diff options
Diffstat (limited to 'src/mesa/drivers/dri/r300/r300_vertprog.c')
| -rw-r--r-- | src/mesa/drivers/dri/r300/r300_vertprog.c | 1684 |
1 files changed, 930 insertions, 754 deletions
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c index 1d90ade2ed1..c4e325e6a76 100644 --- a/src/mesa/drivers/dri/r300/r300_vertprog.c +++ b/src/mesa/drivers/dri/r300/r300_vertprog.c @@ -1,6 +1,7 @@ /************************************************************************** -Copyright (C) 2005 Aapo Tahkola. +Copyright (C) 2005 Aapo Tahkola <[email protected]> +Copyright (C) 2008 Oliver McFadden <[email protected]> All Rights Reserved. @@ -25,16 +26,12 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/** - * \file - * - * \author Aapo Tahkola <[email protected]> - */ +/* Radeon R5xx Acceleration, Revision 1.2 */ -#include "glheader.h" -#include "macros.h" -#include "enums.h" -#include "program.h" +#include "main/glheader.h" +#include "main/macros.h" +#include "main/enums.h" +#include "shader/program.h" #include "shader/prog_instruction.h" #include "shader/prog_parameter.h" #include "shader/prog_statevars.h" @@ -42,64 +39,35 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" -#if SWIZZLE_X != VSF_IN_COMPONENT_X || \ - SWIZZLE_Y != VSF_IN_COMPONENT_Y || \ - SWIZZLE_Z != VSF_IN_COMPONENT_Z || \ - SWIZZLE_W != VSF_IN_COMPONENT_W || \ - SWIZZLE_ZERO != VSF_IN_COMPONENT_ZERO || \ - SWIZZLE_ONE != VSF_IN_COMPONENT_ONE || \ - WRITEMASK_X != VSF_FLAG_X || \ - WRITEMASK_Y != VSF_FLAG_Y || \ - WRITEMASK_Z != VSF_FLAG_Z || \ - WRITEMASK_W != VSF_FLAG_W -#error Cannot change these! -#endif +/* TODO: Get rid of t_src_class call */ +#define CMP_SRCS(a, b) ((a.RelAddr != b.RelAddr) || (a.Index != b.Index && \ + ((t_src_class(a.File) == PVS_SRC_REG_CONSTANT && \ + t_src_class(b.File) == PVS_SRC_REG_CONSTANT) || \ + (t_src_class(a.File) == PVS_SRC_REG_INPUT && \ + t_src_class(b.File) == PVS_SRC_REG_INPUT)))) \ + +/* + * Take an already-setup and valid source then swizzle it appropriately to + * obtain a constant ZERO or ONE source. + */ +#define __CONST(x, y) \ + (PVS_SRC_OPERAND(t_src_index(vp, &src[x]), \ + t_swizzle(y), \ + t_swizzle(y), \ + t_swizzle(y), \ + t_swizzle(y), \ + t_src_class(src[x].File), \ + VSF_FLAG_NONE) | (src[x].RelAddr << 4)) -#define SCALAR_FLAG (1<<31) -#define FLAG_MASK (1<<31) -#define OP_MASK (0xf) /* we are unlikely to have more than 15 */ -#define OPN(operator, ip) {#operator, OPCODE_##operator, ip} - -static struct { - char *name; - int opcode; - unsigned long ip; /* number of input operands and flags */ -} op_names[] = { - /* *INDENT-OFF* */ - OPN(ABS, 1), - OPN(ADD, 2), - OPN(ARL, 1 | SCALAR_FLAG), - OPN(DP3, 2), - OPN(DP4, 2), - OPN(DPH, 2), - OPN(DST, 2), - OPN(EX2, 1 | SCALAR_FLAG), - OPN(EXP, 1 | SCALAR_FLAG), - OPN(FLR, 1), - OPN(FRC, 1), - OPN(LG2, 1 | SCALAR_FLAG), - OPN(LIT, 1), - OPN(LOG, 1 | SCALAR_FLAG), - OPN(MAD, 3), - OPN(MAX, 2), - OPN(MIN, 2), - OPN(MOV, 1), - OPN(MUL, 2), - OPN(POW, 2 | SCALAR_FLAG), - OPN(RCP, 1 | SCALAR_FLAG), - OPN(RSQ, 1 | SCALAR_FLAG), - OPN(SGE, 2), - OPN(SLT, 2), - OPN(SUB, 2), - OPN(SWZ, 1), - OPN(XPD, 2), - OPN(RCC, 0), //extra - OPN(PRINT, 0), - OPN(END, 0) - /* *INDENT-ON* */ -}; - -#undef OPN +#define FREE_TEMPS() \ + do { \ + int u_temp_used = (VSF_MAX_FRAGMENT_TEMPS - 1) - u_temp_i; \ + if((vp->num_temporaries + u_temp_used) > VSF_MAX_FRAGMENT_TEMPS) { \ + WARN_ONCE("Ran out of temps, num temps %d, us %d\n", vp->num_temporaries, u_temp_used); \ + vp->native = GL_FALSE; \ + } \ + u_temp_i=VSF_MAX_FRAGMENT_TEMPS-1; \ + } while (0) int r300VertexProgUpdateParams(GLcontext * ctx, struct r300_vertex_program_cont *vp, float *dst) @@ -133,7 +101,6 @@ int r300VertexProgUpdateParams(GLcontext * ctx, paramList = mesa_vp->Base.Parameters; for (pi = 0; pi < paramList->NumParameters; pi++) { switch (paramList->Parameters[pi].Type) { - case PROGRAM_STATE_VAR: case PROGRAM_NAMED_PARAM: //fprintf(stderr, "%s", vp->Parameters->Parameters[pi].Name); @@ -143,7 +110,6 @@ int r300VertexProgUpdateParams(GLcontext * ctx, *dst++ = paramList->ParameterValues[pi][2]; *dst++ = paramList->ParameterValues[pi][3]; break; - default: _mesa_problem(NULL, "Bad param type in %s", __FUNCTION__); @@ -165,11 +131,11 @@ static unsigned long t_dst_class(enum register_file file) switch (file) { case PROGRAM_TEMPORARY: - return VSF_OUT_CLASS_TMP; + return PVS_DST_REG_TEMPORARY; case PROGRAM_OUTPUT: - return VSF_OUT_CLASS_RESULT; + return PVS_DST_REG_OUT; case PROGRAM_ADDRESS: - return VSF_OUT_CLASS_ADDR; + return PVS_DST_REG_A0; /* case PROGRAM_INPUT: case PROGRAM_LOCAL_PARAM: @@ -197,19 +163,17 @@ static unsigned long t_dst_index(struct r300_vertex_program *vp, static unsigned long t_src_class(enum register_file file) { - switch (file) { case PROGRAM_TEMPORARY: - return VSF_IN_CLASS_TMP; - + return PVS_SRC_REG_TEMPORARY; case PROGRAM_INPUT: - return VSF_IN_CLASS_ATTR; - + return PVS_SRC_REG_INPUT; case PROGRAM_LOCAL_PARAM: case PROGRAM_ENV_PARAM: case PROGRAM_NAMED_PARAM: + case PROGRAM_CONSTANT: case PROGRAM_STATE_VAR: - return VSF_IN_CLASS_PARAM; + return PVS_SRC_REG_CONSTANT; /* case PROGRAM_OUTPUT: case PROGRAM_WRITE_ONLY: @@ -222,7 +186,7 @@ static unsigned long t_src_class(enum register_file file) } } -static __inline unsigned long t_swizzle(GLubyte swizzle) +static INLINE unsigned long t_swizzle(GLubyte swizzle) { /* this is in fact a NOP as the Mesa SWIZZLE_* are all identical to VSF_IN_COMPONENT_* */ return swizzle; @@ -276,13 +240,15 @@ static unsigned long t_src_index(struct r300_vertex_program *vp, } } +/* these two functions should probably be merged... */ + static unsigned long t_src(struct r300_vertex_program *vp, struct prog_src_register *src) { /* src->NegateBase uses the NEGATE_ flags from program_instruction.h, * which equal our VSF_FLAGS_ values, so it's safe to just pass it here. */ - return MAKE_VSF_SOURCE(t_src_index(vp, src), + return