internal/simdgen: (Set|Get)(Lo|Hi)

This CL adds the missing pieces of set/get elements for larger vectors.
It also changes the Set and Get API to be better.

This CL generates CL 693355.

Change-Id: If545221e87776de7946205b41f9a7648a8148b2d
Reviewed-on: https://go-review.googlesource.com/c/arch/+/693335
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: David Chase <drchase@google.com>

internal/simdgen/ops/Moves/categories.yaml

diff --git a/internal/simdgen/ops/Moves/categories.yaml b/internal/simdgen/ops/Moves/categories.yaml
index 5e51bec..d56e4c9 100644
--- a/internal/simdgen/ops/Moves/categories.yaml
+++ b/internal/simdgen/ops/Moves/categories.yaml
@@ -9,16 +9,30 @@
   extension: "AVX.*"
   documentation: !string |-
     // NAME retrieves a single constant-indexed element's value.
-- go: Set128
+- go: SetLo
   commutative: false
+  constImm: 0
   extension: "AVX.*"
   documentation: !string |-
-    // NAME combines a 128-bit vector with a 256-bit vector, where the constant operand specifies whether the low (0) or high (1) half is receives the smaller vector.
-- go: Get128
+    // NAME returns x with its lower half set to y.
+- go: GetLo
   commutative: false
+  constImm: 0
   extension: "AVX.*"
   documentation: !string |-
-    // NAME retrieves the upper (1) or lower (0) half of a 256-bit vector, depending on the constant operand.
+    // NAME returns the lower half of x.
+- go: SetHi
+  commutative: false
+  constImm: 1
+  extension: "AVX.*"
+  documentation: !string |-
+    // NAME returns x with its upper half set to y.
+- go: GetHi
+  commutative: false
+  constImm: 1
+  extension: "AVX.*"
+  documentation: !string |-
+    // NAME returns the upper half of x.
 - go: Permute
   commutative: false
   extension: "AVX.*"

internal/simdgen/ops/Moves/go.yaml

diff --git a/internal/simdgen/ops/Moves/go.yaml b/internal/simdgen/ops/Moves/go.yaml
index 52e6228..b014a7a 100644
--- a/internal/simdgen/ops/Moves/go.yaml
+++ b/internal/simdgen/ops/Moves/go.yaml
@@ -27,156 +27,158 @@
     base: $b
     bits: $e
 
-- go: Set128
-  asm: "VINSERTI128"
+- go: "SetHi|SetLo"
+  asm: "VINSERTI128|VINSERTI64X4"
+  inVariant: []
   in:
-  - &i8x32
+  - &i8x2N
     class: vreg
     base: $t
-    bits: 256
     OverwriteElementBits: 8
-  - &i8x16
+  - &i8xN
     class: vreg
     base: $t
-    bits: 128
     OverwriteElementBits: 8
   - &imm01 # This immediate should be only 0 or 1
     class: immediate
-    immOffset: 0
+    const: 0 # place holder
     name: index
   out:
-  - *i8x32
+  - *i8x2N
 
-- go: Get128
-  asm: "VEXTRACTI128"
+- go: "GetHi|GetLo"
+  asm: "VEXTRACTI128|VEXTRACTI64X4"
+  inVariant: []
   in:
-  - *i8x32
+  - *i8x2N
   - *imm01
   out:
-  - *i8x16
+  - *i8xN
 
-- go: Set128
-  asm: "VINSERTI128"
+- go: "SetHi|SetLo"
+  asm: "VINSERTI128|VINSERTI64X4"
+  inVariant: []
   in:
-  - &i16x16
+  - &i16x2N
     class: vreg
     base: $t
-    bits: 256
     OverwriteElementBits: 16
-  - &i16x8
+  - &i16xN
     class: vreg
     base: $t
-    bits: 128
     OverwriteElementBits: 16
   - *imm01
   out:
-  - *i16x16
+  - *i16x2N
 
