r600g: Implement ARB_derivative_control

Requires Evergreen/Cayman

marek: update release notes

Signed-off-by: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>

diff --git a/docs/GL3.txt b/docs/GL3.txt
index 12b8f62283cb4d1eb4130ee64f025e8f77f8bbee..b38e42c19dc8b6411e0359f5c040f276f40e626f 100644 (file)
--- a/docs/GL3.txt
+++ b/docs/GL3.txt
@@ -188,7 +188,7 @@ GL 4.5, GLSL 4.50:
   GL_ARB_clip_control                                  not started
   GL_ARB_conditional_render_inverted                   not started
   GL_ARB_cull_distance                                 not started
-  GL_ARB_derivative_control                            DONE (i965, nv50, nvc0)
+  GL_ARB_derivative_control                            DONE (i965, nv50, nvc0, r600)
   GL_ARB_direct_state_access                           not started
   GL_ARB_get_texture_sub_image                         started (Brian Paul)
   GL_ARB_shader_texture_image_samples                  not started

diff --git a/docs/relnotes/10.3.html b/docs/relnotes/10.3.html
index 0f22d70ae71d3122468a3376c705d26706f57b66..1afa93294f9e0464bad63d37f0832557d4c181ca 100644 (file)
--- a/docs/relnotes/10.3.html
+++ b/docs/relnotes/10.3.html
@@ -47,7 +47,7 @@ Note: some of the new features are only available with certain drivers.
 <li>GL_ARB_ES3_compatibility on nv50, nvc0, r600, radeonsi, softpipe, llvmpipe</li>
 <li>GL_ARB_clear_texture on i965</li>
 <li>GL_ARB_compressed_texture_pixel_storage on all drivers</li>
-<li>GL_ARB_derivative_control on i965, nv50, nvc0</li>
+<li>GL_ARB_derivative_control on i965, nv50, nvc0, r600</li>
 <li>GL_ARB_draw_indirect on nvc0, radeonsi</li>
 <li>GL_ARB_explicit_uniform_location (all drivers that support GLSL)</li>
 <li>GL_ARB_fragment_layer_viewport on nv50, nvc0, llvmpipe, r600</li>

diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c
index 8a5ba7951bbd487103e32b65bb9431680a5d3c75..bf52a192c9c9ec85065eed7c2e7c2c67a42b8b2a 100644 (file)
--- a/src/gallium/drivers/r600/r600_pipe.c
+++ b/src/gallium/drivers/r600/r600_pipe.c
@@ -305,6 +305,7 @@ static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
        case PIPE_CAP_TGSI_VS_LAYER_VIEWPORT:
        case PIPE_CAP_TEXTURE_GATHER_SM5:
        case PIPE_CAP_TEXTURE_QUERY_LOD:
+       case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
                return family >= CHIP_CEDAR ? 1 : 0;
        case PIPE_CAP_MAX_TEXTURE_GATHER_COMPONENTS:
                return family >= CHIP_CEDAR ? 4 : 0;
@@ -319,7 +320,6 @@ static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
        case PIPE_CAP_SAMPLE_SHADING:
        case PIPE_CAP_TEXTURE_GATHER_OFFSETS:
        case PIPE_CAP_DRAW_INDIRECT:
-       case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE:
                return 0;
 
        /* Stream output. */

diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index 481e9eb29b1bf14331370015233289e0a35f1d02..3f089b4f7d9dbd3569c92b4a2d381edb2664890d 100644 (file)
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -5319,6 +5319,11 @@ static int tgsi_tex(struct r600_shader_ctx *ctx)
        tex.src_gpr = src_gpr;
        tex.dst_gpr = ctx->file_offset[inst->Dst[0].Register.File] + inst->Dst[0].Register.Index;
 
