vc4: Start converting the driver to use vertex shaders.

Note: This is the cutoff point where I switched from developing primarily
on the Pi to developing o the simulator.  As a result, from this point on
the code is untested on the Pi (the kernel code I have currently wasn't
rendering anything at this commit, though the simulator renders
successfully, suggesting kernel bugs).

3 files changed, 177 insertions, 45 deletions

diff --git a/src/gallium/drivers/vc4/vc4_context.h b/src/gallium/drivers/vc4/vc4_context.h
index 73bf05bce19..0c906488756 100644
--- a/src/gallium/drivers/vc4/vc4_context.h
+++ b/src/gallium/drivers/vc4/vc4_context.h
@@ -66,6 +66,8 @@ struct vc4_texture_stateobj {
 struct vc4_shader_state {
         struct pipe_shader_state base;
         struct vc4_bo *bo;
+
+        uint32_t coord_shader_offset;
 };
 
 struct vc4_program_stateobj {
diff --git a/src/gallium/drivers/vc4/vc4_draw.c b/src/gallium/drivers/vc4/vc4_draw.c
index 4c25dabae7d..34977bb6938 100644
--- a/src/gallium/drivers/vc4/vc4_draw.c
+++ b/src/gallium/drivers/vc4/vc4_draw.c
@@ -31,40 +31,11 @@ static struct vc4_bo *
 get_vbo(struct vc4_context *vc4, uint32_t width, uint32_t height)
 {
         struct {
-                uint16_t x, y;
-                float z, rhw, r, g, b;
+                float x, y, z, w;
         } verts[] = {
-                {
-                        // Vertex: Top, red
-                        (-(int)width / 3) << 4, // X in 12.4 fixed point
-                        (-(int)height / 3) << 4, // Y in 12.4 fixed point
-                        1.0f, // Z
-                        1.0f, // 1/W
-                        1.0f, // Varying 0 (Red)
-                        0.0f, // Varying 1 (Green)
-                        0.0f, // Varying 2 (Blue)
-                },
-                {
-                        // Vertex: bottom left, Green
-                        (width / 3) << 4, // X in 12.4 fixed point
-                        (-(int)height / 3) << 4, // Y in 12.4 fixed point
-                        1.0f, // Z
-                        1.0f, // 1/W
-                        0.0f, // Varying 0 (Red)
-                        1.0f, // Varying 1 (Green)
-                        0.0f, // Varying 2 (Blue)
-                },
-
-                {
-                        // Vertex: bottom right, Blue
-                        (width / 3) << 4, // X in 12.4 fixed point
-                        (height / 3) << 4, // Y in 12.4 fixed point
-                        1.0f, // Z
-                        1.0f, // 1/W
-                        0.0f, // Varying 0 (Red)
-                        0.0f, // Varying 1 (Green)
-                        1.0f, // Varying 2 (Blue)
-                },
+                { -1, -1, 1, 1 },
+                {  1, -1, 1, 1 },
+                { -1,  1, 1, 1 },
         };
 
         return vc4_bo_alloc_mem(vc4->screen, verts, sizeof(verts), "verts");
@@ -118,8 +89,22 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
                                                  48 * tilew * tileh, "tilestate");
         struct vc4_bo *ibo = get_ibo(vc4);
 
-        struct vc4_bo *fs_uniform = vc4_bo_alloc(vc4->screen, 0x1000, "fsu");
         struct vc4_bo *vbo = get_vbo(vc4, width, height);
+        static const uint32_t fs_uni[] = { 0 };
+        uint32_t vs_uni[] = {
+                fui(vc4->framebuffer.width * 16.0f / 2.0f),
+                fui(vc4->framebuffer.height * 16.0f / 2.0f),
+        };
+        uint32_t cs_uni[] = {
+                fui(vc4->framebuffer.width * 16.0f / 2.0f),
+                fui(vc4->framebuffer.height * 16.0f / 2.0f),
+        };
+        struct vc4_bo *fs_ubo = vc4_bo_alloc_mem(vc4->screen, fs_uni,
+                                                 sizeof(fs_uni), "fs_ubo");
+        struct vc4_bo *vs_ubo = vc4_bo_alloc_mem(vc4->screen, vs_uni,
+                                                 sizeof(vs_uni), "vs_ubo");
+        struct vc4_bo *cs_ubo = vc4_bo_alloc_mem(vc4->screen, cs_uni,
+                                                 sizeof(cs_uni), "cs_ubo");
 
         vc4->needs_flush = true;
 
