fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-02-09 03:20:26 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

r300: Converted to the new Vector Engine defines.

Browse source
This commit is contained in:
Oliver McFadden 2008-02-25 12:35:55 +00:00
parent 66952fa5ae
commit bd46a482c4
4 changed files with 32 additions and 76 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

45
src/mesa/drivers/dri/r300/r300_reg.h
View file

@ -2499,51 +2499,6 @@ enum {
/*\{*/
/* Every instruction is four dwords long:
* DWORD 0: output and opcode
* DWORD 1: first argument
* DWORD 2: second argument
* DWORD 3: third argument
*
* Notes:
* - ABS r, a is implemented as MAX r, a, -a
* - MOV is implemented as ADD to zero
* - XPD is implemented as MUL + MAD
* - FLR is implemented as FRC + ADD
* - apparently, fglrx tries to schedule instructions so that there is at
* least one instruction between the write to a temporary and the first
* read from said temporary; however, violations of this scheduling are
* allowed
* - register indices seem to be unrelated with OpenGL aliasing to
* conventional state
* - only one attribute and one parameter can be loaded at a time; however,
* the same attribute/parameter can be used for more than one argument
* - the second software argument for POW is the third hardware argument
* (no idea why)
* - MAD with only temporaries as input seems to use VPI_OUT_SELECT_MAD_2
*
* There is some magic surrounding LIT:
* The single argument is replicated across all three inputs, but swizzled:
* First argument: xyzy
* Second argument: xyzx
* Third argument: xyzw
* Whenever the result is used later in the fragment program, fglrx forces
* x and w to be 1.0 in the input selection; I don't know whether this is
* strictly necessary
*/
#define R300_VPI_OUT_OP_DOT (1 << 0)
#define R300_VPI_OUT_OP_MUL (2 << 0)
#define R300_VPI_OUT_OP_ADD (3 << 0)
#define R300_VPI_OUT_OP_MAD (4 << 0)
#define R300_VPI_OUT_OP_DST (5 << 0)
#define R300_VPI_OUT_OP_FRC (6 << 0)
#define R300_VPI_OUT_OP_MAX (7 << 0)
#define R300_VPI_OUT_OP_MIN (8 << 0)
#define R300_VPI_OUT_OP_SGE (9 << 0)
#define R300_VPI_OUT_OP_SLT (10 << 0)
/* Used in GL_POINT_DISTANCE_ATTENUATION_ARB, vector(scalar, vector) */
#define R300_VPI_OUT_OP_UNK12 (12 << 0)
#define R300_VPI_OUT_OP_ARL (13 << 0)
#define R300_VPI_OUT_OP_EXP (65 << 0)
#define R300_VPI_OUT_OP_LOG (66 << 0)
/* Used in fog computations, scalar(scalar) */

2
src/mesa/drivers/dri/r300/r300_state.c
View file

@ -1659,7 +1659,7 @@ static void r300SetupDefaultVertexProgram(r300ContextPtr rmesa)
for (i = VERT_ATTRIB_POS; i < VERT_ATTRIB_MAX; i++) {
if (rmesa->state.sw_tcl_inputs[i] != -1) {
prog->program.body.i[program_end + 0] = MAKE_VSF_OP(R300_VPI_OUT_OP_MUL, o_reg++, VSF_FLAG_ALL, PVS_DST_REG_OUT);
prog->program.body.i[program_end + 0] = MAKE_VSF_OP(VE_MULTIPLY, o_reg++, VSF_FLAG_ALL, PVS_DST_REG_OUT);
prog->program.body.i[program_end + 1] = VSF_REG(rmesa->state.sw_tcl_inputs[i]);
prog->program.body.i[program_end + 2] = VSF_ATTR_UNITY(rmesa->state.sw_tcl_inputs[i]);
prog->program.body.i[program_end + 3] = VSF_UNITY(rmesa->state.sw_tcl_inputs[i]);

