*
* Copyright 2011-2012 Advanced Micro Devices, Inc.
* Copyright 2009 VMware, Inc.
- * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2007-2008 VMware, Inc.
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
* ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
#include "lp_bld_arit.h"
#include "lp_bld_bitarit.h"
#include "lp_bld_const.h"
+#include "lp_bld_conv.h"
#include "lp_bld_gather.h"
#include "lp_bld_logic.h"
+#include "lp_bld_pack.h"
#include "tgsi/tgsi_exec.h"
TGSI_OPCODE_SUB, emit_data->args[0], tmp);
}
-/* TGSI_OPCODE_KIL */
+/* TGSI_OPCODE_KILL_IF */
static void
kil_fetch_args(
emit_data->dst_type = LLVMVoidTypeInContext(bld_base->base.gallivm->context);
}
-/* TGSI_OPCODE_KILP */
+/* TGSI_OPCODE_KILL */
static void
kilp_fetch_args(
LLVMValueRef abs_x, log_abs_x, flr_log_abs_x, ex2_flr_log_abs_x;
/* abs( src0.x) */
- abs_x = lp_build_emit_llvm_unary(bld_base, TGSI_OPCODE_ABS,
- emit_data->args[0] /* src0.x */);
+ abs_x = lp_build_abs(&bld_base->base, emit_data->args[0] /* src0.x */);
/* log( abs( src0.x ) ) */
log_abs_x = lp_build_emit_llvm_unary(bld_base, TGSI_OPCODE_LG2,
log_emit /* emit */
};
+/* TGSI_OPCODE_PK2H */
+
+static void
+pk2h_fetch_args(
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ /* src0.x */
+ emit_data->args[0] = lp_build_emit_fetch(bld_base, emit_data->inst,
+ 0, TGSI_CHAN_X);
+ /* src0.y */
+ emit_data->args[1] = lp_build_emit_fetch(bld_base, emit_data->inst,
+ 0, TGSI_CHAN_Y);
+}
+
+static void
+pk2h_emit(
+ const struct lp_build_tgsi_action *action,
+ struct lp_build_tgsi_context *bld_base,
+ struct lp_build_emit_data *emit_data)
+{
+ struct gallivm_state *gallivm = bld_base->base.gallivm;
+ struct lp_type f16i_t;
+ LLVMValueRef lo, hi, res;
+
+ f16i_t = lp_type_uint_vec(16, bld_base->base.type.length * 32);
+ lo = lp_build_float_to_half(gallivm, emit_data->args[0]);
+ hi = lp_build_float_to_half(gallivm, emit_data->args[1]);
+ /* maybe some interleave doubling vector width would be useful... */
+ lo = lp_build_pad_vector(gallivm, lo, bld_base->base.type.length * 2);
+ hi = lp_build_pad_vector(gallivm, hi, bld_base->base.type.length * 2);
+ res = lp_build_interleave2(gallivm, f16i_t, lo, hi, 0);
+
+ emit_data->output[emit_data->chan] = res;
+}
+
+static struct lp_build_tgsi_action pk2h_action = {
+ pk2h_fetch_args, /* fetch_args */
+ pk2h_emit /* emit */
+};
+
+/* TGSI_OPCODE_UP2H */
+
+static void
+up2h_emit(
+ const struct lp_build_tgsi_action *action,
+ struct lp_build_tgsi_context *bld_base,
+ struct lp_build_emit_data *emit_data)
+{
+ struct gallivm_state *gallivm = bld_base->base.gallivm;
+ LLVMBuilderRef builder = gallivm->builder;
+ LLVMContextRef context = gallivm->context;
+ LLVMValueRef lo, hi, res[2], arg;
+ unsigned nr = bld_base->base.type.length;
+ LLVMTypeRef i16t = LLVMVectorType(LLVMInt16TypeInContext(context), nr * 2);
+
+ arg = LLVMBuildBitCast(builder, emit_data->args[0], i16t, "");
+ lo = lp_build_uninterleave1(gallivm, nr * 2, arg, 0);
+ hi = lp_build_uninterleave1(gallivm, nr * 2, arg, 1);
+ res[0] = lp_build_half_to_float(gallivm, lo);
+ res[1] = lp_build_half_to_float(gallivm, hi);
+
+ emit_data->output[0] = emit_data->output[2] = res[0];
+ emit_data->output[1] = emit_data->output[3] = res[1];
+}
+
+static struct lp_build_tgsi_action up2h_action = {
+ scalar_unary_fetch_args, /* fetch_args */
+ up2h_emit /* emit */
+};
+
/* TGSI_OPCODE_LRP */
static void
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_SUB,
- emit_data->args[1],
- emit_data->args[2]);
- emit_data->output[emit_data->chan] = lp_build_emit_llvm_ternary(bld_base,
- TGSI_OPCODE_MAD, emit_data->args[0], tmp, emit_data->args[2]);
+ struct lp_build_context *bld = &bld_base->base;
+ LLVMValueRef inv, a, b;
+
+ /* This uses the correct version: (1 - t)*a + t*b
+ *
+ * An alternative version is "a + t*(b-a)". The problem is this version
+ * doesn't return "b" for t = 1, because "a + (b-a)" isn't equal to "b"
+ * because of the floating-point rounding.
