[g3dvl] move idct texture addr generation into vertex shader

2 files changed, 143 insertions, 131 deletions

diff --git a/src/gallium/auxiliary/vl/vl_idct.c b/src/gallium/auxiliary/vl/vl_idct.c
index 1a29b91cf51..5a32573dd66 100644
--- a/src/gallium/auxiliary/vl/vl_idct.c
+++ b/src/gallium/auxiliary/vl/vl_idct.c
@@ -58,9 +58,10 @@ enum VS_INPUT
 enum VS_OUTPUT
 {
    VS_O_VPOS,
-   VS_O_BLOCK,
-   VS_O_TEX,
-   VS_O_START
+   VS_O_L_ADDR0,
+   VS_O_L_ADDR1,
+   VS_O_R_ADDR0,
+   VS_O_R_ADDR1
 };
 
 static const float const_matrix[8][8] = {
@@ -74,28 +75,58 @@ static const float const_matrix[8][8] = {
    {  0.0975451f, -0.2777850f,  0.4157350f, -0.4903930f,  0.4903930f, -0.4157350f,  0.277786f, -0.0975458f }
 };
 
+static void
+calc_addr(struct ureg_program *shader, struct ureg_dst addr[2],
+          struct ureg_src tc, struct ureg_src start, bool right_side,
+          bool transposed, float size)
+{
+   unsigned wm_start = (right_side == transposed) ? TGSI_WRITEMASK_X : TGSI_WRITEMASK_Y;
+   unsigned sw_start = right_side ? TGSI_SWIZZLE_Y : TGSI_SWIZZLE_X;
+
+   unsigned wm_tc = (right_side == transposed) ? TGSI_WRITEMASK_Y : TGSI_WRITEMASK_X;
+   unsigned sw_tc = right_side ? TGSI_SWIZZLE_X : TGSI_SWIZZLE_Y;
+
+   /*
+    * addr[0..1].(start) = right_side ? start.x : tc.x
+    * addr[0..1].(tc) = right_side ? tc.y : start.y
+    * addr[0..1].z = tc.z
+    * addr[1].(start) += 1.0f / scale
+    */
+   ureg_MOV(shader, ureg_writemask(addr[0], wm_start), ureg_scalar(start, sw_start));
+   ureg_MOV(shader, ureg_writemask(addr[0], wm_tc), ureg_scalar(tc, sw_tc));
+   ureg_MOV(shader, ureg_writemask(addr[0], TGSI_WRITEMASK_Z), tc);
+
+   ureg_ADD(shader, ureg_writemask(addr[1], wm_start), ureg_scalar(start, sw_start), ureg_imm1f(shader, 1.0f / size));
+   ureg_MOV(shader, ureg_writemask(addr[1], wm_tc), ureg_scalar(tc, sw_tc));
+   ureg_MOV(shader, ureg_writemask(addr[1], TGSI_WRITEMASK_Z), tc);
+}
+
 static void *
-create_vert_shader(struct vl_idct *idct)
+create_vert_shader(struct vl_idct *idct, bool matrix_stage)
 {
    struct ureg_program *shader;
    struct ureg_src scale;
    struct ureg_src vrect, vpos;
-   struct ureg_dst t_vpos;
-   struct ureg_dst o_vpos, o_block, o_tex, o_start;
+   struct ureg_dst t_tex, t_start;
+   struct ureg_dst o_vpos, o_l_addr[2], o_r_addr[2];
 
    shader = ureg_create(TGSI_PROCESSOR_VERTEX);
    if (!shader)
       return NULL;
 
-   t_vpos = ureg_DECL_temporary(shader);
+   t_tex = ureg_DECL_temporary(shader);
+   t_start = ureg_DECL_temporary(shader);
 
    vrect = ureg_DECL_vs_input(shader, VS_I_RECT);
    vpos = ureg_DECL_vs_input(shader, VS_I_VPOS);
 
    o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS);
-   o_block = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK);
-   o_tex = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX);
-   o_start = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_START);
+
+   o_l_addr[0] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR0);
+   o_l_addr[1] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR1);
+
+   o_r_addr[0] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_R_ADDR0);
+   o_r_addr[1] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_R_ADDR1);
 
    /*
     * scale = (BLOCK_WIDTH, BLOCK_HEIGHT) / (dst.width, dst.height)
@@ -104,35 +135,35 @@ create_vert_shader(struct vl_idct *idct)
     * o_vpos.xy = t_vpos * scale
     * o_vpos.zw = vpos
     *
-    * o_block = vrect
-    * o_tex = t_pos
-    * o_start = vpos * scale
+    * o_l_addr = calc_addr(...)
+    * o_r_addr = calc_addr(...)
     *
     */
    scale = ureg_imm2f(shader,
       (float)BLOCK_WIDTH / idct->buffer_width,
       (float)BLOCK_HEIGHT / idct->buffer_height);
 