@@ -137,16 +122,18 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
         cl_u8(&vc4->bcl, VC4_PACKET_START_TILE_BINNING);
 
         cl_u8(&vc4->bcl, VC4_PACKET_PRIMITIVE_LIST_FORMAT);
-        cl_u8(&vc4->bcl, 0x32); // 16 bit triangle
+        cl_u8(&vc4->bcl, 0x12); // 16 bit triangle
 
         vc4_emit_state(pctx);
 
         /* the actual draw call. */
-        cl_u8(&vc4->bcl, VC4_PACKET_NV_SHADER_STATE);
+        uint32_t nr_attributes = 1;
+        cl_u8(&vc4->bcl, VC4_PACKET_GL_SHADER_STATE);
 #ifndef USE_VC4_SIMULATOR
-        cl_u32(&vc4->bcl, 0); /* offset into shader_rec */
+        cl_u32(&vc4->bcl, nr_attributes & 0x7); /* offset into shader_rec */
 #else
-        cl_u32(&vc4->bcl, simpenrose_hw_addr(vc4->shader_rec.next));
+        cl_u32(&vc4->bcl, simpenrose_hw_addr(vc4->shader_rec.next) |
+               (nr_attributes & 0x7));
 #endif
 
         cl_start_reloc(&vc4->bcl, 1);
@@ -161,15 +148,32 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
         cl_u8(&vc4->bcl, VC4_PACKET_HALT);
 
 // Shader Record
-        cl_start_shader_reloc(&vc4->shader_rec, 3);
 
-        cl_u8(&vc4->shader_rec, 0);
-        cl_u8(&vc4->shader_rec, 6*4); // stride
-        cl_u8(&vc4->shader_rec, 0xcc); // num uniforms (not used)
-        cl_u8(&vc4->shader_rec, 3); // num varyings
+        cl_start_shader_reloc(&vc4->shader_rec, 7);
+        cl_u16(&vc4->shader_rec, VC4_SHADER_FLAG_ENABLE_CLIPPING);
+        cl_u8(&vc4->shader_rec, 0); /* fs num uniforms (unused) */
+        cl_u8(&vc4->shader_rec, 0); /* fs num varyings */
         cl_reloc(vc4, &vc4->shader_rec, vc4->prog.fs->bo, 0);
-        cl_reloc(vc4, &vc4->shader_rec, fs_uniform, 0);
+        cl_reloc(vc4, &vc4->shader_rec, fs_ubo, 0);
+
+        cl_u16(&vc4->shader_rec, 0); /* vs num uniforms */
+        cl_u8(&vc4->shader_rec, 1); /* vs attribute array bitfield */
+        cl_u8(&vc4->shader_rec, 16); /* vs total attribute size */
+        cl_reloc(vc4, &vc4->shader_rec, vc4->prog.vs->bo, 0);
+        cl_reloc(vc4, &vc4->shader_rec, vs_ubo, 0);
+
+        cl_u16(&vc4->shader_rec, 0); /* cs num uniforms */
+        cl_u8(&vc4->shader_rec, 1); /* cs attribute array bitfield */
+        cl_u8(&vc4->shader_rec, 16); /* vs total attribute size */
+        cl_reloc(vc4, &vc4->shader_rec, vc4->prog.vs->bo,
+                vc4->prog.vs->coord_shader_offset);
+        cl_reloc(vc4, &vc4->shader_rec, cs_ubo, 0);
+
         cl_reloc(vc4, &vc4->shader_rec, vbo, 0);
+        cl_u8(&vc4->shader_rec, 15); /* bytes - 1 in the attribute*/
+        cl_u8(&vc4->shader_rec, 16); /* attribute stride */
+        cl_u8(&vc4->shader_rec, 0); /* VS VPM offset */
+        cl_u8(&vc4->shader_rec, 0); /* CS VPM offset */
 
