: !string |-
+    DiffWithTruncWithPrecision computes the difference after truncating with specified precision.
+    Const Immediate = 3.
 - go: MaskedDiffWithTruncSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 11
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithTruncSuppressExceptionWithPrecision computes the difference after truncating with specified precision, suppressing exceptions, masked.
+    Const Immediate = 11.
+  docUnmasked: !string |-
+    DiffWithTruncSuppressExceptionWithPrecision computes the difference after truncating with specified precision, suppressing exceptions.
+    Const Immediate = 11.
 
 - go: AddSub
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    AddSub alternatingly adds and subtracts elements of two vectors.
 - go: Average
   commutative: "true"
   extension: "AVX.*" # VPAVGB/W are available across various AVX versions
+  documentation: !string |-
+    Average computes the rounded average of corresponding elements.
 - go: MaskedAverage
   commutative: "true"
   masked: "true"
   extension: "AVX512.*" # Masked operations are typically AVX512
+  documentation: !string |-
+    MaskedAverage computes the rounded average of corresponding elements, masked.
+  docUnmasked: !string |-
+    Average computes the rounded average of corresponding elements.
 
 - go: Absolute
   commutative: "false"
   # Unary operation, not commutative
   extension: "AVX.*" # VPABSB/W/D are AVX, VPABSQ is AVX512
+  documentation: !string |-
+    Absolute computes the absolute value of each element.
 - go: MaskedAbsolute
   commutative: "false"
   masked: "true"
   extension: "AVX512.*"
+  documentation: !string |-
+    MaskedAbsolute computes the absolute value of each element, masked.
+  docUnmasked: !string |-
+    Absolute computes the absolute value of each element.
 
 - go: Sign
   # Applies sign of second operand to first: sign(val, sign_src)
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    Sign applies the sign of the second operand to the first.
   # Sign does not have masked version
 
 - go: MaskedPopCount
   commutative: "false"
   masked: "true"
   extension: "AVX512.*" # VPOPCNT instructions are AVX512 (BITALG or VPOPCNTDQ)
+  documentation: !string |-
+    MaskedPopCount counts the number of set bits in each element, masked.
+  docUnmasked: !string |-
+    PopCount counts the number of set bits in each element.
 - go: PairDotProd
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiply the elements and add the pairs together"
+  documentation: !string |-
+    PairDotProd multiplies and adds adjacent pairs of elements.
 - go: MaskedPairDotProd
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiply the elements and add the pairs together"
+  documentation: !string |-
+    MaskedPairDotProd multiplies and adds adjacent pairs of elements, masked.
+  docUnmasked: !string |-
+    PairDotProd multiplies and adds adjacent pairs of elements.
 - go: SaturatedPairDotProd
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiply the elements and add the pairs together with saturation"
+  documentation: !string |-
+    SaturatedPairDotProd multiplies and adds adjacent pairs of elements with saturation.
 - go: MaskedSaturatedPairDotProd
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiply the elements and add the pairs together with saturation"
+  documentation: !string |-
+    MaskedSaturatedPairDotProd multiplies and adds adjacent pairs of elements with saturation, masked.
+  docUnmasked: !string |-
+    SaturatedPairDotProd multiplies and adds adjacent pairs of elements with saturation.
+
 # QuadDotProd, i.e. VPDPBUSD(S) are operations with src/dst on the same register, we are not supporting this as of now.
 - go: DotProdBroadcast
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiply the elements and add the pairs together; the result is a broadcast of the dot product; imm8 = 127;"
+  documentation: !string |-
+    DotProdBroadcast multiplies all elements and broadcasts the sum.
+    Const Immediate = 127.
 - go: Max
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Max computes the maximum of corresponding elements.
 - go: MaskedMax
   commutative: "true"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedMax computes the maximum of corresponding elements, masked.
+  docUnmasked: !string |-
+    Max computes the maximum of corresponding elements.
 - go: Min
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Min computes the minimum of corresponding elements.
 - go: MaskedMin
   commutative: "true"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedMin computes the minimum of corresponding elements, masked.
+  docUnmasked: !string |-
+    Min computes the minimum of corresponding elements.
 - go: Mul
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Mul multiplies corresponding elements of two vectors.
 - go: MulEvenWiden
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the even index elements from the two sources of size X at index i, store the result of size 2X at index i/2"
+  documentation: !string |-
+    MulEvenWiden multiplies even-indexed elements, widening the result.
+    Result[i] = v1.Even[i] * v2.Even[i].
 - go: MulHigh
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the elements from the two sources of size X at index i, store the high X bits of the result of size 2X at index i"
+  documentation: !string |-
+    MulHigh multiplies elements and stores the high part of the result.
 - go: MulLow
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the elements from the two sources of size X at index i, store the low X bits of the result of size 2X at index i"
+  documentation: !string |-
+    MulLow multiplies elements and stores the low part of the result.
 - go: MaskedMul
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedMul multiplies corresponding elements of two vectors, masked.
+  docUnmasked: !string |-
+    Mul multiplies corresponding elements of two vectors.
 - go: MaskedMulEvenWiden
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the even index elements from the two sources of size X at index i, store the result of size 2X at index i/2"
+  documentation: !string |-
+    MaskedMulEvenWiden multiplies even-indexed elements, widening the result, masked.
+    Result[i] = v1.Even[i] * v2.Even[i].
+  docUnmasked: !string |-
+    MulEvenWiden multiplies even-indexed elements, widening the result.
+    Result[i] = v1.Even[i] * v2.Even[i].
 - go: MaskedMulHigh
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the elements from the two sources of size X at index i, store the high X bits of the result of size 2X at index i"
+  documentation: !string |-
+    MaskedMulHigh multiplies elements and stores the high part of the result, masked.
+  docUnmasked: !string |-
+    MulHigh multiplies elements and stores the high part of the result.
 - go: MaskedMulLow
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the elements from the two sources of size X at index i, store the low X bits of the result of size 2X at index i"
+  documentation: !string |-
+    MaskedMulLow multiplies elements and stores the low part of the result, masked.
+  docUnmasked: !string |-
+    MulLow multiplies elements and stores the low part of the result.
diff --git a/internal/simdgen/gen_simdTypes.go b/internal/simdgen/gen_simdTypes.go
index ed29a75..aac85a4 100644
--- a/internal/simdgen/gen_simdTypes.go
+++ b/internal/simdgen/gen_simdTypes.go
@@ -75,37 +75,55 @@
 
