#include "radeon_reg.h"
#include "r300_reg.h"
#include "r300_cmdbuf.h"
+#include "r300_emit.h"
// Set this to 1 for extremely verbose debugging of command buffers
/**
* Send the current command buffer via ioctl to the hardware.
*/
-int r300FlushCmdBuf(r300ContextPtr r300, const char* caller)
+int r300FlushCmdBufLocked(r300ContextPtr r300, const char* caller)
{
int ret;
int i;
r300->cmdbuf.cmd_buf[i]);
}
- LOCK_HARDWARE(&r300->radeon);
-
cmd.buf = (char*)(r300->cmdbuf.cmd_buf + start);
cmd.bufsz = (r300->cmdbuf.count_used - start) * 4;
cmd.boxes = (drm_clip_rect_t *)r300->radeon.pClipRects;
}
- if (cmd.nbox) {
- ret = drmCommandWrite(r300->radeon.dri.fd,
- DRM_RADEON_CMDBUF, &cmd, sizeof(cmd));
- if (ret) {
- UNLOCK_HARDWARE(&r300->radeon);
- fprintf(stderr, "drmCommandWrite: %d\n", ret);
- exit(-1);
- }
+ ret = drmCommandWrite(r300->radeon.dri.fd,
+ DRM_RADEON_CMDBUF, &cmd, sizeof(cmd));
- if (RADEON_DEBUG & DEBUG_SYNC) {
- fprintf(stderr, "Syncing in %s\n\n", __FUNCTION__);
- radeonWaitForIdleLocked(&r300->radeon);
- }
- } else {
- if (RADEON_DEBUG & DEBUG_IOCTL)
- fprintf(stderr, "%s: No cliprects\n", __FUNCTION__);
+ if (RADEON_DEBUG & DEBUG_SYNC) {
+ fprintf(stderr, "Syncing in %s (from %s)\n\n", __FUNCTION__, caller);
+ radeonWaitForIdleLocked(&r300->radeon);
}
- UNLOCK_HARDWARE(&r300->radeon);
-
+ r300->dma.nr_released_bufs = 0;
r300->cmdbuf.count_used = 0;
r300->cmdbuf.count_reemit = 0;
- return 0;
+ return ret;
+}
+
+
+int r300FlushCmdBuf(r300ContextPtr r300, const char* caller)
+{
+ int ret;
+ int i;
+ drm_radeon_cmd_buffer_t cmd;
+ int start;
+
+ LOCK_HARDWARE(&r300->radeon);
+
+ ret=r300FlushCmdBufLocked(r300, caller);
+
+ UNLOCK_HARDWARE(&r300->radeon);
+
+ if (ret) {
+ fprintf(stderr, "drmRadeonCmdBuffer: %d (exiting)\n", ret);
+ exit(ret);
+ }
+
+ return ret;
}
}
}
-
/**
* Copy dirty hardware state atoms into the command buffer.
*
if (r300->cmdbuf.count_used && !r300->hw.is_dirty && !r300->hw.all_dirty)
return;
-
+
/* To avoid going across the entire set of states multiple times, just check
* for enough space for the case of emitting all state, and inline the
* r300AllocCmdBuf code here without all the checks.