         vc4->shader_rec_count++;
 
diff --git a/src/gallium/drivers/vc4/vc4_program.c b/src/gallium/drivers/vc4/vc4_program.c
index 2fe7c216838..56fe37c5f5f 100644
--- a/src/gallium/drivers/vc4/vc4_program.c
+++ b/src/gallium/drivers/vc4/vc4_program.c
@@ -66,7 +66,7 @@ vc4_fs_state_create(struct pipe_context *pctx,
         uint64_t gen_fsc[100];
         uint64_t cur_inst;
         int gen_fsc_len = 0;
-#if 1
+#if 0
         cur_inst = qpu_load_imm_f(qpu_r5(), 0.0f);
         gen_fsc[gen_fsc_len++] = cur_inst;
 
@@ -176,14 +176,140 @@ vc4_fs_state_create(struct pipe_context *pctx,
         return so;
 }
 
+static int
+gen_vs_cs_code(uint64_t *gen, bool is_vs)
+{
+        uint32_t count = 0;
+        uint64_t cur_inst;
+        struct qpu_reg x = qpu_ra(0);
+        struct qpu_reg y = qpu_ra(1);
+        struct qpu_reg z = qpu_ra(2);
+        struct qpu_reg w = qpu_ra(3);
+        struct qpu_reg xy = qpu_ra(10);
+        struct qpu_reg xs = qpu_ra(12);
+        struct qpu_reg ys = qpu_ra(13);
+        struct qpu_reg vpmread = qpu_ra(QPU_R_VPM);
+        struct qpu_reg vpm = qpu_ra(QPU_W_VPM);
+
+        gen[count++] = qpu_load_imm_ui(qpu_vrsetup(), 0x00401a00);
+        gen[count++] = qpu_load_imm_ui(qpu_vwsetup(), 0x00001a00);
+
+#if 1
+        gen[count++] = qpu_inst(qpu_a_MOV(x, vpmread), qpu_m_NOP());
+        gen[count++] = qpu_inst(qpu_a_MOV(y, vpmread), qpu_m_NOP());
+        gen[count++] = qpu_inst(qpu_a_MOV(z, vpmread), qpu_m_NOP());
+        gen[count++] = qpu_inst(qpu_a_MOV(w, vpmread), qpu_m_NOP());
+
+
+        gen[count++] = qpu_inst(qpu_a_NOP(), qpu_m_FMUL(xs, x,
+                                                        qpu_rb(QPU_R_UNIF)));
+        gen[count++] = qpu_inst(qpu_a_NOP(), qpu_m_FMUL(ys, y,
+                                                        qpu_rb(QPU_R_UNIF)));
+
+        cur_inst = qpu_inst(qpu_a_FTOI(xy, xs), qpu_m_NOP());
+        cur_inst |= QPU_SET_FIELD(QPU_PACK_A_16A, QPU_PACK);
+        gen[count++] = cur_inst;
+        cur_inst = qpu_inst(qpu_a_FTOI(xy, ys), qpu_m_NOP());
+        cur_inst |= QPU_SET_FIELD(QPU_PACK_A_16B, QPU_PACK);
+        gen[count++] = cur_inst;
+
+#else
+
+        struct qpu_reg t = qpu_ra(20);
+        struct qpu_reg hundred = qpu_rb(21);
+        gen[count++] = qpu_inst(qpu_a_NOP(),
+                                qpu_m_MUL24(t,
+                                            qpu_ra(QPU_R_ELEM_QPU),
+                                            qpu_ra(QPU_R_ELEM_QPU)));
+        gen[count++] = qpu_inst(qpu_a_NOP(), qpu_m_NOP());
+
+        gen[count++] = qpu_load_imm_ui(hundred, 400);
+        gen[count++] = qpu_inst(qpu_a_NOP(), qpu_m_NOP());
+
+        struct qpu_reg xm = qpu_ra(22), ym = qpu_ra(23);
+        gen[count++] = qpu_inst(qpu_a_NOP(),
+                                qpu_m_MUL24(xm, hundred, qpu_ra(QPU_R_ELEM_QPU)));
+        gen[count++] = qpu_inst(qpu_a_NOP(), qpu_m_NOP());
