aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Add type suffix to comparison ops in C++ output.

Browse Source 
e.g. "__equal()" -> "__equal_float()", etc.

No functional change; this is necessary groundwork for a forthcoming
peephole optimization that eliminates ANDs of masks in some cases.
This commit is contained in:
Matt Pharr
2012-07-07 07:50:59 -07:00
parent 45e9e0be0b
commit 974b40c8af
6 changed files with 284 additions and 255 deletions
Download Patch File Download Diff File Expand all files Collapse all files

110
examples/intrinsics/generic-32.h
View File

@@ -258,8 +258,8 @@ static FORCEINLINE TYPE NAME(TYPE a, TYPE b) { \
return ret; \
}
#define CMP_OP(TYPE, CAST, NAME, OP) \
static FORCEINLINE __vec32_i1 NAME(TYPE a, TYPE b) { \
#define CMP_OP(TYPE, SUFFIX, CAST, NAME, OP) \
static FORCEINLINE __vec32_i1 NAME##_##SUFFIX(TYPE a, TYPE b) { \
__vec32_i1 ret; \
ret.v = 0; \
for (int i = 0; i < 32; ++i) \
@@ -382,7 +382,7 @@ static FORCEINLINE uint64_t __movmsk(__vec32_i1 mask) {
return (uint64_t)mask.v;
}
static FORCEINLINE __vec32_i1 __equal(__vec32_i1 a, __vec32_i1 b) {
static FORCEINLINE __vec32_i1 __equal_i1(__vec32_i1 a, __vec32_i1 b) {
__vec32_i1 r;
r.v = (a.v & b.v) | (~a.v & ~b.v);
return r;
@@ -491,16 +491,16 @@ SHIFT_UNIFORM(__vec32_i8, uint8_t, __lshr, >>)
SHIFT_UNIFORM(__vec32_i8, int8_t, __ashr, >>)
SHIFT_UNIFORM(__vec32_i8, int8_t, __shl, <<)
CMP_OP(__vec32_i8, int8_t, __equal, ==)
CMP_OP(__vec32_i8, int8_t, __not_equal, !=)
CMP_OP(__vec32_i8, uint8_t, __unsigned_less_equal, <=)
CMP_OP(__vec32_i8, int8_t, __signed_less_equal, <=)
CMP_OP(__vec32_i8, uint8_t, __unsigned_greater_equal, >=)
CMP_OP(__vec32_i8, int8_t, __signed_greater_equal, >=)
CMP_OP(__vec32_i8, uint8_t, __unsigned_less_than, <)
CMP_OP(__vec32_i8, int8_t, __signed_less_than, <)
CMP_OP(__vec32_i8, uint8_t, __unsigned_greater_than, >)
CMP_OP(__vec32_i8, int8_t, __signed_greater_than, >)
CMP_OP(__vec32_i8, i8, int8_t, __equal, ==)
CMP_OP(__vec32_i8, i8, int8_t, __not_equal, !=)
CMP_OP(__vec32_i8, i8, uint8_t, __unsigned_less_equal, <=)
CMP_OP(__vec32_i8, i8, int8_t, __signed_less_equal, <=)
CMP_OP(__vec32_i8, i8, uint8_t, __unsigned_greater_equal, >=)
CMP_OP(__vec32_i8, i8, int8_t, __signed_greater_equal, >=)
CMP_OP(__vec32_i8, i8, uint8_t, __unsigned_less_than, <)
CMP_OP(__vec32_i8, i8, int8_t, __signed_less_than, <)
CMP_OP(__vec32_i8, i8, uint8_t, __unsigned_greater_than, >)
CMP_OP(__vec32_i8, i8, int8_t, __signed_greater_than, >)
SELECT(__vec32_i8)
INSERT_EXTRACT(__vec32_i8, int8_t)
@@ -534,16 +534,16 @@ SHIFT_UNIFORM(__vec32_i16, uint16_t, __lshr, >>)
SHIFT_UNIFORM(__vec32_i16, int16_t, __ashr, >>)
SHIFT_UNIFORM(__vec32_i16, int16_t, __shl, <<)
