- Add some NEON ld / st instruction static encoding.

- Make bits 25-27 for ldrh, etc. explicitly zero. Previously only the JIT uses the encoding information and it's assuming anything not specified to be zero. Making them explicit so the disassembler is happy.
Patch by Sean Callanan.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@75065 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 38 insertions, 4 deletions

diff --git a/lib/Target/ARM/ARMInstrFormats.td b/lib/Target/ARM/ARMInstrFormats.td
index c7ef149e75..89a0e7e02d 100644
--- a/lib/Target/ARM/ARMInstrFormats.td
+++ b/lib/Target/ARM/ARMInstrFormats.td
@@ -451,6 +451,7 @@ class AI3ldh<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 1; // L bit
   let Inst{21}    = 0; // W bit
   let Inst{24}    = 1; // P bit
+  let Inst{27-25} = 0b000;
 }
 class AXI3ldh<dag oops, dag iops, Format f, string asm,
            list<dag> pattern>
@@ -475,6 +476,7 @@ class AI3ldsh<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 1; // L bit
   let Inst{21}    = 0; // W bit
   let Inst{24}    = 1; // P bit
+  let Inst{27-25} = 0b000;
 }
 class AXI3ldsh<dag oops, dag iops, Format f, string asm,
            list<dag> pattern>
@@ -499,6 +501,7 @@ class AI3ldsb<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 1; // L bit
   let Inst{21}    = 0; // W bit
   let Inst{24}    = 1; // P bit
+  let Inst{27-25} = 0b000;
 }
 class AXI3ldsb<dag oops, dag iops, Format f, string asm,
            list<dag> pattern>
@@ -523,6 +526,7 @@ class AI3ldd<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 0; // L bit
   let Inst{21}    = 0; // W bit
   let Inst{24}    = 1; // P bit
+  let Inst{27-25} = 0b000;
 }
 
 // stores
@@ -537,6 +541,7 @@ class AI3sth<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 0; // L bit
   let Inst{21}    = 0; // W bit
   let Inst{24}    = 1; // P bit
+  let Inst{27-25} = 0b000;
 }
 class AXI3sth<dag oops, dag iops, Format f, string asm,
            list<dag> pattern>
@@ -561,6 +566,7 @@ class AI3std<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 0; // L bit
   let Inst{21}    = 0; // W bit
   let Inst{24}    = 1; // P bit
+  let Inst{27-25} = 0b000;
 }
 
 // Pre-indexed loads
@@ -575,6 +581,7 @@ class AI3ldhpr<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 1; // L bit
   let Inst{21}    = 1; // W bit
   let Inst{24}    = 1; // P bit
+  let Inst{27-25} = 0b000;
 }
 class AI3ldshpr<dag oops, dag iops, Format f, string opc,
             string asm, string cstr, list<dag> pattern>
@@ -587,6 +594,7 @@ class AI3ldshpr<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 1; // L bit
   let Inst{21}    = 1; // W bit
   let Inst{24}    = 1; // P bit
+  let Inst{27-25} = 0b000;
 }
 class AI3ldsbpr<dag oops, dag iops, Format f, string opc,
             string asm, string cstr, list<dag> pattern>
@@ -599,6 +607,7 @@ class AI3ldsbpr<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 1; // L bit
   let Inst{21}    = 1; // W bit
   let Inst{24}    = 1; // P bit
+  let Inst{27-25} = 0b000;
 }
 
 // Pre-indexed stores
@@ -613,6 +622,7 @@ class AI3sthpr<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 0; // L bit
   let Inst{21}    = 1; // W bit
   let Inst{24}    = 1; // P bit
+  let Inst{27-25} = 0b000;
 }
 
 // Post-indexed loads
@@ -627,6 +637,7 @@ class AI3ldhpo<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 1; // L bit
   let Inst{21}    = 1; // W bit
   let Inst{24}    = 0; // P bit
+  let Inst{27-25} = 0b000;
 }
 class AI3ldshpo<dag oops, dag iops, Format f, string opc,
             string asm, string cstr, list<dag> pattern>
@@ -639,6 +650,7 @@ class AI3ldshpo<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 1; // L bit
   let Inst{21}    = 1; // W bit
   let Inst{24}    = 0; // P bit
+  let Inst{27-25} = 0b000;
 }
 class AI3ldsbpo<dag oops, dag iops, Format f, string opc,
             string asm, string cstr, list<dag> pattern>
@@ -651,6 +663,7 @@ class AI3ldsbpo<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 1; // L bit
   let Inst{21}    = 1; // W bit
   let Inst{24}    = 0; // P bit
+  let Inst{27-25} = 0b000;
 }
 
 // Post-indexed stores
@@ -665,6 +678,7 @@ class AI3sthpo<dag oops, dag iops, Format f, string opc,
   let Inst{20}    = 0; // L bit
   let Inst{21}    = 1; // W bit
   let Inst{24}    = 0; // P bit
+  let Inst{27-25} = 0b000;
 }
 
 
diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index 3fd63f01b3..2b7192321d 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -91,24 +91,44 @@ let mayLoad = 1 in {
 def VLDMD : NI<(outs),
                (ins addrmode_neonldstm:$addr, reglist:$dst1, variable_ops),
                "vldm${addr:submode} ${addr:base}, $dst1",
-               []>;
+               []> {
+  let Inst{27-25} = 0b110;
+  let Inst{20}    = 1;
+  let Inst{11-9}  = 0b101;
+}
 
 def VLDMS : NI<(outs),
                (ins addrmode_neonldstm:$addr, reglist:$dst1, variable_ops),
                "vldm${addr:submode} ${addr:base}, $dst1",
-               []>;
+               []> {
+  let Inst{27-25} = 0b110;
+  let Inst{20}    = 1;
+  let Inst{11-9}  = 0b101;
+}
 }
 */
 
 // Use vldmia to load a Q register as a D register pair.
 def VLDRQ : NI<(outs QPR:$dst), (ins GPR:$addr),
                "vldmia $addr, ${dst:dregpair}",
-               [(set QPR:$dst, (v2f64 (load GPR:$addr)))]>;
+               [(set QPR:$dst, (v2f64 (load GPR:$addr)))]> {
+  let Inst{27-25} = 0b110;
+  let Inst{24}    = 0; // P bit
+  let Inst{23}    = 1; // U bit
+  let Inst{20}    = 1;
+  let Inst{11-9}  = 0b101;
+}
 
 // Use vstmia to store a Q register as a D register pair.
 def VSTRQ : NI<(outs), (ins QPR:$src, GPR:$addr),
                "vstmia $addr, ${src:dregpair}",
-               [(store (v2f64 QPR:$src), GPR:$addr)]>;
+               [(store (v2f64 QPR:$src), GPR:$addr)]> {
+  let Inst{27-25} = 0b110;
+  let Inst{24}    = 0; // P bit
+  let Inst{23}    = 1; // U bit
+  let Inst{20}    = 0;
+  let Inst{11-9}  = 0b101;
+}
 
 
 //   VLD1     : Vector Load (multiple single elements)