struct r300_vertex_program_compiler compiler;
struct rc_instruction *inst;
+ memset(&compiler, 0, sizeof(compiler));
rc_init(&compiler.Base);
inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev);
compiler.RequiredOutputs = compiler.Base.Program.OutputsWritten = (1 << VERT_RESULT_HPOS) | (1 << VERT_RESULT_TEX0);
compiler.SetHwInputOutput = vp_ins_outs;
compiler.code = &r300->blit.vp_code;
+ compiler.Base.is_r500 = r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515;
+ compiler.Base.disable_optimizations = 0;
+ compiler.Base.has_half_swizzles = 0;
+ compiler.Base.max_temp_regs = 32;
+ compiler.Base.max_constants = 256;
+ compiler.Base.max_alu_insts = compiler.Base.is_r500 ? 1024 : 256;
r3xx_compile_vertex_program(&compiler);
}
struct r300_fragment_program_compiler compiler;
struct rc_instruction *inst;
+ memset(&compiler, 0, sizeof(struct r300_fragment_program_compiler));
rc_init(&compiler.Base);
inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev);
inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW;
compiler.Base.Program.InputsRead = (1 << FRAG_ATTRIB_TEX0);
- compiler.OutputColor = FRAG_RESULT_COLOR;
+ compiler.OutputColor[0] = FRAG_RESULT_COLOR;
compiler.OutputDepth = FRAG_RESULT_DEPTH;
- compiler.is_r500 = (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515);
+ compiler.enable_shadow_ambient = GL_TRUE;
+ compiler.Base.is_r500 = (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515);
+ compiler.Base.disable_optimizations = 0;
+ compiler.Base.has_half_swizzles = 1;
+ compiler.Base.max_temp_regs = (compiler.Base.is_r500) ? 128 : 32;
+ compiler.Base.max_constants = compiler.Base.is_r500 ? 256 : 32;
+ compiler.Base.max_alu_insts = compiler.Base.is_r500 ? 512 : 64;
compiler.code = &r300->blit.fp_code;
compiler.AllocateHwInputs = fp_allocate_hw_inputs;
void r300_blit_init(struct r300_context *r300)
{
- create_vertex_program(r300);
+ if (r300->options.hw_tcl_enabled)
+ create_vertex_program(r300);
create_fragment_program(r300);
}
unsigned height,
unsigned pitch)
{
+ int is_r500 = r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515;
BATCH_LOCALS(&r300->radeon);
- assert(width <= 2048);
- assert(height <= 2048);
+ assert(is_r500 ? width <= 4096 : width <= 2048);
+ assert(is_r500 ? height <= 4096 : height <= 2048);
assert(r300TranslateTexFormat(mesa_format) >= 0);
assert(offset % 32 == 0);
(R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_T_SHIFT) |
(R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_R_SHIFT) |
R300_TX_MIN_FILTER_MIP_NONE |
- R300_TX_MIN_FILTER_LINEAR |
- R300_TX_MAG_FILTER_LINEAR |
+ R300_TX_MIN_FILTER_NEAREST |
+ R300_TX_MAG_FILTER_NEAREST |
(0 << 28));
OUT_BATCH_REGVAL(R300_TX_FILTER1_0, 0);
OUT_BATCH_REGVAL(R300_TX_SIZE_0,
- ((width-1) << R300_TX_WIDTHMASK_SHIFT) |
- ((height-1) << R300_TX_HEIGHTMASK_SHIFT) |
