x86: implement rcpps and rcpss SSE insts

These are packed single-precision approximate reciprocal operations,
vector and scalar versions, respectively.

This code was basically developed by copying the code for
sqrtps and sqrtss.  The mrcp micro-op was simplified relative to
msqrt since there are no double-precision versions of this operation.

1 files changed, 37 insertions, 2 deletions

diff --git a/src/arch/x86/isa/insts/simd128/floating_point/arithmetic/reciprocal_estimation.py b/src/arch/x86/isa/insts/simd128/floating_point/arithmetic/reciprocal_estimation.py
index 6e0d7fbb6..666c45ca1 100644
--- a/src/arch/x86/isa/insts/simd128/floating_point/arithmetic/reciprocal_estimation.py
+++ b/src/arch/x86/isa/insts/simd128/floating_point/arithmetic/reciprocal_estimation.py
@@ -1,4 +1,6 @@
 # Copyright (c) 2007 The Hewlett-Packard Development Company
+# Copyright (c) 2015 Advanced Micro Devices, Inc.
+#
 # All rights reserved.
 #
 # The license below extends only to copyright in the software and shall
@@ -34,8 +36,41 @@
 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #
 # Authors: Gabe Black
+#          Steve Reinhardt
 
 microcode = '''
-# RCPPS
-# RCPSS
+def macroop RCPSS_XMM_XMM {
+    mrcp xmml, xmmlm, size=4, ext=Scalar
+};
+
+def macroop RCPSS_XMM_M {
+    ldfp ufp1, seg, sib, disp, dataSize=8
+    mrcp xmml, ufp1, size=4, ext=Scalar
+};
+
+def macroop RCPSS_XMM_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, disp, dataSize=8
+    mrcp xmml, ufp1, size=4, ext=Scalar
+};
+
+def macroop RCPPS_XMM_XMM {
+    mrcp xmml, xmmlm, size=4, ext=0
+    mrcp xmmh, xmmhm, size=4, ext=0
+};
+
+def macroop RCPPS_XMM_M {
+    ldfp ufp1, seg, sib, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, sib, "DISPLACEMENT + 8", dataSize=8
+    mrcp xmml, ufp1, size=4, ext=0
+    mrcp xmmh, ufp2, size=4, ext=0
+};
+
+def macroop RCPPS_XMM_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, riprel, "DISPLACEMENT + 8", dataSize=8
+    mrcp xmml, ufp1, size=4, ext=0
+    mrcp xmmh, ufp2, size=4, ext=0
+};
 '''