 {{- range .OpsLen1}}
 
-// Asm: {{.Asm}}, Arch: {{.Extension}}{{if .Documentation}}, Doc: {{.Documentation}}{{end}}
+/*{{if .Documentation}}
+{{.Documentation}}{{end}}
+Asm: {{.Asm}}, Arch: {{.Extension}}
+*/
 func (x {{(index .In 0).Go}}) {{.Go}}() {{(index .Out 0).Go}}
 
 {{- end}}
 {{- range .OpsLen2}}
 
-// Asm: {{.Asm}}, Arch: {{.Extension}}{{if .Documentation}}, Doc: {{.Documentation}}{{end}}
+/*{{if .Documentation}}
+{{.Documentation}}{{end}}
+Asm: {{.Asm}}, Arch: {{.Extension}}
+*/
 func (x {{(index .In 0).Go}}) {{.Go}}(y {{(index .In 1).Go}}) {{(index .Out 0).Go}}
 
 {{- end}}
 {{- range .OpsLen3}}
 
-// Asm: {{.Asm}}, Arch: {{.Extension}}{{if .Documentation}}, Doc: {{.Documentation}}{{end}}
+/*{{if .Documentation}}
+{{.Documentation}}{{end}}
+Asm: {{.Asm}}, Arch: {{.Extension}}
+*/
 func (x {{(index .In 0).Go}}) {{.Go}}(y {{(index .In 1).Go}}, z {{(index .In 2).Go}}) {{(index .Out 0).Go}}
 
 {{- end}}
 {{- range .OpsLen1Imm8}}
 
-// Asm: {{.Asm}}, Arch: {{.Extension}}{{if .Documentation}}, Doc: {{.Documentation}}{{end}}
+/*{{if .Documentation}}
+{{.Documentation}}{{end}}
+Asm: {{.Asm}}, Arch: {{.Extension}}
+*/
 func (x {{(index .In 1).Go}}) {{.Go}}(imm8 uint8) {{(index .Out 0).Go}}
 
 {{- end}}
 {{- range .OpsLen2Imm8}}
 
-// Asm: {{.Asm}}, Arch: {{.Extension}}{{if .Documentation}}, Doc: {{.Documentation}}{{end}}
+/*{{if .Documentation}}
+{{.Documentation}}{{end}}
+Asm: {{.Asm}}, Arch: {{.Extension}}
+*/
 func (x {{(index .In 1).Go}}) {{.Go}}(imm uint8, y {{(index .In 2).Go}}) {{(index .Out 0).Go}}
 
 {{- end}}
 {{- range .OpsLen3Imm8}}
 
-// Asm: {{.Asm}}, Arch: {{.Extension}}{{if .Documentation}}, Doc: {{.Documentation}}{{end}}
+/*{{if .Documentation}}
+{{.Documentation}}{{end}}
+Asm: {{.Asm}}, Arch: {{.Extension}}
+*/
 func (x {{(index .In 1).Go}}) {{.Go}}(imm uint8, y {{(index .In 2).Go}}, z {{(index .In 3).Go}}) {{(index .Out 0).Go}}
 
