radv: Split emitting the cache flush out.

So that we can use it without a cmd_buffer.

Signed-off-by: Bas Nieuwenhuizen <basni@google.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>

diff --git a/src/amd/vulkan/si_cmd_buffer.c b/src/amd/vulkan/si_cmd_buffer.c
index e2ba413..1091c7b 100644 (file)
--- a/src/amd/vulkan/si_cmd_buffer.c
+++ b/src/amd/vulkan/si_cmd_buffer.c
@@ -689,37 +689,27 @@ si_get_ia_multi_vgt_param(struct radv_cmd_buffer *cmd_buffer,
 
 }
 
-void
-si_emit_cache_flush(struct radv_cmd_buffer *cmd_buffer)
+static void
+si_cs_emit_cache_flush(struct radeon_winsys_cs *cs,
+                       enum chip_class chip_class,
+                       bool is_mec,
+                       enum radv_cmd_flush_bits flush_bits)
 {
-       enum chip_class chip_class = cmd_buffer->device->physical_device->rad_info.chip_class;
        unsigned cp_coher_cntl = 0;
-       bool is_compute = cmd_buffer->queue_family_index == RADV_QUEUE_COMPUTE;
 
-       if (is_compute)
-               cmd_buffer->state.flush_bits &= ~(RADV_CMD_FLAG_FLUSH_AND_INV_CB |
-                                                 RADV_CMD_FLAG_FLUSH_AND_INV_CB_META |
-                                                 RADV_CMD_FLAG_FLUSH_AND_INV_DB |
-                                                 RADV_CMD_FLAG_FLUSH_AND_INV_DB_META |
-                                                 RADV_CMD_FLAG_PS_PARTIAL_FLUSH |
-                                                 RADV_CMD_FLAG_VS_PARTIAL_FLUSH |
-                                                 RADV_CMD_FLAG_VGT_FLUSH);
-
-       radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 128);
-
-       if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_INV_ICACHE)
+       if (flush_bits & RADV_CMD_FLAG_INV_ICACHE)
                cp_coher_cntl |= S_0085F0_SH_ICACHE_ACTION_ENA(1);
-       if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_INV_SMEM_L1)
+       if (flush_bits & RADV_CMD_FLAG_INV_SMEM_L1)
                cp_coher_cntl |= S_0085F0_SH_KCACHE_ACTION_ENA(1);
-       if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_INV_VMEM_L1)
+       if (flush_bits & RADV_CMD_FLAG_INV_VMEM_L1)
                cp_coher_cntl |= S_0085F0_TCL1_ACTION_ENA(1);
-       if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_INV_GLOBAL_L2) {
+       if (flush_bits & RADV_CMD_FLAG_INV_GLOBAL_L2) {
                cp_coher_cntl |= S_0085F0_TC_ACTION_ENA(1);
                if (chip_class >= VI)
                        cp_coher_cntl |= S_0301F0_TC_WB_ACTION_ENA(1);
        }
 
-       if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_FLUSH_AND_INV_CB) {
+       if (flush_bits & RADV_CMD_FLAG_FLUSH_AND_INV_CB) {
                cp_coher_cntl |= S_0085F0_CB_ACTION_ENA(1) |
                        S_0085F0_CB0_DEST_BASE_ENA(1) |
                        S_0085F0_CB1_DEST_BASE_ENA(1) |
@@ -731,85 +721,108 @@ si_emit_cache_flush(struct radv_cmd_buffer *cmd_buffer)
                        S_0085F0_CB7_DEST_BASE_ENA(1);
 
                /* Necessary for DCC */
-               if (cmd_buffer->device->physical_device->rad_info.chip_class >= VI) {
-                       radeon_emit(cmd_buffer->cs, PKT3(PKT3_EVENT_WRITE_EOP, 4, 0));
-                       radeon_emit(cmd_buffer->cs, EVENT_TYPE(V_028A90_FLUSH_AND_INV_CB_DATA_TS) |
+               if (chip_class >= VI) {
+                       radeon_emit(cs, PKT3(PKT3_EVENT_WRITE_EOP, 4, 0));
+                       radeon_emit(cs, EVENT_TYPE(V_028A90_FLUSH_AND_INV_CB_DATA_TS) |
                                                    EVENT_INDEX(5));
-                       radeon_emit(cmd_buffer->cs, 0);
-                       radeon_emit(cmd_buffer->cs, 0);
-                       radeon_emit(cmd_buffer->cs, 0);
-                       radeon_emit(cmd_buffer->cs, 0);
+                       radeon_emit(cs, 0);
+                       radeon_emit(cs, 0);
+                       radeon_emit(cs, 0);
+                       radeon_emit(cs, 0);
                }
        }
 