CMP_OP(__vec32_i16, int16_t, __equal, ==)
CMP_OP(__vec32_i16, int16_t, __not_equal, !=)
CMP_OP(__vec32_i16, uint16_t, __unsigned_less_equal, <=)
CMP_OP(__vec32_i16, int16_t, __signed_less_equal, <=)
CMP_OP(__vec32_i16, uint16_t, __unsigned_greater_equal, >=)
CMP_OP(__vec32_i16, int16_t, __signed_greater_equal, >=)
CMP_OP(__vec32_i16, uint16_t, __unsigned_less_than, <)
CMP_OP(__vec32_i16, int16_t, __signed_less_than, <)
CMP_OP(__vec32_i16, uint16_t, __unsigned_greater_than, >)
CMP_OP(__vec32_i16, int16_t, __signed_greater_than, >)
CMP_OP(__vec32_i16, i16, int16_t, __equal, ==)
CMP_OP(__vec32_i16, i16, int16_t, __not_equal, !=)
CMP_OP(__vec32_i16, i16, uint16_t, __unsigned_less_equal, <=)
CMP_OP(__vec32_i16, i16, int16_t, __signed_less_equal, <=)
CMP_OP(__vec32_i16, i16, uint16_t, __unsigned_greater_equal, >=)
CMP_OP(__vec32_i16, i16, int16_t, __signed_greater_equal, >=)
CMP_OP(__vec32_i16, i16, uint16_t, __unsigned_less_than, <)
CMP_OP(__vec32_i16, i16, int16_t, __signed_less_than, <)
CMP_OP(__vec32_i16, i16, uint16_t, __unsigned_greater_than, >)
CMP_OP(__vec32_i16, i16, int16_t, __signed_greater_than, >)
SELECT(__vec32_i16)
INSERT_EXTRACT(__vec32_i16, int16_t)
@@ -577,16 +577,16 @@ SHIFT_UNIFORM(__vec32_i32, uint32_t, __lshr, >>)
SHIFT_UNIFORM(__vec32_i32, int32_t, __ashr, >>)
SHIFT_UNIFORM(__vec32_i32, int32_t, __shl, <<)
CMP_OP(__vec32_i32, int32_t, __equal, ==)
CMP_OP(__vec32_i32, int32_t, __not_equal, !=)
CMP_OP(__vec32_i32, uint32_t, __unsigned_less_equal, <=)
CMP_OP(__vec32_i32, int32_t, __signed_less_equal, <=)
CMP_OP(__vec32_i32, uint32_t, __unsigned_greater_equal, >=)
CMP_OP(__vec32_i32, int32_t, __signed_greater_equal, >=)
CMP_OP(__vec32_i32, uint32_t, __unsigned_less_than, <)
CMP_OP(__vec32_i32, int32_t, __signed_less_than, <)
CMP_OP(__vec32_i32, uint32_t, __unsigned_greater_than, >)
CMP_OP(__vec32_i32, int32_t, __signed_greater_than, >)
CMP_OP(__vec32_i32, i32, int32_t, __equal, ==)
CMP_OP(__vec32_i32, i32, int32_t, __not_equal, !=)
CMP_OP(__vec32_i32, i32, uint32_t, __unsigned_less_equal, <=)
CMP_OP(__vec32_i32, i32, int32_t, __signed_less_equal, <=)
CMP_OP(__vec32_i32, i32, uint32_t, __unsigned_greater_equal, >=)
CMP_OP(__vec32_i32, i32, int32_t, __signed_greater_equal, >=)
CMP_OP(__vec32_i32, i32, uint32_t, __unsigned_less_than, <)
CMP_OP(__vec32_i32, i32, int32_t, __signed_less_than, <)
CMP_OP(__vec32_i32, i32, uint32_t, __unsigned_greater_than, >)
CMP_OP(__vec32_i32, i32, int32_t, __signed_greater_than, >)
SELECT(__vec32_i32)
INSERT_EXTRACT(__vec32_i32, int32_t)
@@ -620,16 +620,16 @@ SHIFT_UNIFORM(__vec32_i64, uint64_t, __lshr, >>)
SHIFT_UNIFORM(__vec32_i64, int64_t, __ashr, >>)
SHIFT_UNIFORM(__vec32_i64, int64_t, __shl, <<)
