-}
-
-static void emit_instruction(struct r300_pfs_compile_state *cs, struct prog_instruction *fpi)
-{
- COMPILE_STATE;
- GLuint src[3], dest, temp[2];
- int flags, mask = 0;
- int const_sin[2];
-
- if (fpi->SaturateMode == SATURATE_ZERO_ONE)
- flags = PFS_FLAG_SAT;
- else
- flags = 0;
-
- if (fpi->Opcode != OPCODE_KIL) {
- dest = t_dst(cs, fpi->DstReg);
- mask = fpi->DstReg.WriteMask;
- }
-
- switch (fpi->Opcode) {
- case OPCODE_ADD:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- src[1] = t_src(cs, fpi->SrcReg[1]);
- emit_arith(cs, PFS_OP_MAD, dest, mask,
- src[0], pfs_one, src[1], flags);
- break;
- case OPCODE_CMP:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- src[1] = t_src(cs, fpi->SrcReg[1]);
- src[2] = t_src(cs, fpi->SrcReg[2]);
- /* ARB_f_p - if src0.c < 0.0 ? src1.c : src2.c
- * r300 - if src2.c < 0.0 ? src1.c : src0.c
- */
- emit_arith(cs, PFS_OP_CMP, dest, mask,
- src[2], src[1], src[0], flags);
- break;
- case OPCODE_COS:
- /*
- * cos using a parabola (see SIN):
- * cos(x):
- * x = (x/(2*PI))+0.75
- * x = frac(x)
- * x = (x*2*PI)-PI
- * result = sin(x)
- */
- temp[0] = get_temp_reg(cs);
- const_sin[0] = emit_const4fv(cs, SinCosConsts[0]);
- const_sin[1] = emit_const4fv(cs, SinCosConsts[1]);
- src[0] = t_scalar_src(cs, fpi->SrcReg[0]);
-
- /* add 0.5*PI and do range reduction */
-
- emit_arith(cs, PFS_OP_MAD, temp[0], WRITEMASK_X,
- swizzle(src[0], X, X, X, X),
- swizzle(const_sin[1], Z, Z, Z, Z),
- swizzle(const_sin[1], X, X, X, X), 0);
-
- emit_arith(cs, PFS_OP_FRC, temp[0], WRITEMASK_X,
- swizzle(temp[0], X, X, X, X),
- undef, undef, 0);
-
- emit_arith(cs, PFS_OP_MAD, temp[0], WRITEMASK_Z, swizzle(temp[0], X, X, X, X), swizzle(const_sin[1], W, W, W, W), //2*PI
- negate(swizzle(const_sin[0], Z, Z, Z, Z)), //-PI
- 0);
-
- /* SIN */
-
- emit_arith(cs, PFS_OP_MAD, temp[0],
- WRITEMASK_X | WRITEMASK_Y, swizzle(temp[0],
- Z, Z, Z,
- Z),
- const_sin[0], pfs_zero, 0);
-
- emit_arith(cs, PFS_OP_MAD, temp[0], WRITEMASK_X,
- swizzle(temp[0], Y, Y, Y, Y),
- absolute(swizzle(temp[0], Z, Z, Z, Z)),
- swizzle(temp[0], X, X, X, X), 0);
-
- emit_arith(cs, PFS_OP_MAD, temp[0], WRITEMASK_Y,
- swizzle(temp[0], X, X, X, X),
- absolute(swizzle(temp[0], X, X, X, X)),
- negate(swizzle(temp[0], X, X, X, X)), 0);
-
- emit_arith(cs, PFS_OP_MAD, dest, mask,
- swizzle(temp[0], Y, Y, Y, Y),
- swizzle(const_sin[0], W, W, W, W),
- swizzle(temp[0], X, X, X, X), flags);
-
- free_temp(cs, temp[0]);
- break;
- case OPCODE_DP3:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- src[1] = t_src(cs, fpi->SrcReg[1]);
- emit_arith(cs, PFS_OP_DP3, dest, mask,
- src[0], src[1], undef, flags);
- break;
- case OPCODE_DP4:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- src[1] = t_src(cs, fpi->SrcReg[1]);
