v3d: add shader-db stat to count SFU stalls

SFU operations have a latency of 2 cicles, so if their results
are used in the following cycle to a SFU instruction, the GPU
stalls for an extra cycle until the result is available.

This adds the number of stalls to the shader-db debug mode and
sum of instruction + stalls to evaluate optimizations to schedule
instructions that avoid generating sfu-stalls.

v2: Rename v3d_qpu_generates_sfu_stalls to v3d_qpu_instr_is_sfu (Eric)

Reviewed-by: Eric Anholt <[email protected]>

2 files changed, 23 insertions, 12 deletions

diff --git a/src/broadcom/qpu/qpu_instr.c b/src/broadcom/qpu/qpu_instr.c
index 66e53a6accd..09d06b3fa3e 100644
--- a/src/broadcom/qpu/qpu_instr.c
+++ b/src/broadcom/qpu/qpu_instr.c
@@ -645,19 +645,10 @@ v3d_qpu_uses_tlb(const struct v3d_qpu_instr *inst)
 bool
 v3d_qpu_uses_sfu(const struct v3d_qpu_instr *inst)
 {
-        if (inst->type == V3D_QPU_INSTR_TYPE_ALU) {
-                switch (inst->alu.add.op) {
-                case V3D_QPU_A_RECIP:
-                case V3D_QPU_A_RSQRT:
-                case V3D_QPU_A_EXP:
-                case V3D_QPU_A_LOG:
-                case V3D_QPU_A_SIN:
-                case V3D_QPU_A_RSQRT2:
-                        return true;
-                default:
-                        break;
-                }
+        if (v3d_qpu_instr_is_sfu(inst))
+                return true;
 
+        if (inst->type == V3D_QPU_INSTR_TYPE_ALU) {
                 if (inst->alu.add.magic_write &&
                     v3d_qpu_magic_waddr_is_sfu(inst->alu.add.waddr)) {
                         return true;
@@ -673,6 +664,25 @@ v3d_qpu_uses_sfu(const struct v3d_qpu_instr *inst)
 }
 
 bool
+v3d_qpu_instr_is_sfu(const struct v3d_qpu_instr *inst)
+{
+        if (inst->type == V3D_QPU_INSTR_TYPE_ALU) {
+                switch (inst->alu.add.op) {
+                case V3D_QPU_A_RECIP:
+                case V3D_QPU_A_RSQRT:
+                case V3D_QPU_A_EXP:
+                case V3D_QPU_A_LOG:
+                case V3D_QPU_A_SIN:
+                case V3D_QPU_A_RSQRT2:
+                        return true;
+                default:
+                        return false;
+                }
+        }
+        return false;
+}
+
+bool
 v3d_qpu_writes_tmu(const struct v3d_qpu_instr *inst)
 {
         return (inst->type == V3D_QPU_INSTR_TYPE_ALU &&
diff --git a/src/broadcom/qpu/qpu_instr.h b/src/broadcom/qpu/qpu_instr.h
index 968d0f6fd65..ad2d37b6051 100644
--- a/src/broadcom/qpu/qpu_instr.h
+++ b/src/broadcom/qpu/qpu_instr.h
@@ -447,6 +447,7 @@ bool v3d_qpu_magic_waddr_is_vpm(enum v3d_qpu_waddr waddr) ATTRIBUTE_CONST;
 bool v3d_qpu_magic_waddr_is_tsy(enum v3d_qpu_waddr waddr) ATTRIBUTE_CONST;
 bool v3d_qpu_magic_waddr_loads_unif(enum v3d_qpu_waddr waddr) ATTRIBUTE_CONST;
 bool v3d_qpu_uses_tlb(const struct v3d_qpu_instr *inst) ATTRIBUTE_CONST;
+bool v3d_qpu_instr_is_sfu(const struct v3d_qpu_instr *inst) ATTRIBUTE_CONST;
 bool v3d_qpu_uses_sfu(const struct v3d_qpu_instr *inst) ATTRIBUTE_CONST;
 bool v3d_qpu_writes_tmu(const struct v3d_qpu_instr *inst) ATTRIBUTE_CONST;
 bool v3d_qpu_writes_tmu_not_tmuc(const struct v3d_qpu_instr *inst) ATTRIBUTE_CONST;