#define SCALE_FACTOR_16_TO_9 (32768.0f / 256.0f)
-#define STAGE1_SCALE 4.0f
-#define STAGE2_SCALE (SCALE_FACTOR_16_TO_9 / STAGE1_SCALE)
-
-#define NR_RENDER_TARGETS 1
-
-enum VS_INPUT
-{
- VS_I_RECT,
- VS_I_VPOS,
-
- NUM_VS_INPUTS
-};
+#define NR_RENDER_TARGETS 4
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] = {
{ 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, int color_swizzle)
{
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_src vrect, vpos, vblock, eb[4];
+ struct ureg_src scale, blocks_xy, t_eb;
+ struct ureg_dst t_tex, t_start;
+ struct ureg_dst o_vpos, o_l_addr[2], o_r_addr[2];
+ unsigned label;
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);
+ vblock = ureg_swizzle(vrect, TGSI_SWIZZLE_Z, TGSI_SWIZZLE_W, TGSI_SWIZZLE_X, TGSI_SWIZZLE_X);
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);
+
+ eb[0] = ureg_DECL_vs_input(shader, VS_I_EB_0_0);
+ eb[1] = ureg_DECL_vs_input(shader, VS_I_EB_1_0);
+ eb[2] = ureg_DECL_vs_input(shader, VS_I_EB_0_1);
+ eb[3] = ureg_DECL_vs_input(shader, VS_I_EB_1_1);
+
+ 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)
+ * blocks_xy = (blocks_x, blocks_y)
*
- * t_vpos = vpos + vrect
- * o_vpos.xy = t_vpos * scale
- * o_vpos.zw = vpos
+ * ar = vblock.y * blocks.x + vblock.x
+ * if eb[ar].(color_swizzle)
+ * o_vpos.xy = -1
+ * else
+ * t_tex = vpos * blocks_xy + vblock
+ * t_start = t_tex * scale
+ * t_tex = t_tex + vrect
+ * o_vpos.xy = t_tex * scale
*
- * o_block = vrect
- * o_tex = t_pos
- * o_start = vpos * scale
+ * o_l_addr = calc_addr(...)
+ * o_r_addr = calc_addr(...)
+ * endif
+ * o_vpos.zw = vpos
*
*/
+
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_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), vpos);
+ blocks_xy = ureg_imm2f(shader, idct->blocks_x, idct->blocks_y);
+
+ if (idct->blocks_x > 1 || idct->blocks_y > 1) {
+ struct ureg_dst ar = ureg_DECL_address(shader);
+
+ ureg_MAD(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_X),
+ ureg_scalar(vblock, TGSI_SWIZZLE_Y), blocks_xy, vblock);
+
+ ureg_ARL(shader, ureg_writemask(ar, TGSI_WRITEMASK_X), ureg_src(t_tex));
+ t_eb = ureg_src_indirect(eb[0], ureg_src(ar));
+ } else {
+ t_eb = eb[0];
+ }
+
+ ureg_IF(shader, ureg_scalar(t_eb, color_swizzle), &label);
+
+ ureg_MOV(shader, o_vpos, ureg_imm1f(shader, -1.0f));
- ureg_MOV(shader, ureg_writemask(o_block, TGSI_WRITEMASK_XY), vrect);
- ureg_MOV(shader, ureg_writemask(o_tex, TGSI_WRITEMASK_XY), ureg_src(t_vpos));
- ureg_MUL(shader, ureg_writemask(o_start, TGSI_WRITEMASK_XY), vpos, scale);
+ ureg_fixup_label(shader, label, ureg_get_instruction_number(shader));
+ ureg_ELSE(shader, &label);
+
+ ureg_MAD(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_XY), vpos, blocks_xy, vblock);
+ ureg_MUL(shader, ureg_writemask(t_start, TGSI_WRITEMASK_XY), ureg_src(t_tex), scale);
+
+ ureg_ADD(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_XY), ureg_src(t_tex), vrect);
+
+ ureg_MUL(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_XY), ureg_src(t_tex), scale);