+ (((width - 1) & 0x7ff) << R300_TX_WIDTHMASK_SHIFT) |
+ (((height - 1) & 0x7ff) << R300_TX_HEIGHTMASK_SHIFT) |
(0 << R300_TX_DEPTHMASK_SHIFT) |
(0 << R300_TX_MAX_MIP_LEVEL_SHIFT) |
R300_TX_SIZE_TXPITCH_EN);
OUT_BATCH_REGVAL(R300_TX_FORMAT_0, r300TranslateTexFormat(mesa_format));
- OUT_BATCH_REGVAL(R300_TX_FORMAT2_0, pitch/_mesa_get_format_bytes(mesa_format) - 1);
+ OUT_BATCH_REGVAL(R300_TX_FORMAT2_0,
+ (pitch - 1) |
+ (is_r500 && width > 2048 ? R500_TXWIDTH_BIT11 : 0) |
+ (is_r500 && height > 2048 ? R500_TXHEIGHT_BIT11 : 0));
OUT_BATCH_REGSEQ(R300_TX_OFFSET_0, 1);
OUT_BATCH_RELOC(0, bo, offset, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
{
switch(mesa_format)
{
- case MESA_FORMAT_S8_Z24:
- case MESA_FORMAT_X8_Z24:
case MESA_FORMAT_RGBA8888: // x
return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, B, G, R, 0);
case MESA_FORMAT_RGB565: // x
return EASY_US_FORMAT(R500_OUT_FMT_C4_16, R, G, B, A, 0xf);
default:
- assert(!"Invalid format for US output\n");
+ fprintf(stderr, "Unsupported format %s for US output\n", _mesa_get_format_name(mesa_format));
+ assert(0);
return 0;
}
}
END_BATCH();
}
-static void emit_vap_setup(struct r300_context *r300, unsigned width, unsigned height)
+static void emit_vap_setup(struct r300_context *r300)
{
+ int tex_offset;
BATCH_LOCALS(&r300->radeon);
+ if (r300->options.hw_tcl_enabled)
+ tex_offset = 1;
+ else
+ tex_offset = 6;
+
BEGIN_BATCH(12);
OUT_BATCH_REGSEQ(R300_SE_VTE_CNTL, 2);
OUT_BATCH(R300_VTX_XY_FMT | R300_VTX_Z_FMT);
OUT_BATCH_REGVAL(R300_VAP_PSC_SGN_NORM_CNTL, 0xaaaaaaaa);
OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_0,
((R300_DATA_TYPE_FLOAT_2 | (0 << R300_DST_VEC_LOC_SHIFT)) << 0) |
- (((1 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_2 | R300_LAST_VEC) << 16));
+ (((tex_offset << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_2 | R300_LAST_VEC) << 16));
OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_EXT_0,
((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) |
(R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) |
struct radeon_bo *dst_bo)
{
int ret;
- radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
- src_bo, RADEON_GEM_DOMAIN_VRAM, 0);
- radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
- dst_bo, 0, RADEON_GEM_DOMAIN_VRAM);
+ radeon_cs_space_reset_bos(r300->radeon.cmdbuf.cs);
+
+ ret = radeon_cs_space_check_with_bo(r300->radeon.cmdbuf.cs,
+ src_bo, RADEON_GEM_DOMAIN_VRAM | RADEON_GEM_DOMAIN_GTT, 0);
+ if (ret)
+ return GL_FALSE;
ret = radeon_cs_space_check_with_bo(r300->radeon.cmdbuf.cs,
- first_elem(&r300->radeon.dma.reserved)->bo,
- RADEON_GEM_DOMAIN_GTT, 0);
+ dst_bo, 0, RADEON_GEM_DOMAIN_VRAM | RADEON_GEM_DOMAIN_GTT);
if (ret)
return GL_FALSE;
return GL_TRUE;
}
-static void emit_draw_packet(struct r300_context *r300, float width, float height)
+/**
+ * Calculate texcoords for given image region.
