[03/16] radeonsi: add GFX-IB-size query to the HUD

Submitted by Marek Olšák on Oct. 13, 2017, 12:03 p.m.

Details

Message ID 1507896252-10099-4-git-send-email-maraeo@gmail.com
State New
Headers show
Series "RadeonSI micro-optimizations" ( rev: 1 ) in Mesa

Not browsing as part of any series.

Commit Message

Marek Olšák Oct. 13, 2017, 12:03 p.m.
From: Marek Olšák <marek.olsak@amd.com>

It shows the sum of all IBs per frame.
---
 src/gallium/drivers/radeon/r600_query.c           | 4 ++++
 src/gallium/drivers/radeon/r600_query.h           | 1 +
 src/gallium/drivers/radeon/radeon_winsys.h        | 1 +
 src/gallium/winsys/amdgpu/drm/amdgpu_cs.c         | 1 +
 src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c     | 2 ++
 src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h     | 1 +
 src/gallium/winsys/radeon/drm/radeon_drm_winsys.c | 1 +
 7 files changed, 11 insertions(+)

Patch hide | download patch | download mbox

diff --git a/src/gallium/drivers/radeon/r600_query.c b/src/gallium/drivers/radeon/r600_query.c
index b2809f8..aedf950 100644
--- a/src/gallium/drivers/radeon/r600_query.c
+++ b/src/gallium/drivers/radeon/r600_query.c
@@ -72,20 +72,21 @@  static enum radeon_value_id winsys_id_from_type(unsigned type)
 	switch (type) {
 	case R600_QUERY_REQUESTED_VRAM: return RADEON_REQUESTED_VRAM_MEMORY;
 	case R600_QUERY_REQUESTED_GTT: return RADEON_REQUESTED_GTT_MEMORY;
 	case R600_QUERY_MAPPED_VRAM: return RADEON_MAPPED_VRAM;
 	case R600_QUERY_MAPPED_GTT: return RADEON_MAPPED_GTT;
 	case R600_QUERY_BUFFER_WAIT_TIME: return RADEON_BUFFER_WAIT_TIME_NS;
 	case R600_QUERY_NUM_MAPPED_BUFFERS: return RADEON_NUM_MAPPED_BUFFERS;
 	case R600_QUERY_NUM_GFX_IBS: return RADEON_NUM_GFX_IBS;
 	case R600_QUERY_NUM_SDMA_IBS: return RADEON_NUM_SDMA_IBS;
 	case R600_QUERY_GFX_BO_LIST_SIZE: return RADEON_GFX_BO_LIST_COUNTER;
+	case R600_QUERY_GFX_IB_SIZE: return RADEON_GFX_IB_SIZE_COUNTER;
 	case R600_QUERY_NUM_BYTES_MOVED: return RADEON_NUM_BYTES_MOVED;
 	case R600_QUERY_NUM_EVICTIONS: return RADEON_NUM_EVICTIONS;
 	case R600_QUERY_NUM_VRAM_CPU_PAGE_FAULTS: return RADEON_NUM_VRAM_CPU_PAGE_FAULTS;
 	case R600_QUERY_VRAM_USAGE: return RADEON_VRAM_USAGE;
 	case R600_QUERY_VRAM_VIS_USAGE: return RADEON_VRAM_VIS_USAGE;
 	case R600_QUERY_GTT_USAGE: return RADEON_GTT_USAGE;
 	case R600_QUERY_GPU_TEMPERATURE: return RADEON_GPU_TEMPERATURE;
 	case R600_QUERY_CURRENT_GPU_SCLK: return RADEON_CURRENT_SCLK;
 	case R600_QUERY_CURRENT_GPU_MCLK: return RADEON_CURRENT_MCLK;
 	case R600_QUERY_CS_THREAD_BUSY: return RADEON_CS_THREAD_TIME;
@@ -171,20 +172,21 @@  static bool r600_query_sw_begin(struct r600_common_context *rctx,
