Turn on vzeroupper insertion on call boundaries for AVX; it works as far as I know...

[oota-llvm.git] / lib / Target / X86 / README-SSE.txt

diff --git a/lib/Target/X86/README-SSE.txt b/lib/Target/X86/README-SSE.txt
index b2116e03b148ea222532c956e9dd78f49fea3a06..7d901afae47499b401dc409f664922e6fc1da4ca 100644 (file)
--- a/lib/Target/X86/README-SSE.txt
+++ b/lib/Target/X86/README-SSE.txt
@@ -862,7 +862,7 @@ define float @bar(float %x) nounwind {
 
 This IR (from PR6194):
 
-target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-darwin10.0.0"
 
 %0 = type { double, double }
@@ -923,4 +923,15 @@ The insertps's of $0 are pointless complex copies.
 
 //===---------------------------------------------------------------------===//
 
+If SSE4.1 is available we should inline rounding functions instead of emitting
+a libcall.
 
+floor: roundsd $0x01, %xmm, %xmm
+ceil:  roundsd $0x02, %xmm, %xmm
+
+and likewise for the single precision versions.
+
+Currently, SelectionDAGBuilder doesn't turn calls to these functions into the
+corresponding nodes and some targets (including X86) aren't ready for them.
+
+//===---------------------------------------------------------------------===//