+       if (inst->Instruction.Opcode == TGSI_OPCODE_DDX_FINE ||
+               inst->Instruction.Opcode == TGSI_OPCODE_DDY_FINE) {
+               tex.inst_mod = 1; /* per pixel gradient calculation instead of per 2x2 quad */
+       }
+
        if (inst->Instruction.Opcode == TGSI_OPCODE_TG4) {
                int8_t texture_component_select = ctx->literals[4 * inst->Src[1].Register.Index + inst->Src[1].Register.SwizzleX];
                tex.inst_mod = texture_component_select;
@@ -6789,9 +6794,8 @@ static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction[] = {
        {76,                    0, ALU_OP0_NOP, tgsi_unsupported},
        {TGSI_OPCODE_ELSE,      0, ALU_OP0_NOP, tgsi_else},
        {TGSI_OPCODE_ENDIF,     0, ALU_OP0_NOP, tgsi_endif},
-       /* gap */
-       {79,                    0, ALU_OP0_NOP, tgsi_unsupported},
-       {80,                    0, ALU_OP0_NOP, tgsi_unsupported},
+       {TGSI_OPCODE_DDX_FINE,  0, ALU_OP0_NOP, tgsi_unsupported},
+       {TGSI_OPCODE_DDY_FINE,  0, ALU_OP0_NOP, tgsi_unsupported},
        {TGSI_OPCODE_PUSHA,     0, ALU_OP0_NOP, tgsi_unsupported},
        {TGSI_OPCODE_POPA,      0, ALU_OP0_NOP, tgsi_unsupported},
        {TGSI_OPCODE_CEIL,      0, ALU_OP1_CEIL, tgsi_op2},
@@ -6992,9 +6996,8 @@ static struct r600_shader_tgsi_instruction eg_shader_tgsi_instruction[] = {
        {76,                    0, ALU_OP0_NOP, tgsi_unsupported},
        {TGSI_OPCODE_ELSE,      0, ALU_OP0_NOP, tgsi_else},
        {TGSI_OPCODE_ENDIF,     0, ALU_OP0_NOP, tgsi_endif},
-       /* gap */
-       {79,                    0, ALU_OP0_NOP, tgsi_unsupported},
-       {80,                    0, ALU_OP0_NOP, tgsi_unsupported},
+       {TGSI_OPCODE_DDX_FINE,  0, FETCH_OP_GET_GRADIENTS_H, tgsi_tex},
+       {TGSI_OPCODE_DDY_FINE,  0, FETCH_OP_GET_GRADIENTS_V, tgsi_tex},
        {TGSI_OPCODE_PUSHA,     0, ALU_OP0_NOP, tgsi_unsupported},
        {TGSI_OPCODE_POPA,      0, ALU_OP0_NOP, tgsi_unsupported},
        {TGSI_OPCODE_CEIL,      0, ALU_OP1_CEIL, tgsi_op2},
@@ -7195,9 +7198,8 @@ static struct r600_shader_tgsi_instruction cm_shader_tgsi_instruction[] = {
        {76,                    0, ALU_OP0_NOP, tgsi_unsupported},
        {TGSI_OPCODE_ELSE,      0, ALU_OP0_NOP, tgsi_else},
        {TGSI_OPCODE_ENDIF,     0, ALU_OP0_NOP, tgsi_endif},
-       /* gap */
-       {79,                    0, ALU_OP0_NOP, tgsi_unsupported},
-       {80,                    0, ALU_OP0_NOP, tgsi_unsupported},
+       {TGSI_OPCODE_DDX_FINE,  0, FETCH_OP_GET_GRADIENTS_H, tgsi_tex},
+       {TGSI_OPCODE_DDY_FINE,  0, FETCH_OP_GET_GRADIENTS_V, tgsi_tex},
        {TGSI_OPCODE_PUSHA,     0, ALU_OP0_NOP, tgsi_unsupported},
        {TGSI_OPCODE_POPA,      0, ALU_OP0_NOP, tgsi_unsupported},
        {TGSI_OPCODE_CEIL,      0, ALU_OP1_CEIL, tgsi_op2},