 	case R600_QUERY_VRAM_VIS_USAGE:
 	case R600_QUERY_GTT_USAGE:
 	case R600_QUERY_GPU_TEMPERATURE:
 	case R600_QUERY_CURRENT_GPU_SCLK:
 	case R600_QUERY_CURRENT_GPU_MCLK:
 	case R600_QUERY_BACK_BUFFER_PS_DRAW_RATIO:
 	case R600_QUERY_NUM_MAPPED_BUFFERS:
 		query->begin_result = 0;
 		break;
 	case R600_QUERY_BUFFER_WAIT_TIME:
+	case R600_QUERY_GFX_IB_SIZE:
 	case R600_QUERY_NUM_GFX_IBS:
 	case R600_QUERY_NUM_SDMA_IBS:
 	case R600_QUERY_NUM_BYTES_MOVED:
 	case R600_QUERY_NUM_EVICTIONS:
 	case R600_QUERY_NUM_VRAM_CPU_PAGE_FAULTS: {
 		enum radeon_value_id ws_id = winsys_id_from_type(query->b.type);
 		query->begin_result = rctx->ws->query_value(rctx->ws, ws_id);
 		break;
 	}
 	case R600_QUERY_GFX_BO_LIST_SIZE:
@@ -326,20 +328,21 @@  static bool r600_query_sw_end(struct r600_common_context *rctx,
 	case R600_QUERY_REQUESTED_GTT:
 	case R600_QUERY_MAPPED_VRAM:
 	case R600_QUERY_MAPPED_GTT:
 	case R600_QUERY_VRAM_USAGE:
 	case R600_QUERY_VRAM_VIS_USAGE:
 	case R600_QUERY_GTT_USAGE:
 	case R600_QUERY_GPU_TEMPERATURE:
 	case R600_QUERY_CURRENT_GPU_SCLK:
 	case R600_QUERY_CURRENT_GPU_MCLK:
 	case R600_QUERY_BUFFER_WAIT_TIME:
+	case R600_QUERY_GFX_IB_SIZE:
 	case R600_QUERY_NUM_MAPPED_BUFFERS:
 	case R600_QUERY_NUM_GFX_IBS:
 	case R600_QUERY_NUM_SDMA_IBS:
 	case R600_QUERY_NUM_BYTES_MOVED:
 	case R600_QUERY_NUM_EVICTIONS:
 	case R600_QUERY_NUM_VRAM_CPU_PAGE_FAULTS: {
 		enum radeon_value_id ws_id = winsys_id_from_type(query->b.type);
 		query->end_result = rctx->ws->query_value(rctx->ws, ws_id);
 		break;
 	}
@@ -1926,20 +1929,21 @@  static struct pipe_driver_query_info r600_driver_query_list[] = {
 	X("gallium-thread-busy",	GALLIUM_THREAD_BUSY,	UINT64, AVERAGE),
 	X("requested-VRAM",		REQUESTED_VRAM,		BYTES, AVERAGE),
 	X("requested-GTT",		REQUESTED_GTT,		BYTES, AVERAGE),
 	X("mapped-VRAM",		MAPPED_VRAM,		BYTES, AVERAGE),
 	X("mapped-GTT",			MAPPED_GTT,		BYTES, AVERAGE),
 	X("buffer-wait-time",		BUFFER_WAIT_TIME,	MICROSECONDS, CUMULATIVE),
 	X("num-mapped-buffers",		NUM_MAPPED_BUFFERS,	UINT64, AVERAGE),
 	X("num-GFX-IBs",		NUM_GFX_IBS,		UINT64, AVERAGE),
 	X("num-SDMA-IBs",		NUM_SDMA_IBS,		UINT64, AVERAGE),
 	X("GFX-BO-list-size",		GFX_BO_LIST_SIZE,	UINT64, AVERAGE),
+	X("GFX-IB-size",		GFX_IB_SIZE,		UINT64, AVERAGE),
 	X("num-bytes-moved",		NUM_BYTES_MOVED,	BYTES, CUMULATIVE),
 	X("num-evictions",		NUM_EVICTIONS,		UINT64, CUMULATIVE),
 	X("VRAM-CPU-page-faults",	NUM_VRAM_CPU_PAGE_FAULTS, UINT64, CUMULATIVE),
 	X("VRAM-usage",			VRAM_USAGE,		BYTES, AVERAGE),
 	X("VRAM-vis-usage",		VRAM_VIS_USAGE,		BYTES, AVERAGE),
 	X("GTT-usage",			GTT_USAGE,		BYTES, AVERAGE),
 	X("back-buffer-ps-draw-ratio",	BACK_BUFFER_PS_DRAW_RATIO, UINT64, AVERAGE),
 