r300->hw.all_dirty = GL_FALSE;
}
+#if 0
static __inline__ uint32_t cmducs(int reg, int count)
{
return cmd.u;
}
+#endif
#define CHECK( NM, COUNT ) \
static int check_##NM( r300ContextPtr r300, \
#define ALLOC_STATE( ATOM, CHK, SZ, NM, IDX ) \
do { \
- r300->hw.ATOM.cmd_size = SZ; \
- r300->hw.ATOM.cmd = (uint32_t*)CALLOC(SZ * sizeof(uint32_t)); \
- r300->hw.ATOM.name = NM; \
- r300->hw.ATOM.idx = IDX; \
+ r300->hw.ATOM.cmd_size = (SZ); \
+ r300->hw.ATOM.cmd = (uint32_t*)CALLOC((SZ) * sizeof(uint32_t)); \
+ r300->hw.ATOM.name = (NM); \
+ r300->hw.ATOM.idx = (IDX); \
r300->hw.ATOM.check = check_##CHK; \
r300->hw.ATOM.dirty = GL_FALSE; \
- r300->hw.max_state_size += SZ; \
+ r300->hw.max_state_size += (SZ); \
} while (0)
void r300InitCmdBuf(r300ContextPtr r300)
{
int size, i, mtu;
-
- r300->hw.max_state_size = 0;
+ r300->hw.max_state_size = 0;
+
mtu = r300->radeon.glCtx->Const.MaxTextureUnits;
fprintf(stderr, "Using %d maximum texture units..\n", mtu);
r300->hw.vpt.cmd[R300_VPT_CMD_0] = cmducs(R300_SE_VPORT_XSCALE, 6);
ALLOC_STATE( unk2080, always, 2, "unk2080", 0 );
r300->hw.unk2080.cmd[0] = cmducs(0x2080, 1);
- ALLOC_STATE( ovf, always, R300_OVF_CMDSIZE, "ovf", 0 );
- r300->hw.ovf.cmd[R300_OVF_CMD_0] = cmducs(R300_VAP_OUTPUT_VTX_FMT_0, 2);
ALLOC_STATE( vte, always, 3, "vte", 0 );
r300->hw.vte.cmd[0] = cmducs(R300_SE_VTE_CNTL, 2);
ALLOC_STATE( unk2134, always, 3, "unk2134", 0 );
r300->hw.ps.cmd[0] = cmducs(R300_RE_POINTSIZE, 1);
ALLOC_STATE( unk4230, always, 4, "unk4230", 0 );
r300->hw.unk4230.cmd[0] = cmducs(0x4230, 3);
+ ALLOC_STATE( lcntl, always, 2, "lcntl", 0 );
+ r300->hw.lcntl.cmd[0] = cmducs(R300_RE_LINE_CNT, 1);
+#ifdef EXP_C
+ ALLOC_STATE( lsf, always, 2, "lsf", 0 );
+ r300->hw.lsf.cmd[0] = cmducs(R300_RE_LINE_STIPPLE_FACTOR, 1);
+#endif
ALLOC_STATE( unk4260, always, 4, "unk4260", 0 );
r300->hw.unk4260.cmd[0] = cmducs(0x4260, 3);
ALLOC_STATE( unk4274, always, 5, "unk4274", 0 );
r300->hw.unk4288.cmd[0] = cmducs(0x4288, 5);
ALLOC_STATE( unk42A0, always, 2, "unk42A0", 0 );
r300->hw.unk42A0.cmd[0] = cmducs(0x42A0, 1);
+ ALLOC_STATE( zbs, always, R300_ZBS_CMDSIZE, "zbs", 0 );
+ r300->hw.zbs.cmd[R300_ZBS_CMD_0] = cmducs(R300_RE_ZBIAS_T_FACTOR, 4);
ALLOC_STATE( unk42B4, always, 2, "unk42B4", 0 );
r300->hw.unk42B4.cmd[0] = cmducs(0x42B4, 1);
ALLOC_STATE( cul, always, R300_CUL_CMDSIZE, "cul", 0 );
ALLOC_STATE( fp, always, R300_FP_CMDSIZE, "fp", 0 );
r300->hw.fp.cmd[R300_FP_CMD_0] = cmducs(R300_PFS_CNTL_0, 3);
r300->hw.fp.cmd[R300_FP_CMD_1] = cmducs(R300_PFS_NODE_0, 4);
+ ALLOC_STATE( fpt, variable, R300_FPT_CMDSIZE, "fpt", 0 );
+ r300->hw.fpt.cmd[R300_FPT_CMD_0] = cmducs(R300_PFS_TEXI_0, 0);
ALLOC_STATE( unk46A4, always, 6, "unk46A4", 0 );
r300->hw.unk46A4.cmd[0] = cmducs(0x46A4, 5);
ALLOC_STATE( fpi[0], variable, R300_FPI_CMDSIZE, "fpi/0", 0 );
ALLOC_STATE( unk4BC8, always, 4, "unk4BC8", 0 );
r300->hw.unk4BC8.cmd[0] = cmducs(0x4BC8, 3);
ALLOC_STATE( at, always, R300_AT_CMDSIZE, "at", 0 );