+ * Output values are [minx, maxx, miny, maxy]
+ */
+static void calc_tex_coords(float img_width, float img_height,
+ float x, float y,
+ float reg_width, float reg_height,
+ unsigned flip_y, float *buf)
+{
+ buf[0] = x / img_width;
+ buf[1] = buf[0] + reg_width / img_width;
+ buf[2] = y / img_height;
+ buf[3] = buf[2] + reg_height / img_height;
+ if (flip_y)
+ {
+ buf[2] = 1.0 - buf[2];
+ buf[3] = 1.0 - buf[3];
+ }
+}
+
+static void emit_draw_packet(struct r300_context *r300,
+ unsigned src_width, unsigned src_height,
+ unsigned src_x_offset, unsigned src_y_offset,
+ unsigned dst_x_offset, unsigned dst_y_offset,
+ unsigned reg_width, unsigned reg_height,
+ unsigned flip_y)
{
- float verts[] = { 0.0, 0.0, 0.0, 1.0,
- 0.0, height, 0.0, 0.0,
- width, height, 1.0, 0.0,
- width, 0.0, 1.0, 1.0 };
+ float texcoords[4];
+
+ calc_tex_coords(src_width, src_height,
+ src_x_offset, src_y_offset,
+ reg_width, reg_height,
+ flip_y, texcoords);
+
+ float verts[] = { dst_x_offset, dst_y_offset,
+ texcoords[0], texcoords[2],
+ dst_x_offset, dst_y_offset + reg_height,
+ texcoords[0], texcoords[3],
+ dst_x_offset + reg_width, dst_y_offset + reg_height,
+ texcoords[1], texcoords[3],
+ dst_x_offset + reg_width, dst_y_offset,
+ texcoords[1], texcoords[2] };
BATCH_LOCALS(&r300->radeon);
{
BATCH_LOCALS(&r300->radeon);
- BEGIN_BATCH(15);
+ BEGIN_BATCH(13);
OUT_BATCH_REGVAL(R300_GA_POLY_MODE,
R300_GA_POLY_MODE_FRONT_PTYPE_TRI | R300_GA_POLY_MODE_BACK_PTYPE_TRI);
OUT_BATCH_REGVAL(R300_SU_CULL_MODE, R300_FRONT_FACE_CCW);
OUT_BATCH_REGSEQ(R300_RB3D_CBLEND, 2);
OUT_BATCH(0x0);
OUT_BATCH(0x0);
- OUT_BATCH_REGVAL(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE);
OUT_BATCH_REGVAL(R300_ZB_CNTL, 0);
END_BATCH();
+ if (r300->options.hw_tcl_enabled) {
+ BEGIN_BATCH(2);
+ OUT_BATCH_REGVAL(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE);
+ END_BATCH();
+ }
}
static void emit_cb_setup(struct r300_context *r300,
END_BATCH();
}
-GLboolean r300_blit(struct r300_context *r300,
- struct radeon_bo *src_bo,
- intptr_t src_offset,
- gl_format src_mesaformat,
- unsigned src_pitch,
- unsigned src_width,
- unsigned src_height,
- struct radeon_bo *dst_bo,
- intptr_t dst_offset,
- gl_format dst_mesaformat,
- unsigned dst_pitch,
- unsigned dst_width,
- unsigned dst_height)
+unsigned r300_check_blit(gl_format dst_format)
+{
+ switch (dst_format) {
+ case MESA_FORMAT_RGB565:
+ case MESA_FORMAT_ARGB1555:
+ case MESA_FORMAT_RGBA8888:
+ case MESA_FORMAT_RGBA8888_REV:
+ case MESA_FORMAT_ARGB8888:
+ case MESA_FORMAT_ARGB8888_REV:
+ case MESA_FORMAT_XRGB8888:
+ break;
+ default:
+ return 0;
+ }
+
+ if (_mesa_get_format_bits(dst_format, GL_DEPTH_BITS) > 0)
+ return 0;
+
+ return 1;
+}
+
+/**
+ * Copy a region of [@a width x @a height] pixels from source buffer
+ * to destination buffer.