-   ureg_ADD(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), vpos, vrect);
-   ureg_MUL(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos), scale);
-   ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos));
+   ureg_ADD(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_XY), vpos, vrect);
+   ureg_MUL(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_XY), ureg_src(t_tex), scale);
+   ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), ureg_src(t_tex));
    ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), vpos);
 
-   ureg_MOV(shader, ureg_writemask(o_block, TGSI_WRITEMASK_XY), vrect);
-   ureg_MOV(shader, ureg_writemask(o_block, TGSI_WRITEMASK_Z), ureg_imm1f(shader, 0.0f));
-
-   ureg_MOV(shader, ureg_writemask(o_tex, TGSI_WRITEMASK_XY), ureg_src(t_vpos));
-#if NR_RENDER_TARGETS == 1
-   ureg_MOV(shader, ureg_writemask(o_tex, TGSI_WRITEMASK_Z), ureg_imm1f(shader, 0.0f));
-#else
-   ureg_MUL(shader, ureg_writemask(o_tex, TGSI_WRITEMASK_Z), 
+   ureg_MUL(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_Z),
       ureg_scalar(vrect, TGSI_SWIZZLE_X),
       ureg_imm1f(shader, BLOCK_WIDTH / NR_RENDER_TARGETS));
-#endif
 
-   ureg_MUL(shader, ureg_writemask(o_start, TGSI_WRITEMASK_XY), vpos, scale);
+   ureg_MUL(shader, ureg_writemask(t_start, TGSI_WRITEMASK_XY), vpos, scale);
 
-   ureg_release_temporary(shader, t_vpos);
+   if(matrix_stage) {
+      calc_addr(shader, o_l_addr, ureg_src(t_tex), ureg_src(t_start), false, false, idct->buffer_width / 4);
+      calc_addr(shader, o_r_addr, vrect, ureg_imm1f(shader, 0.0f), true, true, BLOCK_WIDTH / 4);
+   } else {
+      calc_addr(shader, o_l_addr, vrect, ureg_imm1f(shader, 0.0f), false, false, BLOCK_WIDTH / 4);
+      calc_addr(shader, o_r_addr, ureg_src(t_tex), ureg_src(t_start), true, false, idct->buffer_height / 4);
+   }
+
+   ureg_release_temporary(shader, t_tex);
+   ureg_release_temporary(shader, t_start);
 
    ureg_END(shader);
 
@@ -140,40 +171,12 @@ create_vert_shader(struct vl_idct *idct)
 }
 
 static void
-calc_addr(struct ureg_program *shader, struct ureg_dst addr[2],
-          struct ureg_src tc, struct ureg_src start, bool right_side,
-          bool transposed, float size)
-{
-   unsigned wm_start = (right_side == transposed) ? TGSI_WRITEMASK_X : TGSI_WRITEMASK_Y;
-   unsigned wm_tc = (right_side == transposed) ? TGSI_WRITEMASK_Y : TGSI_WRITEMASK_X;
-
-   addr[0] = ureg_DECL_temporary(shader);
-   addr[1] = ureg_DECL_temporary(shader);
-
-   /*
-    * t_tc.x = right_side ? start.x : tc.x
-    * t_tc.y = right_side ? tc.y : start.y
-    * m[0..1] = tex(t_tc++, sampler)
-    */
-   if(!right_side) {
-      ureg_MOV(shader, ureg_writemask(addr[0], wm_start), ureg_scalar(start, TGSI_SWIZZLE_X));
-      ureg_MOV(shader, ureg_writemask(addr[0], wm_tc), ureg_scalar(tc, TGSI_SWIZZLE_Y));
-   } else {
-      ureg_MOV(shader, ureg_writemask(addr[0], wm_start), ureg_scalar(start, TGSI_SWIZZLE_Y));
-      ureg_MOV(shader, ureg_writemask(addr[0], wm_tc), ureg_scalar(tc, TGSI_SWIZZLE_X));
-   }
-   ureg_FRC(shader, ureg_writemask(addr[0], TGSI_WRITEMASK_Z), tc);
-
-   ureg_ADD(shader, ureg_writemask(addr[1], wm_start), ureg_src(addr[0]), ureg_imm1f(shader, 1.0f / size));
-   ureg_MOV(shader, ureg_writemask(addr[1], wm_tc | TGSI_WRITEMASK_Z), ureg_src(addr[0]));
-}
-
-static void
 increment_addr(struct ureg_program *shader, struct ureg_dst addr[2],
                bool right_side, bool transposed, float size)
 {
    unsigned wm_tc = (right_side == transposed) ? TGSI_WRITEMASK_Y : TGSI_WRITEMASK_X;
 
