R600: Fix encoding for R600 family GPUs

Reviewed-by: Vincent Lejeune <vljn@ovi.com>
https://bugs.freedesktop.org/show_bug.cgi?id=64193
https://bugs.freedesktop.org/show_bug.cgi?id=64257
https://bugs.freedesktop.org/show_bug.cgi?id=64320

NOTE: This is a candidate for the 3.3 branch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182113 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/R600/MCTargetDesc/R600MCCodeEmitter.cpp b/lib/Target/R600/MCTargetDesc/R600MCCodeEmitter.cpp
index c5bd01a..cb4cf0c 100644 (file)
--- a/lib/Target/R600/MCTargetDesc/R600MCCodeEmitter.cpp
+++ b/lib/Target/R600/MCTargetDesc/R600MCCodeEmitter.cpp
@@ -179,6 +179,13 @@ void R600MCCodeEmitter::EncodeInstruction(const MCInst &MI, raw_ostream &OS,
     Emit((u_int32_t) 0, OS);
   } else {
     uint64_t Inst = getBinaryCodeForInstr(MI, Fixups);
+    if ((STI.getFeatureBits() & AMDGPU::FeatureR600ALUInst) &&
+       ((Desc.TSFlags & R600_InstFlag::OP1) ||
+         Desc.TSFlags & R600_InstFlag::OP2)) {
+      uint64_t ISAOpCode = Inst & (0x3FFULL << 39);
+      Inst &= ~(0x3FFULL << 39);
+      Inst |= ISAOpCode << 1;
+    }
     Emit(Inst, OS);
   }
 }

diff --git a/test/CodeGen/R600/r600-encoding.ll b/test/CodeGen/R600/r600-encoding.ll
new file mode 100644 (file)
index 0000000..c8040a1
--- /dev/null
+++ b/test/CodeGen/R600/r600-encoding.ll
@@ -0,0 +1,24 @@
+; RUN: llc < %s -march=r600 -show-mc-encoding -mcpu=redwood | FileCheck --check-prefix=EG-CHECK %s
+; RUN: llc < %s -march=r600 -show-mc-encoding -mcpu=rs880 | FileCheck --check-prefix=R600-CHECK %s
+
+; The earliest R600 GPUs have a slightly different encoding than the rest of
+; the VLIW4/5 GPUs.
+
+; EG-CHECK: @test
+; EG-CHECK: MUL_IEEE {{[ *TXYZW.,0-9]+}} ; encoding: [{{0x[0-9a-f]+,0x[0-9a-f]+,0x[0-9a-f]+,0x[0-9a-f]+,0x10,0x01,0x[0-9a-f]+,0x[0-9a-f]+}}]
+
+; R600-CHECK: @test
+; R600-CHECK: MUL_IEEE {{[ *TXYZW.,0-9]+}} ; encoding: [{{0x[0-9a-f]+,0x[0-9a-f]+,0x[0-9a-f]+,0x[0-9a-f]+,0x10,0x02,0x[0-9a-f]+,0x[0-9a-f]+}}]
+
+define void @test() {
+entry:
+  %0 = call float @llvm.R600.load.input(i32 0)
+  %1 = call float @llvm.R600.load.input(i32 1)
+  %2 = fmul float %0, %1
+  call void @llvm.AMDGPU.store.output(float %2, i32 0)
+  ret void
+}
+
+declare float @llvm.R600.load.input(i32) readnone
+
+declare void @llvm.AMDGPU.store.output(float, i32)