Merge branch 'master' into simd-cleanup

[factor.git] / basis / cpu / x86 / x86.factor

diff --git a/basis/cpu/x86/x86.factor b/basis/cpu/x86/x86.factor
index 86006f843ec11f57397d4f9d73222d5a1fa6b06f..302b033a7fd3cfcfe7ca9aa23a6a2742cc214576 100644 (file)
--- a/basis/cpu/x86/x86.factor
+++ b/basis/cpu/x86/x86.factor
@@ -650,7 +650,7 @@ M: x86 %fill-vector-reps
     } available-reps ;
 
 ! M:: x86 %broadcast-vector ( dst src rep -- )
-!     rep unsign-rep {
+!     rep signed-rep {
 !         { float-4-rep [
 !             dst src float-4-rep %copy
 !             dst dst { 0 0 0 0 } SHUFPS
@@ -687,7 +687,7 @@ M: x86 %fill-vector-reps
 !     } available-reps ;
 
 M:: x86 %gather-vector-4 ( dst src1 src2 src3 src4 rep -- )
-    rep unsign-rep {
+    rep signed-rep {
         { float-4-rep [
             dst src1 float-4-rep %copy
             dst src2 UNPCKLPS
@@ -710,7 +710,7 @@ M: x86 %gather-vector-4-reps
     } available-reps ;
 
 M:: x86 %gather-vector-2 ( dst src1 src2 rep -- )
-    rep unsign-rep {
+    rep signed-rep {
         { double-2-rep [
             dst src1 double-2-rep %copy
             dst src2 MOVLHPS
@@ -763,7 +763,7 @@ M: x86 %gather-vector-2-reps
 
 M:: x86 %shuffle-vector-imm ( dst src shuffle rep -- )
     dst src rep %copy
-    dst shuffle rep unsign-rep {
+    dst shuffle rep signed-rep {
         { double-2-rep [ >float-4-shuffle float-4-shuffle ] }
         { float-4-rep [ float-4-shuffle ] }
         { int-4-rep [ int-4-shuffle ] }
@@ -786,7 +786,7 @@ M: x86 %shuffle-vector-reps
 
 M: x86 %merge-vector-head
     [ two-operand ] keep
-    unsign-rep {
+    signed-rep {
         { double-2-rep   [ MOVLHPS ] }
         { float-4-rep    [ UNPCKLPS ] }
         { longlong-2-rep [ PUNPCKLQDQ ] }
@@ -797,7 +797,7 @@ M: x86 %merge-vector-head
 
 M: x86 %merge-vector-tail
     [ two-operand ] keep
-    unsign-rep {
+    signed-rep {
         { double-2-rep   [ UNPCKHPD ] }
         { float-4-rep    [ UNPCKHPS ] }
         { longlong-2-rep [ PUNPCKHQDQ ] }
@@ -826,7 +826,7 @@ M: x86 %signed-pack-vector-reps
 
 M: x86 %unsigned-pack-vector
     [ two-operand ] keep
-    unsign-rep {
+    signed-rep {
         { int-4-rep   [ PACKUSDW ] }
         { short-8-rep [ PACKUSWB ] }
     } case ;
@@ -896,7 +896,7 @@ M: x86 %float>integer-vector-reps
     } case ;
 
 :: (%compare-int-vector) ( dst src rep int64 int32 int16 int8 -- )
-    rep unsign-rep :> rep'
+    rep signed-rep :> rep'
     dst src rep' {
         { longlong-2-rep [ int64 call ] }
         { int-4-rep      [ int32 call ] }
@@ -1162,34 +1162,28 @@ M: x86 %max-vector-reps
 M: x86 %dot-vector
     [ two-operand ] keep
     {
-        { float-4-rep [
-            sse4.1?
-            [ HEX: ff DPPS ]
-            [ [ MULPS ] [ drop dup float-4-rep %horizontal-add-vector ] 2bi ]
-            if
-        ] }
-        { double-2-rep [
-            sse4.1?
-            [ HEX: ff DPPD ]
-            [ [ MULPD ] [ drop dup double-2-rep %horizontal-add-vector ] 2bi ]
-            if
-        ] }
+        { float-4-rep [ HEX: ff DPPS ] }
+        { double-2-rep [ HEX: ff DPPD ] }
     } case ;
 
 M: x86 %dot-vector-reps
     {
-        { sse3? { float-4-rep double-2-rep } }
+        { sse4.1? { float-4-rep double-2-rep } }
     } available-reps ;
 
-M: x86 %horizontal-add-vector ( dst src rep -- )
-    {
-        { float-4-rep [ [ float-4-rep %copy ] [ HADDPS ] [ HADDPS ] 2tri ] }
-        { double-2-rep [ [ double-2-rep %copy ] [ HADDPD ] 2bi ] }
+M: x86 %horizontal-add-vector ( dst src1 src2 rep -- )
+    [ two-operand ] keep
+    signed-rep {
+        { float-4-rep  [ HADDPS ] }
+        { double-2-rep [ HADDPD ] }
+        { int-4-rep    [ PHADDD ] }
+        { short-8-rep  [ PHADDW ] }
     } case ;
 
 M: x86 %horizontal-add-vector-reps
     {
         { sse3? { float-4-rep double-2-rep } }
+        { ssse3? { int-4-rep uint-4-rep short-8-rep ushort-8-rep } }
     } available-reps ;
 
 M: x86 %horizontal-shl-vector-imm ( dst src1 src2 rep -- )
@@ -1197,7 +1191,7 @@ M: x86 %horizontal-shl-vector-imm ( dst src1 src2 rep -- )
 
 M: x86 %horizontal-shl-vector-imm-reps
     {
-        { sse2? { char-16-rep uchar-16-rep short-8-rep ushort-8-rep int-4-rep uint-4-rep longlong-2-rep ulonglong-2-rep } }
+        { sse2? { char-16-rep uchar-16-rep short-8-rep ushort-8-rep int-4-rep uint-4-rep longlong-2-rep ulonglong-2-rep float-4-rep double-2-rep } }
     } available-reps ;
 
 M: x86 %horizontal-shr-vector-imm ( dst src1 src2 rep -- )
@@ -1205,7 +1199,7 @@ M: x86 %horizontal-shr-vector-imm ( dst src1 src2 rep -- )
 
 M: x86 %horizontal-shr-vector-imm-reps
     {
-        { sse2? { char-16-rep uchar-16-rep short-8-rep ushort-8-rep int-4-rep uint-4-rep longlong-2-rep ulonglong-2-rep } }
+        { sse2? { char-16-rep uchar-16-rep short-8-rep ushort-8-rep int-4-rep uint-4-rep longlong-2-rep ulonglong-2-rep float-4-rep double-2-rep } }
     } available-reps ;
 
 M: x86 %abs-vector ( dst src rep -- )
@@ -1329,8 +1323,15 @@ M: x86 %shr-vector-imm-reps %shr-vector-reps ;
 
 M: x86 %integer>scalar drop MOVD ;
 
+! XXX the longlong versions won't work on x86.32
 M:: x86 %scalar>integer ( dst src rep -- )
     rep {
+        { longlong-scalar-rep [
+            dst src MOVD
+        ] }
+        { ulonglong-scalar-rep [
+            dst src MOVD
+        ] }
         { int-scalar-rep [
             dst 32-bit-version-of src MOVD
             dst dst 32-bit-version-of
@@ -1401,7 +1402,6 @@ M: x86 immediate-bitwise? ( n -- ? )
     #! set up by the caller.
     stack-frame get total-size>> + stack@ ;
 
-enable-simd
 enable-min/max
 enable-fixnum-log2