+ ureg_MUL(shader, ureg_writemask(t_tex, TGSI_WRITEMASK_Z),
+ ureg_scalar(vrect, TGSI_SWIZZLE_X),
+ ureg_imm1f(shader, BLOCK_WIDTH / NR_RENDER_TARGETS));
+
+ ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), ureg_src(t_tex));
+
+ 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_fixup_label(shader, label, ureg_get_instruction_number(shader));
+ ureg_ENDIF(shader);
+
+ ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), vpos);
- ureg_release_temporary(shader, t_vpos);
+ ureg_release_temporary(shader, t_tex);
+ ureg_release_temporary(shader, t_start);
ureg_END(shader);
}
static void
-fetch_four(struct ureg_program *shader, struct ureg_dst m[2],
- struct ureg_src tc, struct ureg_src sampler,
- struct ureg_src start, struct ureg_src block,
- bool right_side, bool transposed, float size)
+increment_addr(struct ureg_program *shader, struct ureg_dst daddr[2],
+ struct ureg_src saddr[2], bool right_side, bool transposed,
+ int pos, float size)
{
- struct ureg_dst t_tc;
unsigned wm_start = (right_side == transposed) ? TGSI_WRITEMASK_X : TGSI_WRITEMASK_Y;
unsigned wm_tc = (right_side == transposed) ? TGSI_WRITEMASK_Y : TGSI_WRITEMASK_X;
- t_tc = ureg_DECL_temporary(shader);
- m[0] = ureg_DECL_temporary(shader);
- m[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)
+ * daddr[0..1].(start) = saddr[0..1].(start)
+ * daddr[0..1].(tc) = saddr[0..1].(tc)
*/
- if(!right_side) {
- ureg_MOV(shader, ureg_writemask(t_tc, wm_start), ureg_scalar(start, TGSI_SWIZZLE_X));
- ureg_MOV(shader, ureg_writemask(t_tc, wm_tc), ureg_scalar(tc, TGSI_SWIZZLE_Y));
- } else {
- ureg_MOV(shader, ureg_writemask(t_tc, wm_start), ureg_scalar(start, TGSI_SWIZZLE_Y));
- ureg_MOV(shader, ureg_writemask(t_tc, wm_tc), ureg_scalar(tc, TGSI_SWIZZLE_X));
- }
-#if NR_RENDER_TARGETS == 8
- ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Z), ureg_scalar(block, TGSI_SWIZZLE_X));
-#else
- ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Z), ureg_imm1f(shader, 0.0f));
-#endif
-
- ureg_TEX(shader, m[0], TGSI_TEXTURE_3D, ureg_src(t_tc), sampler);
- ureg_ADD(shader, ureg_writemask(t_tc, wm_start), ureg_src(t_tc), ureg_imm1f(shader, 1.0f / size));
- ureg_TEX(shader, m[1], TGSI_TEXTURE_3D, ureg_src(t_tc), sampler);
+ ureg_MOV(shader, ureg_writemask(daddr[0], wm_start), saddr[0]);
+ ureg_ADD(shader, ureg_writemask(daddr[0], wm_tc), saddr[0], ureg_imm1f(shader, pos / size));
+ ureg_MOV(shader, ureg_writemask(daddr[1], wm_start), saddr[1]);
+ ureg_ADD(shader, ureg_writemask(daddr[1], wm_tc), saddr[1], ureg_imm1f(shader, pos / size));
+}
- ureg_release_temporary(shader, t_tc);
+static void
+fetch_four(struct ureg_program *shader, struct ureg_dst m[2], struct ureg_src addr[2], struct ureg_src 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
matrix_mul(struct ureg_program *shader, struct ureg_dst dst, struct ureg_dst l[2], struct ureg_dst r[2])
{
- struct ureg_dst tmp[2];
- unsigned i;
+ struct ureg_dst tmp;
- for(i = 0; i < 2; ++i) {
- tmp[i] = ureg_DECL_temporary(shader);
- }
+ tmp = ureg_DECL_temporary(shader);
/*
