[PATCH] Use rsqrt (X86) to speed up reciprocal square root calcs (PR20900)

[hfinkel at anl.gov]

================
Comment at: lib/Target/X86/X86ISelLowering.cpp:14347
@@ +14346,3 @@
+  // at least Intel's Core* chips have variable timing based on the number of
+  // sig digs in the divisor and/or sqrt operand.
+  if (!Subtarget->useSqrtEst())
----------------
Please write out "significant digits"

================
Comment at: lib/Target/X86/X86ISelLowering.cpp:14355
@@ +14354,3 @@
+  // TODO: Add support for AVX (v8f32) and AVX512 (v16f32).
+  // TODO: Is it ever worthwhile to use an estimate for f64?
+  if (Subtarget->hasSSE1() && (VT == MVT::f32 || VT == MVT::v4f32)) {
----------------
Why wouldn't it be?

http://reviews.llvm.org/D5658