+ */
+ inv = lp_build_sub(bld, bld_base->base.one, emit_data->args[0]);
+ a = lp_build_mul(bld, emit_data->args[1], emit_data->args[0]);
+ b = lp_build_mul(bld, emit_data->args[2], inv);
+ emit_data->output[emit_data->chan] = lp_build_add(bld, a, b);
}
/* TGSI_OPCODE_MAD */
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->args[0] = lp_build_emit_llvm_unary(bld_base, TGSI_OPCODE_ABS,
- emit_data->args[0]);
if (bld_base->rsq_action.emit) {
bld_base->rsq_action.emit(&bld_base->rsq_action, bld_base, emit_data);
} else {
scs_emit /* emit */
};
-/* TGSI_OPCODE_SFL */
-
-static void
-sfl_emit(
- const struct lp_build_tgsi_action * action,
- struct lp_build_tgsi_context * bld_base,
- struct lp_build_emit_data * emit_data)
-{
- emit_data->output[emit_data->chan] = bld_base->base.zero;
-}
-
-/* TGSI_OPCODE_STR */
-
+/* TGSI_OPCODE_SUB */
static void
-str_emit(
+sub_emit(
const struct lp_build_tgsi_action * action,
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = bld_base->base.one;
+ emit_data->output[emit_data->chan] =
+ LLVMBuildFSub(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ emit_data->args[1], "");
}
-/* TGSI_OPCODE_SUB */
+/* TGSI_OPCODE_F2U */
static void
-sub_emit(
+f2u_emit(
const struct lp_build_tgsi_action * action,
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = LLVMBuildFSub(
- bld_base->base.gallivm->builder,
- emit_data->args[0],
- emit_data->args[1], "");
+ emit_data->output[emit_data->chan] =
+ LLVMBuildFPToUI(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->base.int_vec_type, "");
}
/* TGSI_OPCODE_U2F */
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = LLVMBuildUIToFP(bld_base->base.gallivm->builder,
- emit_data->args[0],
- bld_base->base.vec_type, "");
+ emit_data->output[emit_data->chan] =
+ LLVMBuildUIToFP(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->base.vec_type, "");
}
static void
emit_data->args[0], emit_data->args[1]);
}
+/* TGSI_OPCODE_IMUL_HI */
+static void
+imul_hi_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ struct lp_build_context *int_bld = &bld_base->int_bld;
+ LLVMValueRef hi_bits;
+
+ assert(int_bld->type.width == 32);
+
+ /* low result bits are tossed away */
+ lp_build_mul_32_lohi(int_bld, emit_data->args[0],
+ emit_data->args[1], &hi_bits);
+ emit_data->output[emit_data->chan] = hi_bits;
+}
+
+static void
+imul_hi_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ struct lp_build_context *int_bld = &bld_base->int_bld;
+ LLVMValueRef hi_bits;
+
+ assert(int_bld->type.width == 32);
+
+ /* low result bits are tossed away */
+ lp_build_mul_32_lohi_cpu(int_bld, emit_data->args[0],
+ emit_data->args[1], &hi_bits);
+ emit_data->output[emit_data->chan] = hi_bits;
+}
+
+/* TGSI_OPCODE_UMUL_HI */
+static void
+umul_hi_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ struct lp_build_context *uint_bld = &bld_base->uint_bld;
+ LLVMValueRef hi_bits;
+
+ assert(uint_bld->type.width == 32);
+
+ /* low result bits are tossed away */
+ lp_build_mul_32_lohi(uint_bld, emit_data->args[0],
+ emit_data->args[1], &hi_bits);
+ emit_data->output[emit_data->chan] = hi_bits;
+}
+
+static void
+umul_hi_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ struct lp_build_context *uint_bld = &bld_base->uint_bld;
+ LLVMValueRef hi_bits;
+
+ assert(uint_bld->type.width == 32);
+
+ /* low result bits are tossed away */
+ lp_build_mul_32_lohi_cpu(uint_bld, emit_data->args[0],
+ emit_data->args[1], &hi_bits);
+ emit_data->output[emit_data->chan] = hi_bits;
+}
+
/* TGSI_OPCODE_MAX */
static void fmax_emit(
const struct lp_build_tgsi_action * action,
xpd_emit /* emit */
};
+/* TGSI_OPCODE_D2F */
+static void
+d2f_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildFPTrunc(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->base.vec_type, "");
+}
+
+/* TGSI_OPCODE_D2I */
+static void
+d2i_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildFPToSI(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->base.int_vec_type, "");
+}
+
+/* TGSI_OPCODE_D2U */
+static void
+d2u_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildFPToUI(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->base.int_vec_type, "");
+}
+
+/* TGSI_OPCODE_F2D */
+static void
+f2d_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildFPExt(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->dbl_bld.vec_type, "");
+}
+
+/* TGSI_OPCODE_U2D */
+static void
+u2d_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildUIToFP(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->dbl_bld.vec_type, "");
+}
+
+/* TGSI_OPCODE_I2D */
+static void
+i2d_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildSIToFP(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->dbl_bld.vec_type, "");
+}