- * tmp[0..1] = dot4(m[0][0..1], m[1][0..1])
- * dst = tmp[0] + tmp[1]
+ * tmp.xy = dot4(m[0][0..1], m[1][0..1])
+ * dst = tmp.x + tmp.y
*/
- ureg_DP4(shader, ureg_writemask(tmp[0], TGSI_WRITEMASK_X), ureg_src(l[0]), ureg_src(r[0]));
- ureg_DP4(shader, ureg_writemask(tmp[1], TGSI_WRITEMASK_X), ureg_src(l[1]), ureg_src(r[1]));
+ ureg_DP4(shader, ureg_writemask(tmp, TGSI_WRITEMASK_X), ureg_src(l[0]), ureg_src(r[0]));
+ ureg_DP4(shader, ureg_writemask(tmp, TGSI_WRITEMASK_Y), ureg_src(l[1]), ureg_src(r[1]));
ureg_ADD(shader, dst,
- ureg_scalar(ureg_src(tmp[0]), TGSI_SWIZZLE_X),
- ureg_scalar(ureg_src(tmp[1]), TGSI_SWIZZLE_X));
+ ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_X),
+ ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_Y));
- for(i = 0; i < 2; ++i) {
- ureg_release_temporary(shader, tmp[i]);
- }
+ ureg_release_temporary(shader, tmp);
}
static void *
-create_transpose_frag_shader(struct vl_idct *idct)
+create_matrix_frag_shader(struct vl_idct *idct)
{
struct ureg_program *shader;
- struct ureg_src block, tex, sampler[2];
- struct ureg_src start[2];
+ struct ureg_src l_addr[2], r_addr[2];
- struct ureg_dst l[2], r[2];
- struct ureg_dst tmp, fragment;
+ struct ureg_dst l[4][2], r[2];
+ struct ureg_dst fragment[NR_RENDER_TARGETS];
+
+ unsigned i, j;
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);
+ 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);
- sampler[0] = ureg_DECL_sampler(shader, 0);
- sampler[1] = ureg_DECL_sampler(shader, 1);
+ 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);
- start[0] = ureg_imm1f(shader, 0.0f);
- start[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_START, TGSI_INTERPOLATE_CONSTANT);
+ for (i = 0; i < NR_RENDER_TARGETS; ++i)
+ fragment[i] = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, i);
- fetch_four(shader, l, block, sampler[0], start[0], block, false, false, BLOCK_WIDTH / 4);
- fetch_four(shader, r, tex, sampler[1], start[1], block, true, false, idct->buffer_height / 4);
+ for (i = 0; i < 4; ++i) {
+ l[i][0] = ureg_DECL_temporary(shader);
+ l[i][1] = ureg_DECL_temporary(shader);
+ }
- fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
+ r[0] = ureg_DECL_temporary(shader);
+ r[1] = ureg_DECL_temporary(shader);
- 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));
+ for (i = 1; i < 4; ++i) {
+ increment_addr(shader, l[i], l_addr, false, false, i, idct->buffer_height);
+ }
- ureg_release_temporary(shader, tmp);
- ureg_release_temporary(shader, l[0]);
- ureg_release_temporary(shader, l[1]);
+ for (i = 0; i < 4; ++i) {
+ struct ureg_src s_addr[2];
+ s_addr[0] = i == 0 ? l_addr[0] : ureg_src(l[i][0]);
+ s_addr[1] = i == 0 ? l_addr[1] : ureg_src(l[i][1]);
+ fetch_four(shader, l[i], s_addr, ureg_DECL_sampler(shader, 1));
+ }
+
+ for (i = 0; i < NR_RENDER_TARGETS; ++i) {
+ if(i > 0)
+ increment_addr(shader, r, r_addr, true, true, i, BLOCK_HEIGHT);
+
+ struct ureg_src s_addr[2] = { ureg_src(r[0]), ureg_src(r[1]) };
+ s_addr[0] = i == 0 ? r_addr[0] : ureg_src(r[0]);
+ s_addr[1] = i == 0 ? r_addr[1] : ureg_src(r[1]);
