tfloat = "float"
tint = "int"
tbool = "bool"
-tunsigned = "unsigned"
+tuint = "uint"
commutative = "commutative "
associative = "associative "
unop("frcp", tfloat, "1.0f / src0")
unop("frsq", tfloat, "1.0f / sqrtf(src0)")
unop("fsqrt", tfloat, "sqrtf(src0)")
-unop("fexp", tfloat, "expf(src0)") # < e^x
-unop("flog", tfloat, "logf(src0)") # log base e
unop("fexp2", tfloat, "exp2f(src0)")
unop("flog2", tfloat, "log2f(src0)")
unop_convert("f2i", tfloat, tint, "src0") # Float-to-integer conversion.
-unop_convert("f2u", tfloat, tunsigned, "src0") # Float-to-unsigned conversion
+unop_convert("f2u", tfloat, tuint, "src0") # Float-to-unsigned conversion
unop_convert("i2f", tint, tfloat, "src0") # Integer-to-float conversion.
# Float-to-boolean conversion
-unop_convert("f2b", tfloat, tbool, "src0 == 0.0f")
+unop_convert("f2b", tfloat, tbool, "src0 != 0.0f")
# Boolean-to-float conversion
unop_convert("b2f", tbool, tfloat, "src0 ? 1.0f : 0.0f")
# Int-to-boolean conversion
-unop_convert("i2b", tint, tbool, "src0 == 0")
-unop_convert("b2i", tbool, tint, "src0 ? 0 : -1") # Boolean-to-int conversion
-unop_convert("u2f", tunsigned, tfloat, "src0") #Unsigned-to-float conversion.
-
-unop_reduce("bany", 1, tbool, tbool, "{src}", "{src0} || {src1}", "{src}")
-unop_reduce("ball", 1, tbool, tbool, "{src}", "{src0} && {src1}", "{src}")
-unop_reduce("fany", 1, tfloat, tfloat, "{src} != 0.0f", "{src0} || {src1}",
- "{src} ? 1.0f : 0.0f")
-unop_reduce("fall", 1, tfloat, tfloat, "{src} != 0.0f", "{src0} && {src1}",
- "{src} ? 1.0f : 0.0f")
+unop_convert("i2b", tint, tbool, "src0 != 0")
+unop_convert("b2i", tbool, tint, "src0 ? 1 : 0") # Boolean-to-int conversion
+unop_convert("u2f", tuint, tfloat, "src0") # Unsigned-to-float conversion.
# Unary floating-point rounding operations.
unop("fceil", tfloat, "ceilf(src0)")
unop("ffloor", tfloat, "floorf(src0)")
unop("ffract", tfloat, "src0 - floorf(src0)")
-unop("fround_even", tfloat, "_mesa_round_to_even(src0)")
+unop("fround_even", tfloat, "_mesa_roundevenf(src0)")
# Trigonometric operations.
unop("fsin", tfloat, "sinf(src0)")
unop("fcos", tfloat, "cosf(src0)")
-unop("fsin_reduced", tfloat, "sinf(src0)")
-unop("fcos_reduced", tfloat, "cosf(src0)")