 {{- end}}
diff --git a/internal/simdgen/gen_utility.go b/internal/simdgen/gen_utility.go
index 6ae1cff..5b9a636 100644
--- a/internal/simdgen/gen_utility.go
+++ b/internal/simdgen/gen_utility.go
@@ -357,6 +357,7 @@
 				return nil, fmt.Errorf("simdgen only recognizes masked operations with name starting with 'Masked': %s", op)
 			}
 			op2.Go = strings.ReplaceAll(op2.Go, "Masked", "")
+			op2.Documentation = op2.DocUnmasked // Also change the documentation.
 			splited = append(splited, op2)
 		} else {
 			return nil, fmt.Errorf("simdgen only recognizes masked operations with exactly one mask input: %s", op)
diff --git a/internal/simdgen/godefs.go b/internal/simdgen/godefs.go
index ad20943..c50760b 100644
--- a/internal/simdgen/godefs.go
+++ b/internal/simdgen/godefs.go
@@ -22,6 +22,7 @@
 	Extension     string    // Extension
 	Zeroing       *string   // Zeroing is a flag for asm prefix "Z", if non-nil it will always be "false"
 	Documentation *string   // Documentation will be appended to the stubs comments.
+	DocUnmasked   *string   // Documentation, unmasked version
 	// ConstMask is a hack to reduce the size of defs the user writes for const-immediate
 	// If present, it will be copied to [In[0].Const].
 	ConstImm *string
diff --git a/internal/simdgen/ops/AddSub/categories.yaml b/internal/simdgen/ops/AddSub/categories.yaml
index 1d08a94..8dca0e2 100644
--- a/internal/simdgen/ops/AddSub/categories.yaml
+++ b/internal/simdgen/ops/AddSub/categories.yaml
@@ -2,44 +2,76 @@
 - go: Add
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Add adds corresponding elements of two vectors.
 - go: SaturatedAdd
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    SaturatedAdd adds corresponding elements of two vectors with saturation.
 - go: MaskedAdd
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedAdd adds corresponding elements of two vectors, masked.
+  docUnmasked: !string |-
+    Add adds corresponding elements of two vectors.
 - go: MaskedSaturatedAdd
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedSaturatedAdd adds corresponding elements of two vectors with saturation, masked.
+  docUnmasked: !string |-
+    SaturatedAdd adds corresponding elements of two vectors with saturation.
 - go: Sub
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    Sub subtracts corresponding elements of two vectors.
 - go: SaturatedSub
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    SaturatedSub subtracts corresponding elements of two vectors with saturation.
 - go: MaskedSub
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedSub subtracts corresponding elements of two vectors, masked.
+  docUnmasked: !string |-
+    Sub subtracts corresponding elements of two vectors.
 - go: MaskedSaturatedSub
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedSaturatedSub subtracts corresponding elements of two vectors with saturation, masked.
+  docUnmasked: !string |-
+    SaturatedSub subtracts corresponding elements of two vectors with saturation.
 - go: PairwiseAdd
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Add pairs of elements in vector x and store them in higher half of the target; Add pairs of elements in vector y and store them in lower half of the target"
+  documentation: !string |-
+    PairwiseAdd horizontally adds adjacent pairs of elements.
+    For x = [x0, x1, x2, x3] and y = [y0, y1, y2, y3], the result is [y0+y1, y2+y3, x0+x1, x2+x3].
 - go: PairwiseSub
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Sub pairs of elements in vector x and store them in higher half of the target; Sub pairs of elements in vector y and store them in lower half of the target"
+  documentation: !string |-
+    PairwiseSub horizontally subtracts adjacent pairs of elements.
+    For x = [x0, x1, x2, x3] and y = [y0, y1, y2, y3], the result is [y0-y1, y2-y3, x0-x1, x2-x3].
 - go: SaturatedPairwiseAdd
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Add pairs of elements in vector x and store them in higher half of the target; Add pairs of elements in vector y and store them in lower half of the target; With saturation"
+  documentation: !string |-
+    SaturatedPairwiseAdd horizontally adds adjacent pairs of elements with saturation.
+    For x = [x0, x1, x2, x3] and y = [y0, y1, y2, y3], the result is [y0+y1, y2+y3, x0+x1, x2+x3].
 - go: SaturatedPairwiseSub
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Sub pairs of elements in vector x and store them in higher half of the target; Sub pairs of elements in vector y and store them in lower half of the target; With saturation"
\ No newline at end of file
+  documentation: !string |-
+    SaturatedPairwiseSub horizontally subtracts adjacent pairs of elements with saturation.
+    For x = [x0, x1, x2, x3] and y = [y0, y1, y2, y3], the result is [y0-y1, y2-y3, x0-x1, x2-x3].
\ No newline at end of file
diff --git a/internal/simdgen/ops/BitwiseLogic/categories.yaml b/internal/simdgen/ops/BitwiseLogic/categories.yaml
index bc4eda7..677db09 100644
--- a/internal/simdgen/ops/BitwiseLogic/categories.yaml
+++ b/internal/simdgen/ops/BitwiseLogic/categories.yaml
@@ -2,30 +2,54 @@
 - go: And
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    And performs a bitwise AND operation between two vectors.
 - go: MaskedAnd
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedAnd performs a masked bitwise AND operation between two vectors.
+  docUnmasked: !string |-
+    And performs a bitwise AND operation between two vectors.
 - go: Or
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Or performs a bitwise OR operation between two vectors.
 - go: MaskedOr
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedOr performs a masked bitwise OR operation between two vectors.
+  docUnmasked: !string |-
+    Or performs a bitwise OR operation between two vectors.
 - go: AndNot
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    AndNot performs a bitwise AND NOT operation between two vectors.
 - go: MaskedAndNot
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedAndNot performs a masked bitwise AND NOT operation between two vectors.
+  docUnmasked: !string |-
+    AndNot performs a bitwise AND NOT operation between two vectors.
 - go: Xor
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Xor performs a bitwise XOR operation between two vectors.
 - go: MaskedXor
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedXor performs a masked bitwise XOR operation between two vectors.
+  docUnmasked: !string |-
+    Xor performs a bitwise XOR operation between two vectors.
 # We also have PTEST and VPTERNLOG, those should be hidden from the users
 # and only appear in rewrite rules.
\ No newline at end of file
diff --git a/internal/simdgen/ops/Compares/categories.yaml b/internal/simdgen/ops/Compares/categories.yaml
index 027c8e8..1e01e36 100644
--- a/internal/simdgen/ops/Compares/categories.yaml
+++ b/internal/simdgen/ops/Compares/categories.yaml
@@ -10,77 +10,126 @@
   constImm: 0
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 0 if it has;"
+  documentation: !string |-
+    Equal compares for equality.
+    Const Immediate = 0.
 - go: Less
   constImm: 1
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 1 if it has;"
+  documentation: !string |-
+    Less compares for less than.
+    Const Immediate = 1.
 - go: LessEqual
   constImm: 2
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 2 if it has;"
+  documentation: !string |-
+    LessEqual compares for less than or equal.
+    Const Immediate = 2.
 - go: IsNan # For float only.
   constImm: 3
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 3 if it has; Returns mask element True if either one of the input\\'s element is Nan; Please use this method as x\\.IsNan\\(x\\) to check x only;"
+  documentation: !string |-
+    IsNan checks if elements are NaN. Use as x.IsNan(x).
+    Const Immediate = 3.
 - go: NotEqual
   constImm: 4
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 4 if it has;"
+  documentation: !string |-
+    NotEqual compares for inequality.
+    Const Immediate = 4.
 - go: GreaterEqual
   constImm: 5
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 5 if it has;"
+  documentation: !string |-
+    GreaterEqual compares for greater than or equal.
+    Const Immediate = 5.
 - go: Greater
   constImm: 6
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 6 if it has;"
+  documentation: !string |-
+    Greater compares for greater than.
+    Const Immediate = 6.
 
