AMDGPU/SI: Don't mark VINTRP instructions as mayLoad

Summary:
These instructions technically do read from memory, but the memory
is considered to be out of bounds for normal load/store instructions.

shader-db stats:

SGPRS: 1416075 -> 1413323 (-0.19 %)
VGPRS: 867413 -> 863935 (-0.40 %)
Spilled SGPRs: 1409 -> 1354 (-3.90 %)
Spilled VGPRs: 63 -> 63 (0.00 %)
Private memory VGPRs: 880 -> 880 (0.00 %)
Scratch size: 2648 -> 2632 (-0.60 %) dwords per thread
Code Size: 37889052 -> 37897340 (0.02 %) bytes
LDS: 2147 -> 2147 (0.00 %) blocks
Max Waves: 279243 -> 280369 (0.40 %)
Wait states: 0 -> 0 (0.00 %)

Reviewers: nhaehnle, mareko, arsenm

Subscribers: kzhuravl, wdng, yaxunl, tony-tye

Differential Revision: https://reviews.llvm.org/D27593

llvm-svn: 289219

1 files changed, 13 insertions, 1 deletions

diff --git a/llvm/lib/Target/AMDGPU/SIInstrFormats.td b/llvm/lib/Target/AMDGPU/SIInstrFormats.td
index 5f260ba0c85..16021fff531 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrFormats.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrFormats.td
@@ -229,7 +229,19 @@ let Uses = [EXEC] in {
 
 class VINTRPCommon <dag outs, dag ins, string asm, list<dag> pattern> :
     InstSI <outs, ins, asm, pattern> {
-  let mayLoad = 1;
+
+  // VINTRP instructions read parameter values from LDS, but these parameter
+  // values are stored outside of the LDS memory that is allocated to the
+  // shader for general purpose use.
+  //
+  // While it may be possible for ds_read/ds_write instructions to access
+  // the parameter values in LDS, this would essentially be an out-of-bounds
+  // memory access which we consider to be undefined behavior.
+  //
+  // So even though these instructions read memory, this memory is outside the
+  // addressable memory space for the shader, and we consider these instructions
+  // to be readnone.
+  let mayLoad = 0;
   let mayStore = 0;
   let hasSideEffects = 0;
 }