r300g: do not use TXPITCH_EN if the width is POT and the height is NPOT
[mesa.git] / src / gallium / drivers / r300 / r300_fs.c
index 88303f074cd8a1039adf53e77b473fa91772abde..db5269912e2d1e6db5f648ca6e5c15b92181fd67 100644 (file)
@@ -28,7 +28,9 @@
 #include "tgsi/tgsi_dump.h"
 #include "tgsi/tgsi_ureg.h"
 
+#include "r300_cb.h"
 #include "r300_context.h"
+#include "r300_emit.h"
 #include "r300_screen.h"
 #include "r300_fs.h"
 #include "r300_reg.h"
@@ -137,6 +139,7 @@ static void get_external_state(
 {
     struct r300_textures_state *texstate = r300->textures_state.state;
     unsigned i;
+    unsigned char *swizzle;
 
     for (i = 0; i < texstate->sampler_state_count; i++) {
         struct r300_sampler_state* s = texstate->sampler_states[i];
@@ -146,9 +149,18 @@ static void get_external_state(
         }
 
         if (s->state.compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-            /* XXX Gallium doesn't provide us with any information regarding
-             * this mode, so we are screwed. I'm setting 0 = LUMINANCE. */
-            state->unit[i].depth_texture_mode = 0;
+            state->unit[i].compare_mode_enabled = 1;
+
+            /* Pass depth texture swizzling to the compiler. */
+            if (texstate->sampler_views[i]) {
+                swizzle = texstate->sampler_views[i]->swizzle;
+
+                state->unit[i].depth_texture_swizzle =
+                    RC_MAKE_SWIZZLE(swizzle[0], swizzle[1],
+                                    swizzle[2], swizzle[3]);
+            } else {
+                state->unit[i].depth_texture_swizzle = RC_SWIZZLE_XYZW;
+            }
 
             /* Fortunately, no need to translate this. */
             state->unit[i].texture_compare_func = s->state.compare_func;
@@ -161,7 +173,7 @@ static void get_external_state(
             t = (struct r300_texture*)texstate->sampler_views[i]->base.texture;
 
             /* XXX this should probably take into account STR, not just S. */
-            if (t->uses_pitch) {
+            if (t->desc.is_npot) {
                 switch (s->state.wrap_s) {
                     case PIPE_TEX_WRAP_REPEAT:
                         state->unit[i].wrap_mode = RC_WRAP_REPEAT;
@@ -219,6 +231,123 @@ static void r300_dummy_fragment_shader(
     ureg_destroy(ureg);
 }
 