- emit_arith(cs, PFS_OP_DP4, dest, mask,
- src[0], src[1], undef, flags);
- break;
- case OPCODE_DST:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- src[1] = t_src(cs, fpi->SrcReg[1]);
- /* dest.y = src0.y * src1.y */
- if (mask & WRITEMASK_Y)
- emit_arith(cs, PFS_OP_MAD, dest, WRITEMASK_Y,
- keep(src[0]), keep(src[1]),
- pfs_zero, flags);
- /* dest.z = src0.z */
- if (mask & WRITEMASK_Z)
- emit_arith(cs, PFS_OP_MAD, dest, WRITEMASK_Z,
- src[0], pfs_one, pfs_zero, flags);
- /* result.x = 1.0
- * result.w = src1.w */
- if (mask & WRITEMASK_XW) {
- REG_SET_VSWZ(src[1], SWIZZLE_111); /*Cheat */
- emit_arith(cs, PFS_OP_MAD, dest,
- mask & WRITEMASK_XW,
- src[1], pfs_one, pfs_zero, flags);
- }
- break;
- case OPCODE_EX2:
- src[0] = t_scalar_src(cs, fpi->SrcReg[0]);
- emit_arith(cs, PFS_OP_EX2, dest, mask,
- src[0], undef, undef, flags);
- break;
- case OPCODE_FRC:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- emit_arith(cs, PFS_OP_FRC, dest, mask,
- src[0], undef, undef, flags);
- break;
- case OPCODE_KIL:
- emit_tex(cs, fpi, R300_TEX_OP_KIL);
- break;
- case OPCODE_LG2:
- src[0] = t_scalar_src(cs, fpi->SrcReg[0]);
- emit_arith(cs, PFS_OP_LG2, dest, mask,
- src[0], undef, undef, flags);
- break;
- case OPCODE_LIT:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- emit_lit(cs, dest, mask, src[0], flags);
- break;
- case OPCODE_LRP:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- src[1] = t_src(cs, fpi->SrcReg[1]);
- src[2] = t_src(cs, fpi->SrcReg[2]);
- /* result = tmp0tmp1 + (1 - tmp0)tmp2
- * = tmp0tmp1 + tmp2 + (-tmp0)tmp2
- * MAD temp, -tmp0, tmp2, tmp2
- * MAD result, tmp0, tmp1, temp
- */
- temp[0] = get_temp_reg(cs);
- emit_arith(cs, PFS_OP_MAD, temp[0], mask,
- negate(keep(src[0])), keep(src[2]), src[2],
- 0);
- emit_arith(cs, PFS_OP_MAD, dest, mask,
- src[0], src[1], temp[0], flags);
- free_temp(cs, temp[0]);
- break;
- case OPCODE_MAD:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- src[1] = t_src(cs, fpi->SrcReg[1]);
- src[2] = t_src(cs, fpi->SrcReg[2]);
- emit_arith(cs, PFS_OP_MAD, dest, mask,
- src[0], src[1], src[2], flags);
- break;
- case OPCODE_MAX:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- src[1] = t_src(cs, fpi->SrcReg[1]);
- emit_arith(cs, PFS_OP_MAX, dest, mask,
- src[0], src[1], undef, flags);
- break;
- case OPCODE_MIN:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- src[1] = t_src(cs, fpi->SrcReg[1]);
- emit_arith(cs, PFS_OP_MIN, dest, mask,
- src[0], src[1], undef, flags);
- break;
- case OPCODE_MOV:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- emit_arith(cs, PFS_OP_MAD, dest, mask,
- src[0], pfs_one, pfs_zero, flags);
- break;
- case OPCODE_MUL:
