From ca05a13c62ad7693f8be924c168e400e8139b0d2 Mon Sep 17 00:00:00 2001
From: ReinUsesLisp
Date: Tue, 25 May 2021 02:22:21 -0300
Subject: glasm: Catch more register leaks

Add support for null registers. These are used when an instruction has
no usages.

This comes handy when an instruction is only used for its CC value, with
the caveat of having to invalidate all pseudo-instructions before
defining the instruction itself in the register allocator. This commits
changes this.

Workaround a bug on Nvidia's condition codes conditional execution using
branches.
---
 src/shader_recompiler/backend/glasm/emit_glasm_warp.cpp | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

(limited to 'src/shader_recompiler/backend/glasm/emit_glasm_warp.cpp')

diff --git a/src/shader_recompiler/backend/glasm/emit_glasm_warp.cpp b/src/shader_recompiler/backend/glasm/emit_glasm_warp.cpp
index af0e13d43..6e30790bb 100644
--- a/src/shader_recompiler/backend/glasm/emit_glasm_warp.cpp
+++ b/src/shader_recompiler/backend/glasm/emit_glasm_warp.cpp
@@ -51,6 +51,10 @@ void EmitSubgroupGeMask(EmitContext& ctx, IR::Inst& inst) {
 static void Shuffle(EmitContext& ctx, IR::Inst& inst, ScalarU32 value, ScalarU32 index,
                     const IR::Value& clamp, const IR::Value& segmentation_mask,
                     std::string_view op) {
+    IR::Inst* const in_bounds{inst.GetAssociatedPseudoOperation(IR::Opcode::GetInBoundsFromOp)};
+    if (in_bounds) {
+        in_bounds->Invalidate();
+    }
     std::string mask;
     if (clamp.IsImmediate() && segmentation_mask.IsImmediate()) {
         mask = fmt::to_string(clamp.U32() | (segmentation_mask.U32() << 8));
@@ -61,13 +65,11 @@ static void Shuffle(EmitContext& ctx, IR::Inst& inst, ScalarU32 value, ScalarU32
                 ScalarU32{ctx.reg_alloc.Consume(clamp)});
     }
     const Register value_ret{ctx.reg_alloc.Define(inst)};
-    IR::Inst* const in_bounds{inst.GetAssociatedPseudoOperation(IR::Opcode::GetInBoundsFromOp)};
     if (in_bounds) {
         const Register bounds_ret{ctx.reg_alloc.Define(*in_bounds)};
         ctx.Add("SHF{}.U {},{},{},{};"
                 "MOV.U {}.x,{}.y;",
                 op, bounds_ret, value, index, mask, value_ret, bounds_ret);
-        in_bounds->Invalidate();
     } else {
         ctx.Add("SHF{}.U {},{},{},{};"
                 "MOV.U {}.x,{}.y;",
-- 
cgit v1.2.3