r600g,radeonsi: consolidate variables for CS tracing

Reviewed-by: Michel Dänzer <[email protected]>
Reviewed-by: Tom Stellard <[email protected]>

10 files changed, 42 insertions, 46 deletions

diff --git a/src/gallium/drivers/r600/r600_hw_context.c b/src/gallium/drivers/r600/r600_hw_context.c
index d6b0280affa..9b0c558c29e 100644
--- a/src/gallium/drivers/r600/r600_hw_context.c
+++ b/src/gallium/drivers/r600/r600_hw_context.c
@@ -54,7 +54,7 @@ void r600_need_cs_space(struct r600_context *ctx, unsigned num_dw,
 		for (i = 0; i < R600_NUM_ATOMS; i++) {
 			if (ctx->atoms[i] && ctx->atoms[i]->dirty) {
 				num_dw += ctx->atoms[i]->num_dw;
-				if (ctx->screen->trace_bo) {
+				if (ctx->screen->b.trace_bo) {
 					num_dw += R600_TRACE_CS_DWORDS;
 				}
 			}
@@ -62,7 +62,7 @@ void r600_need_cs_space(struct r600_context *ctx, unsigned num_dw,
 
 		/* The upper-bound of how much space a draw command would take. */
 		num_dw += R600_MAX_FLUSH_CS_DWORDS + R600_MAX_DRAW_CS_DWORDS;
-		if (ctx->screen->trace_bo) {
+		if (ctx->screen->b.trace_bo) {
 			num_dw += R600_TRACE_CS_DWORDS;
 		}
 	}
@@ -270,7 +270,7 @@ void r600_context_flush(struct r600_context *ctx, unsigned flags)
 	}
 
 	/* Flush the CS. */
-	ctx->b.ws->cs_flush(ctx->b.rings.gfx.cs, flags, ctx->screen->cs_count++);
+	ctx->b.ws->cs_flush(ctx->b.rings.gfx.cs, flags, ctx->screen->b.cs_count++);
 }
 
 void r600_begin_new_cs(struct r600_context *ctx)
diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c
index a7138969d96..eec04bfaebb 100644
--- a/src/gallium/drivers/r600/r600_pipe.c
+++ b/src/gallium/drivers/r600/r600_pipe.c
@@ -259,8 +259,8 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void
 		goto fail;
 	}
 
-	if (rscreen->trace_bo) {
-		rctx->b.rings.gfx.cs = rctx->b.ws->cs_create(rctx->b.ws, RING_GFX, rscreen->trace_bo->cs_buf);
+	if (rscreen->b.trace_bo) {
+		rctx->b.rings.gfx.cs = rctx->b.ws->cs_create(rctx->b.ws, RING_GFX, rscreen->b.trace_bo->cs_buf);
 	} else {
 		rctx->b.rings.gfx.cs = rctx->b.ws->cs_create(rctx->b.ws, RING_GFX, NULL);
 	}
@@ -743,9 +743,9 @@ static void r600_destroy_screen(struct pipe_screen* pscreen)
 		compute_memory_pool_delete(rscreen->global_pool);
 	}
 
-	if (rscreen->trace_bo) {
-		rscreen->b.ws->buffer_unmap(rscreen->trace_bo->cs_buf);
-		pipe_resource_reference((struct pipe_resource**)&rscreen->trace_bo, NULL);
+	if (rscreen->b.trace_bo) {
+		rscreen->b.ws->buffer_unmap(rscreen->b.trace_bo->cs_buf);
+		pipe_resource_reference((struct pipe_resource**)&rscreen->b.trace_bo, NULL);
 	}
 
 	rscreen->b.ws->destroy(rscreen->b.ws);
@@ -860,14 +860,14 @@ struct pipe_screen *r600_screen_create(struct radeon_winsys *ws)
 
 	rscreen->global_pool = compute_memory_pool_new(rscreen);
 