- src[0] = t_src(cs, fpi->SrcReg[0]);
- src[1] = t_src(cs, fpi->SrcReg[1]);
- emit_arith(cs, PFS_OP_MAD, dest, mask,
- src[0], src[1], pfs_zero, flags);
- break;
- case OPCODE_RCP:
- src[0] = t_scalar_src(cs, fpi->SrcReg[0]);
- emit_arith(cs, PFS_OP_RCP, dest, mask,
- src[0], undef, undef, flags);
- break;
- case OPCODE_RSQ:
- src[0] = t_scalar_src(cs, fpi->SrcReg[0]);
- emit_arith(cs, PFS_OP_RSQ, dest, mask,
- absolute(src[0]), pfs_zero, pfs_zero, flags);
- break;
- case OPCODE_SCS:
- /*
- * scs using a parabola :
- * scs(x):
- * result.x = sin(-abs(x)+0.5*PI) (cos)
- * result.y = sin(x) (sin)
- *
- */
- temp[0] = get_temp_reg(cs);
- temp[1] = get_temp_reg(cs);
- const_sin[0] = emit_const4fv(cs, SinCosConsts[0]);
- const_sin[1] = emit_const4fv(cs, SinCosConsts[1]);
- src[0] = t_scalar_src(cs, fpi->SrcReg[0]);
-
- /* x = -abs(x)+0.5*PI */
- emit_arith(cs, PFS_OP_MAD, temp[0], WRITEMASK_Z, swizzle(const_sin[0], Z, Z, Z, Z), //PI
- pfs_half,
- negate(abs
- (swizzle(keep(src[0]), X, X, X, X))),
- 0);
-
- /* C*x (sin) */
- emit_arith(cs, PFS_OP_MAD, temp[0], WRITEMASK_W,
- swizzle(const_sin[0], Y, Y, Y, Y),
- swizzle(keep(src[0]), X, X, X, X),
- pfs_zero, 0);
-
- /* B*x, C*x (cos) */
- emit_arith(cs, PFS_OP_MAD, temp[0],
- WRITEMASK_X | WRITEMASK_Y, swizzle(temp[0],
- Z, Z, Z,
- Z),
- const_sin[0], pfs_zero, 0);
-
- /* B*x (sin) */
- emit_arith(cs, PFS_OP_MAD, temp[1], WRITEMASK_W,
- swizzle(const_sin[0], X, X, X, X),
- keep(src[0]), pfs_zero, 0);
-
- /* y = B*x + C*x*abs(x) (sin) */
- emit_arith(cs, PFS_OP_MAD, temp[1], WRITEMASK_Z,
- absolute(src[0]),
- swizzle(temp[0], W, W, W, W),
- swizzle(temp[1], W, W, W, W), 0);
-
- /* y = B*x + C*x*abs(x) (cos) */
- emit_arith(cs, PFS_OP_MAD, temp[1], WRITEMASK_W,
- swizzle(temp[0], Y, Y, Y, Y),
- absolute(swizzle(temp[0], Z, Z, Z, Z)),
- swizzle(temp[0], X, X, X, X), 0);
-
- /* y*abs(y) - y (cos), y*abs(y) - y (sin) */
- emit_arith(cs, PFS_OP_MAD, temp[0],
- WRITEMASK_X | WRITEMASK_Y, swizzle(temp[1],
- W, Z, Y,
- X),
- absolute(swizzle(temp[1], W, Z, Y, X)),
- negate(swizzle(temp[1], W, Z, Y, X)), 0);
-
- /* dest.xy = mad(temp.xy, P, temp2.wz) */
- emit_arith(cs, PFS_OP_MAD, dest,
- mask & (WRITEMASK_X | WRITEMASK_Y), temp[0],
- swizzle(const_sin[0], W, W, W, W),
- swizzle(temp[1], W, Z, Y, X), flags);
-
- free_temp(cs, temp[0]);
- free_temp(cs, temp[1]);
- break;
- case OPCODE_SIN:
- /*
- * using a parabola:
- * sin(x) = 4/pi * x + -4/(pi*pi) * x * abs(x)
- * extra precision is obtained by weighting against
- * itself squared.
- */
-
- temp[0] = get_temp_reg(cs);
- const_sin[0] = emit_const4fv(cs, SinCosConsts[0]);
- const_sin[1] = emit_const4fv(cs, SinCosConsts[1]);
- src[0] = t_scalar_src(cs, fpi->SrcReg[0]);