PVS_SRC_OPERAND(t_src_index(vp, src), t_swizzle(GET_SWZ(src->Swizzle, 0)), t_swizzle(GET_SWZ(src->Swizzle, 1)), t_swizzle(GET_SWZ(src->Swizzle, 2)), @@ -294,8 +260,10 @@ static unsigned long t_src(struct r300_vertex_program *vp, static unsigned long t_src_scalar(struct r300_vertex_program *vp, struct prog_src_register *src) { - - return MAKE_VSF_SOURCE(t_src_index(vp, src), + /* src->NegateBase uses the NEGATE_ flags from program_instruction.h, + * which equal our VSF_FLAGS_ values, so it's safe to just pass it here. + */ + return PVS_SRC_OPERAND(t_src_index(vp, src), t_swizzle(GET_SWZ(src->Swizzle, 0)), t_swizzle(GET_SWZ(src->Swizzle, 0)), t_swizzle(GET_SWZ(src->Swizzle, 0)), @@ -306,128 +274,687 @@ static unsigned long t_src_scalar(struct r300_vertex_program *vp, (src->RelAddr << 4); } -static unsigned long t_opcode(enum prog_opcode opcode) +static GLboolean valid_dst(struct r300_vertex_program *vp, + struct prog_dst_register *dst) { + if (dst->File == PROGRAM_OUTPUT && vp->outputs[dst->Index] == -1) { + return GL_FALSE; + } else if (dst->File == PROGRAM_ADDRESS) { + assert(dst->Index == 0); + } - switch (opcode) { - /* *INDENT-OFF* */ - case OPCODE_ARL: return R300_VPI_OUT_OP_ARL; - case OPCODE_DST: return R300_VPI_OUT_OP_DST; - case OPCODE_EX2: return R300_VPI_OUT_OP_EX2; - case OPCODE_EXP: return R300_VPI_OUT_OP_EXP; - case OPCODE_FRC: return R300_VPI_OUT_OP_FRC; - case OPCODE_LG2: return R300_VPI_OUT_OP_LG2; - case OPCODE_LOG: return R300_VPI_OUT_OP_LOG; - case OPCODE_MAX: return R300_VPI_OUT_OP_MAX; - case OPCODE_MIN: return R300_VPI_OUT_OP_MIN; - case OPCODE_MUL: return R300_VPI_OUT_OP_MUL; - case OPCODE_RCP: return R300_VPI_OUT_OP_RCP; - case OPCODE_RSQ: return R300_VPI_OUT_OP_RSQ; - case OPCODE_SGE: return R300_VPI_OUT_OP_SGE; - case OPCODE_SLT: return R300_VPI_OUT_OP_SLT; - case OPCODE_DP4: return R300_VPI_OUT_OP_DOT; - /* *INDENT-ON* */ + return GL_TRUE; +} - default: - fprintf(stderr, "%s: Should not be called with opcode %d!", - __FUNCTION__, opcode); - } - _mesa_exit(-1); - return 0; +static GLuint *r300TranslateOpcodeABS(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + //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] = PVS_OP_DST_OPERAND(VE_MAXIMUM, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), + t_swizzle(GET_SWZ(src[0].Swizzle, 1)), + t_swizzle(GET_SWZ(src[0].Swizzle, 2)), + t_swizzle(GET_SWZ(src[0].Swizzle, 3)), + t_src_class(src[0].File), + (!src[0]. + NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[0].RelAddr << 4); + inst[3] = 0; + + return inst; } -static unsigned long op_operands(enum prog_opcode opcode) +static GLuint *r300TranslateOpcodeADD(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) { - int i; + inst[0] = PVS_OP_DST_OPERAND(VE_ADD, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = t_src(vp, &src[1]); + inst[3] = __CONST(1, SWIZZLE_ZERO); + + return inst; +} - /* Can we trust mesas opcodes to be in order ? */ - for (i = 0; i < sizeof(op_names) / sizeof(*op_names); i++) - if (op_names[i].opcode == opcode) - return op_names[i].ip; +static GLuint *r300TranslateOpcodeARL(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(VE_FLT2FIX_DX, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + + return inst; +} - fprintf(stderr, "op %d not found in op_names\n", opcode); - _mesa_exit(-1); - return 0; +static GLuint *r300TranslateOpcodeDP3(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + //DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ZERO} PARAM 0{} {X Y Z ZERO} + + inst[0] = PVS_OP_DST_OPERAND(VE_DOT_PRODUCT, + GL_FALSE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), + t_swizzle(GET_SWZ(src[0].Swizzle, 1)), + t_swizzle(GET_SWZ(src[0].Swizzle, 2)), + SWIZZLE_ZERO, + t_src_class(src[0].File), + src[0]. + NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) | + (src[0].RelAddr << 4); + inst[2] = + PVS_SRC_OPERAND(t_src_index(vp, &src[1]), + t_swizzle(GET_SWZ(src[1].Swizzle, 0)), + t_swizzle(GET_SWZ(src[1].Swizzle, 1)), + t_swizzle(GET_SWZ(src[1].Swizzle, 2)), SWIZZLE_ZERO, + t_src_class(src[1].File), + src[1]. + NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) | + (src[1].RelAddr << 4); + inst[3] = __CONST(1, SWIZZLE_ZERO); + + return inst; } -static GLboolean valid_dst(struct r300_vertex_program *vp, - struct prog_dst_register *dst) +static GLuint *r300TranslateOpcodeDP4(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) { - if (dst->File == PROGRAM_OUTPUT && vp->outputs[dst->Index] == -1) { - return GL_FALSE; - } else if (dst->File == PROGRAM_ADDRESS) { - assert(dst->Index == 0); - } + inst[0] = PVS_OP_DST_OPERAND(VE_DOT_PRODUCT, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = t_src(vp, &src[1]); + inst[3] = __CONST(1, SWIZZLE_ZERO); + + return inst; +} - return GL_TRUE; +static GLuint *r300TranslateOpcodeDPH(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + //DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ONE} PARAM 0{} {X Y Z W} + inst[0] = PVS_OP_DST_OPERAND(VE_DOT_PRODUCT, + GL_FALSE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), + t_swizzle(GET_SWZ(src[0].Swizzle, 1)), + t_swizzle(GET_SWZ(src[0].Swizzle, 2)), + PVS_SRC_SELECT_FORCE_1, + t_src_class(src[0].File), + src[0]. + NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) | + (src[0].RelAddr << 4); + inst[2] = t_src(vp, &src[1]); + inst[3] = __CONST(1, SWIZZLE_ZERO); + + return inst; } -/* TODO: Get rid of t_src_class call */ -#define CMP_SRCS(a, b) ((a.RelAddr != b.RelAddr) || (a.Index != b.Index && \ - ((t_src_class(a.File) == VSF_IN_CLASS_PARAM && \ - t_src_class(b.File) == VSF_IN_CLASS_PARAM) || \ - (t_src_class(a.File) == VSF_IN_CLASS_ATTR && \ - t_src_class(b.File) == VSF_IN_CLASS_ATTR)))) \ - -#define ZERO_SRC_0 (MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), \ - SWIZZLE_ZERO, SWIZZLE_ZERO, \ - SWIZZLE_ZERO, SWIZZLE_ZERO, \ - t_src_class(src[0].File), VSF_FLAG_NONE) | (src[0].RelAddr << 4)) - -#define