-	rscreen->cs_count = 0;
+	rscreen->b.cs_count = 0;
 	if (rscreen->b.info.drm_minor >= 28 && (rscreen->b.debug_flags & DBG_TRACE_CS)) {
-		rscreen->trace_bo = (struct r600_resource*)pipe_buffer_create(&rscreen->b.b,
+		rscreen->b.trace_bo = (struct r600_resource*)pipe_buffer_create(&rscreen->b.b,
 										PIPE_BIND_CUSTOM,
 										PIPE_USAGE_STAGING,
 										4096);
-		if (rscreen->trace_bo) {
-			rscreen->trace_ptr = rscreen->b.ws->buffer_map(rscreen->trace_bo->cs_buf, NULL,
+		if (rscreen->b.trace_bo) {
+			rscreen->b.trace_ptr = rscreen->b.ws->buffer_map(rscreen->b.trace_bo->cs_buf, NULL,
 									PIPE_TRANSFER_UNSYNCHRONIZED);
 		}
 	}
diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h
index 101d7405ee1..112cb269d85 100644
--- a/src/gallium/drivers/r600/r600_pipe.h
+++ b/src/gallium/drivers/r600/r600_pipe.h
@@ -204,9 +204,6 @@ struct r600_screen {
 	 * XXX: Not sure if this is the best place for global_pool.  Also,
 	 * it's not thread safe, so it won't work with multiple contexts. */
 	struct compute_memory_pool *global_pool;
-	struct r600_resource		*trace_bo;
-	uint32_t			*trace_ptr;
-	unsigned			cs_count;
 };
 
 struct r600_pipe_sampler_view {
@@ -461,7 +458,7 @@ static INLINE void r600_emit_atom(struct r600_context *rctx, struct r600_atom *a
 {
 	atom->emit(&rctx->b, atom);
 	atom->dirty = false;
-	if (rctx->screen->trace_bo) {
+	if (rctx->screen->b.trace_bo) {
 		r600_trace_emit(rctx);
 	}
 }
diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index 3a6f16753f4..d8fab1083e1 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -1389,7 +1389,7 @@ static void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info
 					(info.count_from_stream_output ? S_0287F0_USE_OPAQUE(1) : 0);
 	}
 
-	if (rctx->screen->trace_bo) {
+	if (rctx->screen->b.trace_bo) {
 		r600_trace_emit(rctx);
 	}
 
@@ -2190,13 +2190,13 @@ void r600_trace_emit(struct r600_context *rctx)
 	uint64_t va;
 	uint32_t reloc;
 
-	va = r600_resource_va(&rscreen->b.b, (void*)rscreen->trace_bo);
-	reloc = r600_context_bo_reloc(&rctx->b, &rctx->b.rings.gfx, rscreen->trace_bo, RADEON_USAGE_READWRITE);
+	va = r600_resource_va(&rscreen->b.b, (void*)rscreen->b.trace_bo);
+	reloc = r600_context_bo_reloc(&rctx->b, &rctx->b.rings.gfx, rscreen->b.trace_bo, RADEON_USAGE_READWRITE);
 	radeon_emit(cs, PKT3(PKT3_MEM_WRITE, 3, 0));
 	radeon_emit(cs, va & 0xFFFFFFFFUL);
 	radeon_emit(cs, (va >> 32UL) & 0xFFUL);
 	radeon_emit(cs, cs->cdw);
-	radeon_emit(cs, rscreen->cs_count);
+	radeon_emit(cs, rscreen->b.cs_count);
 	radeon_emit(cs, PKT3(PKT3_NOP, 0, 0));
 	radeon_emit(cs, reloc);
 }
diff --git a/src/gallium/drivers/radeon/r600_pipe_common.h b/src/gallium/drivers/radeon/r600_pipe_common.h
index 5496a12f2c1..d0e0c436e0b 100644
--- a/src/gallium/drivers/radeon/r600_pipe_common.h
+++ b/src/gallium/drivers/radeon/r600_pipe_common.h
@@ -180,6 +180,10 @@ struct r600_common_screen {
 	 * It must be locked prior to using and flushed before unlocking. */
 	struct pipe_context		*aux_context;
 	pipe_mutex			aux_context_lock;
+
+	struct r600_resource		*trace_bo;
+	uint32_t			*trace_ptr;
+	unsigned			cs_count;
 };
 
