[spectext] Add ext mul instructions

Author: ngzhian

document/core/exec/instructions.rst

@@ -725,6 +725,57 @@ SIMD instructions are defined in terms of generic numeric operators applied lane
    \end{array}
 
 
+.. _exec-simd-extmul:
+
+:math:`t_2\K{x}N\K{.}\EXTMUL\_\K{low}\_t_1\K{x}M\_\sx` and :math:`t_2\K{x}N\K{.}\EXTMUL\_\K{high}\_t_1\K{x}M\_\sx`
+..................................................................................................................
+
+1. Assert: due to :ref:`validation <valid-vitestop>`, two values of :ref:`value type <syntax-valtype>` |V128| is on the top of the stack.
+
+2. Pop the value :math:`\V128.\VCONST~c_2` from the stack.
+
+3. Pop the value :math:`\V128.\VCONST~c_1` from the stack.
+
+4. If :math:`\K{low}` is part of the instruction, then:
+
+   a. Let :math:`i^\ast` be the sequence :math:`\lanes_{t_1\K{x}M}(c_1)[0 \slice N]`.
+
+   b. Let :math:`j^\ast` be the sequence :math:`\lanes_{t_1\K{x}M}(c_2)[0 \slice N]`.
+
+5. Else:
+
+   a. Let :math:`i^\ast` be the sequence :math:`\lanes_{t_1\K{x}M}(c_1)[N \slice N]`.
+
+   b. Let :math:`j^\ast` be the sequence :math:`\lanes_{t_1\K{x}M}(c_2)[N \slice N]`.
+
+6. Let :math:`c` be the result of computing :math:`\lanes^{-1}_{t_2\K{x}N}(\imul_{t_2\K{x}N}(\extend^{\sx}_{M,N}(i^\ast), \extend^{\sx}_{M,N}(j^\ast)))`
+
+8. Push the value :math:`\V128.\VCONST~c` onto the stack.
+
+.. math::
+   \begin{array}{l}
+   \begin{array}{lcl@{\qquad}l}
+   (\V128\K{.}\VCONST~c_1)~(\V128\K{.}\VCONST~c_2)~t_2\K{x}N\K{.}\EXTMUL\_\K{low}\_t_1\K{x}M\_\sx &\stepto& (\V128\K{.}\VCONST~c) \\
+   \end{array}
+   \\ \qquad
+     \begin{array}[t]{@{}r@{~}l@{}}
+     (\iff & i^\ast = \lanes_{t_1\K{x}M}(c_1)[0 \slice N] \\
+     \wedge & j^\ast = \lanes_{t_1\K{x}M}(c_2)[0 \slice N] \\
+     \wedge & c = \lanes^{-1}_{t_2\K{x}N}(\imul_{t_2\K{x}N}(\extend^{\sx}_{M,N}(i^\ast), \extend^{\sx}_{M,N}(j^\ast)))
+     \end{array}
+   \\[1ex]
+   \begin{array}{lcl@{\qquad}l}
+   (\V128\K{.}\VCONST~c_1)~(\V128\K{.}\VCONST~c_2)~t_2\K{x}N\K{.}\EXTMUL\_\K{high}\_t_1\K{x}M\_\sx &\stepto& (\V128\K{.}\VCONST~c) \\
+   \end{array}
+   \\ \qquad
+     \begin{array}[t]{@{}r@{~}l@{}}
+     (\iff & i^\ast = \lanes_{t_1\K{x}M}(c_1)[N \slice N] \\
+     \wedge & j^\ast = \lanes_{t_1\K{x}M}(c_2)[N \slice N] \\
+     \wedge & c = \lanes^{-1}_{t_2\K{x}N}(\imul_{t_2\K{x}N}(\extend^{\sx}_{M,N}(i^\ast), \extend^{\sx}_{M,N}(j^\ast)))
+     \end{array}
+   \end{array}
+
+
 .. index:: parametric instructions, value
    pair: execution; instruction
    single: abstract syntax; instruction