ZERO_SRC_1 (MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), \ - SWIZZLE_ZERO, SWIZZLE_ZERO, \ - SWIZZLE_ZERO, SWIZZLE_ZERO, \ - t_src_class(src[1].File), VSF_FLAG_NONE) | (src[1].RelAddr << 4)) - -#define ZERO_SRC_2 (MAKE_VSF_SOURCE(t_src_index(vp, &src[2]), \ - SWIZZLE_ZERO, SWIZZLE_ZERO, \ - SWIZZLE_ZERO, SWIZZLE_ZERO, \ - t_src_class(src[2].File), VSF_FLAG_NONE) | (src[2].RelAddr << 4)) - -#define ONE_SRC_0 (MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), \ - SWIZZLE_ONE, SWIZZLE_ONE, \ - SWIZZLE_ONE, SWIZZLE_ONE, \ - t_src_class(src[0].File), VSF_FLAG_NONE) | (src[0].RelAddr << 4)) - -#define ONE_SRC_1 (MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), \ - SWIZZLE_ONE, SWIZZLE_ONE, \ - SWIZZLE_ONE, SWIZZLE_ONE, \ - t_src_class(src[1].File), VSF_FLAG_NONE) | (src[1].RelAddr << 4)) - -#define ONE_SRC_2 (MAKE_VSF_SOURCE(t_src_index(vp, &src[2]), \ - SWIZZLE_ONE, SWIZZLE_ONE, \ - SWIZZLE_ONE, SWIZZLE_ONE, \ - t_src_class(src[2].File), VSF_FLAG_NONE) | (src[2].RelAddr << 4)) +static GLuint *r300TranslateOpcodeDST(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(VE_DISTANCE_VECTOR, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = t_src(vp, &src[1]); + inst[3] = __CONST(1, SWIZZLE_ZERO); + + return inst; +} -/* DP4 version seems to trigger some hw peculiarity */ -//#define PREFER_DP4 +static GLuint *r300TranslateOpcodeEX2(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(ME_EXP_BASE2_FULL_DX, + GL_TRUE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + inst[1] = t_src_scalar(vp, &src[0]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + + return inst; +} -#define FREE_TEMPS() \ - do { \ - if(u_temp_i < vp->num_temporaries) { \ - WARN_ONCE("Ran out of temps, num temps %d, us %d\n", vp->num_temporaries, u_temp_i); \ - vp->native = GL_FALSE; \ - } \ - u_temp_i=VSF_MAX_FRAGMENT_TEMPS-1; \ - } while (0) +static GLuint *r300TranslateOpcodeEXP(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(ME_EXP_BASE2_DX, + GL_TRUE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + inst[1] = t_src_scalar(vp, &src[0]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + + return inst; +} -static void r300TranslateVertexShader(struct r300_vertex_program *vp, - struct prog_instruction *vpi) +static GLuint *r300TranslateOpcodeFLR(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3], + int *u_temp_i) { - int i, cur_reg = 0; - VERTEX_SHADER_INSTRUCTION *o_inst; - unsigned long operands; - int are_srcs_scalar; - unsigned long hw_op; - /* Initial value should be last tmp reg that hw supports. - Strangely enough r300 doesnt mind even though these would be out of range. - Smart enough to realize that it doesnt need it? */ - int u_temp_i = VSF_MAX_FRAGMENT_TEMPS - 1; - struct prog_src_register src[3]; + /* FRC TMP 0.X Y Z W PARAM 0{} {X Y Z W} + 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] = PVS_OP_DST_OPERAND(VE_FRACTION, + GL_FALSE, + GL_FALSE, + *u_temp_i, + t_dst_mask(vpi->DstReg.WriteMask), + PVS_DST_REG_TEMPORARY); + inst[1] = t_src(vp, &src[0]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + inst += 4; + + inst[0] = PVS_OP_DST_OPERAND(VE_ADD, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = PVS_SRC_OPERAND(*u_temp_i, + PVS_SRC_SELECT_X, + PVS_SRC_SELECT_Y, + PVS_SRC_SELECT_Z, + PVS_SRC_SELECT_W, PVS_SRC_REG_TEMPORARY, + /* Not 100% sure about this */ + (!src[0]. + NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE + /*VSF_FLAG_ALL */ ); + inst[3] = __CONST(0, SWIZZLE_ZERO); + (*u_temp_i)--; + + return inst; +} - vp->pos_end = 0; /* Not supported yet */ - vp->program.length = 0; - /*vp->num_temporaries=mesa_vp->Base.NumTemporaries; */ +static GLuint *r300TranslateOpcodeFRC(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(VE_FRACTION, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeLG2(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + // LG2 RESULT 1.X Y Z W PARAM 0{} {X X X X} + + inst[0] = PVS_OP_DST_OPERAND(ME_LOG_BASE2_FULL_DX, + GL_TRUE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), + t_src_class(src[0].File), + src[0]. + NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[0].RelAddr << 4); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeLIT(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + //LIT TMP 1.Y Z TMP 1{} {X W Z Y} TMP 1{} {Y W Z X} TMP 1{} {Y X Z W} + + inst[0] = PVS_OP_DST_OPERAND(ME_LIGHT_COEFF_DX, + GL_TRUE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + /* NOTE: Users swizzling might not work. */ + inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // X + t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // W + PVS_SRC_SELECT_FORCE_0, // Z + t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // Y + t_src_class(src[0].File), + src[0]. + NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[0].RelAddr << 4); + inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // Y + t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // W + PVS_SRC_SELECT_FORCE_0, // Z + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // X + t_src_class(src[0].File), + src[0]. + NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[0].RelAddr << 4); + inst[3] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // Y + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // X + PVS_SRC_SELECT_FORCE_0, // Z + t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // W + t_src_class(src[0].File), + src[0]. + NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[0].RelAddr << 4); + + return inst; +} + +static GLuint *r300TranslateOpcodeLOG(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(ME_LOG_BASE2_DX, + GL_TRUE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + inst[1] = t_src_scalar(vp, &src[0]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeMAD(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(PVS_MACRO_OP_2CLK_MADD, + GL_FALSE, + GL_TRUE, + 