v3d: Fix incorrect flagging of ldtmu as writing r4 on v3d 4.x.

Fixes some stalls in 3DMMES's main vertex shader.

total instructions in shared programs: 6280751 -> 6211270 (-1.11%)
instructions in affected programs: 2935050 -> 2865569 (-2.37%)

diff --git a/src/broadcom/qpu/qpu_instr.c b/src/broadcom/qpu/qpu_instr.c
index 12a9c32..22f574c 100644 (file)
--- a/src/broadcom/qpu/qpu_instr.c
+++ b/src/broadcom/qpu/qpu_instr.c
@@ -751,9 +751,6 @@ bool
 v3d_qpu_writes_r4(const struct v3d_device_info *devinfo,
                   const struct v3d_qpu_instr *inst)
 {
-        if (inst->sig.ldtmu)
-                return true;
-
         if (inst->type == V3D_QPU_INSTR_TYPE_ALU) {
                 if (inst->alu.add.magic_write &&
                     (inst->alu.add.waddr == V3D_QPU_WADDR_R4 ||
@@ -768,8 +765,10 @@ v3d_qpu_writes_r4(const struct v3d_device_info *devinfo,
                 }
         }
 
-        if (v3d_qpu_sig_writes_address(devinfo, &inst->sig) &&
-            inst->sig_magic && inst->sig_addr == V3D_QPU_WADDR_R4) {
+        if (v3d_qpu_sig_writes_address(devinfo, &inst->sig)) {
+                if (inst->sig_magic && inst->sig_addr == V3D_QPU_WADDR_R4)
+                        return true;
+        } else if (inst->sig.ldtmu) {
                 return true;
         }