freedreno: Rename vp and fp to vs and fs in fd_program_stateobj
[mesa.git] / src / gallium / drivers / freedreno / a4xx / fd4_gmem.c
index ed2307d0964251dac5a14e7d26314f90cb013bec..663a49c69c88822c9cf2d8d412821186ad3cedb3 100644 (file)
@@ -1,5 +1,3 @@
-/* -*- mode: C; c-file-style: "k&r"; tab-width 4; indent-tabs-mode: t; -*- */
-
 /*
  * Copyright (C) 2014 Rob Clark <robclark@freedesktop.org>
  *
@@ -79,8 +77,9 @@ emit_mrt(struct fd_ringbuffer *ring, unsigned nr_bufs,
                         */
                        if (rsc->stencil) {
                                rsc = rsc->stencil;
-                               pformat = rsc->base.b.format;
-                               bases++;
+                               pformat = rsc->base.format;
+                               if (bases)
+                                       bases++;
                        }
 
                        slice = fd_resource_slice(rsc, psurf->u.tex.level);
@@ -131,13 +130,14 @@ emit_mrt(struct fd_ringbuffer *ring, unsigned nr_bufs,
 }
 
 static bool
-use_hw_binning(struct fd_context *ctx)
+use_hw_binning(struct fd_batch *batch)
 {
-       struct fd_gmem_stateobj *gmem = &ctx->gmem;
-       struct pipe_framebuffer_state *pfb = &ctx->framebuffer;
+       struct fd_gmem_stateobj *gmem = &batch->ctx->gmem;
+
+       if ((gmem->maxpw * gmem->maxph) > 32)
+               return false;
 
-       /* this seems to be a hw bug.. but this hack fixes piglit fbo-maxsize: */
-       if ((pfb->width > 4096) && (pfb->height > 4096))
+       if ((gmem->maxpw > 15) || (gmem->maxph > 15))
                return false;
 
        return fd_binning_enabled && ((gmem->nbins_x * gmem->nbins_y) > 2);
@@ -146,19 +146,22 @@ use_hw_binning(struct fd_context *ctx)
 /* transfer from gmem to system memory (ie. normal RAM) */
 
 static void
-emit_gmem2mem_surf(struct fd_context *ctx, bool stencil,
+emit_gmem2mem_surf(struct fd_batch *batch, bool stencil,
                uint32_t base, struct pipe_surface *psurf)
 {
-       struct fd_ringbuffer *ring = ctx->ring;
+       struct fd_ringbuffer *ring = batch->gmem;
        struct fd_resource *rsc = fd_resource(psurf->texture);
        enum pipe_format pformat = psurf->format;
        struct fd_resource_slice *slice;
        uint32_t offset;
 
+       if (!rsc->valid)
+               return;
+
        if (stencil) {
                debug_assert(rsc->stencil);
                rsc = rsc->stencil;
-               pformat = rsc->base.b.format;
+               pformat = rsc->base.format;
        }
 
        slice = &rsc->slices[psurf->u.tex.level];
@@ -179,19 +182,20 @@ emit_gmem2mem_surf(struct fd_context *ctx, bool stencil,
                        A4XX_RB_COPY_DEST_INFO_ENDIAN(ENDIAN_NONE) |
                        A4XX_RB_COPY_DEST_INFO_SWAP(fd4_pipe2swap(pformat)));
 
-       fd4_draw(ctx, ring, DI_PT_RECTLIST, IGNORE_VISIBILITY,
-                       DI_SRC_SEL_AUTO_INDEX, 2, 1, INDEX_SIZE_IGN, 0, 0, NULL);
+       fd4_draw(batch, ring, DI_PT_RECTLIST, IGNORE_VISIBILITY,
+                       DI_SRC_SEL_AUTO_INDEX, 2, 1, INDEX4_SIZE_8_BIT, 0, 0, NULL);
 }
 
