Not used and not widely supported. Use MIN+MAX instead.
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
bld_base->uint_bld.vec_type, "");
}
-/* TGSI_OPCODE_CLAMP */
-static void
-clamp_emit(
- const struct lp_build_tgsi_action * action,
- struct lp_build_tgsi_context * bld_base,
- struct lp_build_emit_data * emit_data)
-{
- LLVMValueRef tmp;
- tmp = lp_build_emit_llvm_binary(bld_base, TGSI_OPCODE_MAX,
- emit_data->args[0],
- emit_data->args[1]);
- emit_data->output[emit_data->chan] = lp_build_emit_llvm_binary(bld_base,
- TGSI_OPCODE_MIN, tmp, emit_data->args[2]);
-}
-
/* DP* Helper */
static void
bld_base->op_actions[TGSI_OPCODE_ADD].emit = add_emit;
bld_base->op_actions[TGSI_OPCODE_ARR].emit = arr_emit;
- bld_base->op_actions[TGSI_OPCODE_CLAMP].emit = clamp_emit;
bld_base->op_actions[TGSI_OPCODE_END].emit = end_emit;
bld_base->op_actions[TGSI_OPCODE_FRC].emit = frc_emit;
bld_base->op_actions[TGSI_OPCODE_LRP].emit = lrp_emit;
dst0 = lp_build_sub(&bld->bld_base.base, src0, tmp0);
break;
- case TGSI_OPCODE_CLAMP:
- src0 = lp_build_emit_fetch(&bld->bld_base, inst, 0, LP_CHAN_ALL);
- src1 = lp_build_emit_fetch(&bld->bld_base, inst, 1, LP_CHAN_ALL);
- src2 = lp_build_emit_fetch(&bld->bld_base, inst, 2, LP_CHAN_ALL);
- tmp0 = lp_build_max(&bld->bld_base.base, src0, src1);
- dst0 = lp_build_min(&bld->bld_base.base, tmp0, src2);
- break;
-
case TGSI_OPCODE_FLR:
src0 = lp_build_emit_fetch(&bld->bld_base, inst, 0, LP_CHAN_ALL);
dst0 = lp_build_floor(&bld->bld_base.base, src0);
ttn_move_dest(b, dest, nir_slt(b, src[1], src[0]));
}
-static void
-ttn_clamp(nir_builder *b, nir_op op, nir_alu_dest dest, nir_ssa_def **src)
-{
- ttn_move_dest(b, dest, nir_fmin(b, nir_fmax(b, src[0], src[1]), src[2]));
-}
-
static void
ttn_xpd(nir_builder *b, nir_op op, nir_alu_dest dest, nir_ssa_def **src)
{
[TGSI_OPCODE_SQRT] = nir_op_fsqrt,
[TGSI_OPCODE_DP2A] = 0,
[TGSI_OPCODE_FRC] = nir_op_ffract,
- [TGSI_OPCODE_CLAMP] = 0,
[TGSI_OPCODE_FLR] = nir_op_ffloor,
[TGSI_OPCODE_ROUND] = nir_op_fround_even,
[TGSI_OPCODE_EX2] = nir_op_fexp2,
ttn_lit(b, op_trans[tgsi_op], dest, src);
break;
- case TGSI_OPCODE_CLAMP:
- ttn_clamp(b, op_trans[tgsi_op], dest, src);
- break;
-
case TGSI_OPCODE_XPD:
ttn_xpd(b, op_trans[tgsi_op], dest, src);
break;
dst->f[3] = ceilf(src->f[3]);
}
-static void
-micro_clamp(union tgsi_exec_channel *dst,
- const union tgsi_exec_channel *src0,
- const union tgsi_exec_channel *src1,
- const union tgsi_exec_channel *src2)
-{
- dst->f[0] = src0->f[0] < src1->f[0] ? src1->f[0] : src0->f[0] > src2->f[0] ? src2->f[0] : src0->f[0];
- dst->f[1] = src0->f[1] < src1->f[1] ? src1->f[1] : src0->f[1] > src2->f[1] ? src2->f[1] : src0->f[1];
- dst->f[2] = src0->f[2] < src1->f[2] ? src1->f[2] : src0->f[2] > src2->f[2] ? src2->f[2] : src0->f[2];
- dst->f[3] = src0->f[3] < src1->f[3] ? src1->f[3] : src0->f[3] > src2->f[3] ? src2->f[3] : src0->f[3];
-}
-
static void
micro_cmp(union tgsi_exec_channel *dst,
const union tgsi_exec_channel *src0,
exec_vector_unary(mach, inst, micro_frc, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
break;
- case TGSI_OPCODE_CLAMP:
- exec_vector_trinary(mach, inst, micro_clamp, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
- break;
-
case TGSI_OPCODE_FLR:
exec_vector_unary(mach, inst, micro_flr, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
break;
{ 1, 1, 0, 0, 0, 0, 0, COMP, "F2U64", TGSI_OPCODE_F2U64 },
{ 1, 1, 0, 0, 0, 0, 0, COMP, "F2I64", TGSI_OPCODE_F2I64 },
{ 1, 1, 0, 0, 0, 0, 0, COMP, "FRC", TGSI_OPCODE_FRC },
- { 1, 3, 0, 0, 0, 0, 0, COMP, "CLAMP", TGSI_OPCODE_CLAMP },
+ { 1, 3, 0, 0, 0, 0, 0, COMP, "", 25 }, /* removed */