document/core/syntax/instructions.rst

@@ -256,6 +256,12 @@ SIMD instructions provide basic operations over :ref:`values <syntax-value>` of
      \K{i64x2.}\K{mul} \\&&|&
      \K{i8x16.}\AVGR\K{\_u} ~|~
      \K{i16x8.}\AVGR\K{\_u} \\&&|&
+     \K{i16x8.}\EXTMUL\K{\_low}\K{\_i8x16\_}\sx ~|~
+     \K{i16x8.}\EXTMUL\K{\_high}\K{\_i8x16\_}\sx \\&&|&
+     \K{i32x4.}\EXTMUL\K{\_low}\K{\_i16x8\_}\sx ~|~
+     \K{i32x4.}\EXTMUL\K{\_high}\K{\_i16x8\_}\sx \\&&|&
+     \K{i64x2.}\EXTMUL\K{\_low}\K{\_i32x4\_}\sx ~|~
+     \K{i64x2.}\EXTMUL\K{\_high}\K{\_i32x4\_}\sx \\&&|&
      \fshape\K{.}\vfbinop \\&&|&
      \K{i32x4.}\VTRUNC\K{\_sat\_f32x4\_}\sx \\ &&|&
      \K{f32x4.}\VCONVERT\K{\_i32x4\_}\sx \\&&|&
@@ -362,6 +368,7 @@ For the other SIMD instructions, the use of two's complement for the signed inte
 .. _syntax-vunop:
 .. _syntax-vbinop:
 .. _syntax-vwiden:
+.. _syntax-vextmul:
 
 Conventions
 ...........
@@ -387,6 +394,9 @@ Occasionally, it is convenient to group operators together according to the foll
    \production{widen operator} & \vwiden &::=&
      \WIDEN\K{\_low\_}\shape\K{\_}\sx ~|~
      \WIDEN\K{\_high\_}\shape\K{\_}\sx \\
+   \production{extmul operator} & \vextmul &::=&
+     \EXTMUL\K{\_low\_}\ishape\K{\_}\sx ~|~
+     \EXTMUL\K{\_high\_}\ishape\K{\_}\sx \\
    \end{array}
 
 

document/core/util/macros.def

@@ -422,6 +422,7 @@
 .. |NARROW| mathdef:: \xref{syntax/instructions}{syntax-instr-simd}{\K{narrow}}
 .. |WIDEN| mathdef:: \xref{syntax/instructions}{syntax-instr-simd}{\K{widen}}
 .. |AVGR| mathdef:: \xref{syntax/instructions}{syntax-instr-simd}{\K{avgr}}
+.. |EXTMUL| mathdef:: \xref{syntax/instructions}{syntax-instr-simd}{\K{extmul}}
 .. |VTRUNC| mathdef:: \xref{syntax/instructions}{syntax-instr-simd}{\K{trunc}}
 .. |VCONVERT| mathdef:: \xref{syntax/instructions}{syntax-instr-simd}{\K{convert}}
 
@@ -453,6 +454,7 @@
 .. |vternop| mathdef:: \xref{syntax/instructions}{syntax-vternop}{\X{vternop}}
 .. |vwiden| mathdef:: \xref{syntax/instructions}{syntax-vwiden}{\X{vwiden}}
 .. |vcvtop| mathdef:: \xref{syntax/instructions}{syntax-vcvtop}{\X{vcvtop}}
+.. |vextmul| mathdef:: \xref{syntax/instructions}{syntax-vextmul}{\X{vextmul}}
 
 .. |laneidx| mathdef:: \xref{syntax/instructions}{syntax-laneidx}{\X{laneidx}}
 .. |vsunop| mathdef:: \xref{syntax/instructions}{syntax-vsunop}{\X{vsunop}}

document/core/valid/instructions.rst

@@ -450,6 +450,20 @@ We also define an auxiliary function to get number of packed numeric types in a
    }
 
 
+.. _valid-simd-vextmul:
+
+:math:`\ishape\K{.}\vextmul\K{\_}\ishape\K{\_}\sx`
+..................................................
+
+* The instruction is valid with type :math:`[\V128~\V128] \to [\V128]`.
+
+.. math::
+   \frac{
+   }{
+     C \vdashinstr \ishape\K{.}\EXTMUL\K{\_}\ishape\K{\_}\sx : [\V128~\V128] \to [\V128]
+   }
+
+
 .. index:: parametric instructions, value type, polymorphism
    pair: validation; instruction
    single: abstract syntax; instruction