 static void
-fd4_emit_tile_gmem2mem(struct fd_context *ctx, struct fd_tile *tile)
+fd4_emit_tile_gmem2mem(struct fd_batch *batch, struct fd_tile *tile)
 {
-       struct fd4_context *fd4_ctx = fd4_context(ctx);
+       struct fd_context *ctx = batch->ctx;
        struct fd_gmem_stateobj *gmem = &ctx->gmem;
-       struct fd_ringbuffer *ring = ctx->ring;
-       struct pipe_framebuffer_state *pfb = &ctx->framebuffer;
+       struct fd_ringbuffer *ring = batch->gmem;
+       struct pipe_framebuffer_state *pfb = &batch->framebuffer;
        struct fd4_emit emit = {
-                       .vtx = &fd4_ctx->solid_vbuf_state,
+                       .debug = &ctx->debug,
+                       .vtx = &ctx->solid_vbuf_state,
                        .prog = &ctx->solid_prog,
                        .key = {
                                .half_precision = true,
@@ -225,7 +229,7 @@ fd4_emit_tile_gmem2mem(struct fd_context *ctx, struct fd_tile *tile)
        OUT_PKT0(ring, REG_A4XX_GRAS_SU_MODE_CONTROL, 1);
        OUT_RING(ring, A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH(0));
 
-       fd_wfi(ctx, ring);
+       fd_wfi(batch, ring);
 
        OUT_PKT0(ring, REG_A4XX_GRAS_CL_CLIP_CNTL, 1);
        OUT_RING(ring, 0x80000);      /* GRAS_CL_CLIP_CNTL */
@@ -267,22 +271,22 @@ fd4_emit_tile_gmem2mem(struct fd_context *ctx, struct fd_tile *tile)
        fd4_program_emit(ring, &emit, 0, NULL);
        fd4_emit_vertex_bufs(ring, &emit);
 
-       if (ctx->resolve & (FD_BUFFER_DEPTH | FD_BUFFER_STENCIL)) {
+       if (batch->resolve & (FD_BUFFER_DEPTH | FD_BUFFER_STENCIL)) {
                struct fd_resource *rsc = fd_resource(pfb->zsbuf->texture);
-               if (!rsc->stencil || (ctx->resolve & FD_BUFFER_DEPTH))
-                       emit_gmem2mem_surf(ctx, false, ctx->gmem.zsbuf_base[0], pfb->zsbuf);
-               if (rsc->stencil && (ctx->resolve & FD_BUFFER_STENCIL))
-                       emit_gmem2mem_surf(ctx, true, ctx->gmem.zsbuf_base[1], pfb->zsbuf);
+               if (!rsc->stencil || (batch->resolve & FD_BUFFER_DEPTH))
+                       emit_gmem2mem_surf(batch, false, ctx->gmem.zsbuf_base[0], pfb->zsbuf);
+               if (rsc->stencil && (batch->resolve & FD_BUFFER_STENCIL))
+                       emit_gmem2mem_surf(batch, true, ctx->gmem.zsbuf_base[1], pfb->zsbuf);
        }
 
-       if (ctx->resolve & FD_BUFFER_COLOR) {
+       if (batch->resolve & FD_BUFFER_COLOR) {
                unsigned i;
                for (i = 0; i < pfb->nr_cbufs; i++) {
                        if (!pfb->cbufs[i])
                                continue;
-                       if (!(ctx->resolve & (PIPE_CLEAR_COLOR0 << i)))
+                       if (!(batch->resolve & (PIPE_CLEAR_COLOR0 << i)))
                                continue;
-                       emit_gmem2mem_surf(ctx, false, gmem->cbuf_base[i], pfb->cbufs[i]);
+                       emit_gmem2mem_surf(batch, false, gmem->cbuf_base[i], pfb->cbufs[i]);
                }
        }
 
@@ -296,10 +300,10 @@ fd4_emit_tile_gmem2mem(struct fd_context *ctx, struct fd_tile *tile)
 /* transfer from system memory to gmem */
 
 static void
-emit_mem2gmem_surf(struct fd_context *ctx, uint32_t *bases,
+emit_mem2gmem_surf(struct fd_batch *batch, uint32_t *bases,
                struct pipe_surface **bufs, uint32_t nr_bufs, uint32_t bin_w)
 {
-       struct fd_ringbuffer *ring = ctx->ring;
+       struct fd_ringbuffer *ring = batch->gmem;
        struct pipe_surface *zsbufs[2];
 
        emit_mrt(ring, nr_bufs, bufs, bases, bin_w, false);
@@ -316,19 +320,20 @@ emit_mem2gmem_surf(struct fd_context *ctx, uint32_t *bases,
 
        fd4_emit_gmem_restore_tex(ring, nr_bufs, bufs);
 
