small update to simd types

src/simd/simdX86Ops.nim

@@ -17,22 +17,22 @@ template int2mask*(T: typedesc[m512], i: SomeInteger): mmask16 = cvtu32_mask16(u
 proc `[]=`*(r:var m128; i:SomeInteger; x:SomeNumber) {.alwaysInline.} =
   mixin toArray
   var a = toArray(r)
-  a[i] := x
+  a[i] = float32 x
   assign(r, a)
 proc `[]=`*(r:var m128d; i:SomeInteger; x:SomeNumber) {.alwaysInline.} =
   mixin toArray
   var a = toArray(r)
-  a[i] := x
+  a[i] = float x
   assign(r, a)
 proc `[]=`*(r:var m256; i:SomeInteger; x:SomeNumber) {.alwaysInline.} =
   mixin toArray
   var a = toArray(r)
-  a[i] := x
+  a[i] = float32 x
   assign(r, a)
 proc `[]=`*(r:var m256d; i:SomeInteger; x:SomeNumber) {.alwaysInline.} =
   mixin toArray
   var a = toArray(r)
-  a[i] := x
+  a[i] = float x
   assign(r, a)
 #proc `[]=`*(r:var m256d; i:SomeInteger; x:SomeNumber) {.alwaysInline.} =
 #  var a {.noInit.}: m256d

src/simd/simdX86Types.nim

@@ -2,25 +2,52 @@ import simdWrap
 export simdWrap
 
 {.pragma: imm, header:"immintrin.h".}
+{.pragma: imms, header:"immintrin.h", incompleteStruct.}
+type
+  m64*   {.importc: "__m64"  , imms.} = object
+  m128*  {.importc: "__m128" , imms.} = object
+  m128d* {.importc: "__m128d", imms.} = object
+  m128i* {.importc: "__m128i", imms.} = object
+  m128h* = distinct int64
+  m256*  {.importc: "__m256" , imms.} = object
+  m256d* {.importc: "__m256d", imms.} = object
+  m256i* {.importc: "__m256i", imms.} = object
+  m256h* = distinct m128i
+  m512*  {.importc: "__m512" , imms.} = object
+  m512d* {.importc: "__m512d", imms.} = object
+  m512i* {.importc: "__m512i", imms.} = object
+  m512h* = distinct m256i
+  mmask8*  {.importc: "__mmask8" , imms.} = object
+  mmask16* {.importc: "__mmask16", imms.} = object
+  mmask32* {.importc: "__mmask32", imms.} = object
+  mmask64* {.importc: "__mmask64", imms.} = object
+#[
+{.pragma: imm, header:"immintrin.h", incompleteStruct.}
 type
   m64*   {.importc: "__m64"  , imm.} = object
-  m128*  {.importc: "__m128" , imm.} = object
-  m128d* {.importc: "__m128d", imm.} = object
+    a: array[2,float32]
+  m128*  {.importc: "__m128" , imm.} = distinct array[4,float32]
+  m128d* {.importc: "__m128d", imm.} = distinct array[2,float64]
   m128i* {.importc: "__m128i", imm.} = object
+    a: array[4,int32]
   m128h* = distinct int64
   m256*  {.importc: "__m256" , imm.} = object
-  m256d* {.importc: "__m256d", imm.} = object
+    a: array[8,float32]
+  m256d* {.importc: "__m256d", imm.} = distinct array[4,float64]
   m256i* {.importc: "__m256i", imm.} = object
+    a: array[8,int32]
   m256h* = distinct m128i
-  m512*  {.importc: "__m512" , imm.} = object
-  m512d* {.importc: "__m512d", imm.} = object
-  m512i* {.importc: "__m512i", imm.} = object
+  m512*  {.importc: "__m512" , imm.} = distinct array[16,float32]
+  m512d* {.importc: "__m512d", imm.} = distinct array[8,float64]
+  m512i* {.importc: "__m512i", imm.} = distinct array[16,int32]
   m512h* = distinct m256i
   mmask8*  {.importc: "__mmask8" , imm.} = object
   mmask16* {.importc: "__mmask16", imm.} = object
   mmask32* {.importc: "__mmask32", imm.} = object
   mmask64* {.importc: "__mmask64", imm.} = object
+]#
 
+type
   SimdX86S* = m64 | m128 | m256 | m512
   SimdX86D* = m128d | m256d | m512d
   SimdX86* = SimdX86S | SimdX86D