+ * @param[in] r300 r300 context
+ * @param[in] src_bo source radeon buffer object
+ * @param[in] src_offset offset of the source image in the @a src_bo
+ * @param[in] src_mesaformat source image format
+ * @param[in] src_pitch aligned source image width
+ * @param[in] src_width source image width
+ * @param[in] src_height source image height
+ * @param[in] src_x_offset x offset in the source image
+ * @param[in] src_y_offset y offset in the source image
+ * @param[in] dst_bo destination radeon buffer object
+ * @param[in] dst_offset offset of the destination image in the @a dst_bo
+ * @param[in] dst_mesaformat destination image format
+ * @param[in] dst_pitch aligned destination image width
+ * @param[in] dst_width destination image width
+ * @param[in] dst_height destination image height
+ * @param[in] dst_x_offset x offset in the destination image
+ * @param[in] dst_y_offset y offset in the destination image
+ * @param[in] width region width
+ * @param[in] height region height
+ * @param[in] flip_y set if y coords of the source image need to be flipped
+ */
+unsigned r300_blit(struct gl_context *ctx,
+ struct radeon_bo *src_bo,
+ intptr_t src_offset,
+ gl_format src_mesaformat,
+ unsigned src_pitch,
+ unsigned src_width,
+ unsigned src_height,
+ unsigned src_x_offset,
+ unsigned src_y_offset,
+ struct radeon_bo *dst_bo,
+ intptr_t dst_offset,
+ gl_format dst_mesaformat,
+ unsigned dst_pitch,
+ unsigned dst_width,
+ unsigned dst_height,
+ unsigned dst_x_offset,
+ unsigned dst_y_offset,
+ unsigned reg_width,
+ unsigned reg_height,
+ unsigned flip_y)
{
+ r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+ if (!r300_check_blit(dst_mesaformat))
+ return 0;
+
+ /* Make sure that colorbuffer has even width - hw limitation */
+ if (dst_pitch % 2 > 0)
+ ++dst_pitch;
+
+ /* Need to clamp the region size to make sure
+ * we don't read outside of the source buffer
+ * or write outside of the destination buffer.
+ */
+ if (reg_width + src_x_offset > src_width)
+ reg_width = src_width - src_x_offset;
+ if (reg_height + src_y_offset > src_height)
+ reg_height = src_height - src_y_offset;
+ if (reg_width + dst_x_offset > dst_width)
+ reg_width = dst_width - dst_x_offset;
+ if (reg_height + dst_y_offset > dst_height)
+ reg_height = dst_height - dst_y_offset;
+
if (src_bo == dst_bo) {
+ return 0;
+ }
+
+ if (src_offset % 32 || dst_offset % 32) {
return GL_FALSE;
}
if (0) {
- fprintf(stderr, "src: width %d, height %d, pitch %d vs %d, format %s\n",
+ fprintf(stderr, "src: size [%d x %d], pitch %d, "
+ "offset [%d x %d], format %s, bo %p\n",
src_width, src_height, src_pitch,
- _mesa_format_row_stride(src_mesaformat, src_width),
- _mesa_get_format_name(src_mesaformat));
- fprintf(stderr, "dst: width %d, height %d, pitch %d, format %s\n",
- dst_width, dst_height,
- _mesa_format_row_stride(dst_mesaformat, dst_width),
- _mesa_get_format_name(dst_mesaformat));
+ src_x_offset, src_y_offset,
+ _mesa_get_format_name(src_mesaformat),
+ src_bo);
+ fprintf(stderr, "dst: pitch %d, offset[%d x %d], format %s, bo %p\n",
+ dst_pitch, dst_x_offset, dst_y_offset,
+ _mesa_get_format_name(dst_mesaformat), dst_bo);
+ fprintf(stderr, "region: %d x %d\n", reg_width, reg_height);
}
+ /* Flush is needed to make sure that source buffer has correct data */
+ radeonFlush(r300->radeon.glCtx);
+
if (!validate_buffers(r300, src_bo, dst_bo))
- return GL_FALSE;
+ return 0;
rcommonEnsureCmdBufSpace(&r300->radeon, 200, __FUNCTION__);
r300_emit_rs_setup(r300);
}
- emit_pvs_setup(r300, r300->blit.vp_code.body.d, 2);
- emit_vap_setup(r300, dst_width, dst_height);
+ if (r300->options.hw_tcl_enabled)
+ emit_pvs_setup(r300, r300->blit.vp_code.body.d, 2);
+
+ emit_vap_setup(r300);
emit_cb_setup(r300, dst_bo, dst_offset, dst_mesaformat, dst_pitch, dst_width, dst_height);
- emit_draw_packet(r300, dst_width, dst_height);
+ emit_draw_packet(r300, src_width, src_height,
+ src_x_offset, src_y_offset,
+ dst_x_offset, dst_y_offset,
+ reg_width, reg_height,
+ flip_y);
r300EmitCacheFlush(r300);
radeonFlush(r300->radeon.glCtx);
- return GL_TRUE;
-}
\ No newline at end of file
+ return 1;
+}