-       fd4_draw(ctx, ring, DI_PT_RECTLIST, IGNORE_VISIBILITY,
-                       DI_SRC_SEL_AUTO_INDEX, 2, 1, INDEX_SIZE_IGN, 0, 0, NULL);
+       fd4_draw(batch, ring, DI_PT_RECTLIST, IGNORE_VISIBILITY,
+                       DI_SRC_SEL_AUTO_INDEX, 2, 1, INDEX4_SIZE_8_BIT, 0, 0, NULL);
 }
 
 static void
-fd4_emit_tile_mem2gmem(struct fd_context *ctx, struct fd_tile *tile)
+fd4_emit_tile_mem2gmem(struct fd_batch *batch, struct fd_tile *tile)
 {
-       struct fd4_context *fd4_ctx = fd4_context(ctx);
+       struct fd_context *ctx = batch->ctx;
        struct fd_gmem_stateobj *gmem = &ctx->gmem;
-       struct fd_ringbuffer *ring = ctx->ring;
-       struct pipe_framebuffer_state *pfb = &ctx->framebuffer;
+       struct fd_ringbuffer *ring = batch->gmem;
+       struct pipe_framebuffer_state *pfb = &batch->framebuffer;
        struct fd4_emit emit = {
-                       .vtx = &fd4_ctx->blit_vbuf_state,
+                       .debug = &ctx->debug,
+                       .vtx = &ctx->blit_vbuf_state,
                        .sprite_coord_enable = 1,
                        /* NOTE: They all use the same VP, this is for vtx bufs. */
                        .prog = &ctx->blit_prog[0],
@@ -350,7 +355,7 @@ fd4_emit_tile_mem2gmem(struct fd_context *ctx, struct fd_tile *tile)
        y1 = ((float)tile->yoff + bin_h) / ((float)pfb->height);
 
        OUT_PKT3(ring, CP_MEM_WRITE, 5);
-       OUT_RELOCW(ring, fd_resource(fd4_ctx->blit_texcoord_vbuf)->bo, 0, 0, 0);
+       OUT_RELOCW(ring, fd_resource(ctx->blit_texcoord_vbuf)->bo, 0, 0, 0);
        OUT_RING(ring, fui(x0));
        OUT_RING(ring, fui(y0));
        OUT_RING(ring, fui(x1));
@@ -452,14 +457,14 @@ fd4_emit_tile_mem2gmem(struct fd_context *ctx, struct fd_tile *tile)
        bin_w = gmem->bin_w;
        bin_h = gmem->bin_h;
 
-       if (fd_gmem_needs_restore(ctx, tile, FD_BUFFER_COLOR)) {
+       if (fd_gmem_needs_restore(batch, tile, FD_BUFFER_COLOR)) {
                emit.prog = &ctx->blit_prog[pfb->nr_cbufs - 1];
-               emit.fp = NULL;      /* frag shader changed so clear cache */
+               emit.fs = NULL;      /* frag shader changed so clear cache */
                fd4_program_emit(ring, &emit, pfb->nr_cbufs, pfb->cbufs);
-               emit_mem2gmem_surf(ctx, gmem->cbuf_base, pfb->cbufs, pfb->nr_cbufs, bin_w);
+               emit_mem2gmem_surf(batch, gmem->cbuf_base, pfb->cbufs, pfb->nr_cbufs, bin_w);
        }
 
-       if (fd_gmem_needs_restore(ctx, tile, FD_BUFFER_DEPTH | FD_BUFFER_STENCIL)) {
+       if (fd_gmem_needs_restore(batch, tile, FD_BUFFER_DEPTH | FD_BUFFER_STENCIL)) {
                switch (pfb->zsbuf->format) {
                case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT:
                case PIPE_FORMAT_Z32_FLOAT:
@@ -488,9 +493,9 @@ fd4_emit_tile_mem2gmem(struct fd_context *ctx, struct fd_tile *tile)
                        emit.key.half_precision = true;
                        break;
                }
-               emit.fp = NULL;      /* frag shader changed so clear cache */
+               emit.fs = NULL;      /* frag shader changed so clear cache */
                fd4_program_emit(ring, &emit, 1, &pfb->zsbuf);
-               emit_mem2gmem_surf(ctx, gmem->zsbuf_base, &pfb->zsbuf, 1, bin_w);
+               emit_mem2gmem_surf(batch, gmem->zsbuf_base, &pfb->zsbuf, 1, bin_w);
        }
 
