[llvm-commits] [llvm] r160430 - /llvm/trunk/lib/Target/X86/X86InstrSSE.td

[Nadav Rotem]

Author: nadav
Date: Wed Jul 18 03:14:48 2012
New Revision: 160430

URL: http://llvm.org/viewvc/llvm-project?rev=160430&view=rev
Log:
The vbroadcast family of instructions has 'fallback patterns' in case where the
load source operand is used by multiple nodes. The v2i64 broadcast was emulated
by shuffling the two lower i32 elements to the upper two.
We had a bug in the immediate used for the broadcast.
Replacing 0 to 0x44.
0x44 means [01|00|01|00] which corresponds to the correct lane.

Patch by Michael Kuperstein.


Modified:
    llvm/trunk/lib/Target/X86/X86InstrSSE.td

Modified: llvm/trunk/lib/Target/X86/X86InstrSSE.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrSSE.td?rev=160430&r1=160429&r2=160430&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrSSE.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrSSE.td Wed Jul 18 03:14:48 2012
@@ -7781,11 +7781,12 @@
   def : Pat<(v4f64 (X86VBroadcast FR64:$src)),
             (VINSERTF128rr (INSERT_SUBREG (v4f64 (IMPLICIT_DEF)),
               (VPSHUFDri
-                (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd), 0),
-                  sub_xmm),
+                (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd),
+                0x44),
+              sub_xmm),
               (VPSHUFDri
                 (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd),
-              0), 1)>;
+                0x44), 1)>;
 
   def : Pat<(v4i32 (X86VBroadcast GR32:$src)),
             (VPSHUFDri
@@ -7801,11 +7802,12 @@
   def : Pat<(v4i64 (X86VBroadcast GR64:$src)),
             (VINSERTF128rr (INSERT_SUBREG (v4i64 (IMPLICIT_DEF)),
               (VPSHUFDri
-                (INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), GR64:$src, sub_sd), 0),
-                  sub_xmm),
+                (INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), GR64:$src, sub_sd),
+                0x44),
+              sub_xmm),
               (VPSHUFDri
                 (INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), GR64:$src, sub_sd),
-              0), 1)>;
+                0x44), 1)>;
   }
 }