CMP_OP(__vec32_i64, int64_t, __equal, ==)
CMP_OP(__vec32_i64, int64_t, __not_equal, !=)
CMP_OP(__vec32_i64, uint64_t, __unsigned_less_equal, <=)
CMP_OP(__vec32_i64, int64_t, __signed_less_equal, <=)
CMP_OP(__vec32_i64, uint64_t, __unsigned_greater_equal, >=)
CMP_OP(__vec32_i64, int64_t, __signed_greater_equal, >=)
CMP_OP(__vec32_i64, uint64_t, __unsigned_less_than, <)
CMP_OP(__vec32_i64, int64_t, __signed_less_than, <)
CMP_OP(__vec32_i64, uint64_t, __unsigned_greater_than, >)
CMP_OP(__vec32_i64, int64_t, __signed_greater_than, >)
CMP_OP(__vec32_i64, i64, int64_t, __equal, ==)
CMP_OP(__vec32_i64, i64, int64_t, __not_equal, !=)
CMP_OP(__vec32_i64, i64, uint64_t, __unsigned_less_equal, <=)
CMP_OP(__vec32_i64, i64, int64_t, __signed_less_equal, <=)
CMP_OP(__vec32_i64, i64, uint64_t, __unsigned_greater_equal, >=)
CMP_OP(__vec32_i64, i64, int64_t, __signed_greater_equal, >=)
CMP_OP(__vec32_i64, i64, uint64_t, __unsigned_less_than, <)
CMP_OP(__vec32_i64, i64, int64_t, __signed_less_than, <)
CMP_OP(__vec32_i64, i64, uint64_t, __unsigned_greater_than, >)
CMP_OP(__vec32_i64, i64, int64_t, __signed_greater_than, >)
SELECT(__vec32_i64)
INSERT_EXTRACT(__vec32_i64, int64_t)
@@ -647,12 +647,12 @@ BINARY_OP(__vec32_f, __sub, -)
BINARY_OP(__vec32_f, __mul, *)
BINARY_OP(__vec32_f, __div, /)
CMP_OP(__vec32_f, float, __equal, ==)
CMP_OP(__vec32_f, float, __not_equal, !=)
CMP_OP(__vec32_f, float, __less_than, <)
CMP_OP(__vec32_f, float, __less_equal, <=)
CMP_OP(__vec32_f, float, __greater_than, >)
CMP_OP(__vec32_f, float, __greater_equal, >=)
CMP_OP(__vec32_f, float, float, __equal, ==)
CMP_OP(__vec32_f, float, float, __not_equal, !=)
CMP_OP(__vec32_f, float, float, __less_than, <)
CMP_OP(__vec32_f, float, float, __less_equal, <=)
CMP_OP(__vec32_f, float, float, __greater_than, >)
CMP_OP(__vec32_f, float, float, __greater_equal, >=)
static FORCEINLINE __vec32_i1 __ordered(__vec32_f a, __vec32_f b) {
__vec32_i1 ret;
@@ -797,12 +797,12 @@ BINARY_OP(__vec32_d, __sub, -)
BINARY_OP(__vec32_d, __mul, *)
BINARY_OP(__vec32_d, __div, /)
CMP_OP(__vec32_d, double, __equal, ==)
CMP_OP(__vec32_d, double, __not_equal, !=)
CMP_OP(__vec32_d, double, __less_than, <)
CMP_OP(__vec32_d, double, __less_equal, <=)
CMP_OP(__vec32_d, double, __greater_than, >)
CMP_OP(__vec32_d, double, __greater_equal, >=)
CMP_OP(__vec32_d, double, double, __equal, ==)
CMP_OP(__vec32_d, double, double, __not_equal, !=)
CMP_OP(__vec32_d, double, double, __less_than, <)
CMP_OP(__vec32_d, double, double, __less_equal, <=)
CMP_OP(__vec32_d, double, double, __greater_than, >)
CMP_OP(__vec32_d, double, double, __greater_equal, >=)
static FORCEINLINE __vec32_i1 __ordered(__vec32_d a, __vec32_d b) {
__vec32_i1 ret;