-       if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_FLUSH_AND_INV_DB) {
+       if (flush_bits & RADV_CMD_FLAG_FLUSH_AND_INV_DB) {
                cp_coher_cntl |= S_0085F0_DB_ACTION_ENA(1) |
                        S_0085F0_DB_DEST_BASE_ENA(1);
        }
 
-       if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_FLUSH_AND_INV_CB_META) {
-               radeon_emit(cmd_buffer->cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
-               radeon_emit(cmd_buffer->cs, EVENT_TYPE(V_028A90_FLUSH_AND_INV_CB_META) | EVENT_INDEX(0));
+       if (flush_bits & RADV_CMD_FLAG_FLUSH_AND_INV_CB_META) {
+               radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
+               radeon_emit(cs, EVENT_TYPE(V_028A90_FLUSH_AND_INV_CB_META) | EVENT_INDEX(0));
        }
 
-       if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_FLUSH_AND_INV_DB_META) {
-               radeon_emit(cmd_buffer->cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
-               radeon_emit(cmd_buffer->cs, EVENT_TYPE(V_028A90_FLUSH_AND_INV_DB_META) | EVENT_INDEX(0));
+       if (flush_bits & RADV_CMD_FLAG_FLUSH_AND_INV_DB_META) {
+               radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
+               radeon_emit(cs, EVENT_TYPE(V_028A90_FLUSH_AND_INV_DB_META) | EVENT_INDEX(0));
        }
 
-       if (!(cmd_buffer->state.flush_bits & (RADV_CMD_FLAG_FLUSH_AND_INV_CB |
+       if (!(flush_bits & (RADV_CMD_FLAG_FLUSH_AND_INV_CB |
                                              RADV_CMD_FLAG_FLUSH_AND_INV_DB))) {
-               if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_PS_PARTIAL_FLUSH) {
-                       radeon_emit(cmd_buffer->cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
-                       radeon_emit(cmd_buffer->cs, EVENT_TYPE(V_028A90_PS_PARTIAL_FLUSH) | EVENT_INDEX(4));
-               } else if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_VS_PARTIAL_FLUSH) {
-                       radeon_emit(cmd_buffer->cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
-                       radeon_emit(cmd_buffer->cs, EVENT_TYPE(V_028A90_VS_PARTIAL_FLUSH) | EVENT_INDEX(4));
+               if (flush_bits & RADV_CMD_FLAG_PS_PARTIAL_FLUSH) {
+                       radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
+                       radeon_emit(cs, EVENT_TYPE(V_028A90_PS_PARTIAL_FLUSH) | EVENT_INDEX(4));
+               } else if (flush_bits & RADV_CMD_FLAG_VS_PARTIAL_FLUSH) {
+                       radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
+                       radeon_emit(cs, EVENT_TYPE(V_028A90_VS_PARTIAL_FLUSH) | EVENT_INDEX(4));
                }
        }
 
-       if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_CS_PARTIAL_FLUSH) {
-               radeon_emit(cmd_buffer->cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
-               radeon_emit(cmd_buffer->cs, EVENT_TYPE(V_028A90_CS_PARTIAL_FLUSH) | EVENT_INDEX(4));
+       if (flush_bits & RADV_CMD_FLAG_CS_PARTIAL_FLUSH) {
+               radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
+               radeon_emit(cs, EVENT_TYPE(V_028A90_CS_PARTIAL_FLUSH) | EVENT_INDEX(4));
        }
 
        /* VGT state sync */
-       if (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_VGT_FLUSH) {
-               radeon_emit(cmd_buffer->cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
-               radeon_emit(cmd_buffer->cs, EVENT_TYPE(V_028A90_VGT_FLUSH) | EVENT_INDEX(0));
+       if (flush_bits & RADV_CMD_FLAG_VGT_FLUSH) {
+               radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
+               radeon_emit(cs, EVENT_TYPE(V_028A90_VGT_FLUSH) | EVENT_INDEX(0));
        }
 
        /* Make sure ME is idle (it executes most packets) before continuing.
         * This prevents read-after-write hazards between PFP and ME.
         */
-       if ((cp_coher_cntl || (cmd_buffer->state.flush_bits & RADV_CMD_FLAG_CS_PARTIAL_FLUSH)) &&
-           !radv_cmd_buffer_uses_mec(cmd_buffer)) {
-               radeon_emit(cmd_buffer->cs, PKT3(PKT3_PFP_SYNC_ME, 0, 0));
-               radeon_emit(cmd_buffer->cs, 0);
+       if ((cp_coher_cntl || (flush_bits & RADV_CMD_FLAG_CS_PARTIAL_FLUSH)) &&
+           !is_mec) {
+               radeon_emit(cs, PKT3(PKT3_PFP_SYNC_ME, 0, 0));
+               radeon_emit(cs, 0);
        }
 