+   /* addr[0..1]++ */
    ureg_ADD(shader, ureg_writemask(addr[0], wm_tc),
       ureg_src(addr[0]), ureg_imm1f(shader, 1.0f / size));
    ureg_ADD(shader, ureg_writemask(addr[1], wm_tc),
@@ -181,13 +184,13 @@ increment_addr(struct ureg_program *shader, struct ureg_dst addr[2],
 }
 
 static void
-fetch_four(struct ureg_program *shader, struct ureg_dst m[2], struct ureg_dst addr[2], struct ureg_src sampler)
+fetch_four(struct ureg_program *shader, struct ureg_dst m[2], struct ureg_src addr[2], struct ureg_src sampler)
 {
    m[0] = ureg_DECL_temporary(shader);
    m[1] = ureg_DECL_temporary(shader);
 
-   ureg_TEX(shader, m[0], TGSI_TEXTURE_3D, ureg_src(addr[0]), sampler);
-   ureg_TEX(shader, m[1], TGSI_TEXTURE_3D, ureg_src(addr[1]), sampler);
+   ureg_TEX(shader, m[0], TGSI_TEXTURE_3D, addr[0], sampler);
+   ureg_TEX(shader, m[1], TGSI_TEXTURE_3D, addr[1], sampler);
 }
 
 static void
@@ -211,58 +214,14 @@ matrix_mul(struct ureg_program *shader, struct ureg_dst dst, struct ureg_dst l[2
 }
 
 static void *
-create_transpose_frag_shader(struct vl_idct *idct)
-{
-   struct ureg_program *shader;
-
-   struct ureg_src block, tex, start;
-
-   struct ureg_dst laddr[2], l[2], r[2], raddr[2];
-   struct ureg_dst tmp, fragment;
-
-   shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
-   if (!shader)
-      return NULL;
-
-   block = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK, TGSI_INTERPOLATE_LINEAR);
-   tex = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX, TGSI_INTERPOLATE_CONSTANT);
-   start = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_START, TGSI_INTERPOLATE_CONSTANT);
-
-   calc_addr(shader, laddr, block, ureg_imm1f(shader, 0.0f), false, false, BLOCK_WIDTH / 4);
-   calc_addr(shader, raddr, tex, start, true, false, idct->buffer_height / 4);
-   fetch_four(shader, l, laddr, ureg_DECL_sampler(shader, 0));
-   fetch_four(shader, r, raddr, ureg_DECL_sampler(shader, 1));
-
-   fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
-
-   tmp = ureg_DECL_temporary(shader);
-   matrix_mul(shader, ureg_writemask(tmp, TGSI_WRITEMASK_X), l, r);
-   ureg_MUL(shader, fragment, ureg_src(tmp), ureg_imm1f(shader, STAGE2_SCALE));
-
-   ureg_release_temporary(shader, tmp);
-   ureg_release_temporary(shader, laddr[0]);
-   ureg_release_temporary(shader, laddr[1]);
-   ureg_release_temporary(shader, l[0]);
-   ureg_release_temporary(shader, l[1]);
-   ureg_release_temporary(shader, raddr[0]);
-   ureg_release_temporary(shader, raddr[1]);
-   ureg_release_temporary(shader, r[0]);
-   ureg_release_temporary(shader, r[1]);
-
-   ureg_END(shader);
-
-   return ureg_create_shader_and_destroy(shader, idct->pipe);
-}
-
-static void *
 create_matrix_frag_shader(struct vl_idct *idct)
 {
    struct ureg_program *shader;
 
-   struct ureg_src tex, block, start;
+   struct ureg_src l_addr[2], r_addr[2], saddr[2];
 
-   struct ureg_dst l[4][2], r[2];
-   struct ureg_dst addr[2], fragment[NR_RENDER_TARGETS];
+   struct ureg_dst addr[2], l[4][2], r[2];
+   struct ureg_dst fragment[NR_RENDER_TARGETS];
 
    unsigned i, j;
 
@@ -270,31 +229,39 @@ create_matrix_frag_shader(struct vl_idct *idct)
    if (!shader)
       return NULL;
 