+ fetch_four(shader, r, s_addr, 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);
+ }
+ }
+
+ for (i = 0; i < 4; ++i) {
+ ureg_release_temporary(shader, l[i][0]);
+ ureg_release_temporary(shader, l[i][1]);
+ }
ureg_release_temporary(shader, r[0]);
ureg_release_temporary(shader, r[1]);
}
static void *
-create_matrix_frag_shader(struct vl_idct *idct)
+create_transpose_frag_shader(struct vl_idct *idct)
{
struct ureg_program *shader;
- struct ureg_src tex, block, sampler[2];
- struct ureg_src start[2];
+ struct ureg_src l_addr[2], r_addr[2];
- struct ureg_dst l[4][2], r[2];
- struct ureg_dst t_tc, tmp, fragment[NR_RENDER_TARGETS];
-
- unsigned i, j;
+ struct ureg_dst l[2], r[2];
+ struct ureg_dst fragment;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return NULL;
- t_tc = ureg_DECL_temporary(shader);
- tmp = ureg_DECL_temporary(shader);
-
- 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);
+ 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);
- sampler[0] = ureg_DECL_sampler(shader, 1);
- sampler[1] = ureg_DECL_sampler(shader, 0);
-
- start[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_START, TGSI_INTERPOLATE_CONSTANT);
- start[1] = ureg_imm1f(shader, 0.0f);
-
- for (i = 0; i < NR_RENDER_TARGETS; ++i)
- fragment[i] = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, i);
+ 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);
- ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Y), tex);
- for (i = 0; i < 4; ++i) {
- fetch_four(shader, l[i], ureg_src(t_tc), sampler[0], start[0], block, false, false, idct->buffer_width / 4);
- ureg_MUL(shader, l[i][0], ureg_src(l[i][0]), ureg_imm1f(shader, STAGE1_SCALE));
- ureg_MUL(shader, l[i][1], ureg_src(l[i][1]), ureg_imm1f(shader, STAGE1_SCALE));
- if(i != 3)
- ureg_ADD(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Y),
- ureg_src(t_tc), ureg_imm1f(shader, 1.0f / idct->buffer_height));
- }
-
- for (i = 0; i < NR_RENDER_TARGETS; ++i) {
+ l[0] = ureg_DECL_temporary(shader);
+ l[1] = ureg_DECL_temporary(shader);
+ r[0] = ureg_DECL_temporary(shader);
+ r[1] = ureg_DECL_temporary(shader);
-#if NR_RENDER_TARGETS == 8
- ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_X), ureg_imm1f(shader, 1.0f / BLOCK_WIDTH * i));
- fetch_four(shader, r, ureg_src(t_tc), sampler[1], start[1], block, true, true, BLOCK_WIDTH / 4);
-#elif NR_RENDER_TARGETS == 1
- fetch_four(shader, r, block, sampler[1], start[1], block, true, true, BLOCK_WIDTH / 4);
-#else
-#error invalid number of render targets
-#endif
+ fetch_four(shader, l, l_addr, ureg_DECL_sampler(shader, 0));
+ fetch_four(shader, r, r_addr, ureg_DECL_sampler(shader, 1));
- for (j = 0; j < 4; ++j) {
- matrix_mul(shader, ureg_writemask(fragment[i], TGSI_WRITEMASK_X << j), l[j], r);
- }
- ureg_release_temporary(shader, r[0]);
- ureg_release_temporary(shader, r[1]);
- }
+ fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
- ureg_release_temporary(shader, t_tc);
- ureg_release_temporary(shader, tmp);