        /* When one of the DEST_BASE flags is set, SURFACE_SYNC waits for idle.
         * Therefore, it should be last. Done in PFP.
         */
        if (cp_coher_cntl) {
-               if (radv_cmd_buffer_uses_mec(cmd_buffer)) {
-                       radeon_emit(cmd_buffer->cs, PKT3(PKT3_ACQUIRE_MEM, 5, 0) |
+               if (is_mec) {
+                       radeon_emit(cs, PKT3(PKT3_ACQUIRE_MEM, 5, 0) |
                                                    PKT3_SHADER_TYPE_S(1));
-                       radeon_emit(cmd_buffer->cs, cp_coher_cntl);   /* CP_COHER_CNTL */
-                       radeon_emit(cmd_buffer->cs, 0xffffffff);      /* CP_COHER_SIZE */
-                       radeon_emit(cmd_buffer->cs, 0xff);            /* CP_COHER_SIZE_HI */
-                       radeon_emit(cmd_buffer->cs, 0);               /* CP_COHER_BASE */
-                       radeon_emit(cmd_buffer->cs, 0);               /* CP_COHER_BASE_HI */
-                       radeon_emit(cmd_buffer->cs, 0x0000000A);      /* POLL_INTERVAL */
+                       radeon_emit(cs, cp_coher_cntl);   /* CP_COHER_CNTL */
+                       radeon_emit(cs, 0xffffffff);      /* CP_COHER_SIZE */
+                       radeon_emit(cs, 0xff);            /* CP_COHER_SIZE_HI */
+                       radeon_emit(cs, 0);               /* CP_COHER_BASE */
+                       radeon_emit(cs, 0);               /* CP_COHER_BASE_HI */
+                       radeon_emit(cs, 0x0000000A);      /* POLL_INTERVAL */
                } else {
                        /* ACQUIRE_MEM is only required on a compute ring. */
-                       radeon_emit(cmd_buffer->cs, PKT3(PKT3_SURFACE_SYNC, 3, 0));
-                       radeon_emit(cmd_buffer->cs, cp_coher_cntl);   /* CP_COHER_CNTL */
-                       radeon_emit(cmd_buffer->cs, 0xffffffff);      /* CP_COHER_SIZE */
-                       radeon_emit(cmd_buffer->cs, 0);               /* CP_COHER_BASE */
-                       radeon_emit(cmd_buffer->cs, 0x0000000A);      /* POLL_INTERVAL */
+                       radeon_emit(cs, PKT3(PKT3_SURFACE_SYNC, 3, 0));
+                       radeon_emit(cs, cp_coher_cntl);   /* CP_COHER_CNTL */
+                       radeon_emit(cs, 0xffffffff);      /* CP_COHER_SIZE */
+                       radeon_emit(cs, 0);               /* CP_COHER_BASE */
+                       radeon_emit(cs, 0x0000000A);      /* POLL_INTERVAL */
                }
        }
+}
+
+void
+si_emit_cache_flush(struct radv_cmd_buffer *cmd_buffer)
+{
+       bool is_compute = cmd_buffer->queue_family_index == RADV_QUEUE_COMPUTE;
+
+       if (is_compute)
+               cmd_buffer->state.flush_bits &= ~(RADV_CMD_FLAG_FLUSH_AND_INV_CB |
+                                                 RADV_CMD_FLAG_FLUSH_AND_INV_CB_META |
+                                                 RADV_CMD_FLAG_FLUSH_AND_INV_DB |
+                                                 RADV_CMD_FLAG_FLUSH_AND_INV_DB_META |
+                                                 RADV_CMD_FLAG_PS_PARTIAL_FLUSH |
+                                                 RADV_CMD_FLAG_VS_PARTIAL_FLUSH |
+                                                 RADV_CMD_FLAG_VGT_FLUSH);
+
+       radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 128);
+
+       si_cs_emit_cache_flush(cmd_buffer->cs,
+                              cmd_buffer->device->physical_device->rad_info.chip_class,
+                              radv_cmd_buffer_uses_mec(cmd_buffer),
+                              cmd_buffer->state.flush_bits);
+
 
        if (cmd_buffer->state.flush_bits)
                radv_cmd_buffer_trace_emit(cmd_buffer);