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]); + inst[2] = t_src(vp, &src[1]); + inst[3] = t_src(vp, &src[2]); + + return inst; +} + +static GLuint *r300TranslateOpcodeMAX(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(VE_MAXIMUM, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = t_src(vp, &src[1]); + inst[3] = __CONST(1, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeMIN(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(VE_MINIMUM, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = t_src(vp, &src[1]); + inst[3] = __CONST(1, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeMOV(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + //ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{} {ZERO ZERO ZERO ZERO} + + inst[0] = PVS_OP_DST_OPERAND(VE_ADD, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeMUL(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(VE_MULTIPLY, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = t_src(vp, &src[1]); + inst[3] = __CONST(1, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodePOW(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(ME_POWER_FUNC_FF, + GL_TRUE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + inst[1] = t_src_scalar(vp, &src[0]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = t_src_scalar(vp, &src[1]); + + return inst; +} + +static GLuint *r300TranslateOpcodeRCP(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(ME_RECIP_DX, + GL_TRUE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + inst[1] = t_src_scalar(vp, &src[0]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeRSQ(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(ME_RECIP_SQRT_DX, + GL_TRUE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + inst[1] = t_src_scalar(vp, &src[0]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeSGE(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(VE_SET_GREATER_THAN_EQUAL, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = t_src(vp, &src[1]); + inst[3] = __CONST(1, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeSLT(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + inst[0] = PVS_OP_DST_OPERAND(VE_SET_LESS_THAN, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = t_src(vp, &src[1]); + inst[3] = __CONST(1, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeSUB(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + //ADD RESULT 1.X Y Z W TMP 0{} {X Y Z W} PARAM 1{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W + +#if 0 + inst[0] = PVS_OP_DST_OPERAND(VE_ADD, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]), + t_swizzle(GET_SWZ(src[1].Swizzle, 0)), + t_swizzle(GET_SWZ(src[1].Swizzle, 1)), + t_swizzle(GET_SWZ(src[1].Swizzle, 2)), + t_swizzle(GET_SWZ(src[1].Swizzle, 3)), + t_src_class(src[1].File), + (!src[1]. + NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[1].RelAddr << 4); + inst[3] = 0; +#else + inst[0] = + PVS_OP_DST_OPERAND(VE_MULTIPLY_ADD, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = __CONST(0, SWIZZLE_ONE); + inst[3] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]), + t_swizzle(GET_SWZ(src[1].Swizzle, 0)), + t_swizzle(GET_SWZ(src[1].Swizzle, 1)), + t_swizzle(GET_SWZ(src[1].Swizzle, 2)), + t_swizzle(GET_SWZ(src[1].Swizzle, 3)), + t_src_class(src[1].File), + (!src[1]. + NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[1].RelAddr << 4); +#endif + + return inst; +} + +static GLuint *r300TranslateOpcodeSWZ(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3]) +{ + //ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{} {ZERO ZERO ZERO ZERO} + + inst[0] = PVS_OP_DST_OPERAND(VE_ADD, + GL_FALSE, + GL_FALSE, + 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]); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + + return inst; +} + +static GLuint *r300TranslateOpcodeXPD(struct r300_vertex_program *vp, + struct prog_instruction *vpi, + GLuint * inst, + struct prog_src_register src[3], + int *u_temp_i) +{ + /* mul r0, r1.yzxw, r2.zxyw + mad r0, -r2.yzxw, r1.zxyw, r0 + */ + + inst[0] = PVS_OP_DST_OPERAND(VE_MULTIPLY_ADD, + GL_FALSE, + GL_FALSE, + *u_temp_i, + t_dst_mask(vpi->DstReg.WriteMask), + PVS_DST_REG_TEMPORARY); + inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // Y + t_swizzle(GET_SWZ(src[0].Swizzle, 2)), // Z + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // X + t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // W + t_src_class(src[0].File), + src[0]. + NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[0].RelAddr << 4); + inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 2)), // Z + t_swizzle(GET_SWZ(src[1].Swizzle, 0)), // X + t_swizzle(GET_SWZ(src[1].Swizzle, 1)), // Y + t_swizzle(GET_SWZ(src[1].Swizzle, 3)), // W + t_src_class(src[1].File), + src[1]. + NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[1].RelAddr << 4); + inst[3] = __CONST(1, SWIZZLE_ZERO); + inst += 4; + + inst[0] = PVS_OP_DST_OPERAND(VE_MULTIPLY_ADD, + GL_FALSE, + GL_FALSE, + t_dst_index(vp, &vpi->DstReg), + t_dst_mask(vpi->DstReg.WriteMask), + t_dst_class(vpi->DstReg.File)); + inst[1] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 1)), // Y + t_swizzle(GET_SWZ(src[1].Swizzle, 2)), // Z + t_swizzle(GET_SWZ(src[1].Swizzle, 0)), // X + t_swizzle(GET_SWZ(src[1].Swizzle, 3)), // W + t_src_class(src[1].File), + (!src[1]. + NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[1].RelAddr << 4); + inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 2)), // Z + t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // X + t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // Y + t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // W + t_src_class(src[0].File), + src[0]. + NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) | + (src[0].RelAddr << 4); + inst[3] = + PVS_SRC_OPERAND(*u_temp_i, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, + PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, + PVS_SRC_REG_TEMPORARY, VSF_FLAG_NONE); + + (*u_temp_i)--; + + return