59
src/mesa/drivers/dri/r300/r300_vertprog.c
View file

@ -376,7 +376,7 @@ static GLuint *t_opcode_abs(struct r300_vertex_program *vp,
//MAX RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_MAX, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_MAXIMUM, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -406,7 +406,7 @@ static GLuint *t_opcode_add(struct r300_vertex_program *vp,
hw_op = (src[0].File == PROGRAM_TEMPORARY
&& src[1].File ==
PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
R300_VPI_OUT_OP_MAD;
VE_MULTIPLY_ADD;
inst[0] =
MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
@ -417,7 +417,7 @@ static GLuint *t_opcode_add(struct r300_vertex_program *vp,
inst[3] = t_src(vp, &src[1]);
#else
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_ADD, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
inst[1] = t_src(vp, &src[0]);
@ -434,7 +434,7 @@ static GLuint *t_opcode_arl(struct r300_vertex_program *vp,
struct prog_src_register src[3])
{
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_ARL, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_FLT2FIX_DX, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -452,7 +452,7 @@ static GLuint *t_opcode_dp3(struct r300_vertex_program *vp,
//DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ZERO} PARAM 0{} {X Y Z ZERO}
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_DOT, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_DOT_PRODUCT, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -486,7 +486,7 @@ static GLuint *t_opcode_dp4(struct r300_vertex_program *vp,
struct prog_src_register src[3])
{
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_DOT, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_DOT_PRODUCT, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -503,7 +503,7 @@ static GLuint *t_opcode_dph(struct r300_vertex_program *vp,
{
//DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ONE} PARAM 0{} {X Y Z W}
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_DOT, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_DOT_PRODUCT, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -527,7 +527,7 @@ static GLuint *t_opcode_dst(struct r300_vertex_program *vp,
struct prog_src_register src[3])
{
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_DST, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_DISTANCE_VECTOR, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -578,7 +578,7 @@ static GLuint *t_opcode_flr(struct r300_vertex_program *vp,
ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} TMP 0{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W */
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_FRC, *u_temp_i,
MAKE_VSF_OP(VE_FRACTION, *u_temp_i,
t_dst_mask(vpi->DstReg.WriteMask),
PVS_DST_REG_TEMPORARY);
@ -588,7 +588,7 @@ static GLuint *t_opcode_flr(struct r300_vertex_program *vp,
inst += 4;
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_ADD, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -613,7 +613,7 @@ static GLuint *t_opcode_frc(struct r300_vertex_program *vp,
struct prog_src_register src[3])
{
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_FRC, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_FRACTION, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -716,7 +716,7 @@ static GLuint *t_opcode_mad(struct r300_vertex_program *vp,
&& src[1].File == PROGRAM_TEMPORARY
&& src[2].File ==
PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
R300_VPI_OUT_OP_MAD;
VE_MULTIPLY_ADD;
inst[0] =
MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
@ -734,7 +734,7 @@ static GLuint *t_opcode_max(struct r300_vertex_program *vp,
struct prog_src_register src[3])
{
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_MAX, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_MAXIMUM, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -750,7 +750,7 @@ static GLuint *t_opcode_min(struct r300_vertex_program *vp,
struct prog_src_register src[3])
{
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_MIN, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_MINIMUM, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -769,7 +769,7 @@ static GLuint *t_opcode_mov(struct r300_vertex_program *vp,