- r300->hw.at.cmd[R300_AT_CMD_0] = cmducs(R300_PP_ALPHA_TEST, 1);
+ r300->hw.at.cmd[R300_AT_CMD_0] = cmducs(R300_PP_ALPHA_TEST, 2);
ALLOC_STATE( unk4BD8, always, 2, "unk4BD8", 0 );
r300->hw.unk4BD8.cmd[0] = cmducs(0x4BD8, 1);
+ ALLOC_STATE( fpp, variable, R300_FPP_CMDSIZE, "fpp", 0 );
+ r300->hw.fpp.cmd[R300_FPP_CMD_0] = cmducs(R300_PFS_PARAM_0_X, 0);
ALLOC_STATE( unk4E00, always, 2, "unk4E00", 0 );
r300->hw.unk4E00.cmd[0] = cmducs(0x4E00, 1);
ALLOC_STATE( bld, always, R300_BLD_CMDSIZE, "bld", 0 );
r300->hw.unk4E88.cmd[0] = cmducs(0x4E88, 1);
ALLOC_STATE( unk4EA0, always, 3, "unk4EA0 R350 only", 0 );
r300->hw.unk4EA0.cmd[0] = cmducs(0x4EA0, 2);
- ALLOC_STATE( zc, always, R300_ZC_CMDSIZE, "zc", 0 );
- r300->hw.zc.cmd[R300_ZC_CMD_0] = cmducs(R300_RB3D_ZCNTL_0, 2);
- ALLOC_STATE( unk4F08, always, 2, "unk4F08", 0 );
- r300->hw.unk4F08.cmd[0] = cmducs(0x4F08, 1);
+ ALLOC_STATE( zs, always, R300_ZS_CMDSIZE, "zstencil", 0 );
+ r300->hw.zs.cmd[R300_ZS_CMD_0] = cmducs(R300_RB3D_ZSTENCIL_CNTL_0, 3);
ALLOC_STATE( unk4F10, always, 5, "unk4F10", 0 );
r300->hw.unk4F10.cmd[0] = cmducs(0x4F10, 4);
ALLOC_STATE( zb, always, R300_ZB_CMDSIZE, "zb", 0 );
r300->hw.vps.cmd[R300_VPS_CMD_0] = cmdvpu(R300_PVS_UPLOAD_POINTSIZE, 1);
/* Textures */
- ALLOC_STATE( tex.filter, always, mtu, "tex_filter", 0 );
- r300->hw.tex.filter.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_FILTER_0, mtu-1);
-
- ALLOC_STATE( tex.unknown1, always, mtu, "tex_unknown1", 0 );
- r300->hw.tex.unknown1.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_UNK1_0, mtu-1);
-
- ALLOC_STATE( tex.size, always, mtu, "tex_size", 0 );
- r300->hw.tex.size.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_SIZE_0, mtu-1);
-
- ALLOC_STATE( tex.format, always, mtu, "tex_format", 0 );
- r300->hw.tex.format.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_FORMAT_0, mtu-1);
-
- ALLOC_STATE( tex.offset, always, mtu, "tex_offset", 0 );
- r300->hw.tex.offset.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_OFFSET_0, mtu-1);
-
- ALLOC_STATE( tex.unknown4, always, mtu, "tex_unknown4", 0 );
- r300->hw.tex.unknown4.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_UNK4_0, mtu-1);
-
- ALLOC_STATE( tex.unknown5, always, mtu, "tex_unknown5", 0 );
- r300->hw.tex.unknown5.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_UNK5_0, mtu-1);
-
+ ALLOC_STATE( tex.filter, variable, mtu+1, "tex_filter", 0 );
+ r300->hw.tex.filter.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_FILTER_0, 0);
+
+ ALLOC_STATE( tex.unknown1, variable, mtu+1, "tex_unknown1", 0 );
+ r300->hw.tex.unknown1.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_UNK1_0, 0);
+
+ ALLOC_STATE( tex.size, variable, mtu+1, "tex_size", 0 );
+ r300->hw.tex.size.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_SIZE_0, 0);
+
+ ALLOC_STATE( tex.format, variable, mtu+1, "tex_format", 0 );
+ r300->hw.tex.format.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_FORMAT_0, 0);