 	/* GPIN queries are for the benefit of old versions of GPUPerfStudio,
 	 * which use it as a fallback path to detect the GPU type.
diff --git a/src/gallium/drivers/radeon/r600_query.h b/src/gallium/drivers/radeon/r600_query.h
index a20da07..04943da 100644
--- a/src/gallium/drivers/radeon/r600_query.h
+++ b/src/gallium/drivers/radeon/r600_query.h
@@ -65,20 +65,21 @@  enum {
 	R600_QUERY_GALLIUM_THREAD_BUSY,
 	R600_QUERY_REQUESTED_VRAM,
 	R600_QUERY_REQUESTED_GTT,
 	R600_QUERY_MAPPED_VRAM,
 	R600_QUERY_MAPPED_GTT,
 	R600_QUERY_BUFFER_WAIT_TIME,
 	R600_QUERY_NUM_MAPPED_BUFFERS,
 	R600_QUERY_NUM_GFX_IBS,
 	R600_QUERY_NUM_SDMA_IBS,
 	R600_QUERY_GFX_BO_LIST_SIZE,
+	R600_QUERY_GFX_IB_SIZE,
 	R600_QUERY_NUM_BYTES_MOVED,
 	R600_QUERY_NUM_EVICTIONS,
 	R600_QUERY_NUM_VRAM_CPU_PAGE_FAULTS,
 	R600_QUERY_VRAM_USAGE,
 	R600_QUERY_VRAM_VIS_USAGE,
 	R600_QUERY_GTT_USAGE,
 	R600_QUERY_GPU_TEMPERATURE,
 	R600_QUERY_CURRENT_GPU_SCLK,
 	R600_QUERY_CURRENT_GPU_MCLK,
 	R600_QUERY_GPU_LOAD,
diff --git a/src/gallium/drivers/radeon/radeon_winsys.h b/src/gallium/drivers/radeon/radeon_winsys.h
index 2438ec2..2d3f646 100644
--- a/src/gallium/drivers/radeon/radeon_winsys.h
+++ b/src/gallium/drivers/radeon/radeon_winsys.h
@@ -85,20 +85,21 @@  enum radeon_value_id {
     RADEON_REQUESTED_VRAM_MEMORY,
     RADEON_REQUESTED_GTT_MEMORY,
     RADEON_MAPPED_VRAM,
     RADEON_MAPPED_GTT,
     RADEON_BUFFER_WAIT_TIME_NS,
     RADEON_NUM_MAPPED_BUFFERS,
     RADEON_TIMESTAMP,
     RADEON_NUM_GFX_IBS,
     RADEON_NUM_SDMA_IBS,
     RADEON_GFX_BO_LIST_COUNTER, /* number of BOs submitted in gfx IBs */
+    RADEON_GFX_IB_SIZE_COUNTER,
     RADEON_NUM_BYTES_MOVED,
     RADEON_NUM_EVICTIONS,
     RADEON_NUM_VRAM_CPU_PAGE_FAULTS,
     RADEON_VRAM_USAGE,
     RADEON_VRAM_VIS_USAGE,
     RADEON_GTT_USAGE,
     RADEON_GPU_TEMPERATURE, /* DRM 2.42.0 */
     RADEON_CURRENT_SCLK,
     RADEON_CURRENT_MCLK,
     RADEON_GPU_RESET_COUNTER, /* DRM 2.43.0 */
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c b/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c
index 8fbe8ae..8f43e93 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c
@@ -1416,20 +1416,21 @@  static int amdgpu_cs_flush(struct radeon_winsys_cs *rcs,
       break;
    case RING_GFX:
       /* pad GFX ring to 8 DWs to meet CP fetch alignment requirements */
       if (ws->info.gfx_ib_pad_with_type2) {
          while (rcs->current.cdw & 7)
             radeon_emit(rcs, 0x80000000); /* type2 nop packet */
       } else {
          while (rcs->current.cdw & 7)
             radeon_emit(rcs, 0xffff1000); /* type3 nop packet */
       }
+      ws->gfx_ib_size_counter += (rcs->prev_dw + rcs->current.cdw) * 4;
       break;
    case RING_UVD:
       while (rcs->current.cdw & 15)
          radeon_emit(rcs, 0x80000000); /* type2 nop packet */
       break;
    case RING_VCN_DEC:
       while (rcs->current.cdw & 15)
          radeon_emit(rcs, 0x81ff); /* nop packet */
       break;
    default:
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c
index 5e0c1fd..a210a27 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c
@@ -138,20 +138,22 @@  static uint64_t amdgpu_query_value(struct radeon_winsys *rws,
       return ws->num_mapped_buffers;
    case RADEON_TIMESTAMP:
       amdgpu_query_info(ws->dev, AMDGPU_INFO_TIMESTAMP, 8, &retval);
       return retval;
    case RADEON_NUM_GFX_IBS:
       return ws->num_gfx_IBs;
    case RADEON_NUM_SDMA_IBS:
       return ws->num_sdma_IBs;
    case RADEON_GFX_BO_LIST_COUNTER:
       return ws->gfx_bo_list_counter;
+   case RADEON_GFX_IB_SIZE_COUNTER:
+      return ws->gfx_ib_size_counter;
    case RADEON_NUM_BYTES_MOVED:
       amdgpu_query_info(ws->dev, AMDGPU_INFO_NUM_BYTES_MOVED, 8, &retval);
       return retval;
    case RADEON_NUM_EVICTIONS:
       amdgpu_query_info(ws->dev, AMDGPU_INFO_NUM_EVICTIONS, 8, &retval);
       return retval;
    case RADEON_NUM_VRAM_CPU_PAGE_FAULTS:
       amdgpu_query_info(ws->dev, AMDGPU_INFO_NUM_VRAM_CPU_PAGE_FAULTS, 8, &retval);
       return retval;
    case RADEON_VRAM_USAGE:
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h
index de54470..8b62e2d 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.h
@@ -62,20 +62,21 @@  struct amdgpu_winsys {
    uint32_t next_bo_unique_id;
    uint64_t allocated_vram;
    uint64_t allocated_gtt;
    uint64_t mapped_vram;
    uint64_t mapped_gtt;
    uint64_t buffer_wait_time; /* time spent in buffer_wait in ns */
    uint64_t num_gfx_IBs;
    uint64_t num_sdma_IBs;
    uint64_t num_mapped_buffers;
    uint64_t gfx_bo_list_counter;
+   uint64_t gfx_ib_size_counter;
 