 /* This encapsulates a state or an operation which can emitted into the GPU
diff --git a/src/gallium/drivers/radeonsi/si_hw_context.c b/src/gallium/drivers/radeonsi/si_hw_context.c
index a7536e3fbbf..bb3d0809e22 100644
--- a/src/gallium/drivers/radeonsi/si_hw_context.c
+++ b/src/gallium/drivers/radeonsi/si_hw_context.c
@@ -71,7 +71,7 @@ void si_need_cs_space(struct si_context *ctx, unsigned num_dw,
 	num_dw += ctx->atoms.cache_flush->num_dw;
 
 #if SI_TRACE_CS
-	if (ctx->screen->trace_bo) {
+	if (ctx->screen->b.trace_bo) {
 		num_dw += SI_TRACE_CS_DWORDS;
 	}
 #endif
@@ -118,14 +118,14 @@ void si_context_flush(struct si_context *ctx, unsigned flags)
 	flags |= RADEON_FLUSH_KEEP_TILING_FLAGS;
 
 #if SI_TRACE_CS
-	if (ctx->screen->trace_bo) {
+	if (ctx->screen->b.trace_bo) {
 		struct si_screen *sscreen = ctx->screen;
 		unsigned i;
 
 		for (i = 0; i < cs->cdw; i++) {
-			fprintf(stderr, "[%4d] [%5d] 0x%08x\n", sscreen->cs_count, i, cs->buf[i]);
+			fprintf(stderr, "[%4d] [%5d] 0x%08x\n", sscreen->b.cs_count, i, cs->buf[i]);
 		}
-		sscreen->cs_count++;
+		sscreen->b.cs_count++;
 	}
 #endif
 
@@ -133,21 +133,21 @@ void si_context_flush(struct si_context *ctx, unsigned flags)
 	ctx->b.ws->cs_flush(ctx->b.rings.gfx.cs, flags, 0);
 
 #if SI_TRACE_CS
-	if (ctx->screen->trace_bo) {
+	if (ctx->screen->b.trace_bo) {
 		struct si_screen *sscreen = ctx->screen;
 		unsigned i;
 
 		for (i = 0; i < 10; i++) {
 			usleep(5);
-			if (!ctx->ws->buffer_is_busy(sscreen->trace_bo->buf, RADEON_USAGE_READWRITE)) {
+			if (!ctx->ws->buffer_is_busy(sscreen->b.trace_bo->buf, RADEON_USAGE_READWRITE)) {
 				break;
 			}
 		}
 		if (i == 10) {
 			fprintf(stderr, "timeout on cs lockup likely happen at cs %d dw %d\n",
-				sscreen->trace_ptr[1], sscreen->trace_ptr[0]);
+				sscreen->b.trace_ptr[1], sscreen->b.trace_ptr[0]);
 		} else {
-			fprintf(stderr, "cs %d executed in %dms\n", sscreen->trace_ptr[1], i * 5);
+			fprintf(stderr, "cs %d executed in %dms\n", sscreen->b.trace_ptr[1], i * 5);
 		}
 	}
 #endif
@@ -193,8 +193,8 @@ void si_trace_emit(struct si_context *sctx)
 	struct radeon_winsys_cs *cs = sctx->cs;
 	uint64_t va;
 
-	va = r600_resource_va(&sscreen->screen, (void*)sscreen->trace_bo);
-	r600_context_bo_reloc(sctx, sscreen->trace_bo, RADEON_USAGE_READWRITE);
+	va = r600_resource_va(&sscreen->screen, (void*)sscreen->b.trace_bo);
+	r600_context_bo_reloc(sctx, sscreen->b.trace_bo, RADEON_USAGE_READWRITE);
 	cs->buf[cs->cdw++] = PKT3(PKT3_WRITE_DATA, 4, 0);
 	cs->buf[cs->cdw++] = PKT3_WRITE_DATA_DST_SEL(PKT3_WRITE_DATA_DST_SEL_MEM_SYNC) |
 				PKT3_WRITE_DATA_WR_CONFIRM |
@@ -202,6 +202,6 @@ void si_trace_emit(struct si_context *sctx)
 	cs->buf[cs->cdw++] = va & 0xFFFFFFFFUL;
 	cs->buf[cs->cdw++] = (va >> 32UL) & 0xFFFFFFFFUL;
 	cs->buf[cs->cdw++] = cs->cdw;
-	cs->buf[cs->cdw++] = sscreen->cs_count;
+	cs->buf[cs->cdw++] = sscreen->b.cs_count;
 }
 #endif
diff --git a/src/gallium/drivers/radeonsi/si_pipe.c b/src/gallium/drivers/radeonsi/si_pipe.c
index 8666b5b071c..6aaf59568c4 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.c
+++ b/src/gallium/drivers/radeonsi/si_pipe.c
@@ -564,9 +564,9 @@ static void si_destroy_screen(struct pipe_screen* pscreen)
 	r600_common_screen_cleanup(&sscreen->b);
 