{ 1, 1, 0, 0, 0, 0, 0, COMP, "FLR", TGSI_OPCODE_FLR },
{ 1, 1, 0, 0, 0, 0, 0, COMP, "ROUND", TGSI_OPCODE_ROUND },
{ 1, 1, 0, 0, 0, 0, 0, REPL, "EX2", TGSI_OPCODE_EX2 },
OP11(SQRT)
OP13(DP2A)
OP11(FRC)
-OP13(CLAMP)
OP11(FLR)
OP11(ROUND)
OP11(EX2)
case TGSI_OPCODE_FMA:
case TGSI_OPCODE_FRC:
case TGSI_OPCODE_CEIL:
- case TGSI_OPCODE_CLAMP:
case TGSI_OPCODE_FLR:
case TGSI_OPCODE_ROUND:
case TGSI_OPCODE_POW:
dst.w = src.w - \lfloor src.w\rfloor
-.. opcode:: CLAMP - Clamp
-
-.. math::
-
- dst.x = clamp(src0.x, src1.x, src2.x)
-
- dst.y = clamp(src0.y, src1.y, src2.y)
-
- dst.z = clamp(src0.z, src1.z, src2.z)
-
- dst.w = clamp(src0.w, src1.w, src2.w)
-
-
.. opcode:: FLR - Floor
.. math::
mkCvt(OP_CVT, TYPE_F32, dst0[c], TYPE_F32, fetchSrc(0, c))
->rnd = ROUND_NI;
break;
- case TGSI_OPCODE_CLAMP:
- FOR_EACH_DST_ENABLED_CHANNEL(0, c, tgsi) {
- src0 = fetchSrc(0, c);
- src1 = fetchSrc(1, c);
- src2 = fetchSrc(2, c);
- val0 = getScratch();
- mkOp2(OP_MIN, TYPE_F32, val0, src0, src1);
- mkOp2(OP_MAX, TYPE_F32, dst0[c], val0, src2);
- }
- break;
case TGSI_OPCODE_SLT:
case TGSI_OPCODE_SGE:
case TGSI_OPCODE_SEQ:
case TGSI_OPCODE_LRP: return RC_OPCODE_LRP;
/* case TGSI_OPCODE_DP2A: return RC_OPCODE_DP2A; */
case TGSI_OPCODE_FRC: return RC_OPCODE_FRC;
- case TGSI_OPCODE_CLAMP: return RC_OPCODE_CLAMP;
case TGSI_OPCODE_FLR: return RC_OPCODE_FLR;
case TGSI_OPCODE_ROUND: return RC_OPCODE_ROUND;
case TGSI_OPCODE_EX2: return RC_OPCODE_EX2;
[22] = { ALU_OP0_NOP, tgsi_unsupported},
[23] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FRC] = { ALU_OP1_FRACT, tgsi_op2},
- [TGSI_OPCODE_CLAMP] = { ALU_OP0_NOP, tgsi_unsupported},
+ [25] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FLR] = { ALU_OP1_FLOOR, tgsi_op2},
[TGSI_OPCODE_ROUND] = { ALU_OP1_RNDNE, tgsi_op2},
[TGSI_OPCODE_EX2] = { ALU_OP1_EXP_IEEE, tgsi_trans_srcx_replicate},
[22] = { ALU_OP0_NOP, tgsi_unsupported},
[23] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FRC] = { ALU_OP1_FRACT, tgsi_op2},
- [TGSI_OPCODE_CLAMP] = { ALU_OP0_NOP, tgsi_unsupported},
+ [25] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FLR] = { ALU_OP1_FLOOR, tgsi_op2},
[TGSI_OPCODE_ROUND] = { ALU_OP1_RNDNE, tgsi_op2},
[TGSI_OPCODE_EX2] = { ALU_OP1_EXP_IEEE, tgsi_trans_srcx_replicate},
[22] = { ALU_OP0_NOP, tgsi_unsupported},
[23] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FRC] = { ALU_OP1_FRACT, tgsi_op2},
- [TGSI_OPCODE_CLAMP] = { ALU_OP0_NOP, tgsi_unsupported},
+ [25] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FLR] = { ALU_OP1_FLOOR, tgsi_op2},
[TGSI_OPCODE_ROUND] = { ALU_OP1_RNDNE, tgsi_op2},
[TGSI_OPCODE_EX2] = { ALU_OP1_EXP_IEEE, cayman_emit_float_instr},
HAVE_LLVM >= 0x0308 ? "llvm.bitreverse.i32" : "llvm.AMDGPU.brev";
bld_base->op_actions[TGSI_OPCODE_CEIL].emit = build_tgsi_intrinsic_nomem;
bld_base->op_actions[TGSI_OPCODE_CEIL].intr_name = "llvm.ceil.f32";
- bld_base->op_actions[TGSI_OPCODE_CLAMP].emit = build_tgsi_intrinsic_nomem;
- bld_base->op_actions[TGSI_OPCODE_CLAMP].intr_name =
- HAVE_LLVM >= 0x0308 ? "llvm.AMDGPU.clamp." : "llvm.AMDIL.clamp.";
bld_base->op_actions[TGSI_OPCODE_CMP].emit = emit_cmp;
bld_base->op_actions[TGSI_OPCODE_COS].emit = build_tgsi_intrinsic_nomem;
bld_base->op_actions[TGSI_OPCODE_COS].intr_name = "llvm.cos.f32";
/* These aren't actually used by any of the frontends we care
* about:
*/
- case TGSI_OPCODE_CLAMP:
case TGSI_OPCODE_AND:
case TGSI_OPCODE_OR:
case TGSI_OPCODE_I2F:
#define TGSI_OPCODE_F2U64 22
#define TGSI_OPCODE_F2I64 23
#define TGSI_OPCODE_FRC 24
-#define TGSI_OPCODE_CLAMP 25
+/* gap */
#define TGSI_OPCODE_FLR 26
#define TGSI_OPCODE_ROUND 27
#define TGSI_OPCODE_EX2 28