 - go: MaskedEqual
   constImm: 0
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 0 if it has;"
+  documentation: !string |-
+    MaskedEqual compares for equality, masked.
+    Const Immediate = 0.
+  docUnmasked: !string |-
+    Equal compares for equality.
+    Const Immediate = 0.
 - go: MaskedLess
   constImm: 1
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 1 if it has;"
+  documentation: !string |-
+    MaskedLess compares for less than, masked.
+    Const Immediate = 1.
+  docUnmasked: !string |-
+    Less compares for less than.
+    Const Immediate = 1.
 - go: MaskedLessEqual
   constImm: 2
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 2 if it has;"
+  documentation: !string |-
+    MaskedLessEqual compares for less than or equal, masked.
+    Const Immediate = 2.
+  docUnmasked: !string |-
+    LessEqual compares for less than or equal.
+    Const Immediate = 2.
 - go: MaskedIsNan # For float only.
   constImm: 3
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 3 if it has; Returns mask element True if either one of the input\\'s element is Nan; Please use this method as x\\.IsNan\\(x\\) to check x only;"
+  documentation: !string |-
+    MaskedIsNan checks if elements are NaN, masked. Use as x.IsNan(x).
+    Const Immediate = 3.
+  docUnmasked: !string |-
+    IsNan checks if elements are NaN. Use as x.IsNan(x).
+    Const Immediate = 3.
 - go: MaskedNotEqual
   constImm: 4
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 4 if it has;"
+  documentation: !string |-
+    MaskedNotEqual compares for inequality, masked.
+    Const Immediate = 4.
+  docUnmasked: !string |-
+    NotEqual compares for inequality.
+    Const Immediate = 4.
 - go: MaskedGreaterEqual
   constImm: 5
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 5 if it has;"
+  documentation: !string |-
+    MaskedGreaterEqual compares for greater than or equal, masked.
+    Const Immediate = 5.
+  docUnmasked: !string |-
+    GreaterEqual compares for greater than or equal.
+    Const Immediate = 5.
 - go: MaskedGreater
   constImm: 6
   masked: "true"
   commutative: "false"
   extension: "AVX.*"
-  documentation: "Predicate immediate is 6 if it has;"
\ No newline at end of file
+  documentation: !string |-
+    MaskedGreater compares for greater than, masked.
+    Const Immediate = 6.
+  docUnmasked: !string |-
+    Greater compares for greater than.
+    Const Immediate = 6.
\ No newline at end of file
diff --git a/internal/simdgen/ops/FPonlyArith/categories.yaml b/internal/simdgen/ops/FPonlyArith/categories.yaml
index e486225..3f60e84 100644
--- a/internal/simdgen/ops/FPonlyArith/categories.yaml
+++ b/internal/simdgen/ops/FPonlyArith/categories.yaml
@@ -2,136 +2,274 @@
 - go: Div
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    Div divides elements of two vectors.
 - go: MaskedDiv
   commutative: "false"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedDiv divides elements of two vectors, masked.
+  docUnmasked: !string |-
+    Div divides elements of two vectors.
 - go: Sqrt
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    Sqrt computes the square root of each element.
 - go: MaskedSqrt
   commutative: "false"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedSqrt computes the square root of each element, masked.
+  docUnmasked: !string |-
+    Sqrt computes the square root of each element.
 - go: ApproximateReciprocal
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    ApproximateReciprocal computes an approximate reciprocal of each element.
 - go: MaskedApproximateReciprocal
   commutative: "false"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedApproximateReciprocal computes an approximate reciprocal of each element, masked.
+  docUnmasked: !string |-
+    ApproximateReciprocal computes an approximate reciprocal of each element.
 - go: ApproximateReciprocalOfSqrt
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    ApproximateReciprocalOfSqrt computes an approximate reciprocal of the square root of each element.
 - go: MaskedApproximateReciprocalOfSqrt
   commutative: "false"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedApproximateReciprocalOfSqrt computes an approximate reciprocal of the square root of each element, masked.
+  docUnmasked: !string |-
+    ApproximateReciprocalOfSqrt computes an approximate reciprocal of the square root of each element.
 - go: MaskedMulByPowOf2 # This operation is all after AVX512, the unmasked version will be generated.
   commutative: "false"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedMulByPowOf2 multiplies elements by a power of 2, masked.
+  docUnmasked: !string |-
+    MulByPowOf2 multiplies elements by a power of 2.
 