        OUT_PKT0(ring, REG_A4XX_GRAS_SC_CONTROL, 1);
@@ -505,24 +510,24 @@ fd4_emit_tile_mem2gmem(struct fd_context *ctx, struct fd_tile *tile)
 }
 
 static void
-patch_draws(struct fd_context *ctx, enum pc_di_vis_cull_mode vismode)
+patch_draws(struct fd_batch *batch, enum pc_di_vis_cull_mode vismode)
 {
        unsigned i;
-       for (i = 0; i < fd_patch_num_elements(&ctx->draw_patches); i++) {
-               struct fd_cs_patch *patch = fd_patch_element(&ctx->draw_patches, i);
+       for (i = 0; i < fd_patch_num_elements(&batch->draw_patches); i++) {
+               struct fd_cs_patch *patch = fd_patch_element(&batch->draw_patches, i);
                *patch->cs = patch->val | DRAW4(0, 0, 0, vismode);
        }
-       util_dynarray_resize(&ctx->draw_patches, 0);
+       util_dynarray_clear(&batch->draw_patches);
 }
 
 /* for rendering directly to system memory: */
 static void
-fd4_emit_sysmem_prep(struct fd_context *ctx)
+fd4_emit_sysmem_prep(struct fd_batch *batch)
 {
-       struct pipe_framebuffer_state *pfb = &ctx->framebuffer;
-       struct fd_ringbuffer *ring = ctx->ring;
+       struct pipe_framebuffer_state *pfb = &batch->framebuffer;
+       struct fd_ringbuffer *ring = batch->gmem;
 
-       fd4_emit_restore(ctx);
+       fd4_emit_restore(batch, ring);
 
        OUT_PKT0(ring, REG_A4XX_RB_FRAME_BUFFER_DIMENSION, 1);
        OUT_RING(ring, A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH(pfb->width) |
@@ -549,14 +554,15 @@ fd4_emit_sysmem_prep(struct fd_context *ctx)
        OUT_PKT0(ring, REG_A4XX_RB_RENDER_CONTROL, 1);
        OUT_RING(ring, 0x8);
 
-       patch_draws(ctx, IGNORE_VISIBILITY);
+       patch_draws(batch, IGNORE_VISIBILITY);
 }
 
 static void
-update_vsc_pipe(struct fd_context *ctx)
+update_vsc_pipe(struct fd_batch *batch)
 {
+       struct fd_context *ctx = batch->ctx;
        struct fd4_context *fd4_ctx = fd4_context(ctx);
-       struct fd_ringbuffer *ring = ctx->ring;
+       struct fd_ringbuffer *ring = batch->gmem;
        int i;
 
        OUT_PKT0(ring, REG_A4XX_VSC_SIZE_ADDRESS, 1);
@@ -564,7 +570,7 @@ update_vsc_pipe(struct fd_context *ctx)
 
        OUT_PKT0(ring, REG_A4XX_VSC_PIPE_CONFIG_REG(0), 8);
        for (i = 0; i < 8; i++) {
-               struct fd_vsc_pipe *pipe = &ctx->pipe[i];
+               struct fd_vsc_pipe *pipe = &ctx->vsc_pipe[i];
                OUT_RING(ring, A4XX_VSC_PIPE_CONFIG_REG_X(pipe->x) |
                                A4XX_VSC_PIPE_CONFIG_REG_Y(pipe->y) |
                                A4XX_VSC_PIPE_CONFIG_REG_W(pipe->w) |
@@ -573,27 +579,28 @@ update_vsc_pipe(struct fd_context *ctx)
 
        OUT_PKT0(ring, REG_A4XX_VSC_PIPE_DATA_ADDRESS_REG(0), 8);
        for (i = 0; i < 8; i++) {
-               struct fd_vsc_pipe *pipe = &ctx->pipe[i];
+               struct fd_vsc_pipe *pipe = &ctx->vsc_pipe[i];
                if (!pipe->bo) {
                        pipe->bo = fd_bo_new(ctx->dev, 0x40000,
-                                       DRM_FREEDRENO_GEM_TYPE_KMEM);
+                                       DRM_FREEDRENO_GEM_TYPE_KMEM, "vsc_pipe[%u]", i);
                }
                OUT_RELOCW(ring, pipe->bo, 0, 0, 0);       /* VSC_PIPE_DATA_ADDRESS[i] */
        }
 