-- go: Get128
-  asm: "VEXTRACTI128"
+- go: "GetHi|GetLo"
+  asm: "VEXTRACTI128|VEXTRACTI64X4"
+  inVariant: []
   in:
-  - *i16x16
+  - *i16x2N
   - *imm01
   out:
-  - *i16x8
+  - *i16xN
 
-- go: Set128
-  asm: "VINSERTI128"
+- go: "SetHi|SetLo"
+  asm: "VINSERTI128|VINSERTI64X4"
+  inVariant: []
   in:
-  - &i32x8
+  - &i32x2N
     class: vreg
     base: $t
-    bits: 256
     OverwriteElementBits: 32
-  - &i32x4
+  - &i32xN
     class: vreg
     base: $t
-    bits: 128
     OverwriteElementBits: 32
   - *imm01
   out:
-  - *i32x8
+  - *i32x2N
 
-- go: Get128
-  asm: "VEXTRACTI128"
+- go: "GetHi|GetLo"
+  asm: "VEXTRACTI128|VEXTRACTI64X4"
+  inVariant: []
   in:
-  - *i32x8
+  - *i32x2N
   - *imm01
   out:
-  - *i32x4
+  - *i32xN
 
-- go: Set128
-  asm: "VINSERTI128"
+- go: "SetHi|SetLo"
+  asm: "VINSERTI128|VINSERTI64X4"
+  inVariant: []
   in:
-  - &i64x4
+  - &i64x2N
     class: vreg
     base: $t
-    bits: 256
     OverwriteElementBits: 64
-  - &i64x2
+  - &i64xN
     class: vreg
     base: $t
-    bits: 128
     OverwriteElementBits: 64
   - *imm01
   out:
-  - *i64x4
+  - *i64x2N
 
-- go: Get128
-  asm: "VEXTRACTI128"
+- go: "GetHi|GetLo"
+  asm: "VEXTRACTI128|VEXTRACTI64X4"
+  inVariant: []
   in:
-  - *i64x4
+  - *i64x2N
   - *imm01
   out:
-  - *i64x2
+  - *i64xN
 
-- go: Set128
-  asm: "VINSERTF128"
+- go: "SetHi|SetLo"
+  asm: "VINSERTF128|VINSERTF64X4"
+  inVariant: []
   in:
-  - &f32x8
+  - &f32x2N
     class: vreg
     base: $t
-    bits: 256
     OverwriteElementBits: 32
-  - &f32x4
+  - &f32xN
     class: vreg
     base: $t
-    bits: 128
     OverwriteElementBits: 32
   - *imm01
   out:
-  - *f32x8
+  - *f32x2N
 
-- go: Get128
-  asm: "VEXTRACTF128"
+- go: "GetHi|GetLo"
+  asm: "VEXTRACTF128|VEXTRACTF64X4"
+  inVariant: []
   in:
-  - *f32x8
+  - *f32x2N
   - *imm01
   out:
-  - *f32x4
+  - *f32xN
 
-- go: Set128
-  asm: "VINSERTF128"
+- go: "SetHi|SetLo"
+  asm: "VINSERTF128|VINSERTF64X4"
+  inVariant: []
   in:
-  - &f64x4
+  - &f64x2N
     class: vreg
     base: $t
-    bits: 256
-  - &f64x2
+    OverwriteElementBits: 64
+  - &f64xN
     class: vreg
     base: $t
-    bits: 128
+    OverwriteElementBits: 64
   - *imm01
   out:
-  - *f64x4
+  - *f64x2N
 
-- go: Get128
-  asm: "VEXTRACTF128"
+- go: "GetHi|GetLo"
+  asm: "VEXTRACTF128|VEXTRACTF64X4"
+  inVariant: []
   in:
-  - *f64x4
+  - *f64x2N
   - *imm01
   out:
-  - *f64x2
+  - *f64xN
 
 - go: Permute
   asm: "VPERM[BWDQ]|VPERMP[SD]"