-   tex = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX, TGSI_INTERPOLATE_LINEAR);
-   block = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK, TGSI_INTERPOLATE_LINEAR);
-   start = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_START, TGSI_INTERPOLATE_CONSTANT);
+   addr[0] = ureg_DECL_temporary(shader);
+   addr[1] = ureg_DECL_temporary(shader);
+
+   saddr[0] = ureg_src(addr[0]);
+   saddr[1] = ureg_src(addr[1]);
+
+   l_addr[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR0, TGSI_INTERPOLATE_LINEAR);
+   l_addr[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR1, TGSI_INTERPOLATE_LINEAR);
+
+   r_addr[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_R_ADDR0, TGSI_INTERPOLATE_LINEAR);
+   r_addr[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_R_ADDR1, TGSI_INTERPOLATE_LINEAR);
 
    for (i = 0; i < NR_RENDER_TARGETS; ++i)
        fragment[i] = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, i);
 
    for (i = 0; i < 4; ++i) {
-      if(i == 0)
-         calc_addr(shader, addr, tex, start, false, false, idct->buffer_width / 4);
-      else
+      if(i == 0) {
+         ureg_MOV(shader, addr[0], l_addr[0]);
+         ureg_MOV(shader, addr[1], l_addr[1]);
+      } else
          increment_addr(shader, addr, false, false, idct->buffer_height);
 
-      fetch_four(shader, l[i], addr, ureg_DECL_sampler(shader, 1));
+      fetch_four(shader, l[i], saddr, ureg_DECL_sampler(shader, 1));
    }
-   ureg_release_temporary(shader, addr[0]);
-   ureg_release_temporary(shader, addr[1]);
    
    for (i = 0; i < NR_RENDER_TARGETS; ++i) {
-      if (i == 0)
-         calc_addr(shader, addr, block, ureg_imm1f(shader, 0.0f), true, true, BLOCK_WIDTH / 4);
-      else
+      if(i == 0) {
+         ureg_MOV(shader, addr[0], r_addr[0]);
+         ureg_MOV(shader, addr[1], r_addr[1]);
+      } else
          increment_addr(shader, addr, true, true, BLOCK_HEIGHT);
 
-      fetch_four(shader, r, addr, ureg_DECL_sampler(shader, 0));
+      fetch_four(shader, r, saddr, ureg_DECL_sampler(shader, 0));
 
       for (j = 0; j < 4; ++j) {
          matrix_mul(shader, ureg_writemask(fragment[i], TGSI_WRITEMASK_X << j), l[j], r);
@@ -302,13 +269,53 @@ create_matrix_frag_shader(struct vl_idct *idct)
       ureg_release_temporary(shader, r[0]);
       ureg_release_temporary(shader, r[1]);
    }
-   ureg_release_temporary(shader, addr[0]);
-   ureg_release_temporary(shader, addr[1]);
 
    for (i = 0; i < 4; ++i) {
       ureg_release_temporary(shader, l[i][0]);
       ureg_release_temporary(shader, l[i][1]);
    }
+   ureg_release_temporary(shader, addr[0]);
+   ureg_release_temporary(shader, addr[1]);
+
+   ureg_END(shader);
+
+   return ureg_create_shader_and_destroy(shader, idct->pipe);
+}
+
+static void *
+create_transpose_frag_shader(struct vl_idct *idct)
+{
+   struct ureg_program *shader;
+
+   struct ureg_src l_addr[2], r_addr[2];
+
+   struct ureg_dst l[2], r[2];
+   struct ureg_dst tmp, fragment;
+
+   shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
+   if (!shader)
+      return NULL;
+
+   l_addr[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR0, TGSI_INTERPOLATE_LINEAR);
+   l_addr[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_L_ADDR1, TGSI_INTERPOLATE_LINEAR);
+
+   r_addr[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_R_ADDR0, TGSI_INTERPOLATE_LINEAR);
+   r_addr[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_R_ADDR1, TGSI_INTERPOLATE_LINEAR);
+
+   fetch_four(shader, l, l_addr, ureg_DECL_sampler(shader, 0));
+   fetch_four(shader, r, r_addr, ureg_DECL_sampler(shader, 1));
+
+   fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
+
+   tmp = ureg_DECL_temporary(shader);
+   matrix_mul(shader, ureg_writemask(tmp, TGSI_WRITEMASK_X), l, r);
+   ureg_MUL(shader, fragment, ureg_src(tmp), ureg_imm1f(shader, STAGE2_SCALE));
+
+   ureg_release_temporary(shader, tmp);
+   ureg_release_temporary(shader, l[0]);
+   ureg_release_temporary(shader, l[1]);
+   ureg_release_temporary(shader, r[0]);
+   ureg_release_temporary(shader, r[1]);
 
    ureg_END(shader);
 