+ matrix_mul(shader, ureg_writemask(fragment, TGSI_WRITEMASK_X), l, r);
- for (i = 0; i < 4; ++i) {
- ureg_release_temporary(shader, l[i][0]);
- ureg_release_temporary(shader, l[i][1]);
- }
+ 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);
}
static bool
-init_shaders(struct vl_idct *idct)
+init_shaders(struct vl_idct *idct, int color_swizzle)
{
- idct->vs = create_vert_shader(idct);
+ idct->matrix_vs = create_vert_shader(idct, true, color_swizzle);
idct->matrix_fs = create_matrix_frag_shader(idct);
+
+ idct->transpose_vs = create_vert_shader(idct, false, color_swizzle);
idct->transpose_fs = create_transpose_frag_shader(idct);
- return
- idct->vs != NULL &&
- idct->transpose_fs != NULL &&
- idct->matrix_fs != NULL;
+ return
+ 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);
}
static bool
init_state(struct vl_idct *idct)
{
- struct pipe_vertex_element vertex_elems[NUM_VS_INPUTS];
struct pipe_sampler_state sampler;
struct pipe_rasterizer_state rs_state;
unsigned i;
assert(idct);
- idct->quad = vl_vb_upload_quads(idct->pipe, idct->max_blocks);
-
- if(idct->quad.buffer == NULL)
- return false;
-
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;
rs_state.gl_rasterization_rules = false;
idct->rs_state = idct->pipe->create_rasterizer_state(idct->pipe, &rs_state);
- vertex_elems[VS_I_RECT] = vl_vb_get_quad_vertex_element();
-
- /* Pos element */
- vertex_elems[VS_I_VPOS].src_format = PIPE_FORMAT_R32G32_FLOAT;
-
- idct->vertex_buffer_stride = vl_vb_element_helper(&vertex_elems[VS_I_VPOS], 1, 1);
- idct->vertex_elems_state = idct->pipe->create_vertex_elements_state(idct->pipe, 2, vertex_elems);
-
return true;
}
idct->pipe->delete_sampler_state(idct->pipe, idct->samplers.all[i]);
idct->pipe->delete_rasterizer_state(idct->pipe, idct->rs_state);
- idct->pipe->delete_vertex_elements_state(idct->pipe, idct->vertex_elems_state);
}
static bool
/* create textures */
memset(&template, 0, sizeof(struct pipe_resource));
template.last_level = 0;
- template.depth0 = 1;
template.bind = PIPE_BIND_SAMPLER_VIEW;
template.flags = 0;
template.width0 = idct->buffer_width / 4;
template.height0 = idct->buffer_height;
template.depth0 = 1;
+ template.array_size = 1;
template.usage = PIPE_USAGE_STREAM;
buffer->textures.individual.source = idct->pipe->screen->resource_create(idct->pipe->screen, &template);
}
}
-static bool
-init_vertex_buffers(struct vl_idct *idct, struct vl_idct_buffer *buffer)
-{
- assert(idct && buffer);
-
- buffer->vertex_bufs.individual.quad.stride = idct->quad.stride;
- buffer->vertex_bufs.individual.quad.max_index = idct->quad.max_index;
- buffer->vertex_bufs.individual.quad.buffer_offset = idct->quad.buffer_offset;
- pipe_resource_reference(&buffer->vertex_bufs.individual.quad.buffer, idct->quad.buffer);
-
- buffer->vertex_bufs.individual.pos = vl_vb_create_buffer(idct->pipe, idct->max_blocks, idct->vertex_buffer_stride);
-
- if(buffer->vertex_bufs.individual.pos.buffer == NULL)
- return false;
-
- if (!vl_vb_init(&buffer->blocks, idct->max_blocks, 2))
- return false;
-
- return true;
-}
-