+
+/* TGSI_OPCODE_DMAD */
+static void
+dmad_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_DMUL,
+ emit_data->args[0],
+ emit_data->args[1]);
+ emit_data->output[emit_data->chan] = lp_build_emit_llvm_binary(bld_base,
+ TGSI_OPCODE_DADD, tmp, emit_data->args[2]);
+}
+
+/*.TGSI_OPCODE_DRCP.*/
+static void drcp_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ LLVMValueRef one;
+ one = lp_build_const_vec(bld_base->dbl_bld.gallivm, bld_base->dbl_bld.type, 1.0f);
+ emit_data->output[emit_data->chan] = LLVMBuildFDiv(
+ bld_base->base.gallivm->builder,
+ one, emit_data->args[0], "");
+}
+
+/* TGSI_OPCODE_DFRAC */
+static void dfrac_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_floor(&bld_base->dbl_bld,
+ emit_data->args[0]);
+ emit_data->output[emit_data->chan] = LLVMBuildFSub(bld_base->base.gallivm->builder,
+ emit_data->args[0], tmp, "");
+}
+
+static void
+u64mul_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_mul(&bld_base->uint64_bld,
+ emit_data->args[0], emit_data->args[1]);
+}
+
+static void
+u64mod_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef div_mask = lp_build_cmp(&bld_base->uint64_bld,
+ PIPE_FUNC_EQUAL, emit_data->args[1],
+ bld_base->uint64_bld.zero);
+ /* We want to make sure that we never divide/mod by zero to not
+ * generate sigfpe. We don't want to crash just because the
+ * shader is doing something weird. */
+ LLVMValueRef divisor = LLVMBuildOr(builder,
+ div_mask,
+ emit_data->args[1], "");
+ LLVMValueRef result = lp_build_mod(&bld_base->uint64_bld,
+ emit_data->args[0], divisor);
+ /* umod by zero doesn't have a guaranteed return value chose -1 for now. */
+ emit_data->output[emit_data->chan] = LLVMBuildOr(builder,
+ div_mask,
+ result, "");
+}
+
+static void
+i64mod_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef div_mask = lp_build_cmp(&bld_base->uint64_bld,
+ PIPE_FUNC_EQUAL, emit_data->args[1],
+ bld_base->uint64_bld.zero);
+ /* We want to make sure that we never divide/mod by zero to not
+ * generate sigfpe. We don't want to crash just because the
+ * shader is doing something weird. */
+ LLVMValueRef divisor = LLVMBuildOr(builder,
+ div_mask,
+ emit_data->args[1], "");
+ LLVMValueRef result = lp_build_mod(&bld_base->int64_bld,
+ emit_data->args[0], divisor);
+ /* umod by zero doesn't have a guaranteed return value chose -1 for now. */
+ emit_data->output[emit_data->chan] = LLVMBuildOr(builder,
+ div_mask,
+ result, "");
+}
+
+static void
+u64div_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef div_mask = lp_build_cmp(&bld_base->uint64_bld,
+ PIPE_FUNC_EQUAL, emit_data->args[1],
+ bld_base->uint64_bld.zero);
+ /* We want to make sure that we never divide/mod by zero to not
+ * generate sigfpe. We don't want to crash just because the
+ * shader is doing something weird. */
+ LLVMValueRef divisor = LLVMBuildOr(builder,
+ div_mask,
+ emit_data->args[1], "");
+ LLVMValueRef result = LLVMBuildUDiv(builder,
+ emit_data->args[0], divisor, "");
+ /* udiv by zero is guaranteed to return 0xffffffff at least with d3d10 */
+ emit_data->output[emit_data->chan] = LLVMBuildOr(builder,
+ div_mask,
+ result, "");
+}
+
+static void
+i64div_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef div_mask = lp_build_cmp(&bld_base->int64_bld,
+ PIPE_FUNC_EQUAL, emit_data->args[1],
+ bld_base->int64_bld.zero);
+ /* We want to make sure that we never divide/mod by zero to not
+ * generate sigfpe. We don't want to crash just because the
+ * shader is doing something weird. */
+ LLVMValueRef divisor = LLVMBuildOr(builder,
+ div_mask,
+ emit_data->args[1], "");
+ LLVMValueRef result = LLVMBuildSDiv(builder,
+ emit_data->args[0], divisor, "");
+ /* udiv by zero is guaranteed to return 0xffffffff at least with d3d10 */
+ emit_data->output[emit_data->chan] = LLVMBuildOr(builder,
+ div_mask,
+ result, "");
+}
+
+static void
+f2u64_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildFPToUI(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->uint64_bld.vec_type, "");
+}
+
+static void
+f2i64_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildFPToSI(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->int64_bld.vec_type, "");
+}
+
+static void
+u2i64_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildZExt(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->uint64_bld.vec_type, "");
+}
+
+static void
+i2i64_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildSExt(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->int64_bld.vec_type, "");
+}
+
+static void
+i642f_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildSIToFP(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->base.vec_type, "");
+}
+
+static void