# Partial derivatives.
# Floating point pack and unpack operations.
def pack_2x16(fmt):
- unop_horiz("pack_" + fmt + "_2x16", 1, tunsigned, 2, tfloat, """
+ unop_horiz("pack_" + fmt + "_2x16", 1, tuint, 2, tfloat, """
dst.x = (uint32_t) pack_fmt_1x16(src0.x);
dst.x |= ((uint32_t) pack_fmt_1x16(src0.y)) << 16;
""".replace("fmt", fmt))
def pack_4x8(fmt):
- unop_horiz("pack_" + fmt + "_4x8", 1, tunsigned, 4, tfloat, """
+ unop_horiz("pack_" + fmt + "_4x8", 1, tuint, 4, tfloat, """
dst.x = (uint32_t) pack_fmt_1x8(src0.x);
dst.x |= ((uint32_t) pack_fmt_1x8(src0.y)) << 8;
dst.x |= ((uint32_t) pack_fmt_1x8(src0.z)) << 16;
""".replace("fmt", fmt))
def unpack_2x16(fmt):
- unop_horiz("unpack_" + fmt + "_2x16", 2, tfloat, 1, tunsigned, """
+ unop_horiz("unpack_" + fmt + "_2x16", 2, tfloat, 1, tuint, """
dst.x = unpack_fmt_1x16((uint16_t)(src0.x & 0xffff));
dst.y = unpack_fmt_1x16((uint16_t)(src0.x << 16));
""".replace("fmt", fmt))
def unpack_4x8(fmt):
- unop_horiz("unpack_" + fmt + "_4x8", 4, tfloat, 1, tunsigned, """
+ unop_horiz("unpack_" + fmt + "_4x8", 4, tfloat, 1, tuint, """
dst.x = unpack_fmt_1x8((uint8_t)(src0.x & 0xff));
dst.y = unpack_fmt_1x8((uint8_t)((src0.x >> 8) & 0xff));
dst.z = unpack_fmt_1x8((uint8_t)((src0.x >> 16) & 0xff));
# Lowered floating point unpacking operations.
-unop_horiz("unpack_half_2x16_split_x", 1, tfloat, 1, tunsigned,
+unop_horiz("unpack_half_2x16_split_x", 1, tfloat, 1, tuint,
"unpack_half_1x16((uint16_t)(src0.x & 0xffff))")
-unop_horiz("unpack_half_2x16_split_y", 1, tfloat, 1, tunsigned,
+unop_horiz("unpack_half_2x16_split_y", 1, tfloat, 1, tuint,
"unpack_half_1x16((uint16_t)(src0.x >> 16))")
# Bit operations, part of ARB_gpu_shader5.
-unop("bitfield_reverse", tunsigned, """
+unop("bitfield_reverse", tuint, """
/* we're not winning any awards for speed here, but that's ok */
dst = 0;
for (unsigned bit = 0; bit < 32; bit++)
dst |= ((src0 >> bit) & 1) << (31 - bit);
""")
-unop("bit_count", tunsigned, """
+unop("bit_count", tuint, """
dst = 0;
for (unsigned bit = 0; bit < 32; bit++) {
if ((src0 >> bit) & 1)
}
""")
-unop_convert("ufind_msb", tunsigned, tint, """
+unop_convert("ufind_msb", tuint, tint, """
dst = -1;
for (int bit = 31; bit > 0; bit--) {
if ((src0 >> bit) & 1) {
binop("imul_high", tint, commutative,
"(int32_t)(((int64_t) src0 * (int64_t) src1) >> 32)")
# high 32-bits of unsigned integer multiply
-binop("umul_high", tunsigned, commutative,
+binop("umul_high", tuint, commutative,
"(uint32_t)(((uint64_t) src0 * (uint64_t) src1) >> 32)")
binop("fdiv", tfloat, "", "src0 / src1")
binop("idiv", tint, "", "src0 / src1")
-binop("udiv", tunsigned, "", "src0 / src1")
+binop("udiv", tuint, "", "src0 / src1")
# returns a boolean representing the carry resulting from the addition of
# the two unsigned arguments.
-binop_convert("uadd_carry", tbool, tunsigned, commutative, "src0 + src1 < src0")
+binop_convert("uadd_carry", tuint, tuint, commutative, "src0 + src1 < src0")