 - go: Round
   commutative: "false"
   extension: "AVX.*"
   constImm: 0
+  documentation: !string |-
+    Round rounds elements to the nearest integer.
+    Const Immediate = 0.
 - go: MaskedRoundWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 0
   masked: "true"
+  documentation: !string |-
+    MaskedRoundWithPrecision rounds elements with specified precision, masked.
+    Const Immediate = 0.
+  docUnmasked: !string |-
+    RoundWithPrecision rounds elements with specified precision.
+    Const Immediate = 0.
 - go: MaskedRoundSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 8
   masked: "true"
+  documentation: !string |-
+    MaskedRoundSuppressExceptionWithPrecision rounds elements with specified precision, suppressing exceptions, masked.
+    Const Immediate = 8.
+  docUnmasked: !string |-
+    RoundSuppressExceptionWithPrecision rounds elements with specified precision, suppressing exceptions.
+    Const Immediate = 8.
 - go: MaskedDiffWithRoundWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 0
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithRoundWithPrecision computes the difference after rounding with specified precision, masked.
+    Const Immediate = 0.
+  docUnmasked: !string |-
+    DiffWithRoundWithPrecision computes the difference after rounding with specified precision.
+    Const Immediate = 0.
 - go: MaskedDiffWithRoundSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 8
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithRoundSuppressExceptionWithPrecision computes the difference after rounding with specified precision, suppressing exceptions, masked.
+    Const Immediate = 8.
+  docUnmasked: !string |-
+    DiffWithRoundSuppressExceptionWithPrecision computes the difference after rounding with specified precision, suppressing exceptions.
+    Const Immediate = 8.
 