+u642f_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildUIToFP(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->base.vec_type, "");
+}
+
+static void
+i642d_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildSIToFP(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->dbl_bld.vec_type, "");
+}
+
+static void
+u642d_emit(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ LLVMBuildUIToFP(bld_base->base.gallivm->builder,
+ emit_data->args[0],
+ bld_base->dbl_bld.vec_type, "");
+}
+
void
lp_set_default_actions(struct lp_build_tgsi_context * bld_base)
{
bld_base->op_actions[TGSI_OPCODE_EXP] = exp_action;
bld_base->op_actions[TGSI_OPCODE_LIT] = lit_action;
bld_base->op_actions[TGSI_OPCODE_LOG] = log_action;
+ bld_base->op_actions[TGSI_OPCODE_PK2H] = pk2h_action;
bld_base->op_actions[TGSI_OPCODE_RSQ] = rsq_action;
bld_base->op_actions[TGSI_OPCODE_SQRT] = sqrt_action;
bld_base->op_actions[TGSI_OPCODE_POW] = pow_action;
bld_base->op_actions[TGSI_OPCODE_SCS] = scs_action;
+ bld_base->op_actions[TGSI_OPCODE_UP2H] = up2h_action;
bld_base->op_actions[TGSI_OPCODE_XPD] = xpd_action;
bld_base->op_actions[TGSI_OPCODE_BREAKC].fetch_args = scalar_unary_fetch_args;
+ bld_base->op_actions[TGSI_OPCODE_SWITCH].fetch_args = scalar_unary_fetch_args;
+ bld_base->op_actions[TGSI_OPCODE_CASE].fetch_args = scalar_unary_fetch_args;
bld_base->op_actions[TGSI_OPCODE_COS].fetch_args = scalar_unary_fetch_args;
bld_base->op_actions[TGSI_OPCODE_EX2].fetch_args = scalar_unary_fetch_args;
bld_base->op_actions[TGSI_OPCODE_IF].fetch_args = scalar_unary_fetch_args;
- bld_base->op_actions[TGSI_OPCODE_KIL].fetch_args = kil_fetch_args;
- bld_base->op_actions[TGSI_OPCODE_KILP].fetch_args = kilp_fetch_args;
+ bld_base->op_actions[TGSI_OPCODE_UIF].fetch_args = scalar_unary_fetch_args;
+ bld_base->op_actions[TGSI_OPCODE_KILL_IF].fetch_args = kil_fetch_args;
+ bld_base->op_actions[TGSI_OPCODE_KILL].fetch_args = kilp_fetch_args;
bld_base->op_actions[TGSI_OPCODE_RCP].fetch_args = scalar_unary_fetch_args;
bld_base->op_actions[TGSI_OPCODE_SIN].fetch_args = scalar_unary_fetch_args;
bld_base->op_actions[TGSI_OPCODE_LG2].fetch_args = scalar_unary_fetch_args;
bld_base->op_actions[TGSI_OPCODE_MUL].emit = mul_emit;
bld_base->op_actions[TGSI_OPCODE_DIV].emit = fdiv_emit;
bld_base->op_actions[TGSI_OPCODE_RCP].emit = rcp_emit;
- bld_base->op_actions[TGSI_OPCODE_SFL].emit = sfl_emit;
- bld_base->op_actions[TGSI_OPCODE_STR].emit = str_emit;
bld_base->op_actions[TGSI_OPCODE_SUB].emit = sub_emit;
bld_base->op_actions[TGSI_OPCODE_UARL].emit = mov_emit;
+ bld_base->op_actions[TGSI_OPCODE_F2U].emit = f2u_emit;
bld_base->op_actions[TGSI_OPCODE_U2F].emit = u2f_emit;
bld_base->op_actions[TGSI_OPCODE_UMAD].emit = umad_emit;
bld_base->op_actions[TGSI_OPCODE_UMUL].emit = umul_emit;
+ bld_base->op_actions[TGSI_OPCODE_IMUL_HI].emit = imul_hi_emit;
+ bld_base->op_actions[TGSI_OPCODE_UMUL_HI].emit = umul_hi_emit;
bld_base->op_actions[TGSI_OPCODE_MAX].emit = fmax_emit;
bld_base->op_actions[TGSI_OPCODE_MIN].emit = fmin_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_DADD].emit = add_emit;
+ bld_base->op_actions[TGSI_OPCODE_DMAX].emit = fmax_emit;
+ bld_base->op_actions[TGSI_OPCODE_DMIN].emit = fmin_emit;
+ bld_base->op_actions[TGSI_OPCODE_DMUL].emit = mul_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_D2F].emit = d2f_emit;
+ bld_base->op_actions[TGSI_OPCODE_D2I].emit = d2i_emit;
+ bld_base->op_actions[TGSI_OPCODE_D2U].emit = d2u_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_F2D].emit = f2d_emit;
+ bld_base->op_actions[TGSI_OPCODE_I2D].emit = i2d_emit;
+ bld_base->op_actions[TGSI_OPCODE_U2D].emit = u2d_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_DMAD].emit = dmad_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_DRCP].emit = drcp_emit;
+ bld_base->op_actions[TGSI_OPCODE_DFRAC].emit = dfrac_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_U64MUL].emit = u64mul_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_F2I64].emit = f2i64_emit;
+ bld_base->op_actions[TGSI_OPCODE_F2U64].emit = f2u64_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_D2I64].emit = f2i64_emit;
+ bld_base->op_actions[TGSI_OPCODE_D2U64].emit = f2u64_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_I2I64].emit = i2i64_emit;
+ bld_base->op_actions[TGSI_OPCODE_U2I64].emit = u2i64_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_I642F].emit = i642f_emit;
+ bld_base->op_actions[TGSI_OPCODE_U642F].emit = u642f_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_I642F].emit = i642f_emit;
+ bld_base->op_actions[TGSI_OPCODE_U642F].emit = u642f_emit;
+
+ bld_base->op_actions[TGSI_OPCODE_I642D].emit = i642d_emit;
+ bld_base->op_actions[TGSI_OPCODE_U642D].emit = u642d_emit;
+
}
/* CPU Only default actions */
* intrinsics.