+
+        gen[count++] = qpu_inst(qpu_a_NOP(),
+                                qpu_m_MUL24(ym, hundred, t));
+        gen[count++] = qpu_inst(qpu_a_NOP(), qpu_m_NOP());
+
+        cur_inst = qpu_inst(qpu_a_MOV(xy, xm), qpu_m_NOP());
+        cur_inst |= QPU_SET_FIELD(QPU_PACK_A_16A, QPU_PACK);
+        gen[count++] = cur_inst;
+        cur_inst = qpu_inst(qpu_a_MOV(xy, ym), qpu_m_NOP());
+        cur_inst |= QPU_SET_FIELD(QPU_PACK_A_16B, QPU_PACK);
+        gen[count++] = cur_inst;
+#endif
+
+        gen[count++] = qpu_inst(qpu_a_NOP(), qpu_m_NOP());
+
+        if (is_vs) {
+                gen[count++] = qpu_inst(qpu_a_MOV(vpm, xy), qpu_m_NOP());
+
+                /* XXX */
+                gen[count++] = qpu_inst(qpu_a_MOV(vpm, z), qpu_m_NOP());
+                gen[count++] = qpu_inst(qpu_a_MOV(vpm, w), qpu_m_NOP());
+
+        } else {
+                gen[count++] = qpu_inst(qpu_a_MOV(vpm, x), qpu_m_NOP());
+                gen[count++] = qpu_inst(qpu_a_MOV(vpm, y), qpu_m_NOP());
+                gen[count++] = qpu_inst(qpu_a_MOV(vpm, z), qpu_m_NOP());
+                gen[count++] = qpu_inst(qpu_a_MOV(vpm, w), qpu_m_NOP());
+                gen[count++] = qpu_inst(qpu_a_MOV(vpm, xy), qpu_m_NOP());
+
+                /* XXX */
+                gen[count++] = qpu_inst(qpu_a_MOV(vpm, z), qpu_m_NOP());
+                gen[count++] = qpu_inst(qpu_a_MOV(vpm, w), qpu_m_NOP());
+        }
+
+        /* PROGRAM END */
+        cur_inst = qpu_inst(qpu_a_NOP(), qpu_m_NOP());
+        cur_inst = (cur_inst & ~QPU_SIG_MASK) | QPU_SET_FIELD(QPU_SIG_PROG_END, QPU_SIG);
+        gen[count++] = cur_inst;
+
+        cur_inst = qpu_inst(qpu_a_NOP(), qpu_m_NOP());
+        gen[count++] = cur_inst;
+
+        cur_inst = qpu_inst(qpu_a_NOP(), qpu_m_NOP());
+        gen[count++] = cur_inst;
+
+        vc4_qpu_validate(gen, count);
+
+        return count;
+}
+
 static void *
 vc4_vs_state_create(struct pipe_context *pctx,
                     const struct pipe_shader_state *cso)
 {
+        struct vc4_context *vc4 = vc4_context(pctx);
         struct vc4_shader_state *so = vc4_shader_state_create(pctx, cso);
         if (!so)
                 return NULL;
 
+        uint64_t gen[100];
+        uint64_t count = 0;
+        uint64_t *vsc = gen;
+
+        /* VS */
+        count += gen_vs_cs_code(gen + count, true);
+        fprintf(stderr, "VS:\n");
+        vc4_dump_program(vsc, count);
+
+        /* CS */
+
+        /* XXX alignment? */
+        uint64_t *csc = gen + count;
+        so->coord_shader_offset = count * sizeof(uint64_t);
+        count += gen_vs_cs_code(gen + count, false);
+
+        fprintf(stderr, "CS:\n");
+        vc4_dump_program(csc, count - (csc - gen));
+
+        so->bo = vc4_bo_alloc_mem(vc4->screen, gen, count * sizeof(uint64_t),
+                                  "vs_code");
 
         return so;
 }