#if 1
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_ADD, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
inst[1] = t_src(vp, &src[0]);
@ -779,7 +779,7 @@ static GLuint *t_opcode_mov(struct r300_vertex_program *vp,
hw_op =
(src[0].File ==
PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
R300_VPI_OUT_OP_MAD;
VE_MULTIPLY_ADD;
inst[0] =
MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
@ -804,7 +804,7 @@ static GLuint *t_opcode_mul(struct r300_vertex_program *vp,
hw_op = (src[0].File == PROGRAM_TEMPORARY
&& src[1].File ==
PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
R300_VPI_OUT_OP_MAD;
VE_MULTIPLY_ADD;
inst[0] =
MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
@ -870,7 +870,7 @@ static GLuint *t_opcode_sge(struct r300_vertex_program *vp,
struct prog_src_register src[3])
{
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_SGE, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_SET_GREATER_THAN_EQUAL, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -886,7 +886,7 @@ static GLuint *t_opcode_slt(struct r300_vertex_program *vp,
struct prog_src_register src[3])
{
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_SLT, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_SET_LESS_THAN, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -909,7 +909,7 @@ static GLuint *t_opcode_sub(struct r300_vertex_program *vp,
hw_op = (src[0].File == PROGRAM_TEMPORARY
&& src[1].File ==
PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
R300_VPI_OUT_OP_MAD;
VE_MULTIPLY_ADD;
inst[0] =
MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
@ -929,7 +929,7 @@ static GLuint *t_opcode_sub(struct r300_vertex_program *vp,
(src[1].RelAddr << 4);
#else
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_ADD, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -958,7 +958,7 @@ static GLuint *t_opcode_swz(struct r300_vertex_program *vp,
#if 1
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_ADD, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
inst[1] = t_src(vp, &src[0]);
@ -968,7 +968,7 @@ static GLuint *t_opcode_swz(struct r300_vertex_program *vp,
hw_op =
(src[0].File ==
PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
R300_VPI_OUT_OP_MAD;
VE_MULTIPLY_ADD;
inst[0] =
MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
@ -992,7 +992,7 @@ static GLuint *t_opcode_xpd(struct r300_vertex_program *vp,
*/
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_MAD, *u_temp_i,
MAKE_VSF_OP(VE_MULTIPLY_ADD, *u_temp_i,
t_dst_mask(vpi->DstReg.WriteMask),
PVS_DST_REG_TEMPORARY);
@ -1019,7 +1019,7 @@ static GLuint *t_opcode_xpd(struct r300_vertex_program *vp,
(*u_temp_i)--;
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_MAD, t_dst_index(vp, &vpi->DstReg),
MAKE_VSF_OP(VE_MULTIPLY_ADD, t_dst_index(vp, &vpi->DstReg),
t_dst_mask(vpi->DstReg.WriteMask),
t_dst_class(vpi->DstReg.File));
@ -1147,7 +1147,7 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
if (CMP_SRCS(src[1], src[2])
|| CMP_SRCS(src[0], src[2])) {
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_ADD,
MAKE_VSF_OP(VE_ADD,
u_temp_i, VSF_FLAG_ALL,
PVS_DST_REG_TEMPORARY);
@ -1175,7 +1175,7 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
if (num_operands >= 2) {
if (CMP_SRCS(src[1], src[0])) {
inst[0] =
MAKE_VSF_OP(R300_VPI_OUT_OP_ADD,
MAKE_VSF_OP(VE_ADD,
u_temp_i, VSF_FLAG_ALL,
PVS_DST_REG_TEMPORARY);
@ -1296,8 +1296,9 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
of the fragment program. Blank the outputs here. */
for (i = 0; i < VERT_RESULT_MAX; i++) {
if (vp->key.OutputsAdded & (1 << i)) {
inst[0] = MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, vp->outputs[i],
VSF_FLAG_ALL, VSF_OUT_CLASS_RESULT);
inst[0] = MAKE_VSF_OP(VE_ADD, vp->outputs[i],
VSF_FLAG_ALL,
VSF_OUT_CLASS_RESULT);
inst[1] = ZERO_SRC_0;
inst[2] = ZERO_SRC_0;
inst[3] = ZERO_SRC_0;

2
src/mesa/drivers/dri/r300/r300_vertprog.h
View file

@ -102,7 +102,7 @@
#define VP_OUTMASK_ZW (VP_OUTMASK_Z|VP_OUTMASK_W)
#define VP_OUT(instr,outclass,outidx,outmask) \
(R300_VPI_OUT_OP_##instr | \
(VE_##instr | \
((outidx) << R300_VPI_OUT_REG_INDEX_SHIFT) | \
VP_OUTCLASS_##outclass | \
VP_OUTMASK_##outmask)