[PATCH] D23364: AMDGPU: Set sizes of spill pseudos

[Matt Arsenault via llvm-commits]

arsenm created this revision.
arsenm added a subscriber: llvm-commits.
Herald added a reviewer: tstellarAMD.
Herald added subscribers: kzhuravl, arsenm.

https://reviews.llvm.org/D23364

Files:
  lib/Target/AMDGPU/SIInstrInfo.cpp
  lib/Target/AMDGPU/SIInstructions.td
  lib/Target/AMDGPU/SIRegisterInfo.td

Index: lib/Target/AMDGPU/SIRegisterInfo.td
===================================================================
--- lib/Target/AMDGPU/SIRegisterInfo.td
+++ lib/Target/AMDGPU/SIRegisterInfo.td
@@ -193,6 +193,7 @@
 def VGPR_32 : RegisterClass<"AMDGPU", [i32, f32], 32,
                             (add (sequence "VGPR%u", 0, 255))> {
   let AllocationPriority = 1;
+  let Size = 32;
 }
 
 // VGPR 64-bit registers
@@ -306,6 +307,8 @@
 
 // Register class for all vector registers (VGPRs + Interploation Registers)
 def VReg_64 : RegisterClass<"AMDGPU", [i64, f64, v2i32, v2f32], 32, (add VGPR_64)> {
+  let Size = 64;
+
   // Requires 2 v_mov_b32 to copy
   let CopyCost = 2;
   let AllocationPriority = 2;
@@ -320,17 +323,21 @@
 }
 
 def VReg_128 : RegisterClass<"AMDGPU", [v4i32, v4f32, v2i64, v2f64], 32, (add VGPR_128)> {
+  let Size = 128;
+
   // Requires 4 v_mov_b32 to copy
   let CopyCost = 4;
   let AllocationPriority = 4;
 }
 
 def VReg_256 : RegisterClass<"AMDGPU", [v8i32, v8f32], 32, (add VGPR_256)> {
+  let Size = 256;
   let CopyCost = 8;
   let AllocationPriority = 5;
 }
 
 def VReg_512 : RegisterClass<"AMDGPU", [v16i32, v16f32], 32, (add VGPR_512)> {
+  let Size = 512;
   let CopyCost = 16;
   let AllocationPriority = 6;
 }
Index: lib/Target/AMDGPU/SIInstructions.td
===================================================================
--- lib/Target/AMDGPU/SIInstructions.td
+++ lib/Target/AMDGPU/SIInstructions.td
@@ -1948,14 +1948,19 @@
            SReg_32:$scratch_offset, i32imm:$offset)> {
       let mayStore = 1;
       let mayLoad = 0;
+      // (2 * 4) + (8 * num_subregs) bytes maximum
+      let Size = !add(!shl(!srl(vgpr_class.Size, 5), 3), 8);
     }
 
     def _RESTORE : PseudoInstSI <
       (outs vgpr_class:$dst),
       (ins i32imm:$frame_idx, SReg_128:$scratch_rsrc, SReg_32:$scratch_offset,
            i32imm:$offset)> {
       let mayStore = 0;
       let mayLoad = 1;
+
+      // (2 * 4) + (8 * num_subregs) bytes maximum
+      let Size = !add(!shl(!srl(vgpr_class.Size, 5), 3), 8);
     }
   } // End UseNamedOperandTable = 1, VGPRSpill = 1
 }
Index: lib/Target/AMDGPU/SIInstrInfo.cpp
===================================================================
--- lib/Target/AMDGPU/SIInstrInfo.cpp
+++ lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -3146,11 +3146,9 @@
 
   // If we have a definitive size, we can use it. Otherwise we need to inspect
   // the operands to know the size.
-  if (DescSize == 8 || DescSize == 4)
+  if (DescSize != 0)
     return DescSize;
 
-  assert(DescSize == 0);
-
   // 4-byte instructions may have a 32-bit literal encoded after them. Check
   // operands that coud ever be literals.
   if (isVALU(MI) || isSALU(MI)) {


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D23364.67576.patch
Type: text/x-patch
Size: 2689 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160810/72f59591/attachment.bin>