# returns a boolean representing the borrow resulting from the subtraction
# of the two unsigned arguments.
-binop_convert("usub_borrow", tbool, tunsigned, "", "src1 < src0")
+binop_convert("usub_borrow", tuint, tuint, "", "src0 < src1")
binop("fmod", tfloat, "", "src0 - src1 * floorf(src0 / src1)")
-binop("umod", tunsigned, "", "src1 == 0 ? 0 : src0 % src1")
+binop("umod", tuint, "", "src1 == 0 ? 0 : src0 % src1")
#
# Comparisons
binop_compare("ige", tint, "", "src0 >= src1")
binop_compare("ieq", tint, commutative, "src0 == src1")
binop_compare("ine", tint, commutative, "src0 != src1")
-binop_compare("ult", tunsigned, "", "src0 < src1")
-binop_compare("uge", tunsigned, "", "src0 >= src1")
+binop_compare("ult", tuint, "", "src0 < src1")
+binop_compare("uge", tuint, "", "src0 >= src1")
# integer-aware GLSL-style comparisons that compare floats and ints
binop("ishl", tint, "", "src0 << src1")
binop("ishr", tint, "", "src0 >> src1")
-binop("ushr", tunsigned, "", "src0 >> src1")
+binop("ushr", tuint, "", "src0 >> src1")
# bitwise logic operators
#
# integers.
-binop("iand", tunsigned, commutative + associative, "src0 & src1")
-binop("ior", tunsigned, commutative + associative, "src0 | src1")
-binop("ixor", tunsigned, commutative + associative, "src0 ^ src1")
+binop("iand", tuint, commutative + associative, "src0 & src1")
+binop("ior", tuint, commutative + associative, "src0 | src1")
+binop("ixor", tuint, commutative + associative, "src0 ^ src1")
# floating point logic operators
binop_reduce("fdot", 1, tfloat, tfloat, "{src0} * {src1}", "{src0} + {src1}",
"{src}")
+binop_reduce("fdot_replicated", 4, tfloat, tfloat,
+ "{src0} * {src1}", "{src0} + {src1}", "{src}")
+
+opcode("fdph", 1, tfloat, [3, 4], [tfloat, tfloat], "",
+ "src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w")
+opcode("fdph_replicated", 4, tfloat, [3, 4], [tfloat, tfloat], "",
+ "src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w")
+
binop("fmin", tfloat, "", "fminf(src0, src1)")
binop("imin", tint, commutative + associative, "src1 > src0 ? src0 : src1")
-binop("umin", tunsigned, commutative + associative, "src1 > src0 ? src0 : src1")
+binop("umin", tuint, commutative + associative, "src1 > src0 ? src0 : src1")
binop("fmax", tfloat, "", "fmaxf(src0, src1)")
binop("imax", tint, commutative + associative, "src1 > src0 ? src1 : src0")
-binop("umax", tunsigned, commutative + associative, "src1 > src0 ? src1 : src0")
+binop("umax", tuint, commutative + associative, "src1 > src0 ? src1 : src0")
+
+# Saturated vector add for 4 8bit ints.
+binop("usadd_4x8", tint, commutative + associative, """
+dst = 0;
+for (int i = 0; i < 32; i += 8) {
+ dst |= MIN2(((src0 >> i) & 0xff) + ((src1 >> i) & 0xff), 0xff) << i;
+}
+""")
+
+# Saturated vector subtract for 4 8bit ints.
+binop("ussub_4x8", tint, "", """
+dst = 0;
+for (int i = 0; i < 32; i += 8) {
+ int src0_chan = (src0 >> i) & 0xff;
+ int src1_chan = (src1 >> i) & 0xff;
+ if (src0_chan > src1_chan)
+ dst |= (src0_chan - src1_chan) << i;
+}
+""")
+
+# vector min for 4 8bit ints.
+binop("umin_4x8", tint, commutative + associative, """
+dst = 0;
+for (int i = 0; i < 32; i += 8) {
+ dst |= MIN2((src0 >> i) & 0xff, (src1 >> i) & 0xff) << i;
+}
+""")
+
+# vector max for 4 8bit ints.
+binop("umax_4x8", tint, commutative + associative, """
+dst = 0;
+for (int i = 0; i < 32; i += 8) {
+ dst |= MAX2((src0 >> i) & 0xff, (src1 >> i) & 0xff) << i;
+}
+""")
+
+# unorm multiply: (a * b) / 255.