 - go: Floor
   commutative: "false"
   extension: "AVX.*"
   constImm: 1
+  documentation: !string |-
+    Floor rounds elements down to the nearest integer.
+    Const Immediate = 1.
 - go: MaskedFloorWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 1
   masked: "true"
+  documentation: !string |-
+    MaskedFloorWithPrecision rounds elements down with specified precision, masked.
+    Const Immediate = 1.
+  docUnmasked: !string |-
+    FloorWithPrecision rounds elements down with specified precision.
+    Const Immediate = 1.
 - go: MaskedFloorSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 9
   masked: "true"
+  documentation: !string |-
+    MaskedFloorSuppressExceptionWithPrecision rounds elements down with specified precision, suppressing exceptions, masked.
+    Const Immediate = 9.
+  docUnmasked: !string |-
+    FloorSuppressExceptionWithPrecision rounds elements down with specified precision, suppressing exceptions.
+    Const Immediate = 9.
 - go: MaskedDiffWithFloorWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 1
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithFloorWithPrecision computes the difference after flooring with specified precision, masked.
+    Const Immediate = 1.
+  docUnmasked: !string |-
+    DiffWithFloorWithPrecision computes the difference after flooring with specified precision.
+    Const Immediate = 1.
 - go: MaskedDiffWithFloorSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 9
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithFloorSuppressExceptionWithPrecision computes the difference after flooring with specified precision, suppressing exceptions, masked.
+    Const Immediate = 9.
+  docUnmasked: !string |-
+    DiffWithFloorSuppressExceptionWithPrecision computes the difference after flooring with specified precision, suppressing exceptions.
+    Const Immediate = 9.
 
 - go: Ceil
   commutative: "false"
   extension: "AVX.*"
   constImm: 2
+  documentation: !string |-
+    Ceil rounds elements up to the nearest integer.
+    Const Immediate = 2.
 - go: MaskedCeilWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 2
   masked: "true"
+  documentation: !string |-
+    MaskedCeilWithPrecision rounds elements up with specified precision, masked.
+    Const Immediate = 2.
+  docUnmasked: !string |-
+    CeilWithPrecision rounds elements up with specified precision.
+    Const Immediate = 2.
 - go: MaskedCeilSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 10
   masked: "true"
+  documentation: !string |-
+    MaskedCeilSuppressExceptionWithPrecision rounds elements up with specified precision, suppressing exceptions, masked.
+    Const Immediate = 10.
+  docUnmasked: !string |-
+    CeilSuppressExceptionWithPrecision rounds elements up with specified precision, suppressing exceptions.
+    Const Immediate = 10.
 - go: MaskedDiffWithCeilWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 2
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithCeilWithPrecision computes the difference after ceiling with specified precision, masked.
+    Const Immediate = 2.
+  docUnmasked: !string |-
+    DiffWithCeilWithPrecision computes the difference after ceiling with specified precision.
+    Const Immediate = 2.
 - go: MaskedDiffWithCeilSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 10
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithCeilSuppressExceptionWithPrecision computes the difference after ceiling with specified precision, suppressing exceptions, masked.
+    Const Immediate = 10.
+  docUnmasked: !string |-
+    DiffWithCeilSuppressExceptionWithPrecision computes the difference after ceiling with specified precision, suppressing exceptions.
+    Const Immediate = 10.
 
 - go: Trunc
   commutative: "false"
   extension: "AVX.*"
   constImm: 3
+  documentation: !string |-
+    Trunc truncates elements towards zero.
+    Const Immediate = 3.
 - go: MaskedTruncWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 3
   masked: "true"
+  documentation: !string |-
+    MaskedTruncWithPrecision truncates elements with specified precision, masked.
+    Const Immediate = 3.
+  docUnmasked: !string |-
+    TruncWithPrecision truncates elements with specified precision.
+    Const Immediate = 3.
 - go: MaskedTruncSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 11
   masked: "true"
+  documentation: !string |-
+    MaskedTruncSuppressExceptionWithPrecision truncates elements with specified precision, suppressing exceptions, masked.
+    Const Immediate = 11.
+  docUnmasked: !string |-
+    TruncSuppressExceptionWithPrecision truncates elements with specified precision, suppressing exceptions.
+    Const Immediate = 11.
 - go: MaskedDiffWithTruncWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 3
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithTruncWithPrecision computes the difference after truncating with specified precision, masked.
+    Const Immediate = 3.
+  docUnmasked: !string |-
+    DiffWithTruncWithPrecision computes the difference after truncating with specified precision.
+    Const Immediate = 3.
 - go: MaskedDiffWithTruncSuppressExceptionWithPrecision
   commutative: "false"
   extension: "AVX.*"
   constImm: 11
   masked: "true"
+  documentation: !string |-
+    MaskedDiffWithTruncSuppressExceptionWithPrecision computes the difference after truncating with specified precision, suppressing exceptions, masked.
+    Const Immediate = 11.
+  docUnmasked: !string |-
+    DiffWithTruncSuppressExceptionWithPrecision computes the difference after truncating with specified precision, suppressing exceptions.
+    Const Immediate = 11.
 