*/
-/* TGSI_OPCODE_ABS (CPU Only)*/
-
-static void
-abs_emit_cpu(
- const struct lp_build_tgsi_action * action,
- struct lp_build_tgsi_context * bld_base,
- struct lp_build_emit_data * emit_data)
-{
- emit_data->output[emit_data->chan] = lp_build_abs(&bld_base->base,
- emit_data->args[0]);
-}
-
/* TGSI_OPCODE_ADD (CPU Only) */
static void
add_emit_cpu(
cond, emit_data->args[1], emit_data->args[2]);
}
-/* TGSI_OPCODE_CND (CPU Only) */
+/* TGSI_OPCODE_UCMP (CPU Only) */
static void
-cnd_emit_cpu(
+ucmp_emit_cpu(
const struct lp_build_tgsi_action * action,
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- LLVMValueRef half, tmp;
- half = lp_build_const_vec(bld_base->base.gallivm, bld_base->base.type, 0.5);
- tmp = lp_build_cmp(&bld_base->base, PIPE_FUNC_GREATER,
- emit_data->args[2], half);
- emit_data->output[emit_data->chan] = lp_build_select(&bld_base->base,
- tmp,
- emit_data->args[0],
- emit_data->args[1]);
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ struct lp_build_context *uint_bld = &bld_base->uint_bld;
+ LLVMValueRef unsigned_cond =
+ LLVMBuildBitCast(builder, emit_data->args[0], uint_bld->vec_type, "");
+ LLVMValueRef cond = lp_build_cmp(uint_bld, PIPE_FUNC_NOTEQUAL,
+ unsigned_cond,
+ uint_bld->zero);
+ emit_data->output[emit_data->chan] =
+ lp_build_select(&bld_base->base,
+ cond, emit_data->args[1], emit_data->args[2]);
}
/* TGSI_OPCODE_COS (CPU Only) */
emit_data->args[0]);
}
-/* TGSI_OPCODE_EXP (CPU Only) */
+/* TGSI_OPCODE_F2I (CPU Only) */
static void
-exp_emit_cpu(
+f2i_emit_cpu(
const struct lp_build_tgsi_action * action,
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- lp_build_exp2_approx(&bld_base->base, emit_data->args[0],
- &emit_data->output[TGSI_CHAN_X],
- &emit_data->output[TGSI_CHAN_Y],
- &emit_data->output[TGSI_CHAN_Z]);
- emit_data->output[TGSI_CHAN_W] = bld_base->base.one;
+ emit_data->output[emit_data->chan] = lp_build_itrunc(&bld_base->base,
+ emit_data->args[0]);
+}
+
+/* TGSI_OPCODE_FSET Helper (CPU Only) */
+static void
+fset_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data,
+ unsigned pipe_func)
+{
+ LLVMValueRef cond;
+
+ if (pipe_func != PIPE_FUNC_NOTEQUAL) {
+ cond = lp_build_cmp_ordered(&bld_base->base, pipe_func,
+ emit_data->args[0], emit_data->args[1]);
+ }
+ else {
+ cond = lp_build_cmp(&bld_base->base, pipe_func,
+ emit_data->args[0], emit_data->args[1]);
+
+ }
+ emit_data->output[emit_data->chan] = cond;
+}
+
+
+/* TGSI_OPCODE_FSEQ (CPU Only) */
+static void
+fseq_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ fset_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_EQUAL);
+}
+
+/* TGSI_OPCODE_ISGE (CPU Only) */
+static void
+fsge_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ fset_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_GEQUAL);
}
-/* TGSI_OPCODE_F2I (CPU Only) */
+/* TGSI_OPCODE_ISLT (CPU Only) */
static void
-f2i_emit_cpu(
+fslt_emit_cpu(
const struct lp_build_tgsi_action * action,
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_itrunc(&bld_base->base,
- emit_data->args[0]);
+ fset_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_LESS);
}
-/* TGSI_OPCODE_F2U (CPU Only) */
+/* TGSI_OPCODE_USNE (CPU Only) */
+
static void
-f2u_emit_cpu(
+fsne_emit_cpu(
const struct lp_build_tgsi_action * action,
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- /* FIXME: implement and use lp_build_utrunc() */
- emit_data->output[emit_data->chan] = lp_build_itrunc(&bld_base->base,
- emit_data->args[0]);
+ fset_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_NOTEQUAL);
}
/* TGSI_OPCODE_FLR (CPU Only) */
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_div(&bld_base->int_bld,
- emit_data->args[0], emit_data->args[1]);
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef div_mask = lp_build_cmp(&bld_base->uint_bld,
+ PIPE_FUNC_EQUAL, emit_data->args[1],
+ bld_base->uint_bld.zero);
+ /* We want to make sure that we never divide/mod by zero to not
+ * generate sigfpe. We don't want to crash just because the
+ * shader is doing something weird. */
+ LLVMValueRef divisor = LLVMBuildOr(builder,
+ div_mask,
+ emit_data->args[1], "");
+ LLVMValueRef result = lp_build_div(&bld_base->int_bld,
+ emit_data->args[0], divisor);
+ LLVMValueRef not_div_mask = LLVMBuildNot(builder,
+ div_mask,"");
+ /* idiv by zero doesn't have a guaranteed return value chose 0 for now. */
+ emit_data->output[emit_data->chan] = LLVMBuildAnd(builder,
+ not_div_mask,
+ result, "");
}
/* TGSI_OPCODE_INEG (CPU Only) */
struct lp_build_emit_data * emit_data,
unsigned pipe_func)
{
- LLVMValueRef nz = lp_build_const_vec(bld_base->base.gallivm,
- bld_base->int_bld.type, ~0U);
LLVMValueRef cond = lp_build_cmp(&bld_base->int_bld, pipe_func,
emit_data->args[0], emit_data->args[1]);
- emit_data->output[emit_data->chan] = lp_build_select(&bld_base->int_bld,
- cond,
- nz,
- bld_base->int_bld.zero);
+ emit_data->output[emit_data->chan] = cond;
}
/* TGSI_OPCODE_IMAX (CPU Only) */
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_shr(&bld_base->int_bld,
- emit_data->args[0], emit_data->args[1]);
+ struct lp_build_context *int_bld = &bld_base->int_bld;
+ LLVMValueRef mask = lp_build_const_vec(int_bld->gallivm, int_bld->type,
+ int_bld->type.width - 1);
+ LLVMValueRef masked_count = lp_build_and(int_bld, emit_data->args[1], mask);
+ emit_data->output[emit_data->chan] = lp_build_shr(int_bld, emit_data->args[0],
+ masked_count);
}
/* TGSI_OPCODE_ISLT (CPU Only) */
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_log2(&bld_base->base,
- emit_data->args[0]);
+ emit_data->output[emit_data->chan] = lp_build_log2_safe(&bld_base->base,
+ emit_data->args[0]);
}
/* TGSI_OPCODE_LOG (CPU Only) */
LLVMValueRef src0 = emit_data->args[0];