        OUT_PKT0(ring, REG_A4XX_VSC_PIPE_DATA_LENGTH_REG(0), 8);
        for (i = 0; i < 8; i++) {
-               struct fd_vsc_pipe *pipe = &ctx->pipe[i];
+               struct fd_vsc_pipe *pipe = &ctx->vsc_pipe[i];
                OUT_RING(ring, fd_bo_size(pipe->bo) - 32); /* VSC_PIPE_DATA_LENGTH[i] */
        }
 }
 
 static void
-emit_binning_pass(struct fd_context *ctx)
+emit_binning_pass(struct fd_batch *batch)
 {
+       struct fd_context *ctx = batch->ctx;
        struct fd_gmem_stateobj *gmem = &ctx->gmem;
-       struct pipe_framebuffer_state *pfb = &ctx->framebuffer;
-       struct fd_ringbuffer *ring = ctx->ring;
+       struct pipe_framebuffer_state *pfb = &batch->framebuffer;
+       struct fd_ringbuffer *ring = batch->gmem;
        int i;
 
        uint32_t x1 = gmem->minx;
@@ -632,10 +639,10 @@ emit_binning_pass(struct fd_context *ctx)
        }
 
        /* emit IB to binning drawcmds: */
-       ctx->emit_ib(ring, ctx->binning_start, ctx->binning_end);
+       fd4_emit_ib(ring, batch->binning);
 
-       fd_reset_wfi(ctx);
-       fd_wfi(ctx, ring);
+       fd_reset_wfi(batch);
+       fd_wfi(batch, ring);
 
        /* and then put stuff back the way it was: */
 
@@ -648,26 +655,32 @@ emit_binning_pass(struct fd_context *ctx)
                        A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES(MSAA_ONE) |
                        A4XX_GRAS_SC_CONTROL_RASTER_MODE(0));
 
-       fd_event_write(ctx, ring, CACHE_FLUSH);
-       fd_wfi(ctx, ring);
+       fd_event_write(batch, ring, CACHE_FLUSH);
+       fd_wfi(batch, ring);
 }
 
 /* before first tile */
 static void
-fd4_emit_tile_init(struct fd_context *ctx)
+fd4_emit_tile_init(struct fd_batch *batch)
 {
-       struct fd_ringbuffer *ring = ctx->ring;
-       struct fd_gmem_stateobj *gmem = &ctx->gmem;
+       struct fd_ringbuffer *ring = batch->gmem;
+       struct pipe_framebuffer_state *pfb = &batch->framebuffer;
+       struct fd_gmem_stateobj *gmem = &batch->ctx->gmem;
 
-       fd4_emit_restore(ctx);
+       fd4_emit_restore(batch, ring);
 
        OUT_PKT0(ring, REG_A4XX_VSC_BIN_SIZE, 1);
        OUT_RING(ring, A4XX_VSC_BIN_SIZE_WIDTH(gmem->bin_w) |
                        A4XX_VSC_BIN_SIZE_HEIGHT(gmem->bin_h));
 
-       update_vsc_pipe(ctx);
+       update_vsc_pipe(batch);
 
-       if (use_hw_binning(ctx)) {
+       fd_wfi(batch, ring);
+       OUT_PKT0(ring, REG_A4XX_RB_FRAME_BUFFER_DIMENSION, 1);
+       OUT_RING(ring, A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH(pfb->width) |
+                       A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT(pfb->height));
+
+       if (use_hw_binning(batch)) {
                OUT_PKT0(ring, REG_A4XX_RB_MODE_CONTROL, 1);
                OUT_RING(ring, A4XX_RB_MODE_CONTROL_WIDTH(gmem->bin_w) |
                                A4XX_RB_MODE_CONTROL_HEIGHT(gmem->bin_h));
@@ -678,11 +691,11 @@ fd4_emit_tile_init(struct fd_context *ctx)
                                0x8);
 
                /* emit hw binning pass: */
-               emit_binning_pass(ctx);
+               emit_binning_pass(batch);
 
-               patch_draws(ctx, USE_VISIBILITY);
+               patch_draws(batch, USE_VISIBILITY);
        } else {
-               patch_draws(ctx, IGNORE_VISIBILITY);
+               patch_draws(batch, IGNORE_VISIBILITY);
        }
 