inst; +} + +static void t_inputs_outputs(struct r300_vertex_program *vp) +{ + int i; + int cur_reg = 0; for (i = 0; i < VERT_ATTRIB_MAX; i++) vp->inputs[i] = -1; @@ -437,39 +964,71 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp, assert(vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS)); - /* Assign outputs */ - if (vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS)) + if (vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS)) { vp->outputs[VERT_RESULT_HPOS] = cur_reg++; + } - if (vp->key.OutputsWritten & (1 << VERT_RESULT_PSIZ)) + if (vp->key.OutputsWritten & (1 << VERT_RESULT_PSIZ)) { vp->outputs[VERT_RESULT_PSIZ] = cur_reg++; + } - if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL0)) + if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL0)) { vp->outputs[VERT_RESULT_COL0] = cur_reg++; + } - if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL1)) - vp->outputs[VERT_RESULT_COL1] = cur_reg++; - -#if 0 /* Not supported yet */ - if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0)) - vp->outputs[VERT_RESULT_BFC0] = cur_reg++; + if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL1)) { + vp->outputs[VERT_RESULT_COL1] = + vp->outputs[VERT_RESULT_COL0] + 1; + cur_reg = vp->outputs[VERT_RESULT_COL1] + 1; + } - if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) - vp->outputs[VERT_RESULT_BFC1] = cur_reg++; + if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0)) { + vp->outputs[VERT_RESULT_BFC0] = + vp->outputs[VERT_RESULT_COL0] + 2; + cur_reg = vp->outputs[VERT_RESULT_BFC0] + 2; + } - if (vp->key.OutputsWritten & (1 << VERT_RESULT_FOGC)) + if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) { + vp->outputs[VERT_RESULT_BFC1] = + vp->outputs[VERT_RESULT_COL0] + 3; + cur_reg = vp->outputs[VERT_RESULT_BFC1] + 1; + } +#if 0 + if (vp->key.OutputsWritten & (1 << VERT_RESULT_FOGC)) { vp->outputs[VERT_RESULT_FOGC] = cur_reg++; + } #endif - for (i = VERT_RESULT_TEX0; i <= VERT_RESULT_TEX7; i++) - if (vp->key.OutputsWritten & (1 << i)) + for (i = VERT_RESULT_TEX0; i <= VERT_RESULT_TEX7; i++) { + if (vp->key.OutputsWritten & (1 << i)) { vp->outputs[i] = cur_reg++; + } + } +} + +static void r300TranslateVertexShader(struct r300_vertex_program *vp, + struct prog_instruction *vpi) +{ + int i; + GLuint *inst; + unsigned long num_operands; + /* Initial value should be last tmp reg that hw supports. + Strangely enough r300 doesnt mind even though these would be out of range. + Smart enough to realize that it doesnt need it? */ + int u_temp_i = VSF_MAX_FRAGMENT_TEMPS - 1; + struct prog_src_register src[3]; + vp->pos_end = 0; /* Not supported yet */ + vp->program.length = 0; + /*vp->num_temporaries=mesa_vp->Base.NumTemporaries; */ vp->translated = GL_TRUE; vp->native = GL_TRUE; - o_inst = vp->program.body.i; - for (; vpi->Opcode != OPCODE_END; vpi++, o_inst++) { + t_inputs_outputs(vp); + + for (inst = vp->program.body.i; vpi->Opcode != OPCODE_END; + vpi++, inst += 4) { + FREE_TEMPS(); if (!valid_dst(vp, &vpi->DstReg)) { @@ -478,61 +1037,64 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp, vpi->DstReg.Index = u_temp_i; } - operands = op_operands(vpi->Opcode); - are_srcs_scalar = operands & SCALAR_FLAG; - operands &= OP_MASK; + num_operands = _mesa_num_inst_src_regs(vpi->Opcode); - for (i = 0; i < operands; i++) + /* copy the sources (src) from mesa into a local variable... is this needed? */ + for (i = 0; i < num_operands; i++) { src[i] = vpi->SrcReg[i]; + } - if (operands == 3) { /* TODO: scalars */ + if (num_operands == 3) { /* TODO: scalars */ if (CMP_SRCS(src[1], src[2]) || CMP_SRCS(src[0], src[2])) { - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, u_temp_i, - VSF_FLAG_ALL, - VSF_OUT_CLASS_TMP); - - o_inst->src[0] = - MAKE_VSF_SOURCE(t_src_index(vp, &src[2]), - SWIZZLE_X, SWIZZLE_Y, - SWIZZLE_Z, SWIZZLE_W, + inst[0] = PVS_OP_DST_OPERAND(VE_ADD, + GL_FALSE, + GL_FALSE, + u_temp_i, + VSF_FLAG_ALL, + PVS_DST_REG_TEMPORARY); + inst[1] = + PVS_SRC_OPERAND(t_src_index(vp, &src[2]), + SWIZZLE_X, + SWIZZLE_Y, + SWIZZLE_Z, + SWIZZLE_W, t_src_class(src[2].File), VSF_FLAG_NONE) | (src[2]. RelAddr << 4); - - o_inst->src[1] = ZERO_SRC_2; - o_inst->src[2] = ZERO_SRC_2; - o_inst++; + inst[2] = __CONST(2, SWIZZLE_ZERO); + inst[3] = __CONST(2, SWIZZLE_ZERO); + inst += 4; src[2].File = PROGRAM_TEMPORARY; src[2].Index = u_temp_i; src[2].RelAddr = 0; u_temp_i--; } - } - if (operands >= 2) { + if (num_operands >= 2) { if (CMP_SRCS(src[1], src[0])) { - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, u_temp_i, - VSF_FLAG_ALL, - VSF_OUT_CLASS_TMP); - - o_inst->src[0] = - MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), - SWIZZLE_X, SWIZZLE_Y, - SWIZZLE_Z, SWIZZLE_W, + inst[0] = PVS_OP_DST_OPERAND(VE_ADD, + GL_FALSE, + GL_FALSE, + u_temp_i, + VSF_FLAG_ALL, + PVS_DST_REG_TEMPORARY); + inst[1] = + PVS_SRC_OPERAND(t_src_index(vp, &src[0]), + SWIZZLE_X, + SWIZZLE_Y, + SWIZZLE_Z, + SWIZZLE_W, t_src_class(src[0].File), VSF_FLAG_NONE) | (src[0]. RelAddr << 4); - - o_inst->src[1] = ZERO_SRC_0; - o_inst->src[2] = ZERO_SRC_0; - o_inst++; + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + inst += 4; src[0].File = PROGRAM_TEMPORARY; src[0].Index = u_temp_i; @@ -541,517 +1103,118 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp, } } - /* These ops need special handling. */ switch (vpi->Opcode) { - case OPCODE_POW: - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_POW, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - o_inst->src[0] = t_src_scalar(vp, &src[0]); - o_inst->src[1] = ZERO_SRC_0; - o_inst->src[2] = t_src_scalar(vp, &src[1]); - goto next; - - case OPCODE_MOV: //ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{} {ZERO ZERO ZERO ZERO} - case OPCODE_SWZ: -#if 1 - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = ZERO_SRC_0; - o_inst->src[2] = ZERO_SRC_0; -#else - hw_op = - (src[0].File == - PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 : - R300_VPI_OUT_OP_MAD; - - o_inst->op = - MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = ONE_SRC_0; - o_inst->src[2] = ZERO_SRC_0; -#endif - - goto next; - + case OPCODE_ABS: + inst = r300TranslateOpcodeABS(vp, vpi, inst, src); + break; case OPCODE_ADD: -#if 1 - hw_op = (src[0].File == PROGRAM_TEMPORARY && - src[1].File == - PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 : - R300_VPI_OUT_OP_MAD; - - o_inst->op = - MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - o_inst->src[0] = ONE_SRC_0; - o_inst->src[1] = t_src(vp, &src[0]); - o_inst->src[2] = t_src(vp, &src[1]); -#else - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = t_src(vp, &src[1]); - o_inst->src[2] = ZERO_SRC_1; - -#endif - goto next; - - case OPCODE_MAD: - hw_op = (src[0].File == PROGRAM_TEMPORARY && - src[1].File == PROGRAM_TEMPORARY && - src[2].File == - PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 : - R300_VPI_OUT_OP_MAD; - - o_inst->op = - MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = t_src(vp, &src[1]); - o_inst->src[2] = t_src(vp, &src[2]); - goto next; - - case OPCODE_MUL: /* HW mul can take third arg but appears to have some other limitations. */ - hw_op = (src[0].File == PROGRAM_TEMPORARY && - src[1].File == - PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 : - R300_VPI_OUT_OP_MAD; - - o_inst->op = - MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = t_src(vp, &src[1]); - - o_inst->src[2] = ZERO_SRC_1; - goto next; - - case OPCODE_DP3: //DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ZERO} PARAM 0{} {X Y Z ZERO} - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_DOT, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - - o_inst->src[0] = - MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), - t_swizzle(GET_SWZ - (src[0].Swizzle, 0)), - t_swizzle(GET_SWZ - (src[0].Swizzle, 1)), - t_swizzle(GET_SWZ - (src[0].Swizzle, 2)), - SWIZZLE_ZERO, - t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_XYZ : - VSF_FLAG_NONE) | (src[0]. - RelAddr << 4); - - o_inst->src[1] = - MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), - t_swizzle(GET_SWZ - (src[1].Swizzle, 0)), - t_swizzle(GET_SWZ - (src[1].Swizzle, 1)), - t_swizzle(GET_SWZ - (src[1].Swizzle, 2)), - SWIZZLE_ZERO, - t_src_class(src[1].File), - src[1]. - NegateBase ? VSF_FLAG_XYZ : - VSF_FLAG_NONE) | (src[1]. - RelAddr << 4); - - o_inst->src[2] = ZERO_SRC_1; - goto next; - - case OPCODE_SUB: //ADD RESULT 1.X Y Z W TMP 0{} {X Y Z W} PARAM 1{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W -#if 1 - hw_op = (src[0].File == PROGRAM_TEMPORARY && - src[1].File == - PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 : - R300_VPI_OUT_OP_MAD; - - o_inst->op = - MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = ONE_SRC_0; - o_inst->src[2] = - MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), - t_swizzle(GET_SWZ - (src[1].Swizzle, 0)), - t_swizzle(GET_SWZ - (src[1].Swizzle, 1)), - t_swizzle(GET_SWZ - (src[1].Swizzle, 2)), - t_swizzle(GET_SWZ - (src[1].Swizzle, 3)), - t_src_class(src[1].File), - (!src[1]. - NegateBase) ? VSF_FLAG_ALL : - VSF_FLAG_NONE) | (src[1]. - RelAddr << 4); -#else - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = - MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), - t_swizzle(GET_SWZ - (src[1].Swizzle, 0)), - t_swizzle(GET_SWZ - (src[1].Swizzle, 1)), - t_swizzle(GET_SWZ - (src[1].Swizzle, 2)), - t_swizzle(GET_SWZ - (src[1].Swizzle, 3)), - t_src_class(src[1].File), - (!src[1]. - NegateBase) ? VSF_FLAG_ALL : - VSF_FLAG_NONE) | (src[1]. - RelAddr << 4); - o_inst->src[2] = 0; -#endif - goto next; - - case OPCODE_ABS: //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 - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_MAX, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = - MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), - t_swizzle(GET_SWZ - (src[0].Swizzle, 0)), - t_swizzle(GET_SWZ - (src[0].Swizzle, 1)), - t_swizzle(GET_SWZ - (src[0].Swizzle, 2)), - t_swizzle(GET_SWZ - (src[0].Swizzle, 3)), - t_src_class(src[0].File), - (!src[0]. - NegateBase) ? VSF_FLAG_ALL : - VSF_FLAG_NONE) | (src[0]. - RelAddr << 4); - o_inst->src[2] = 0; - goto next; - + inst = r300TranslateOpcodeADD(vp, vpi, inst, src); + break; + case OPCODE_ARL: + inst = r300TranslateOpcodeARL(vp, vpi, inst, src); + break; + case OPCODE_DP3: + inst = r300TranslateOpcodeDP3(vp, vpi, inst, src); + break; + case OPCODE_DP4: + inst = r300TranslateOpcodeDP4(vp, vpi, inst, src); + break; + case OPCODE_DPH: + inst = r300TranslateOpcodeDPH(vp, vpi, inst, src); + break; + case OPCODE_DST: + inst = r300TranslateOpcodeDST(vp, vpi, inst, src); + break; + case OPCODE_EX2: + inst = r300TranslateOpcodeEX2(vp, vpi, inst, src); + break; + case OPCODE_EXP: + inst = r300TranslateOpcodeEXP(vp, vpi, inst, src); + break; case OPCODE_FLR: - /* FRC TMP 0.X Y Z W PARAM 0{} {X Y Z W} - 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 */ - - o_inst->op = MAKE_VSF_OP(R300_VPI_OUT_OP_FRC, u_temp_i, - t_dst_mask(vpi->DstReg. - WriteMask), - VSF_OUT_CLASS_TMP); - - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = ZERO_SRC_0; - o_inst->src[2] = ZERO_SRC_0; - o_inst++; - - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = MAKE_VSF_SOURCE(u_temp_i, - VSF_IN_COMPONENT_X, - VSF_IN_COMPONENT_Y, - VSF_IN_COMPONENT_Z, - VSF_IN_COMPONENT_W, - VSF_IN_CLASS_TMP, - /* Not 100% sure about this */ - (!src[0]. - NegateBase) ? - VSF_FLAG_ALL : - VSF_FLAG_NONE - /*VSF_FLAG_ALL */ ); - - o_inst->src[2] = ZERO_SRC_0; - u_temp_i--; - goto next; - - case OPCODE_LG2: // LG2 RESULT 1.X Y Z W PARAM 0{} {X X X X} - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_LG2, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - - o_inst->src[0] = - MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), - t_swizzle(GET_SWZ - (src[0].Swizzle, 0)), - t_swizzle(GET_SWZ - (src[0].Swizzle, 0)), - t_swizzle(GET_SWZ - (src[0].Swizzle, 0)), - t_swizzle(GET_SWZ - (src[0].Swizzle, 0)), - t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_ALL : - VSF_FLAG_NONE) | (src[0]. - RelAddr << 4); - o_inst->src[1] = ZERO_SRC_0; - o_inst->src[2] = ZERO_SRC_0; - goto next; - - case OPCODE_LIT: //LIT TMP 1.Y Z TMP 1{} {X W Z Y} TMP 1{} {Y W Z X} TMP 1{} {Y X Z W} - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_LIT, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - /* NOTE: Users swizzling might not work. */ - o_inst->src[0] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // x - t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // w - VSF_IN_COMPONENT_ZERO, // z - t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // y - t_src_class(src[0]. - File), - src[0]. - NegateBase ? - VSF_FLAG_ALL : - VSF_FLAG_NONE) | - (src[0].RelAddr << 4); - o_inst->src[1] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // y - t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // w - VSF_IN_COMPONENT_ZERO, // z - t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // x - t_src_class(src[0]. - File), - src[0]. - NegateBase ? - VSF_FLAG_ALL : - VSF_FLAG_NONE) | - (src[0].RelAddr << 4); - o_inst->src[2] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // y - t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // x - VSF_IN_COMPONENT_ZERO, // z - t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // w - t_src_class(src[0]. - File), - src[0]. - NegateBase ? - VSF_FLAG_ALL : - VSF_FLAG_NONE) | - (src[0].RelAddr << 4); - goto next; - - case OPCODE_DPH: //DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ONE} PARAM 0{} {X Y Z W} - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_DOT, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - - o_inst->src[0] = - MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), - t_swizzle(GET_SWZ - (src[0].Swizzle, 0)), - t_swizzle(GET_SWZ - (src[0].Swizzle, 1)), - t_swizzle(GET_SWZ - (src[0].Swizzle, 2)), - VSF_IN_COMPONENT_ONE, - t_src_class(src[0].File), - src[0]. - NegateBase ? VSF_FLAG_XYZ : - VSF_FLAG_NONE) | (src[0]. - RelAddr << 4); - o_inst->src[1] = t_src(vp, &src[1]); - o_inst->src[2] = ZERO_SRC_1; - goto next; - - case OPCODE_XPD: - /* mul r0, r1.yzxw, r2.zxyw - mad r0, -r2.yzxw, r1.zxyw, r0 - NOTE: might need MAD_2 - */ - - o_inst->op = MAKE_VSF_OP(R300_VPI_OUT_OP_MAD, u_temp_i, - t_dst_mask(vpi->DstReg. - WriteMask), - VSF_OUT_CLASS_TMP); - - o_inst->src[0] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // y - t_swizzle(GET_SWZ(src[0].Swizzle, 2)), // z - t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // x - t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // w - t_src_class(src[0]. - File), - src[0]. - NegateBase ? - VSF_FLAG_ALL : - VSF_FLAG_NONE) | - (src[0].RelAddr << 4); - - o_inst->src[1] = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 2)), // z - t_swizzle(GET_SWZ(src[1].Swizzle, 0)), // x - t_swizzle(GET_SWZ(src[1].Swizzle, 1)), // y - t_swizzle(GET_SWZ(src[1].Swizzle, 3)), // w - t_src_class(src[1]. - File), - src[1]. - NegateBase ? - VSF_FLAG_ALL : - VSF_FLAG_NONE) | - (src[1].RelAddr << 4); - - o_inst->src[2] = ZERO_SRC_1; - o_inst++; - u_temp_i--; - - o_inst->op = - MAKE_VSF_OP(R300_VPI_OUT_OP_MAD, - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - - o_inst->src[0] = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 1)), // y - t_swizzle(GET_SWZ(src[1].Swizzle, 2)), // z - t_swizzle(GET_SWZ(src[1].Swizzle, 0)), // x - t_swizzle(GET_SWZ(src[1].Swizzle, 3)), // w - t_src_class(src[1]. - File), - (!src[1]. - NegateBase) ? - VSF_FLAG_ALL : - VSF_FLAG_NONE) | - (src[1].RelAddr << 4); - - o_inst->src[1] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 2)), // z - t_swizzle(GET_SWZ(src[0].Swizzle, 0)), // x - t_swizzle(GET_SWZ(src[0].Swizzle, 1)), // y - t_swizzle(GET_SWZ(src[0].Swizzle, 3)), // w - t_src_class(src[0]. - File), - src[0]. - NegateBase ? - VSF_FLAG_ALL : - VSF_FLAG_NONE) | - (src[0].RelAddr << 4); - - o_inst->src[2] = MAKE_VSF_SOURCE(u_temp_i + 1, - VSF_IN_COMPONENT_X, - VSF_IN_COMPONENT_Y, - VSF_IN_COMPONENT_Z, - VSF_IN_COMPONENT_W, - VSF_IN_CLASS_TMP, - VSF_FLAG_NONE); - - goto next; - - case OPCODE_RCC: - fprintf(stderr, "Dont know how to handle op %d yet\n", - vpi->Opcode); - _mesa_exit(-1); + inst = r300TranslateOpcodeFLR(vp, vpi, inst, src, /* FIXME */ + &u_temp_i); + break; + case OPCODE_FRC: + inst = r300TranslateOpcodeFRC(vp, vpi, inst, src); + break; + case OPCODE_LG2: + inst = r300TranslateOpcodeLG2(vp, vpi, inst, src); + break; + case OPCODE_LIT: + inst = r300TranslateOpcodeLIT(vp, vpi, inst, src); + break; + case OPCODE_LOG: + inst = r300TranslateOpcodeLOG(vp, vpi, inst, src); + break; + case OPCODE_MAD: + inst = r300TranslateOpcodeMAD(vp, vpi, inst, src); + break; + case OPCODE_MAX: + inst = r300TranslateOpcodeMAX(vp, vpi, inst, src); + break; + case OPCODE_MIN: + inst = r300TranslateOpcodeMIN(vp, vpi, inst, src); + break; + case OPCODE_MOV: + inst = r300TranslateOpcodeMOV(vp, vpi, inst, src); + break; + case OPCODE_MUL: + inst = r300TranslateOpcodeMUL(vp, vpi, inst, src); + break; + case OPCODE_POW: + inst = r300TranslateOpcodePOW(vp, vpi, inst, src); break; - case OPCODE_END: + case OPCODE_RCP: + inst = r300TranslateOpcodeRCP(vp, vpi, inst, src); + break; + case OPCODE_RSQ: + inst = r300TranslateOpcodeRSQ(vp, vpi, inst, src); + break; + case OPCODE_SGE: + inst = r300TranslateOpcodeSGE(vp, vpi, inst, src); + break; + case OPCODE_SLT: + inst = r300TranslateOpcodeSLT(vp, vpi, inst, src); + break; + case OPCODE_SUB: + inst = r300TranslateOpcodeSUB(vp, vpi, inst, src); + break; + case OPCODE_SWZ: + inst = r300TranslateOpcodeSWZ(vp, vpi, inst, src); + break; + case OPCODE_XPD: + inst = r300TranslateOpcodeXPD(vp, vpi, inst, src, /* FIXME */ + &u_temp_i); break; default: + assert(0); break; } + } - o_inst->op = - MAKE_VSF_OP(t_opcode(vpi->Opcode), - t_dst_index(vp, &vpi->DstReg), - t_dst_mask(vpi->DstReg.WriteMask), - t_dst_class(vpi->DstReg.File)); - - if (are_srcs_scalar) { - switch (operands) { - case 1: - o_inst->src[0] = t_src_scalar(vp, &src[0]); - o_inst->src[1] = ZERO_SRC_0; - o_inst->src[2] = ZERO_SRC_0; - break; - - case 2: - o_inst->src[0] = t_src_scalar(vp, &src[0]); - o_inst->src[1] = t_src_scalar(vp, &src[1]); - o_inst->src[2] = ZERO_SRC_1; - break; - - case 3: - o_inst->src[0] = t_src_scalar(vp, &src[0]); - o_inst->src[1] = t_src_scalar(vp, &src[1]); - o_inst->src[2] = t_src_scalar(vp, &src[2]); - break; - - default: - fprintf(stderr, - "scalars and op RCC not handled yet"); - _mesa_exit(-1); - break; - } - } else { - switch (operands) { - case 1: - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = ZERO_SRC_0; - o_inst->src[2] = ZERO_SRC_0; - break; - - case 2: - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = t_src(vp, &src[1]); - o_inst->src[2] = ZERO_SRC_1; - break; - - case 3: - o_inst->src[0] = t_src(vp, &src[0]); - o_inst->src[1] = t_src(vp, &src[1]); - o_inst->src[2] = t_src(vp, &src[2]); - break; - - default: - fprintf(stderr, - "scalars and op RCC not handled yet"); - _mesa_exit(-1); - break; - } + /* Some outputs may be artificially added, to match the inputs + of the fragment program. Blank the outputs here. */ + for (i = 0; i < VERT_RESULT_MAX; i++) { + if (vp->key.OutputsAdded & (1 << i)) { + inst[0] = PVS_OP_DST_OPERAND(VE_ADD, + GL_FALSE, + GL_FALSE, + vp->outputs[i], + VSF_FLAG_ALL, + PVS_DST_REG_OUT); + inst[1] = __CONST(0, SWIZZLE_ZERO); + inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[3] = __CONST(0, SWIZZLE_ZERO); + inst += 4; } - next:; } - /* Will most likely segfault before we get here... fix later. */ - if (o_inst - vp->program.body.i >= VSF_MAX_FRAGMENT_LENGTH / 4) { + vp->program.length = (inst - vp->program.body.i); + if (vp->program.length >= VSF_MAX_FRAGMENT_LENGTH) { vp->program.length = 0; vp->native = GL_FALSE; - return; } - vp->program.length = (o_inst - vp->program.body.i) * 4; #if 0 fprintf(stderr, "hw program:\n"); for (i = 0; i < vp->program.length; i++) @@ -1059,6 +1222,9 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp, #endif } +/* DP4 version seems to trigger some hw peculiarity */ +//#define PREFER_DP4 + static void position_invariant(struct gl_program *prog) { struct prog_instruction *vpi; @@ -1145,8 +1311,8 @@ static void position_invariant(struct gl_program *prog) assert(vpi->Opcode == OPCODE_END); } -static void insert_wpos(struct r300_vertex_program *vp, - struct gl_program *prog, GLuint temp_index) +static void insert_wpos(struct r300_vertex_program *vp, struct gl_program *prog, + GLuint temp_index) { struct prog_instruction *vpi; struct prog_instruction *vpi_insert; @@ -1206,8 +1372,8 @@ static void pos_as_texcoord(struct r300_vertex_program *vp, prog->NumTemporaries++; for (vpi = prog->Instructions; vpi->Opcode != OPCODE_END; vpi++) { - if (vpi->DstReg.File == PROGRAM_OUTPUT && - vpi->DstReg.Index == VERT_RESULT_HPOS) { + if (vpi->DstReg.File == PROGRAM_OUTPUT + && vpi->DstReg.Index == VERT_RESULT_HPOS) { vpi->DstReg.File = PROGRAM_TEMPORARY; vpi->DstReg.Index = tempregi; } @@ -1223,25 +1389,32 @@ static struct r300_vertex_program *build_program(struct r300_vertex_program_key vp = _mesa_calloc(sizeof(*vp)); _mesa_memcpy(&vp->key, wanted_key, sizeof(vp->key)); - vp->wpos_idx = wpos_idx; if (mesa_vp->IsPositionInvariant) { position_invariant(&mesa_vp->Base); } - if (wpos_idx > -1) + if (wpos_idx > -1) { pos_as_texcoord(vp, &mesa_vp->Base); + } assert(mesa_vp->Base.NumInstructions); - vp->num_temporaries = mesa_vp->Base.NumTemporaries; - r300TranslateVertexShader(vp, mesa_vp->Base.Instructions); return vp; } +static void add_outputs(struct r300_vertex_program_key *key, GLint vert) +{ + if (key->OutputsWritten & (1 << vert)) + return; + + key->OutputsWritten |= 1 << vert; + key->OutputsAdded |= 1 << vert; +} + void r300SelectVertexShader(r300ContextPtr r300) { GLcontext *ctx = ctx = r300->radeon.glCtx; @@ -1253,10 +1426,10 @@ void r300SelectVertexShader(r300ContextPtr r300) GLint wpos_idx; vpc = (struct r300_vertex_program_cont *)ctx->VertexProgram._Current; + wanted_key.InputsRead = vpc->mesa_program.Base.InputsRead; + wanted_key.OutputsWritten = vpc->mesa_program.Base.OutputsWritten; InputsRead = ctx->FragmentProgram._Current->Base.InputsRead; - wanted_key.OutputsWritten |= 1 << VERT_RESULT_HPOS; - wpos_idx = -1; if (InputsRead & FRAG_BIT_WPOS) { for (i = 0; i < ctx->Const.MaxTextureUnits; i++) @@ -1268,31 +1441,34 @@ void r300SelectVertexShader(r300ContextPtr r300) _mesa_exit(-1); } - InputsRead |= (FRAG_BIT_TEX0 << i); + wanted_key.OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i); wpos_idx = i; } - if (InputsRead & FRAG_BIT_COL0) - wanted_key.OutputsWritten |= 1 << VERT_RESULT_COL0; + add_outputs(&wanted_key, VERT_RESULT_HPOS); - if ((InputsRead & FRAG_BIT_COL1) /*|| - (InputsRead & FRAG_BIT_FOGC) */ ) - wanted_key.OutputsWritten |= 1 << VERT_RESULT_COL1; + if (InputsRead & FRAG_BIT_COL0) { + add_outputs(&wanted_key, VERT_RESULT_COL0); + } - for (i = 0; i < ctx->Const.MaxTextureUnits; i++) - if (InputsRead & (FRAG_BIT_TEX0 << i)) - wanted_key.OutputsWritten |= - 1 << (VERT_RESULT_TEX0 + i); + if (InputsRead & FRAG_BIT_COL1) { + add_outputs(&wanted_key, VERT_RESULT_COL1); + } + + for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { + if (InputsRead & (FRAG_BIT_TEX0 << i)) { + add_outputs(&wanted_key, VERT_RESULT_TEX0 + i); + } + } - wanted_key.InputsRead = vpc->mesa_program.Base.InputsRead; if (vpc->mesa_program.IsPositionInvariant) { /* we wan't position don't we ? */ wanted_key.InputsRead |= (1 << VERT_ATTRIB_POS); } for (vp = vpc->progs; vp; vp = vp->next) - if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key)) == - 0) { + if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key)) + == 0) { r300->selected_vp = vp; return; } |