 - go: AddSub
   commutative: "false"
-  extension: "AVX.*"
\ No newline at end of file
+  extension: "AVX.*"
+  documentation: !string |-
+    AddSub alternatingly adds and subtracts elements of two vectors.
\ No newline at end of file
diff --git a/internal/simdgen/ops/IntOnlyArith/categories.yaml b/internal/simdgen/ops/IntOnlyArith/categories.yaml
index c74b57c..01a55cc 100644
--- a/internal/simdgen/ops/IntOnlyArith/categories.yaml
+++ b/internal/simdgen/ops/IntOnlyArith/categories.yaml
@@ -2,27 +2,45 @@
 - go: Average
   commutative: "true"
   extension: "AVX.*" # VPAVGB/W are available across various AVX versions
+  documentation: !string |-
+    Average computes the rounded average of corresponding elements.
 - go: MaskedAverage
   commutative: "true"
   masked: "true"
   extension: "AVX512.*" # Masked operations are typically AVX512
+  documentation: !string |-
+    MaskedAverage computes the rounded average of corresponding elements, masked.
+  docUnmasked: !string |-
+    Average computes the rounded average of corresponding elements.
 
 - go: Absolute
   commutative: "false"
   # Unary operation, not commutative
   extension: "AVX.*" # VPABSB/W/D are AVX, VPABSQ is AVX512
+  documentation: !string |-
+    Absolute computes the absolute value of each element.
 - go: MaskedAbsolute
   commutative: "false"
   masked: "true"
   extension: "AVX512.*"
+  documentation: !string |-
+    MaskedAbsolute computes the absolute value of each element, masked.
+  docUnmasked: !string |-
+    Absolute computes the absolute value of each element.
 
 - go: Sign
   # Applies sign of second operand to first: sign(val, sign_src)
   commutative: "false"
   extension: "AVX.*"
+  documentation: !string |-
+    Sign applies the sign of the second operand to the first.
   # Sign does not have masked version
 
 - go: MaskedPopCount
   commutative: "false"
   masked: "true"
-  extension: "AVX512.*" # VPOPCNT instructions are AVX512 (BITALG or VPOPCNTDQ)
\ No newline at end of file
+  extension: "AVX512.*" # VPOPCNT instructions are AVX512 (BITALG or VPOPCNTDQ)
+  documentation: !string |-
+    MaskedPopCount counts the number of set bits in each element, masked.
+  docUnmasked: !string |-
+    PopCount counts the number of set bits in each element.
\ No newline at end of file
diff --git a/internal/simdgen/ops/MLOps/categories.yaml b/internal/simdgen/ops/MLOps/categories.yaml
index 01eaf0d..075f836 100644
--- a/internal/simdgen/ops/MLOps/categories.yaml
+++ b/internal/simdgen/ops/MLOps/categories.yaml
@@ -2,24 +2,34 @@
 - go: PairDotProd
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiply the elements and add the pairs together"
+  documentation: !string |-
+    PairDotProd multiplies and adds adjacent pairs of elements.
 - go: MaskedPairDotProd
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiply the elements and add the pairs together"
+  documentation: !string |-
+    MaskedPairDotProd multiplies and adds adjacent pairs of elements, masked.
+  docUnmasked: !string |-
+    PairDotProd multiplies and adds adjacent pairs of elements.
 - go: SaturatedPairDotProd
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiply the elements and add the pairs together with saturation"
+  documentation: !string |-
+    SaturatedPairDotProd multiplies and adds adjacent pairs of elements with saturation.
 - go: MaskedSaturatedPairDotProd
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiply the elements and add the pairs together with saturation"
+  documentation: !string |-
+    MaskedSaturatedPairDotProd multiplies and adds adjacent pairs of elements with saturation, masked.
+  docUnmasked: !string |-
+    SaturatedPairDotProd multiplies and adds adjacent pairs of elements with saturation.
 