lp_build_log2_approx(&bld_base->base, src0,
- &p_exp, &p_floor_log2, &p_log2);
+ &p_exp, &p_floor_log2, &p_log2, FALSE);
emit_data->output[TGSI_CHAN_X] = p_floor_log2;
}
+/* TGSI_OPCODE_MAD (CPU Only) */
+
+static void
+mad_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] =
+ lp_build_mad(&bld_base->base,
+ emit_data->args[0], emit_data->args[1], emit_data->args[2]);
+}
+
/* TGSI_OPCODE_MAX (CPU Only) */
static void
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_max(&bld_base->base,
- emit_data->args[0], emit_data->args[1]);
+ emit_data->output[emit_data->chan] =
+ lp_build_max_ext(&bld_base->base,
+ emit_data->args[0], emit_data->args[1],
+ GALLIVM_NAN_RETURN_OTHER);
}
/* TGSI_OPCODE_MIN (CPU Only) */
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_min(&bld_base->base,
- emit_data->args[0], emit_data->args[1]);
+ emit_data->output[emit_data->chan] =
+ lp_build_min_ext(&bld_base->base,
+ emit_data->args[0], emit_data->args[1],
+ GALLIVM_NAN_RETURN_OTHER);
}
/* TGSI_OPCODE_MOD (CPU Only) */
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_mod(&bld_base->int_bld,
- emit_data->args[0], emit_data->args[1]);
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef div_mask = lp_build_cmp(&bld_base->uint_bld,
+ PIPE_FUNC_EQUAL, emit_data->args[1],
+ bld_base->uint_bld.zero);
+ /* We want to make sure that we never divide/mod by zero to not
+ * generate sigfpe. We don't want to crash just because the
+ * shader is doing something weird. */
+ LLVMValueRef divisor = LLVMBuildOr(builder,
+ div_mask,
+ emit_data->args[1], "");
+ LLVMValueRef result = lp_build_mod(&bld_base->int_bld,
+ emit_data->args[0], divisor);
+ /* umod by zero doesn't have a guaranteed return value chose -1 for now. */
+ emit_data->output[emit_data->chan] = LLVMBuildOr(builder,
+ div_mask,
+ result, "");
}
/* TGSI_OPCODE_NOT */
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_not(&bld_base->base,
+ emit_data->output[emit_data->chan] = lp_build_not(&bld_base->uint_bld,
emit_data->args[0]);
}
}
/* Reciprical squareroot (CPU Only) */
-
-/* This is not the same as TGSI_OPCODE_RSQ, which requres the argument to be
- * greater than or equal to 0 */
static void
recip_sqrt_emit_cpu(
const struct lp_build_tgsi_action * action,
struct lp_build_emit_data * emit_data,
unsigned pipe_func)
{
- LLVMValueRef cond = lp_build_cmp(&bld_base->base, pipe_func,
- emit_data->args[0], emit_data->args[1]);
+ LLVMValueRef cond;
+
+ if (pipe_func != PIPE_FUNC_NOTEQUAL) {
+ cond = lp_build_cmp_ordered(&bld_base->base, pipe_func,
+ emit_data->args[0], emit_data->args[1]);
+ }
+ else {
+ cond = lp_build_cmp(&bld_base->base, pipe_func,
+ emit_data->args[0], emit_data->args[1]);
+
+ }
emit_data->output[emit_data->chan] = lp_build_select(&bld_base->base,
cond,
bld_base->base.one,
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_shl(&bld_base->uint_bld,
- emit_data->args[0], emit_data->args[1]);
+ struct lp_build_context *uint_bld = &bld_base->uint_bld;
+ LLVMValueRef mask = lp_build_const_vec(uint_bld->gallivm, uint_bld->type,
+ uint_bld->type.width - 1);
+ LLVMValueRef masked_count = lp_build_and(uint_bld, emit_data->args[1], mask);
+ emit_data->output[emit_data->chan] = lp_build_shl(uint_bld, emit_data->args[0],
+ masked_count);
}
/* TGSI_OPCODE_SIN (CPU Only) */
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_div(&bld_base->uint_bld,
- emit_data->args[0], emit_data->args[1]);
+
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef div_mask = lp_build_cmp(&bld_base->uint_bld,
+ PIPE_FUNC_EQUAL, emit_data->args[1],
+ bld_base->uint_bld.zero);
+ /* We want to make sure that we never divide/mod by zero to not
+ * generate sigfpe. We don't want to crash just because the
+ * shader is doing something weird. */
+ LLVMValueRef divisor = LLVMBuildOr(builder,
+ div_mask,
+ emit_data->args[1], "");
+ LLVMValueRef result = lp_build_div(&bld_base->uint_bld,
+ emit_data->args[0], divisor);
+ /* udiv by zero is guaranteed to return 0xffffffff at least with d3d10 */
+ emit_data->output[emit_data->chan] = LLVMBuildOr(builder,
+ div_mask,
+ result, "");
}
/* TGSI_OPCODE_UMAX (CPU Only) */
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_mod(&bld_base->uint_bld,
- emit_data->args[0], emit_data->args[1]);
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef div_mask = lp_build_cmp(&bld_base->uint_bld,
+ PIPE_FUNC_EQUAL, emit_data->args[1],
+ bld_base->uint_bld.zero);
+ /* We want to make sure that we never divide/mod by zero to not
+ * generate sigfpe. We don't want to crash just because the
+ * shader is doing something weird. */
+ LLVMValueRef divisor = LLVMBuildOr(builder,
+ div_mask,
+ emit_data->args[1], "");
+ LLVMValueRef result = lp_build_mod(&bld_base->uint_bld,
+ emit_data->args[0], divisor);
+ /* umod by zero is guaranteed to return 0xffffffff */
+ emit_data->output[emit_data->chan] = LLVMBuildOr(builder,
+ div_mask,
+ result, "");
}
/* TGSI_OPCODE_USET Helper (CPU Only) */
struct lp_build_emit_data * emit_data,
unsigned pipe_func)
{
- LLVMValueRef nz = lp_build_const_vec(bld_base->base.gallivm,
- bld_base->uint_bld.type, ~0U);
LLVMValueRef cond = lp_build_cmp(&bld_base->uint_bld, pipe_func,
emit_data->args[0], emit_data->args[1]);
- emit_data->output[emit_data->chan] = lp_build_select(&bld_base->uint_bld,
- cond,
- nz,
- bld_base->uint_bld.zero);
+ emit_data->output[emit_data->chan] = cond;
}
struct lp_build_tgsi_context * bld_base,
struct lp_build_emit_data * emit_data)
{
- emit_data->output[emit_data->chan] = lp_build_shr(&bld_base->uint_bld,
- emit_data->args[0], emit_data->args[1]);
+ struct lp_build_context *uint_bld = &bld_base->uint_bld;
+ LLVMValueRef mask = lp_build_const_vec(uint_bld->gallivm, uint_bld->type,
+ uint_bld->type.width - 1);