+binop("umul_unorm_4x8", tint, commutative + associative, """
+dst = 0;
+for (int i = 0; i < 32; i += 8) {
+ int src0_chan = (src0 >> i) & 0xff;
+ int src1_chan = (src1 >> i) & 0xff;
+ dst |= ((src0_chan * src1_chan) / 255) << i;
+}
+""")
binop("fpow", tfloat, "", "powf(src0, src1)")
-binop_horiz("pack_half_2x16_split", 1, tunsigned, 1, tfloat, 1, tfloat,
+binop_horiz("pack_half_2x16_split", 1, tuint, 1, tfloat, 1, tfloat,
"pack_half_1x16(src0.x) | (pack_half_1x16(src1.x) << 16)")
-binop_convert("bfm", tunsigned, tint, "", """
+binop_convert("bfm", tuint, tint, "", """
int offset = src0, bits = src1;
if (offset < 0 || bits < 0 || offset + bits > 32)
dst = 0; /* undefined per the spec */
dst = ((1 << bits)- 1) << offset;
""")
-opcode("ldexp", 0, tunsigned, [0, 0], [tfloat, tint], "", """
-dst = ldexp(src0, src1);
+opcode("ldexp", 0, tfloat, [0, 0], [tfloat, tint], "", """
+dst = ldexpf(src0, src1);
/* flush denormals to zero. */
-if (fpclassify(dst) != FP_NORMAL)
- dst = copysign(0.0f, src0);
+if (!isnormal(dst))
+ dst = copysignf(0.0f, src0);
""")
# Combines the first component of each input to make a 2-component vector.
-binop_horiz("vec2", 2, tunsigned, 1, tunsigned, 1, tunsigned, """
+binop_horiz("vec2", 2, tuint, 1, tuint, 1, tuint, """
dst.x = src0.x;
dst.y = src1.x;
""")
def triop(name, ty, const_expr):
opcode(name, 0, ty, [0, 0, 0], [ty, ty, ty], "", const_expr)
def triop_horiz(name, output_size, src1_size, src2_size, src3_size, const_expr):
- opcode(name, output_size, tunsigned,
+ opcode(name, output_size, tuint,
[src1_size, src2_size, src3_size],
- [tunsigned, tunsigned, tunsigned], "", const_expr)
+ [tuint, tuint, tuint], "", const_expr)
triop("ffma", tfloat, "src0 * src1 + src2")
triop("fcsel", tfloat, "(src0 != 0.0f) ? src1 : src2")
-opcode("bcsel", 0, tunsigned, [0, 0, 0],
- [tbool, tunsigned, tunsigned], "", "src0 ? src1 : src2")
+opcode("bcsel", 0, tuint, [0, 0, 0],
+ [tbool, tuint, tuint], "", "src0 ? src1 : src2")
-triop("bfi", tunsigned, """
-unsigned mask = src0, insert = src1 & mask, base = src2;
+triop("bfi", tuint, """
+unsigned mask = src0, insert = src1, base = src2;
if (mask == 0) {
dst = base;
} else {
tmp >>= 1;
insert <<= 1;
}
- dst = (base & ~mask) | insert;
+ dst = (base & ~mask) | (insert & mask);
}
""")
-opcode("ubitfield_extract", 0, tunsigned,
- [0, 1, 1], [tunsigned, tint, tint], "", """
+opcode("ubitfield_extract", 0, tuint,
+ [0, 1, 1], [tuint, tint, tint], "", """
unsigned base = src0;
int offset = src1.x, bits = src2.x;
if (bits == 0) {
def quadop_horiz(name, output_size, src1_size, src2_size, src3_size,
src4_size, const_expr):
- opcode(name, output_size, tunsigned,
+ opcode(name, output_size, tuint,
[src1_size, src2_size, src3_size, src4_size],
- [tunsigned, tunsigned, tunsigned, tunsigned],
+ [tuint, tuint, tuint, tuint],
"", const_expr)
-opcode("bitfield_insert", 0, tunsigned, [0, 0, 1, 1],
- [tunsigned, tunsigned, tint, tint], "", """
+opcode("bitfield_insert", 0, tuint, [0, 0, 1, 1],
+ [tuint, tuint, tint, tint], "", """
unsigned base = src0, insert = src1;
int offset = src2.x, bits = src3.x;
if (bits == 0) {