freedreno: per-generation OUT_IB packet

Some a4xx firmware doesn't implement the "PFD" (prefetch-disabled)
version of the CP_INDIRECT_BUFFER packet.  So allow for PFD vs PFE per
generation.  Switch a3xx and a4xx over to using prefetch-enabled version
(which is also what blob does.. it seems only on a2xx we cannot use
PFE).

Signed-off-by: Rob Clark <[email protected]>

9 files changed, 43 insertions, 6 deletions

diff --git a/src/gallium/drivers/freedreno/a2xx/fd2_context.c b/src/gallium/drivers/freedreno/a2xx/fd2_context.c
index 3bed73573a6..058f8219ed5 100644
--- a/src/gallium/drivers/freedreno/a2xx/fd2_context.c
+++ b/src/gallium/drivers/freedreno/a2xx/fd2_context.c
@@ -109,6 +109,7 @@ fd2_context_create(struct pipe_screen *pscreen, void *priv, unsigned flags)
 	fd2_gmem_init(pctx);
 	fd2_texture_init(pctx);
 	fd2_prog_init(pctx);
+	fd2_emit_init(pctx);
 
 	pctx = fd_context_init(&fd2_ctx->base, pscreen,
 			(screen->gpu_id >= 220) ? a22x_primtypes : a20x_primtypes,
diff --git a/src/gallium/drivers/freedreno/a2xx/fd2_emit.c b/src/gallium/drivers/freedreno/a2xx/fd2_emit.c
index cc0ed59f300..4f667ab7d57 100644
--- a/src/gallium/drivers/freedreno/a2xx/fd2_emit.c
+++ b/src/gallium/drivers/freedreno/a2xx/fd2_emit.c
@@ -446,3 +446,17 @@ fd2_emit_setup(struct fd_context *ctx)
 	fd_ringbuffer_flush(ring);
 	fd_ringmarker_mark(ctx->draw_start);
 }
+
+static void
+fd2_emit_ib(struct fd_ringbuffer *ring, struct fd_ringmarker *start,
+		struct fd_ringmarker *end)
+{
+	__OUT_IB(ring, false, start, end);
+}
+
+void
+fd2_emit_init(struct pipe_context *pctx)
+{
+	struct fd_context *ctx = fd_context(pctx);
+	ctx->emit_ib = fd2_emit_ib;
+}
diff --git a/src/gallium/drivers/freedreno/a2xx/fd2_emit.h b/src/gallium/drivers/freedreno/a2xx/fd2_emit.h
index 8ee04632091..3c146c17151 100644
--- a/src/gallium/drivers/freedreno/a2xx/fd2_emit.h
+++ b/src/gallium/drivers/freedreno/a2xx/fd2_emit.h
@@ -45,4 +45,6 @@ void fd2_emit_vertex_bufs(struct fd_ringbuffer *ring, uint32_t val,
 void fd2_emit_state(struct fd_context *ctx, uint32_t dirty);
 void fd2_emit_setup(struct fd_context *ctx);
 
+void fd2_emit_init(struct pipe_context *pctx);
+
 #endif /* FD2_EMIT_H */
diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_emit.c b/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
index e65a352e7f6..811f58bbba2 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_emit.c
@@ -891,10 +891,18 @@ fd3_emit_restore(struct fd_context *ctx)
 	ctx->needs_rb_fbd = true;
 }
 
+static void
+fd3_emit_ib(struct fd_ringbuffer *ring, struct fd_ringmarker *start,
+		struct fd_ringmarker *end)
+{
+	__OUT_IB(ring, true, start, end);
+}
+
 void
 fd3_emit_init(struct pipe_context *pctx)
 {
 	struct fd_context *ctx = fd_context(pctx);
 	ctx->emit_const = fd3_emit_const;
 	ctx->emit_const_bo = fd3_emit_const_bo;
+	ctx->emit_ib = fd3_emit_ib;
 }
diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_gmem.c b/src/gallium/drivers/freedreno/a3xx/fd3_gmem.c
index 21fb59e450d..2ce393a41ae 100644
--- a/src/gallium/drivers/freedreno/a3xx/fd3_gmem.c
+++ b/src/gallium/drivers/freedreno/a3xx/fd3_gmem.c
@@ -853,7 +853,7 @@ emit_binning_pass(struct fd_context *ctx)
 			A3XX_PC_VSTREAM_CONTROL_N(0));
 