 #if SI_TRACE_CS
-	if (sscreen->trace_bo) {
-		sscreen->ws->buffer_unmap(sscreen->trace_bo->cs_buf);
-		pipe_resource_reference((struct pipe_resource**)&sscreen->trace_bo, NULL);
+	if (sscreen->b.trace_bo) {
+		sscreen->ws->buffer_unmap(sscreen->b.trace_bo->cs_buf);
+		pipe_resource_reference((struct pipe_resource**)&sscreen->b.trace_bo, NULL);
 	}
 #endif
 
@@ -611,14 +611,14 @@ struct pipe_screen *radeonsi_screen_create(struct radeon_winsys *ws)
 		sscreen->b.debug_flags |= DBG_FS | DBG_VS | DBG_GS | DBG_PS | DBG_CS;
 
 #if SI_TRACE_CS
-	sscreen->cs_count = 0;
+	sscreen->b.cs_count = 0;
 	if (sscreen->info.drm_minor >= 28) {
-		sscreen->trace_bo = (struct r600_resource*)pipe_buffer_create(&sscreen->screen,
+		sscreen->b.trace_bo = (struct r600_resource*)pipe_buffer_create(&sscreen->screen,
 										PIPE_BIND_CUSTOM,
 										PIPE_USAGE_STAGING,
 										4096);
-		if (sscreen->trace_bo) {
-			sscreen->trace_ptr = sscreen->ws->buffer_map(sscreen->trace_bo->cs_buf, NULL,
+		if (sscreen->b.trace_bo) {
+			sscreen->b.trace_ptr = sscreen->ws->buffer_map(sscreen->b.trace_bo->cs_buf, NULL,
 									PIPE_TRANSFER_UNSYNCHRONIZED);
 		}
 	}
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index aa62852b093..e137bfc58d4 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -55,11 +55,6 @@ struct si_pipe_compute;
 
 struct si_screen {
 	struct r600_common_screen	b;
-#if SI_TRACE_CS
-	struct r600_resource		*trace_bo;
-	uint32_t			*trace_ptr;
-	unsigned			cs_count;
-#endif
 };
 
 struct si_pipe_sampler_view {
diff --git a/src/gallium/drivers/radeonsi/si_pm4.c b/src/gallium/drivers/radeonsi/si_pm4.c
index 5b132850272..c8769370db0 100644
--- a/src/gallium/drivers/radeonsi/si_pm4.c
+++ b/src/gallium/drivers/radeonsi/si_pm4.c
@@ -202,7 +202,7 @@ unsigned si_pm4_dirty_dw(struct si_context *sctx)
 		count += state->ndw;
 #if SI_TRACE_CS
 		/* for tracing each states */
-		if (sctx->screen->trace_bo) {
+		if (sctx->screen->b.trace_bo) {
 			count += SI_TRACE_CS_DWORDS;
 		}
 #endif
@@ -228,7 +228,7 @@ void si_pm4_emit(struct si_context *sctx, struct si_pm4_state *state)
 	cs->cdw += state->ndw;
 
 #if SI_TRACE_CS
-	if (sctx->screen->trace_bo) {
+	if (sctx->screen->b.trace_bo) {
 		si_trace_emit(sctx);
 	}
 #endif
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index 5c2ff8a249e..45dd4baa470 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -750,7 +750,7 @@ void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info)
 	sctx->pm4_dirty_cdwords = 0;
 
 #if SI_TRACE_CS
-	if (sctx->screen->trace_bo) {
+	if (sctx->screen->b.trace_bo) {
 		si_trace_emit(sctx);
 	}
 #endif