target-i386: add SSE4a instruction support

This adds support for the AMD Phenom/Barcelona's SSE4a instructions.
Those include insertq and extrq, which are doing shift and mask on
XMM registers, in two versions (immediate shift/length values and
stored in another XMM register).
Additionally it implements movntss, movntsd, which are scalar
non-temporal stores (avoiding cache trashing). These are implemented
as normal stores, though.
SSE4a is guarded by the SSE4A CPUID bit (Fn8000_0001:ECX[6]).

Signed-off-by: Andre Przywara <andre.przywara@amd.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

1 files changed, 44 insertions, 0 deletions

diff --git a/target-i386/ops_sse.h b/target-i386/ops_sse.h
index 709732a..3232abd 100644
--- a/target-i386/ops_sse.h
+++ b/target-i386/ops_sse.h
@@ -802,6 +802,50 @@ void helper_rcpss(XMMReg *d, XMMReg *s)
     d->XMM_S(0) = approx_rcp(s->XMM_S(0));
 }
 
+static inline uint64_t helper_extrq(uint64_t src, int shift, int len)
+{
+    uint64_t mask;
+
+    if (len == 0) {
+        mask = ~0LL;
+    } else {
+        mask = (1ULL << len) - 1;
+    }
+    return (src >> shift) & mask;
+}
+
+void helper_extrq_r(XMMReg *d, XMMReg *s)
+{
+    d->XMM_Q(0) = helper_extrq(d->XMM_Q(0), s->XMM_B(1), s->XMM_B(0));
+}
+
+void helper_extrq_i(XMMReg *d, int index, int length)
+{
+    d->XMM_Q(0) = helper_extrq(d->XMM_Q(0), index, length);
+}
+
+static inline uint64_t helper_insertq(uint64_t src, int shift, int len)
+{
+    uint64_t mask;
+
+    if (len == 0) {
+        mask = ~0ULL;
+    } else {
+        mask = (1ULL << len) - 1;
+    }
+    return (src & ~(mask << shift)) | ((src & mask) << shift);
+}
+
+void helper_insertq_r(XMMReg *d, XMMReg *s)
+{
+    d->XMM_Q(0) = helper_insertq(s->XMM_Q(0), s->XMM_B(9), s->XMM_B(8));
+}
+
+void helper_insertq_i(XMMReg *d, int index, int length)
+{
+    d->XMM_Q(0) = helper_insertq(d->XMM_Q(0), index, length);
+}
+
 void helper_haddps(XMMReg *d, XMMReg *s)
 {
     XMMReg r;