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 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 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 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 GLboolean evergreenTryDrawPrims(GLcontext *ctx,
const struct gl_client_array *arrays[],
const struct _mesa_prim *prim,
GLuint nr_prims,
const struct _mesa_index_buffer *ib,
GLuint min_index,
GLuint max_index )
{
context_t *context = EVERGREEN_CONTEXT(ctx);
radeonContextPtr radeon = &context->radeon;
GLuint i, id = 0;
struct radeon_renderbuffer *rrb;
if (ctx->NewState)
_mesa_update_state( ctx );
if (evergreen_check_fallbacks(ctx))
return GL_FALSE;
_tnl_UpdateFixedFunctionProgram(ctx);
evergreenSetVertexFormat(ctx, arrays, max_index + 1);
/* shaders need to be updated before buffers are validated */
evergreenUpdateShaders(ctx);
if (!evergreenValidateBuffers(ctx))
return GL_FALSE;
/* always emit CB base to prevent
* lock ups on some chips.
*/
EVERGREEN_STATECHANGE(context, cb);
/* mark vtx as dirty since it changes per-draw */
EVERGREEN_STATECHANGE(context, vtx);
evergreenSetScissor(context);
evergreenSetupVertexProgram(ctx);
evergreenSetupFragmentProgram(ctx);
evergreenUpdateShaderStates(ctx);
GLuint emit_end = evergreenPredictRenderSize(ctx, prim, ib, nr_prims)
+ context->radeon.cmdbuf.cs->cdw;
/* evergreenPredictRenderSize will call radeonReleaseDmaRegions, so update VP/FP const buf after it. */
evergreenSetupVPconstants(ctx);
evergreenSetupFPconstants(ctx);
evergreenSetupIndexBuffer(ctx, ib);
evergreenSetupStreams(ctx, arrays, max_index + 1);
radeonEmitState(radeon);
radeon_debug_add_indent();
for (i = 0; i < nr_prims; ++i)
{
if (context->ind_buf.bo)
evergreenRunRenderPrimitive(ctx,
prim[i].start,
prim[i].start + prim[i].count,
prim[i].mode,
prim[i].basevertex);
else
evergreenRunRenderPrimitiveImmediate(ctx,
prim[i].start,
prim[i].start + prim[i].count,
prim[i].mode);
}
radeon_debug_remove_indent();
/* Flush render op cached for last several quads. */
/* XXX drm should handle this in fence submit */
//evergreeWaitForIdleClean(context);
rrb = radeon_get_colorbuffer(&context->radeon);
if (rrb && rrb->bo)
r700SyncSurf(context, rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM,
CB_ACTION_ENA_bit | (1 << (id + 6)));
rrb = radeon_get_depthbuffer(&context->radeon);
if (rrb && rrb->bo)
r700SyncSurf(context, rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM,
DB_ACTION_ENA_bit | DB_DEST_BASE_ENA_bit);
evergreenFreeData(ctx);
if (emit_end < context->radeon.cmdbuf.cs->cdw)
{
WARN_ONCE("Rendering was %d commands larger than predicted size."
" We might overflow command buffer.\n", context->radeon.cmdbuf.cs->cdw - emit_end);
}
return GL_TRUE;
}