        OUT_PKT0(ring, REG_A4XX_RB_MODE_CONTROL, 1);
@@ -693,10 +706,11 @@ fd4_emit_tile_init(struct fd_context *ctx)
 
 /* before mem2gmem */
 static void
-fd4_emit_tile_prep(struct fd_context *ctx, struct fd_tile *tile)
+fd4_emit_tile_prep(struct fd_batch *batch, struct fd_tile *tile)
 {
-       struct fd_ringbuffer *ring = ctx->ring;
-       struct pipe_framebuffer_state *pfb = &ctx->framebuffer;
+       struct fd_context *ctx = batch->ctx;
+       struct fd_ringbuffer *ring = batch->gmem;
+       struct pipe_framebuffer_state *pfb = &batch->framebuffer;
        struct fd_gmem_stateobj *gmem = &ctx->gmem;
 
        if (pfb->zsbuf) {
@@ -736,45 +750,38 @@ fd4_emit_tile_prep(struct fd_context *ctx, struct fd_tile *tile)
        } else {
                OUT_RING(ring, A4XX_GRAS_DEPTH_CONTROL_FORMAT(DEPTH4_NONE));
        }
-
-       if (ctx->needs_rb_fbd) {
-               fd_wfi(ctx, ring);
-               OUT_PKT0(ring, REG_A4XX_RB_FRAME_BUFFER_DIMENSION, 1);
-               OUT_RING(ring, A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH(pfb->width) |
-                               A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT(pfb->height));
-               ctx->needs_rb_fbd = false;
-       }
 }
 
 /* before IB to rendering cmds: */
 static void
-fd4_emit_tile_renderprep(struct fd_context *ctx, struct fd_tile *tile)
+fd4_emit_tile_renderprep(struct fd_batch *batch, struct fd_tile *tile)
 {
+       struct fd_context *ctx = batch->ctx;
        struct fd4_context *fd4_ctx = fd4_context(ctx);
-       struct fd_ringbuffer *ring = ctx->ring;
+       struct fd_ringbuffer *ring = batch->gmem;
        struct fd_gmem_stateobj *gmem = &ctx->gmem;
-       struct pipe_framebuffer_state *pfb = &ctx->framebuffer;
+       struct pipe_framebuffer_state *pfb = &batch->framebuffer;
 
        uint32_t x1 = tile->xoff;
        uint32_t y1 = tile->yoff;
        uint32_t x2 = tile->xoff + tile->bin_w - 1;
        uint32_t y2 = tile->yoff + tile->bin_h - 1;
 
-       if (use_hw_binning(ctx)) {
-               struct fd_vsc_pipe *pipe = &ctx->pipe[tile->p];
+       if (use_hw_binning(batch)) {
+               struct fd_vsc_pipe *pipe = &ctx->vsc_pipe[tile->p];
 
-               assert(pipe->w * pipe->h);
+               assert(pipe->w && pipe->h);
 
-               fd_event_write(ctx, ring, HLSQ_FLUSH);
-               fd_wfi(ctx, ring);
+               fd_event_write(batch, ring, HLSQ_FLUSH);
+               fd_wfi(batch, ring);
 
                OUT_PKT0(ring, REG_A4XX_PC_VSTREAM_CONTROL, 1);
                OUT_RING(ring, A4XX_PC_VSTREAM_CONTROL_SIZE(pipe->w * pipe->h) |
                                A4XX_PC_VSTREAM_CONTROL_N(tile->n));
 
                OUT_PKT3(ring, CP_SET_BIN_DATA, 2);
-               OUT_RELOC(ring, pipe->bo, 0, 0, 0);    /* BIN_DATA_ADDR <- VSC_PIPE[p].DATA_ADDRESS */
-               OUT_RELOC(ring, fd4_ctx->vsc_size_mem, /* BIN_SIZE_ADDR <- VSC_SIZE_ADDRESS + (p * 4) */
+               OUT_RELOCW(ring, pipe->bo, 0, 0, 0);    /* BIN_DATA_ADDR <- VSC_PIPE[p].DATA_ADDRESS */
+               OUT_RELOCW(ring, fd4_ctx->vsc_size_mem, /* BIN_SIZE_ADDR <- VSC_SIZE_ADDRESS + (p * 4) */
                                (tile->p * 4), 0, 0);
        } else {
                OUT_PKT0(ring, REG_A4XX_PC_VSTREAM_CONTROL, 1);