-static void
-cleanup_vertex_buffers(struct vl_idct *idct, struct vl_idct_buffer *buffer)
-{
- assert(idct && buffer);
-
- pipe_resource_reference(&buffer->vertex_bufs.individual.quad.buffer, NULL);
- pipe_resource_reference(&buffer->vertex_bufs.individual.pos.buffer, NULL);
-
- vl_vb_cleanup(&buffer->blocks);
-}
-
struct pipe_resource *
vl_idct_upload_matrix(struct pipe_context *pipe)
{
struct pipe_box rect =
{
0, 0, 0,
- BLOCK_WIDTH,
+ BLOCK_WIDTH / 4,
BLOCK_HEIGHT,
1
};
template.width0 = 2;
template.height0 = 8;
template.depth0 = 1;
+ template.array_size = 1;
template.usage = PIPE_USAGE_IMMUTABLE;
template.bind = PIPE_BIND_SAMPLER_VIEW;
template.flags = 0;
buf_transfer = pipe->get_transfer
(
pipe, matrix,
- u_subresource(0, 0),
- PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
+ 0, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
&rect
);
pitch = buf_transfer->stride / sizeof(float);
f = pipe->transfer_map(pipe, buf_transfer);
for(i = 0; i < BLOCK_HEIGHT; ++i)
for(j = 0; j < BLOCK_WIDTH; ++j)
- f[i * pitch + j] = const_matrix[j][i]; // transpose
+ // transpose and scale
+ f[i * pitch + j] = const_matrix[j][i] * sqrtf(SCALE_FACTOR_16_TO_9);
pipe->transfer_unmap(pipe, buf_transfer);
pipe->transfer_destroy(pipe, buf_transfer);
return matrix;
}
-bool vl_idct_init(struct vl_idct *idct, struct pipe_context *pipe,
+bool vl_idct_init(struct vl_idct *idct, struct pipe_context *pipe,
unsigned buffer_width, unsigned buffer_height,
- struct pipe_resource *matrix)
+ unsigned blocks_x, unsigned blocks_y,
+ int color_swizzle, struct pipe_resource *matrix)
{
assert(idct && pipe && matrix);
idct->pipe = pipe;
idct->buffer_width = buffer_width;
idct->buffer_height = buffer_height;
+ idct->blocks_x = blocks_x;
+ idct->blocks_y = blocks_y;
pipe_resource_reference(&idct->matrix, matrix);
- idct->max_blocks =
- align(buffer_width, BLOCK_WIDTH) / BLOCK_WIDTH *
- align(buffer_height, BLOCK_HEIGHT) / BLOCK_HEIGHT;
-
- if(!init_shaders(idct))
+ if(!init_shaders(idct, color_swizzle))
return false;
if(!init_state(idct)) {
}
bool
-vl_idct_init_buffer(struct vl_idct *idct, struct vl_idct_buffer *buffer, struct pipe_resource *dst)
+vl_idct_init_buffer(struct vl_idct *idct, struct vl_idct_buffer *buffer,
+ struct pipe_resource *dst)
{
+ struct pipe_surface template;
+
unsigned i;
assert(buffer);
if (!init_textures(idct, buffer))
return false;
- if (!init_vertex_buffers(idct, buffer))
- return false;
-
/* init state */
buffer->viewport[0].scale[0] = buffer->textures.individual.intermediate->width0;
buffer->viewport[0].scale[1] = buffer->textures.individual.intermediate->height0;
buffer->fb_state[0].nr_cbufs = NR_RENDER_TARGETS;
for(i = 0; i < NR_RENDER_TARGETS; ++i) {
- buffer->fb_state[0].cbufs[i] = idct->pipe->screen->get_tex_surface(
- idct->pipe->screen, buffer->textures.individual.intermediate, 0, 0, i,
- PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET);
+ memset(&template, 0, sizeof(template));
+ template.format = buffer->textures.individual.intermediate->format;