+
+ ALLOC_STATE( tex.offset, variable, mtu+1, "tex_offset", 0 );
+ r300->hw.tex.offset.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_OFFSET_0, 0);
+
+ ALLOC_STATE( tex.unknown4, variable, mtu+1, "tex_unknown4", 0 );
+ r300->hw.tex.unknown4.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_UNK4_0, 0);
+
+ ALLOC_STATE( tex.border_color, variable, mtu+1, "tex_border_color", 0 );
+ r300->hw.tex.border_color.cmd[R300_TEX_CMD_0] = cmducs(R300_TX_BORDER_COLOR_0, 0);
+
/* Setup the atom linked list */
make_empty_list(&r300->hw.atomlist);
insert_at_tail(&r300->hw.atomlist, &r300->hw.vpt);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk2080);
- insert_at_tail(&r300->hw.atomlist, &r300->hw.ovf);
insert_at_tail(&r300->hw.atomlist, &r300->hw.vte);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk2134);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk2140);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4214);
insert_at_tail(&r300->hw.atomlist, &r300->hw.ps);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4230);
+ insert_at_tail(&r300->hw.atomlist, &r300->hw.lcntl);
+#ifdef EXP_C
+ insert_at_tail(&r300->hw.atomlist, &r300->hw.lsf);
+#endif
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4260);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4274);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4288);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk42A0);
+ insert_at_tail(&r300->hw.atomlist, &r300->hw.zbs);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk42B4);
insert_at_tail(&r300->hw.atomlist, &r300->hw.cul);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk42C0);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk43A4);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk43E8);
insert_at_tail(&r300->hw.atomlist, &r300->hw.fp);
+ insert_at_tail(&r300->hw.atomlist, &r300->hw.fpt);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk46A4);
insert_at_tail(&r300->hw.atomlist, &r300->hw.fpi[0]);
insert_at_tail(&r300->hw.atomlist, &r300->hw.fpi[1]);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4BC8);
insert_at_tail(&r300->hw.atomlist, &r300->hw.at);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4BD8);
+ insert_at_tail(&r300->hw.atomlist, &r300->hw.fpp);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4E00);
insert_at_tail(&r300->hw.atomlist, &r300->hw.bld);
insert_at_tail(&r300->hw.atomlist, &r300->hw.cmk);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4E50);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4E88);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4EA0);
- insert_at_tail(&r300->hw.atomlist, &r300->hw.zc);
- insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4F08);
+ insert_at_tail(&r300->hw.atomlist, &r300->hw.zs);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4F10);
insert_at_tail(&r300->hw.atomlist, &r300->hw.zb);
insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4F28);
insert_at_tail(&r300->hw.atomlist, &r300->hw.vpp);
insert_at_tail(&r300->hw.atomlist, &r300->hw.vps);
- insert_at_tail(&r300->hw.atomlist, &r300->hw.tex.filter);
+ insert_at_tail(&r300->hw.atomlist, &r300->hw.tex.filter);
insert_at_tail(&r300->hw.atomlist, &r300->hw.tex.unknown1);
insert_at_tail(&r300->hw.atomlist, &r300->hw.tex.size);
insert_at_tail(&r300->hw.atomlist, &r300->hw.tex.format);
insert_at_tail(&r300->hw.atomlist, &r300->hw.tex.offset);
insert_at_tail(&r300->hw.atomlist, &r300->hw.tex.unknown4);
- insert_at_tail(&r300->hw.atomlist, &r300->hw.tex.unknown5);
+ insert_at_tail(&r300->hw.atomlist, &r300->hw.tex.border_color);
r300->hw.is_dirty = GL_TRUE;
r300->hw.all_dirty = GL_TRUE;
/* Initialize command buffer */
size = 256 * driQueryOptioni(&r300->radeon.optionCache, "command_buffer_size");
- if (size < 2*r300->hw.max_state_size)
- size = 2*r300->hw.max_state_size;
+ if (size < 2*r300->hw.max_state_size){
+ size = 2*r300->hw.max_state_size+65535;
+ }
- if (RADEON_DEBUG & DEBUG_IOCTL)
+ if (1 || RADEON_DEBUG & DEBUG_IOCTL){
+ fprintf(stderr, "sizeof(drm_r300_cmd_header_t)=%d\n",
+ sizeof(drm_r300_cmd_header_t));
+ fprintf(stderr, "sizeof(drm_radeon_cmd_buffer_t)=%d\n",
+ sizeof(drm_radeon_cmd_buffer_t));
fprintf(stderr,
"Allocating %d bytes command buffer (max state is %d bytes)\n",
size*4, r300->hw.max_state_size*4);
+ }
r300->cmdbuf.size = size;
r300->cmdbuf.cmd_buf = (uint32_t*)CALLOC(size*4);
assert(h < (1 << 16));
cmd =
- (drm_radeon_cmd_header_t *) r200AllocCmdBuf(rmesa, 8 * sizeof(int),
+ (drm_radeon_cmd_header_t *) r300AllocCmdBuf(rmesa, 8 * sizeof(int),
__FUNCTION__);
cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
assert(!(flags & ~(RADEON_WAIT_2D | RADEON_WAIT_3D)));
cmd =
- (drm_radeon_cmd_header_t *) r200AllocCmdBuf(rmesa,
+ (drm_radeon_cmd_header_t *) r300AllocCmdBuf(rmesa,
1 * sizeof(int),
__FUNCTION__);
cmd[0].i = 0;
- cmd[0].wait.cmd_type = RADEON_CMD_WAIT;
+ cmd[0].wait.cmd_type = R300_CMD_WAIT;
cmd[0].wait.flags = flags;
}
}
+
+void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
+{
+ if (RADEON_DEBUG & DEBUG_VERTS)
+ fprintf(stderr, "%s: nr=%d, ofs=0x%08x\n", __func__, nr, offset);
+ int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2;
+ int i;
+ LOCAL_VARS
+
+ start_packet3(RADEON_CP_PACKET3_3D_LOAD_VBPNTR, sz-1);
+ e32(nr);
+ for(i=0;i+1<nr;i+=2){
+ e32( (rmesa->state.aos[i].aos_size << 0)
+ |(rmesa->state.aos[i].aos_stride << 8)
+ |(rmesa->state.aos[i+1].aos_size << 16)
+ |(rmesa->state.aos[i+1].aos_stride << 24)
+ );
+ e32(rmesa->state.aos[i].aos_offset+offset*4*rmesa->state.aos[i].aos_stride);
+ e32(rmesa->state.aos[i+1].aos_offset+offset*4*rmesa->state.aos[i+1].aos_stride);
+ }
+ if(nr & 1){
+ e32( (rmesa->state.aos[nr-1].aos_size << 0)
+ |(rmesa->state.aos[nr-1].aos_stride << 8)
+ );
+ e32(rmesa->state.aos[nr-1].aos_offset+offset*4*rmesa->state.aos[nr-1].aos_stride);
+ }
+
+}
+