 	/* emit IB to binning drawcmds: */
-	OUT_IB(ring, ctx->binning_start, ctx->binning_end);
+	ctx->emit_ib(ring, ctx->binning_start, ctx->binning_end);
 	fd_reset_wfi(ctx);
 
 	fd_wfi(ctx, ring);
diff --git a/src/gallium/drivers/freedreno/a4xx/fd4_emit.c b/src/gallium/drivers/freedreno/a4xx/fd4_emit.c
index bc62a5d9a4b..4a3f1da30ed 100644
--- a/src/gallium/drivers/freedreno/a4xx/fd4_emit.c
+++ b/src/gallium/drivers/freedreno/a4xx/fd4_emit.c
@@ -885,10 +885,18 @@ fd4_emit_restore(struct fd_context *ctx)
 	ctx->needs_rb_fbd = true;
 }
 
+static void
+fd4_emit_ib(struct fd_ringbuffer *ring, struct fd_ringmarker *start,
+		struct fd_ringmarker *end)
+{
+	__OUT_IB(ring, true, start, end);
+}
+
 void
 fd4_emit_init(struct pipe_context *pctx)
 {
 	struct fd_context *ctx = fd_context(pctx);
 	ctx->emit_const = fd4_emit_const;
 	ctx->emit_const_bo = fd4_emit_const_bo;
+	ctx->emit_ib = fd4_emit_ib;
 }
diff --git a/src/gallium/drivers/freedreno/freedreno_context.h b/src/gallium/drivers/freedreno/freedreno_context.h
index 418b71b95de..9e7130ab915 100644
--- a/src/gallium/drivers/freedreno/freedreno_context.h
+++ b/src/gallium/drivers/freedreno/freedreno_context.h
@@ -386,6 +386,10 @@ struct fd_context {
 			const uint32_t *dwords, struct pipe_resource *prsc);
 	void (*emit_const_bo)(struct fd_ringbuffer *ring, enum shader_t type, boolean write,
 			uint32_t regid, uint32_t num, struct fd_bo **bos, uint32_t *offsets);
+
+	/* indirect-branch emit: */
+	void (*emit_ib)(struct fd_ringbuffer *ring, struct fd_ringmarker *start,
+			struct fd_ringmarker *end);
 };
 
 static inline struct fd_context *
diff --git a/src/gallium/drivers/freedreno/freedreno_gmem.c b/src/gallium/drivers/freedreno/freedreno_gmem.c
index 648db9baee5..0d73349057c 100644
--- a/src/gallium/drivers/freedreno/freedreno_gmem.c
+++ b/src/gallium/drivers/freedreno/freedreno_gmem.c
@@ -331,7 +331,7 @@ render_tiles(struct fd_context *ctx)
 		fd_hw_query_prepare_tile(ctx, i, ctx->ring);
 
 		/* emit IB to drawcmds: */
-		OUT_IB(ctx->ring, ctx->draw_start, ctx->draw_end);
+		ctx->emit_ib(ctx->ring, ctx->draw_start, ctx->draw_end);
 		fd_reset_wfi(ctx);
 
 		/* emit gmem2mem to transfer tile back to system memory: */
@@ -349,7 +349,7 @@ render_sysmem(struct fd_context *ctx)
 	fd_hw_query_prepare_tile(ctx, 0, ctx->ring);
 
 	/* emit IB to drawcmds: */
-	OUT_IB(ctx->ring, ctx->draw_start, ctx->draw_end);
+	ctx->emit_ib(ctx->ring, ctx->draw_start, ctx->draw_end);
 	fd_reset_wfi(ctx);
 }
 
diff --git a/src/gallium/drivers/freedreno/freedreno_util.h b/src/gallium/drivers/freedreno/freedreno_util.h
index 0d2418e1e00..47dd467f498 100644
--- a/src/gallium/drivers/freedreno/freedreno_util.h
+++ b/src/gallium/drivers/freedreno/freedreno_util.h
@@ -265,8 +265,8 @@ OUT_WFI(struct fd_ringbuffer *ring)
 }
 
 static inline void
-OUT_IB(struct fd_ringbuffer *ring, struct fd_ringmarker *start,
-		struct fd_ringmarker *end)
+__OUT_IB(struct fd_ringbuffer *ring, bool prefetch,
+		struct fd_ringmarker *start, struct fd_ringmarker *end)
 {
 	uint32_t dwords = fd_ringmarker_dwords(start, end);
 
@@ -280,7 +280,7 @@ OUT_IB(struct fd_ringbuffer *ring, struct fd_ringmarker *start,
 	 */
 	emit_marker(ring, 6);
 
-	OUT_PKT3(ring, CP_INDIRECT_BUFFER_PFD, 2);
+	OUT_PKT3(ring, prefetch ? CP_INDIRECT_BUFFER_PFE : CP_INDIRECT_BUFFER_PFD, 2);
 	fd_ringbuffer_emit_reloc_ring(ring, start, end);
 	OUT_RING(ring, dwords);