+ template.u.tex.first_layer = i;
+ template.u.tex.last_layer = i;
+ template.usage = PIPE_BIND_RENDER_TARGET;
+ buffer->fb_state[0].cbufs[i] = idct->pipe->create_surface(
+ idct->pipe, buffer->textures.individual.intermediate,
+ &template);
}
buffer->fb_state[1].width = buffer->destination->width0;
buffer->fb_state[1].height = buffer->destination->height0;
buffer->fb_state[1].nr_cbufs = 1;
- buffer->fb_state[1].cbufs[0] = idct->pipe->screen->get_tex_surface(
- idct->pipe->screen, buffer->destination, 0, 0, 0,
- PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET);
+
+ memset(&template, 0, sizeof(template));
+ template.format = buffer->destination->format;
+ template.usage = PIPE_BIND_RENDER_TARGET;
+ buffer->fb_state[1].cbufs[0] = idct->pipe->create_surface(
+ idct->pipe, buffer->destination, &template);
for(i = 0; i < 2; ++i) {
buffer->viewport[i].scale[2] = 1;
assert(buffer);
for(i = 0; i < NR_RENDER_TARGETS; ++i) {
- idct->pipe->screen->tex_surface_destroy(buffer->fb_state[0].cbufs[i]);
+ idct->pipe->surface_destroy(idct->pipe, buffer->fb_state[0].cbufs[i]);
}
- idct->pipe->screen->tex_surface_destroy(buffer->fb_state[1].cbufs[0]);
+ idct->pipe->surface_destroy(idct->pipe, buffer->fb_state[1].cbufs[0]);
cleanup_textures(idct, buffer);
- cleanup_vertex_buffers(idct, buffer);
}
void
buffer->tex_transfer = idct->pipe->get_transfer
(
idct->pipe, buffer->textures.individual.source,
- u_subresource(0, 0),
- PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
+ 0, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
&rect
);
void
vl_idct_add_block(struct vl_idct_buffer *buffer, unsigned x, unsigned y, short *block)
{
- struct vertex2f v;
+ //struct vertex2s v;
unsigned tex_pitch;
short *texels;
for (i = 0; i < BLOCK_HEIGHT; ++i)
memcpy(texels + i * tex_pitch, block + i * BLOCK_WIDTH, BLOCK_WIDTH * sizeof(short));
-
- v.x = x;
- v.y = y;
- vl_vb_add_block(&buffer->blocks, (float*)&v);
}
void
}
void
-vl_idct_flush(struct vl_idct *idct, struct vl_idct_buffer *buffer)
+vl_idct_flush(struct vl_idct *idct, struct vl_idct_buffer *buffer, unsigned num_instances)
{
- struct pipe_transfer *vec_transfer;
- void *vectors;
unsigned num_verts;
assert(idct);
+ assert(buffer);
- vectors = pipe_buffer_map
- (
- idct->pipe,
- buffer->vertex_bufs.individual.pos.buffer,
- PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
- &vec_transfer
- );
-
- num_verts = vl_vb_upload(&buffer->blocks, vectors);
-
- pipe_buffer_unmap(idct->pipe, buffer->vertex_bufs.individual.pos.buffer, vec_transfer);
-
- if(num_verts > 0) {
+ if(num_instances > 0) {
+ num_verts = idct->blocks_x * idct->blocks_y * 4;
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);
+ util_draw_arrays_instanced(idct->pipe, PIPE_PRIM_QUADS, 0, num_verts, 0, num_instances);
/* second stage */
idct->pipe->set_framebuffer_state(idct->pipe, &buffer->fb_state[1]);
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);
+ util_draw_arrays_instanced(idct->pipe, PIPE_PRIM_QUADS, 0, num_verts, 0, num_instances);
}
}
projects / mesa.git / blobdiff