+static void r300_emit_fs_code_to_buffer(
+    struct r300_context *r300,
+    struct r300_fragment_shader_code *shader)
+{
+    struct rX00_fragment_program_code *generic_code = &shader->code;
+    unsigned imm_count = shader->immediates_count;
+    unsigned imm_first = shader->externals_count;
+    unsigned imm_end = generic_code->constants.Count;
+    struct rc_constant *constants = generic_code->constants.Constants;
+    unsigned i;
+    CB_LOCALS;
+
+    if (r300->screen->caps.is_r500) {
+        struct r500_fragment_program_code *code = &generic_code->code.r500;
+
+        shader->cb_code_size = 19 +
+                               ((code->inst_end + 1) * 6) +
+                               imm_count * 7;
+
+        NEW_CB(shader->cb_code, shader->cb_code_size);
+        OUT_CB_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
+        OUT_CB_REG(R500_US_PIXSIZE, code->max_temp_idx);
+        OUT_CB_REG(R500_US_FC_CTRL, code->us_fc_ctrl);
+        OUT_CB_REG(R500_US_CODE_RANGE,
+                   R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(code->inst_end));
+        OUT_CB_REG(R500_US_CODE_OFFSET, 0);
+        OUT_CB_REG(R500_US_CODE_ADDR,
+                   R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(code->inst_end));
+
+        OUT_CB_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_INSTR);
+        OUT_CB_ONE_REG(R500_GA_US_VECTOR_DATA, (code->inst_end + 1) * 6);
+        for (i = 0; i <= code->inst_end; i++) {
+            OUT_CB(code->inst[i].inst0);
+            OUT_CB(code->inst[i].inst1);
+            OUT_CB(code->inst[i].inst2);
+            OUT_CB(code->inst[i].inst3);
+            OUT_CB(code->inst[i].inst4);
+            OUT_CB(code->inst[i].inst5);
+        }
+
+        /* Emit immediates. */
+        if (imm_count) {
+            for(i = imm_first; i < imm_end; ++i) {
+                if (constants[i].Type == RC_CONSTANT_IMMEDIATE) {
+                    const float *data = constants[i].u.Immediate;
+
+                    OUT_CB_REG(R500_GA_US_VECTOR_INDEX,
+                               R500_GA_US_VECTOR_INDEX_TYPE_CONST |
+                               (i & R500_GA_US_VECTOR_INDEX_MASK));
+                    OUT_CB_ONE_REG(R500_GA_US_VECTOR_DATA, 4);
+                    OUT_CB_TABLE(data, 4);
+                }
+            }
+        }
+    } else { /* r300 */
+        struct r300_fragment_program_code *code = &generic_code->code.r300;
+
+        shader->cb_code_size = 19 +
+                               (r300->screen->caps.is_r400 ? 2 : 0) +
+                               code->alu.length * 4 +
+                               (code->tex.length ? (1 + code->tex.length) : 0) +
+                               imm_count * 5;
+
+        NEW_CB(shader->cb_code, shader->cb_code_size);
+
+        if (r300->screen->caps.is_r400)
+            OUT_CB_REG(R400_US_CODE_BANK, 0);
+
+        OUT_CB_REG(R300_US_CONFIG, code->config);
+        OUT_CB_REG(R300_US_PIXSIZE, code->pixsize);
+        OUT_CB_REG(R300_US_CODE_OFFSET, code->code_offset);
+
+        OUT_CB_REG_SEQ(R300_US_CODE_ADDR_0, 4);
+        OUT_CB_TABLE(code->code_addr, 4);
+
+        OUT_CB_REG_SEQ(R300_US_ALU_RGB_INST_0, code->alu.length);
+        for (i = 0; i < code->alu.length; i++)
+            OUT_CB(code->alu.inst[i].rgb_inst);
+
+        OUT_CB_REG_SEQ(R300_US_ALU_RGB_ADDR_0, code->alu.length);
+        for (i = 0; i < code->alu.length; i++)
+            OUT_CB(code->alu.inst[i].rgb_addr);
+
+        OUT_CB_REG_SEQ(R300_US_ALU_ALPHA_INST_0, code->alu.length);
+        for (i = 0; i < code->alu.length; i++)
+            OUT_CB(code->alu.inst[i].alpha_inst);
+
+        OUT_CB_REG_SEQ(R300_US_ALU_ALPHA_ADDR_0, code->alu.length);
+        for (i = 0; i < code->alu.length; i++)
+            OUT_CB(code->alu.inst[i].alpha_addr);
+
+        if (code->tex.length) {
+            OUT_CB_REG_SEQ(R300_US_TEX_INST_0, code->tex.length);
+            OUT_CB_TABLE(code->tex.inst, code->tex.length);
+        }
+
+        /* Emit immediates. */
+        if (imm_count) {
+            for(i = imm_first; i < imm_end; ++i) {
+                if (constants[i].Type == RC_CONSTANT_IMMEDIATE) {
+                    const float *data = constants[i].u.Immediate;
+
+                    OUT_CB_REG_SEQ(R300_PFS_PARAM_0_X + i * 16, 4);
+                    OUT_CB(pack_float24(data[0]));
+                    OUT_CB(pack_float24(data[1]));
+                    OUT_CB(pack_float24(data[2]));
+                    OUT_CB(pack_float24(data[3]));
+                }
+            }
+        }
+    }
+
+    OUT_CB_REG(R300_FG_DEPTH_SRC, shader->fg_depth_src);
+    OUT_CB_REG(R300_US_W_FMT, shader->us_out_w);
+    END_CB;
+}
+
 static void r300_translate_fragment_shader(
     struct r300_context* r300,
     struct r300_fragment_shader_code* shader,
@@ -241,8 +370,8 @@ static void r300_translate_fragment_shader(
 
     compiler.code = &shader->code;
     compiler.state = shader->compare_state;
-    compiler.is_r500 = r300->screen->caps.is_r500;
-    compiler.max_temp_regs = compiler.is_r500 ? 128 : 32;
+    compiler.Base.is_r500 = r300->screen->caps.is_r500;
+    compiler.Base.max_temp_regs = compiler.Base.is_r500 ? 128 : 32;
     compiler.AllocateHwInputs = &allocate_hardware_inputs;
     compiler.UserData = &shader->inputs;
 
@@ -285,7 +414,8 @@ static void r300_translate_fragment_shader(
 
     if (compiler.Base.Error) {
         fprintf(stderr, "r300 FP: Compiler Error:\n%sUsing a dummy shader"
-                " instead.\n", compiler.Base.ErrorMsg);
+                " instead.\nIf there's an 'unknown opcode' message, please"
+                " file a bug report and attach this log.\n", compiler.Base.ErrorMsg);
 
         if (shader->dummy) {
             fprintf(stderr, "r300 FP: Cannot compile the dummy shader! "
@@ -327,6 +457,9 @@ static void r300_translate_fragment_shader(
 
     /* And, finally... */
     rc_destroy(&compiler.Base);
+
+    /* Build the command buffer. */
+    r300_emit_fs_code_to_buffer(r300, shader);
 }
 
 boolean r300_pick_fragment_shader(struct r300_context* r300)