+ LLVMValueRef masked_count = lp_build_and(uint_bld, emit_data->args[1], mask);
+ emit_data->output[emit_data->chan] = lp_build_shr(uint_bld, emit_data->args[0],
+ masked_count);
}
/* TGSI_OPCODE_ISLT (CPU Only) */
emit_data->args[1]);
}
+/* TGSI_OPCODE_DABS (CPU Only) */
+static void
+dabs_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_abs(&bld_base->dbl_bld,
+ emit_data->args[0]);
+}
+
+/* TGSI_OPCODE_DNEG (CPU Only) */
+static void
+dneg_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_sub(&bld_base->dbl_bld,
+ bld_base->dbl_bld.zero,
+ emit_data->args[0]);
+}
+
+/* TGSI_OPCODE_DSET Helper (CPU Only) */
+static void
+dset_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data,
+ unsigned pipe_func)
+{
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef cond = lp_build_cmp(&bld_base->dbl_bld, pipe_func,
+ emit_data->args[0], emit_data->args[1]);
+ /* arguments were 64 bit but store as 32 bit */
+ cond = LLVMBuildTrunc(builder, cond, bld_base->int_bld.int_vec_type, "");
+ emit_data->output[emit_data->chan] = cond;
+}
+
+/* TGSI_OPCODE_DSEQ (CPU Only) */
+static void
+dseq_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ dset_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_EQUAL);
+}
+
+/* TGSI_OPCODE_DSGE (CPU Only) */
+static void
+dsge_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ dset_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_GEQUAL);
+}
+
+/* TGSI_OPCODE_DSLT (CPU Only) */
+static void
+dslt_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ dset_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_LESS);
+}
+
+/* TGSI_OPCODE_DSNE (CPU Only) */
+static void
+dsne_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ dset_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_NOTEQUAL);
+}
+
+/* Double Reciprocal squareroot (CPU Only) */
+static void
+drecip_sqrt_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_rsqrt(&bld_base->dbl_bld,
+ emit_data->args[0]);
+}
+
+/* Double Squareroot (CPU Only) */
+static void
+dsqrt_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_sqrt(&bld_base->dbl_bld,
+ emit_data->args[0]);
+}
+
+static void
+i64abs_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_abs(&bld_base->int64_bld,
+ emit_data->args[0]);
+}
+
+static void
+i64ssg_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_sgn(&bld_base->int64_bld,
+ emit_data->args[0]);
+}
+
+static void
+i64neg_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_sub(&bld_base->int64_bld,
+ bld_base->int64_bld.zero,
+ emit_data->args[0]);
+}
+
+static void
+u64set_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data,
+ unsigned pipe_func)
+{
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef cond = lp_build_cmp(&bld_base->uint64_bld, pipe_func,
+ emit_data->args[0], emit_data->args[1]);
+ /* arguments were 64 bit but store as 32 bit */
+ cond = LLVMBuildTrunc(builder, cond, bld_base->int_bld.int_vec_type, "");
+ emit_data->output[emit_data->chan] = cond;
+}
+
+static void
+u64seq_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ u64set_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_EQUAL);
+}
+
+static void
+u64sne_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ u64set_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_NOTEQUAL);
+}
+
+static void
+u64slt_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ u64set_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_LESS);
+}
+
+static void
+u64sge_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ u64set_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_GEQUAL);
+}
+
+static void
+i64set_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data,
+ unsigned pipe_func)
+{
+ LLVMBuilderRef builder = bld_base->base.gallivm->builder;
+ LLVMValueRef cond = lp_build_cmp(&bld_base->int64_bld, pipe_func,
+ emit_data->args[0], emit_data->args[1]);
+ /* arguments were 64 bit but store as 32 bit */
+ cond = LLVMBuildTrunc(builder, cond, bld_base->int_bld.int_vec_type, "");
+ emit_data->output[emit_data->chan] = cond;
+}
+
+static void
+i64slt_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ i64set_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_LESS);
+}
+
+static void
+i64sge_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ i64set_emit_cpu(action, bld_base, emit_data, PIPE_FUNC_GEQUAL);
+}
+
+static void
+u64max_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_max(&bld_base->uint64_bld,
+ emit_data->args[0], emit_data->args[1]);
+}
+
+static void
+u64min_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_min(&bld_base->uint64_bld,
+ emit_data->args[0], emit_data->args[1]);
+}
+
+static void
+i64max_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_max(&bld_base->int64_bld,
+ emit_data->args[0], emit_data->args[1]);
+}
+
+static void
+i64min_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_min(&bld_base->int64_bld,
+ emit_data->args[0], emit_data->args[1]);
+}
+
+static void
+u64add_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ emit_data->output[emit_data->chan] = lp_build_add(&bld_base->uint64_bld,
+ emit_data->args[0], emit_data->args[1]);
+}
+
+static void
+u64shl_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ struct lp_build_context *uint_bld = &bld_base->uint64_bld;
+ LLVMValueRef mask = lp_build_const_vec(uint_bld->gallivm, uint_bld->type,
+ uint_bld->type.width - 1);
+ LLVMValueRef masked_count = lp_build_and(uint_bld, emit_data->args[1], mask);
+ emit_data->output[emit_data->chan] = lp_build_shl(uint_bld, emit_data->args[0],
+ masked_count);
+}
+
+static void
+i64shr_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ struct lp_build_context *int_bld = &bld_base->int64_bld;
+ LLVMValueRef mask = lp_build_const_vec(int_bld->gallivm, int_bld->type,