    struct radeon_info info;
 
    /* multithreaded IB submission */
    struct util_queue cs_queue;
 
    struct amdgpu_gpu_info amdinfo;
    ADDR_HANDLE addrlib;
 
    bool check_vm;
diff --git a/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c b/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c
index 542e549..c61f3d9 100644
--- a/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c
+++ b/src/gallium/winsys/radeon/drm/radeon_drm_winsys.c
@@ -625,20 +625,21 @@  static uint64_t radeon_query_value(struct radeon_winsys *rws,
     case RADEON_NUM_SDMA_IBS:
         return ws->num_sdma_IBs;
     case RADEON_NUM_BYTES_MOVED:
         radeon_get_drm_value(ws->fd, RADEON_INFO_NUM_BYTES_MOVED,
                              "num-bytes-moved", (uint32_t*)&retval);
         return retval;
     case RADEON_NUM_EVICTIONS:
     case RADEON_NUM_VRAM_CPU_PAGE_FAULTS:
     case RADEON_VRAM_VIS_USAGE:
     case RADEON_GFX_BO_LIST_COUNTER:
+    case RADEON_GFX_IB_SIZE_COUNTER:
         return 0; /* unimplemented */
     case RADEON_VRAM_USAGE:
         radeon_get_drm_value(ws->fd, RADEON_INFO_VRAM_USAGE,
                              "vram-usage", (uint32_t*)&retval);
         return retval;
     case RADEON_GTT_USAGE:
         radeon_get_drm_value(ws->fd, RADEON_INFO_GTT_USAGE,
                              "gtt-usage", (uint32_t*)&retval);
         return retval;
     case RADEON_GPU_TEMPERATURE: