static void r700SendQueryBegin(GLcontext *ctx, struct radeon_state_atom *atom)
{
radeonContextPtr radeon = RADEON_CONTEXT(ctx);
struct radeon_query_object *query = radeon->query.current;
BATCH_LOCALS(radeon);
radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
/* clear the buffer */
radeon_bo_map(query->bo, GL_FALSE);
memset(query->bo->ptr, 0, 4 * 2 * sizeof(uint64_t)); /* 4 DBs, 2 qwords each */
radeon_bo_unmap(query->bo);
radeon_cs_space_check_with_bo(radeon->cmdbuf.cs,
query->bo,
0, RADEON_GEM_DOMAIN_GTT);
BEGIN_BATCH_NO_AUTOSTATE(4 + 2);
R600_OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 2));
R600_OUT_BATCH(ZPASS_DONE);
R600_OUT_BATCH(query->curr_offset); /* hw writes qwords */
R600_OUT_BATCH(0x00000000);
R600_OUT_BATCH_RELOC(VGT_EVENT_INITIATOR, query->bo, 0, 0, RADEON_GEM_DOMAIN_GTT, 0);
END_BATCH();
query->emitted_begin = GL_TRUE;
}
static void evergreen_emit_query_finish(radeonContextPtr radeon)
{
context_t *context = (context_t*) radeon;
BATCH_LOCALS(&context->radeon);
struct radeon_query_object *query = radeon->query.current;
BEGIN_BATCH_NO_AUTOSTATE(4 + 2);
R600_OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 2));
R600_OUT_BATCH(R600_EVENT_TYPE(ZPASS_DONE) | R600_EVENT_INDEX(1));
R600_OUT_BATCH(query->curr_offset + 8); /* hw writes qwords */
R600_OUT_BATCH(0x00000000);
R600_OUT_BATCH_RELOC(VGT_EVENT_INITIATOR, query->bo, 0, 0, RADEON_GEM_DOMAIN_GTT, 0);
END_BATCH();
assert(query->curr_offset < RADEON_QUERY_PAGE_SIZE);
query->emitted_begin = GL_FALSE;
}
static void r700SendFSState(GLcontext *ctx, struct radeon_state_atom *atom)
{
context_t *context = R700_CONTEXT(ctx);
R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
struct radeon_bo * pbo;
BATCH_LOCALS(&context->radeon);
radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
/* XXX fixme
* R6xx chips require a FS be emitted, even if it's not used.
* since we aren't using FS yet, just send the VS address to make
* the kernel command checker happy
*/
pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context));
r700->fs.SQ_PGM_START_FS.u32All = r700->vs.SQ_PGM_START_VS.u32All;
r700->fs.SQ_PGM_RESOURCES_FS.u32All = 0;
r700->fs.SQ_PGM_CF_OFFSET_FS.u32All = 0;
/* XXX */
if (!pbo)
return;
r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
R600_OUT_BATCH_REGSEQ(SQ_PGM_START_FS, 1);
R600_OUT_BATCH(r700->fs.SQ_PGM_START_FS.u32All);
R600_OUT_BATCH_RELOC(r700->fs.SQ_PGM_START_FS.u32All,
pbo,
r700->fs.SQ_PGM_START_FS.u32All,
RADEON_GEM_DOMAIN_GTT, 0, 0);
END_BATCH();
BEGIN_BATCH_NO_AUTOSTATE(6);
R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_FS, r700->fs.SQ_PGM_RESOURCES_FS.u32All);
R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_FS, r700->fs.SQ_PGM_CF_OFFSET_FS.u32All);
END_BATCH();
COMMIT_BATCH();
}
static void r700SendDepthTargetState(GLcontext *ctx, struct radeon_state_atom *atom)
{
context_t *context = R700_CONTEXT(ctx);
R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
struct radeon_renderbuffer *rrb;
BATCH_LOCALS(&context->radeon);
radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
rrb = radeon_get_depthbuffer(&context->radeon);
if (!rrb || !rrb->bo) {
return;
}
r700SetDepthTarget(context);
BEGIN_BATCH_NO_AUTOSTATE(8 + 2);
R600_OUT_BATCH_REGSEQ(DB_DEPTH_SIZE, 2);
R600_OUT_BATCH(r700->DB_DEPTH_SIZE.u32All);
R600_OUT_BATCH(r700->DB_DEPTH_VIEW.u32All);
R600_OUT_BATCH_REGSEQ(DB_DEPTH_BASE, 2);
R600_OUT_BATCH(r700->DB_DEPTH_BASE.u32All);
R600_OUT_BATCH(r700->DB_DEPTH_INFO.u32All);
R600_OUT_BATCH_RELOC(r700->DB_DEPTH_BASE.u32All,
rrb->bo,
r700->DB_DEPTH_BASE.u32All,
0, RADEON_GEM_DOMAIN_VRAM, 0);
END_BATCH();
if ((context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) &&
(context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)) {
BEGIN_BATCH_NO_AUTOSTATE(2);
R600_OUT_BATCH(CP_PACKET3(R600_IT_SURFACE_BASE_UPDATE, 0));
R600_OUT_BATCH(1 << 0);
END_BATCH();
}
COMMIT_BATCH();
}
static void r700SendPSState(GLcontext *ctx, struct radeon_state_atom *atom)
{
context_t *context = R700_CONTEXT(ctx);
R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
struct radeon_bo * pbo;
BATCH_LOCALS(&context->radeon);
radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context));
if (!pbo)
return;
r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1);
R600_OUT_BATCH(r700->ps.SQ_PGM_START_PS.u32All);
R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All,
pbo,
r700->ps.SQ_PGM_START_PS.u32All,
RADEON_GEM_DOMAIN_GTT, 0, 0);
END_BATCH();
BEGIN_BATCH_NO_AUTOSTATE(9);
R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All);
R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All);
R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All);
END_BATCH();
BEGIN_BATCH_NO_AUTOSTATE(3);
R600_OUT_BATCH_REGVAL(SQ_LOOP_CONST_0, 0x01000FFF);
END_BATCH();
COMMIT_BATCH();
}
static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom *atom)
{
context_t *context = R700_CONTEXT(ctx);
R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
struct radeon_renderbuffer *rrb;
BATCH_LOCALS(&context->radeon);
int id = 0;
radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
rrb = radeon_get_colorbuffer(&context->radeon);
if (!rrb || !rrb->bo) {
return;
}
r700SetRenderTarget(context, 0);
if (id > R700_MAX_RENDER_TARGETS)
return;
if (!r700->render_target[id].enabled)
return;
BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
R600_OUT_BATCH_REGSEQ(CB_COLOR0_BASE + (4 * id), 1);
R600_OUT_BATCH(r700->render_target[id].CB_COLOR0_BASE.u32All);
R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_BASE.u32All,
rrb->bo,
r700->render_target[id].CB_COLOR0_BASE.u32All,
0, RADEON_GEM_DOMAIN_VRAM, 0);
END_BATCH();
if ((context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) &&
(context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)) {
BEGIN_BATCH_NO_AUTOSTATE(2);
R600_OUT_BATCH(CP_PACKET3(R600_IT_SURFACE_BASE_UPDATE, 0));
R600_OUT_BATCH((2 << id));
END_BATCH();
}
/* Set CMASK & TILE buffer to the offset of color buffer as
* we don't use those this shouldn't cause any issue and we
* then have a valid cmd stream
*/
BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
R600_OUT_BATCH_REGSEQ(CB_COLOR0_TILE + (4 * id), 1);
R600_OUT_BATCH(r700->render_target[id].CB_COLOR0_TILE.u32All);
R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_BASE.u32All,
rrb->bo,
r700->render_target[id].CB_COLOR0_BASE.u32All,
0, RADEON_GEM_DOMAIN_VRAM, 0);
END_BATCH();
BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
R600_OUT_BATCH_REGSEQ(CB_COLOR0_FRAG + (4 * id), 1);
R600_OUT_BATCH(r700->render_target[id].CB_COLOR0_FRAG.u32All);
R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_BASE.u32All,
rrb->bo,
r700->render_target[id].CB_COLOR0_BASE.u32All,
0, RADEON_GEM_DOMAIN_VRAM, 0);
END_BATCH();
BEGIN_BATCH_NO_AUTOSTATE(12);
R600_OUT_BATCH_REGVAL(CB_COLOR0_SIZE + (4 * id), r700->render_target[id].CB_COLOR0_SIZE.u32All);
R600_OUT_BATCH_REGVAL(CB_COLOR0_VIEW + (4 * id), r700->render_target[id].CB_COLOR0_VIEW.u32All);
R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), r700->render_target[id].CB_COLOR0_INFO.u32All);
R600_OUT_BATCH_REGVAL(CB_COLOR0_MASK + (4 * id), r700->render_target[id].CB_COLOR0_MASK.u32All);
END_BATCH();
COMMIT_BATCH();
}
static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom)
{
context_t *context = R700_CONTEXT(ctx);
R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
struct r700_vertex_program *vp = context->selected_vp;
struct radeon_bo *bo = NULL;
unsigned int i;
BATCH_LOCALS(&context->radeon);
radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__);
for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) {
if (ctx->Texture.Unit[i]._ReallyEnabled) {
radeonTexObj *t = r700->textures[i];
if (t) {
if (!t->image_override) {
bo = t->mt->bo;
} else {
bo = t->bo;
}
if (bo) {
r700SyncSurf(context, bo,
RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM,
0, TC_ACTION_ENA_bit);
BEGIN_BATCH_NO_AUTOSTATE(9 + 4);
R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7));
if( (1<<i) & vp->r700AsmCode.unVetTexBits )
{ /* vs texture */
R600_OUT_BATCH((i + VERT_ATTRIB_MAX + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE);
}
else
{
R600_OUT_BATCH(i * 7);
}
R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE0);
R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE1);
R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE2);
R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE3);
R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE4);
R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE5);
R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE6);
R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE2,
bo,
r700->textures[i]->SQ_TEX_RESOURCE2,
RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE3,
bo,
r700->textures[i]->SQ_TEX_RESOURCE3,
RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
END_BATCH();
COMMIT_BATCH();
}
}
}
}
}
static void r700SetupVTXConstants(GLcontext * ctx,
void * pAos,
StreamDesc * pStreamDesc)
{
context_t *context = R700_CONTEXT(ctx);
struct radeon_aos * paos = (struct radeon_aos *)pAos;
unsigned int nVBsize;
BATCH_LOCALS(&context->radeon);
unsigned int uSQ_VTX_CONSTANT_WORD0_0;
unsigned int uSQ_VTX_CONSTANT_WORD1_0;
unsigned int uSQ_VTX_CONSTANT_WORD2_0 = 0;
unsigned int uSQ_VTX_CONSTANT_WORD3_0 = 0;
unsigned int uSQ_VTX_CONSTANT_WORD6_0 = 0;
if (!paos->bo)
return;
if ((context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV610) ||
(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV620) ||
(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS780) ||
(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS880) ||
(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV710))
r700SyncSurf(context, paos->bo, RADEON_GEM_DOMAIN_GTT, 0, TC_ACTION_ENA_bit);
else
r700SyncSurf(context, paos->bo, RADEON_GEM_DOMAIN_GTT, 0, VC_ACTION_ENA_bit);
if(0 == pStreamDesc->stride)
{
nVBsize = paos->count * pStreamDesc->size * getTypeSize(pStreamDesc->type);
}
else
{
nVBsize = (paos->count - 1) * pStreamDesc->stride
+ pStreamDesc->size * getTypeSize(pStreamDesc->type);
}
uSQ_VTX_CONSTANT_WORD0_0 = paos->offset;
uSQ_VTX_CONSTANT_WORD1_0 = nVBsize - 1;
SETfield(uSQ_VTX_CONSTANT_WORD2_0, 0, BASE_ADDRESS_HI_shift, BASE_ADDRESS_HI_mask); /* TODO */
SETfield(uSQ_VTX_CONSTANT_WORD2_0, pStreamDesc->stride, SQ_VTX_CONSTANT_WORD2_0__STRIDE_shift,
SQ_VTX_CONSTANT_WORD2_0__STRIDE_mask);
SETfield(uSQ_VTX_CONSTANT_WORD2_0, GetSurfaceFormat(pStreamDesc->type, pStreamDesc->size, NULL),
SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_shift,
SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_mask); /* TODO : trace back api for initial data type, not only GL_FLOAT */
if(GL_TRUE == pStreamDesc->normalize)
{
SETfield(uSQ_VTX_CONSTANT_WORD2_0, SQ_NUM_FORMAT_NORM,
SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift, SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_mask);
}
else
{
SETfield(uSQ_VTX_CONSTANT_WORD2_0, SQ_NUM_FORMAT_SCALED,
SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift, SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_mask);
}
if(1 == pStreamDesc->_signed)
{
SETbit(uSQ_VTX_CONSTANT_WORD2_0, SQ_VTX_CONSTANT_WORD2_0__FORMAT_COMP_ALL_bit);
}
SETfield(uSQ_VTX_CONSTANT_WORD3_0, 1, MEM_REQUEST_SIZE_shift, MEM_REQUEST_SIZE_mask);
SETfield(uSQ_VTX_CONSTANT_WORD6_0, SQ_TEX_VTX_VALID_BUFFER,
SQ_TEX_RESOURCE_WORD6_0__TYPE_shift, SQ_TEX_RESOURCE_WORD6_0__TYPE_mask);
BEGIN_BATCH_NO_AUTOSTATE(9 + 2);
R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7));
R600_OUT_BATCH((pStreamDesc->element + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE);
R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD0_0);
R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD1_0);
R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD2_0);
R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD3_0);
R600_OUT_BATCH(0);
R600_OUT_BATCH(0);
R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD6_0);
R600_OUT_BATCH_RELOC(uSQ_VTX_CONSTANT_WORD0_0,
paos->bo,
uSQ_VTX_CONSTANT_WORD0_0,
RADEON_GEM_DOMAIN_GTT, 0, 0);
END_BATCH();
COMMIT_BATCH();
}
static void evergreenRunRenderPrimitive(GLcontext * ctx, int start, int end, int prim,
GLint basevertex) //same
{
context_t *context = EVERGREEN_CONTEXT(ctx);
BATCH_LOCALS(&context->radeon);
int type, total_emit;
int num_indices;
uint32_t vgt_draw_initiator = 0;
uint32_t vgt_index_type = 0;
uint32_t vgt_primitive_type = 0;
uint32_t vgt_num_indices = 0;
type = evergreenPrimitiveType(prim);
num_indices = evergreenNumVerts(end - start, prim);
radeon_print(RADEON_RENDER, RADEON_TRACE,
"%s type %x num_indices %d\n",
__func__, type, num_indices);
if (type < 0 || num_indices <= 0)
return;
SETfield(vgt_primitive_type, type,
VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift, VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask);
SETfield(vgt_index_type, DI_INDEX_SIZE_32_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask);
if(GL_TRUE != context->ind_buf.is_32bit)
{
SETfield(vgt_index_type, DI_INDEX_SIZE_16_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask);
}
vgt_num_indices = num_indices;
SETfield(vgt_draw_initiator, DI_SRC_SEL_DMA, SOURCE_SELECT_shift, SOURCE_SELECT_mask);
SETfield(vgt_draw_initiator, DI_MAJOR_MODE_0, MAJOR_MODE_shift, MAJOR_MODE_mask);
total_emit = 3 /* VGT_PRIMITIVE_TYPE */
+ 2 /* VGT_INDEX_TYPE */
+ 2 /* NUM_INSTANCES */
+ 4 /* VTX_BASE_VTX_LOC + VTX_START_INST_LOC */
+ 5 + 2; /* DRAW_INDEX */
BEGIN_BATCH_NO_AUTOSTATE(total_emit);
// prim
R600_OUT_BATCH_REGSEQ(VGT_PRIMITIVE_TYPE, 1);
R600_OUT_BATCH(vgt_primitive_type);
// index type
R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0));
R600_OUT_BATCH(vgt_index_type);
// num instances
R600_OUT_BATCH(CP_PACKET3(R600_IT_NUM_INSTANCES, 0));
R600_OUT_BATCH(1);
/* offset */
R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 2));
R600_OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX);
R600_OUT_BATCH(basevertex); //VTX_BASE_VTX_LOC
R600_OUT_BATCH(0); //VTX_START_INST_LOC
// draw packet
R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX, 3));
R600_OUT_BATCH(context->ind_buf.bo_offset);
R600_OUT_BATCH(0);
R600_OUT_BATCH(vgt_num_indices);
R600_OUT_BATCH(vgt_draw_initiator);
R600_OUT_BATCH_RELOC(context->ind_buf.bo_offset,
context->ind_buf.bo,
context->ind_buf.bo_offset,
RADEON_GEM_DOMAIN_GTT, 0, 0);
END_BATCH();
COMMIT_BATCH();
}