@@ -318,21 +325,25 @@ create_matrix_frag_shader(struct vl_idct *idct)
 static bool
 init_shaders(struct vl_idct *idct)
 {
-   idct->vs = create_vert_shader(idct);
+   idct->matrix_vs = create_vert_shader(idct, true);
    idct->matrix_fs = create_matrix_frag_shader(idct);
+
+   idct->transpose_vs = create_vert_shader(idct, false);
    idct->transpose_fs = create_transpose_frag_shader(idct);
 
    return 
-      idct->vs != NULL &&
-      idct->transpose_fs != NULL &&
-      idct->matrix_fs != NULL;
+      idct->matrix_vs != NULL &&
+      idct->matrix_fs != NULL &&
+      idct->transpose_vs != NULL &&
+      idct->transpose_fs != NULL;
 }
 
 static void
 cleanup_shaders(struct vl_idct *idct)
 {
-   idct->pipe->delete_vs_state(idct->pipe, idct->vs);
+   idct->pipe->delete_vs_state(idct->pipe, idct->matrix_vs);
    idct->pipe->delete_fs_state(idct->pipe, idct->matrix_fs);
+   idct->pipe->delete_vs_state(idct->pipe, idct->transpose_vs);
    idct->pipe->delete_fs_state(idct->pipe, idct->transpose_fs);
 }
 
@@ -353,9 +364,9 @@ init_state(struct vl_idct *idct)
 
    for (i = 0; i < 4; ++i) {
       memset(&sampler, 0, sizeof(sampler));
-      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_s = PIPE_TEX_WRAP_REPEAT;
+      sampler.wrap_t = PIPE_TEX_WRAP_REPEAT;
+      sampler.wrap_r = PIPE_TEX_WRAP_REPEAT;
       sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
       sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
       sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
@@ -730,13 +741,13 @@ vl_idct_flush(struct vl_idct *idct, struct vl_idct_buffer *buffer)
       idct->pipe->bind_rasterizer_state(idct->pipe, idct->rs_state);
       idct->pipe->set_vertex_buffers(idct->pipe, 2, buffer->vertex_bufs.all);
       idct->pipe->bind_vertex_elements_state(idct->pipe, idct->vertex_elems_state);
-      idct->pipe->bind_vs_state(idct->pipe, idct->vs);
 
       /* first stage */
       idct->pipe->set_framebuffer_state(idct->pipe, &buffer->fb_state[0]);
       idct->pipe->set_viewport_state(idct->pipe, &buffer->viewport[0]);
       idct->pipe->set_fragment_sampler_views(idct->pipe, 2, buffer->sampler_views.stage[0]);
       idct->pipe->bind_fragment_sampler_states(idct->pipe, 2, idct->samplers.stage[0]);
+      idct->pipe->bind_vs_state(idct->pipe, idct->matrix_vs);
       idct->pipe->bind_fs_state(idct->pipe, idct->matrix_fs);
       util_draw_arrays(idct->pipe, PIPE_PRIM_QUADS, 0, num_verts);
 
@@ -745,6 +756,7 @@ vl_idct_flush(struct vl_idct *idct, struct vl_idct_buffer *buffer)
       idct->pipe->set_viewport_state(idct->pipe, &buffer->viewport[1]);
       idct->pipe->set_fragment_sampler_views(idct->pipe, 2, buffer->sampler_views.stage[1]);
       idct->pipe->bind_fragment_sampler_states(idct->pipe, 2, idct->samplers.stage[1]);
+      idct->pipe->bind_vs_state(idct->pipe, idct->transpose_vs);
       idct->pipe->bind_fs_state(idct->pipe, idct->transpose_fs);
       util_draw_arrays(idct->pipe, PIPE_PRIM_QUADS, 0, num_verts);
    }
diff --git a/src/gallium/auxiliary/vl/vl_idct.h b/src/gallium/auxiliary/vl/vl_idct.h
index 6076bdec46c..fcba75a7607 100644
--- a/src/gallium/auxiliary/vl/vl_idct.h
+++ b/src/gallium/auxiliary/vl/vl_idct.h
@@ -53,7 +53,7 @@ struct vl_idct
       } individual;
    } samplers;
 
-   void *vs;
+   void *matrix_vs, *transpose_vs;
    void *matrix_fs, *transpose_fs;
 
    struct pipe_resource *matrix;