+ int_bld->type.width - 1);
+ LLVMValueRef masked_count = lp_build_and(int_bld, emit_data->args[1], mask);
+ emit_data->output[emit_data->chan] = lp_build_shr(int_bld, emit_data->args[0],
+ masked_count);
+}
+
+static void
+u64shr_emit_cpu(
+ const struct lp_build_tgsi_action * action,
+ struct lp_build_tgsi_context * bld_base,
+ struct lp_build_emit_data * emit_data)
+{
+ struct lp_build_context *uint_bld = &bld_base->uint64_bld;
+ LLVMValueRef mask = lp_build_const_vec(uint_bld->gallivm, uint_bld->type,
+ uint_bld->type.width - 1);
+ LLVMValueRef masked_count = lp_build_and(uint_bld, emit_data->args[1], mask);
+ emit_data->output[emit_data->chan] = lp_build_shr(uint_bld, emit_data->args[0],
+ masked_count);
+}
+
void
lp_set_default_actions_cpu(
struct lp_build_tgsi_context * bld_base)
{
lp_set_default_actions(bld_base);
- bld_base->op_actions[TGSI_OPCODE_ABS].emit = abs_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_ADD].emit = add_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_AND].emit = and_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_ARL].emit = arl_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_ARR].emit = arr_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_CEIL].emit = ceil_emit_cpu;
- bld_base->op_actions[TGSI_OPCODE_CND].emit = cnd_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_COS].emit = cos_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_CMP].emit = cmp_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_DIV].emit = div_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_EX2].emit = ex2_emit_cpu;
- bld_base->op_actions[TGSI_OPCODE_EXP].emit = exp_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_F2I].emit = f2i_emit_cpu;
- bld_base->op_actions[TGSI_OPCODE_F2U].emit = f2u_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_FLR].emit = flr_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_FSEQ].emit = fseq_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_FSGE].emit = fsge_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_FSLT].emit = fslt_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_FSNE].emit = fsne_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_I2F].emit = i2f_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_IABS].emit = iabs_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_ISHR].emit = ishr_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_ISLT].emit = islt_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_ISSG].emit = issg_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_IMUL_HI].emit = imul_hi_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_UMUL_HI].emit = umul_hi_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_LG2].emit = lg2_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_LOG].emit = log_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_MAD].emit = mad_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_MAX].emit = max_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_MIN].emit = min_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_MOD].emit = mod_emit_cpu;
bld_base->sqrt_action.emit = sqrt_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_UADD].emit = uadd_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_UCMP].emit = ucmp_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_UDIV].emit = udiv_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_UMAX].emit = umax_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_UMIN].emit = umin_emit_cpu;
bld_base->op_actions[TGSI_OPCODE_XOR].emit = xor_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_DABS].emit = dabs_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_DNEG].emit = dneg_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_DSEQ].emit = dseq_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_DSGE].emit = dsge_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_DSLT].emit = dslt_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_DSNE].emit = dsne_emit_cpu;
+
+ bld_base->op_actions[TGSI_OPCODE_DRSQ].emit = drecip_sqrt_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_DSQRT].emit = dsqrt_emit_cpu;
+
+ bld_base->op_actions[TGSI_OPCODE_I64ABS].emit = i64abs_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_I64SSG].emit = i64ssg_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_I64NEG].emit = i64neg_emit_cpu;
+
+ bld_base->op_actions[TGSI_OPCODE_U64SEQ].emit = u64seq_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_U64SNE].emit = u64sne_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_U64SLT].emit = u64slt_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_U64SGE].emit = u64sge_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_I64SLT].emit = i64slt_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_I64SGE].emit = i64sge_emit_cpu;
+
+ bld_base->op_actions[TGSI_OPCODE_U64MIN].emit = u64min_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_U64MAX].emit = u64max_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_I64MIN].emit = i64min_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_I64MAX].emit = i64max_emit_cpu;
+
+ bld_base->op_actions[TGSI_OPCODE_U64ADD].emit = u64add_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_U64MOD].emit = u64mod_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_I64MOD].emit = i64mod_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_U64DIV].emit = u64div_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_I64DIV].emit = i64div_emit_cpu;
+
+ bld_base->op_actions[TGSI_OPCODE_U64SHL].emit = u64shl_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_I64SHR].emit = i64shr_emit_cpu;
+ bld_base->op_actions[TGSI_OPCODE_U64SHR].emit = u64shr_emit_cpu;
}
projects / mesa.git / blobdiff