 # QuadDotProd, i.e. VPDPBUSD(S) are operations with src/dst on the same register, we are not supporting this as of now.
 - go: DotProdBroadcast
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiply all the elements and add them together; the result is a broadcast of the dot product; imm8 = 127;"
+  documentation: !string |-
+    DotProdBroadcast multiplies all elements and broadcasts the sum.
+    Const Immediate = 127.
diff --git a/internal/simdgen/ops/MinMax/categories.yaml b/internal/simdgen/ops/MinMax/categories.yaml
index d513195..b61a85d 100644
--- a/internal/simdgen/ops/MinMax/categories.yaml
+++ b/internal/simdgen/ops/MinMax/categories.yaml
@@ -2,14 +2,26 @@
 - go: Max
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Max computes the maximum of corresponding elements.
 - go: MaskedMax
   commutative: "true"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedMax computes the maximum of corresponding elements, masked.
+  docUnmasked: !string |-
+    Max computes the maximum of corresponding elements.
 - go: Min
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Min computes the minimum of corresponding elements.
 - go: MaskedMin
   commutative: "true"
   masked: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedMin computes the minimum of corresponding elements, masked.
+  docUnmasked: !string |-
+    Min computes the minimum of corresponding elements.
diff --git a/internal/simdgen/ops/Mul/categories.yaml b/internal/simdgen/ops/Mul/categories.yaml
index 0ef6cf5..0fe6a1d 100644
--- a/internal/simdgen/ops/Mul/categories.yaml
+++ b/internal/simdgen/ops/Mul/categories.yaml
@@ -2,34 +2,55 @@
 - go: Mul
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    Mul multiplies corresponding elements of two vectors.
 - go: MulEvenWiden
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the even index elements from the two sources of size X at index i, store the result of size 2X at index i/2"
+  documentation: !string |-
+    MulEvenWiden multiplies even-indexed elements, widening the result.
+    Result[i] = v1.Even[i] * v2.Even[i].
 - go: MulHigh
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the elements from the two sources of size X at index i, store the high X bits of the result of size 2X at index i"
+  documentation: !string |-
+    MulHigh multiplies elements and stores the high part of the result.
 - go: MulLow
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the elements from the two sources of size X at index i, store the low X bits of the result of size 2X at index i"
+  documentation: !string |-
+    MulLow multiplies elements and stores the low part of the result.
 - go: MaskedMul
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
+  documentation: !string |-
+    MaskedMul multiplies corresponding elements of two vectors, masked.
+  docUnmasked: !string |-
+    Mul multiplies corresponding elements of two vectors.
 - go: MaskedMulEvenWiden
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the even index elements from the two sources of size X at index i, store the result of size 2X at index i/2"
+  documentation: !string |-
+    MaskedMulEvenWiden multiplies even-indexed elements, widening the result, masked.
+    Result[i] = v1.Even[i] * v2.Even[i].
+  docUnmasked: !string |-
+    MulEvenWiden multiplies even-indexed elements, widening the result.
+    Result[i] = v1.Even[i] * v2.Even[i].
 - go: MaskedMulHigh
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the elements from the two sources of size X at index i, store the high X bits of the result of size 2X at index i"
+  documentation: !string |-
+    MaskedMulHigh multiplies elements and stores the high part of the result, masked.
+  docUnmasked: !string |-
+    MulHigh multiplies elements and stores the high part of the result.
 - go: MaskedMulLow
   masked: "true"
   commutative: "true"
   extension: "AVX.*"
-  documentation: "Multiplies the elements from the two sources of size X at index i, store the low X bits of the result of size 2X at index i"
\ No newline at end of file
+  documentation: !string |-
+    MaskedMulLow multiplies elements and stores the low part of the result, masked.
+  docUnmasked: !string |-
+    MulLow multiplies elements and stores the low part of the result.
\ No newline at end of file

• M internal/simdgen/categories.yaml
• M internal/simdgen/gen_simdTypes.go
• M internal/simdgen/gen_utility.go
• M internal/simdgen/godefs.go
• M internal/simdgen/ops/AddSub/categories.yaml
• M internal/simdgen/ops/BitwiseLogic/categories.yaml
• M internal/simdgen/ops/Compares/categories.yaml
• M internal/simdgen/ops/FPonlyArith/categories.yaml
• M internal/simdgen/ops/IntOnlyArith/categories.yaml
• M internal/simdgen/ops/MLOps/categories.yaml
• M internal/simdgen/ops/MinMax/categories.yaml
• M internal/simdgen/ops/Mul/categories.yaml

• Code-Review
• No-Unresolved-Comments
• Review-Enforcement
• TryBots-Pass

• David Chase