* [PATCH, rs6000] Signed builtin support
@ 2017-06-28 15:37 Carl Love
2017-06-28 22:05 ` Segher Boessenkool
0 siblings, 1 reply; 3+ messages in thread
From: Carl Love @ 2017-06-28 15:37 UTC (permalink / raw)
To: gcc-patches, David Edelsohn, Segher Boessenkool; +Cc: Bill Schmidt, cel
GCC Maintainers:
The following patch adds support for the vec_signed, vec_signede,
vec_signedo and vec_signed2 builtins. The patch has been tested on
powerpc64le-unknown-linux-gnu (Power 8 LE) and
powerpc64-unknown-linux-gnu(Power 8 BE).
Is the fix OK for gcc mainline?
Carl Love
----------------------------------------------------------------------
gcc/ChangeLog:
2017-06-28 Carl Love <cel@us.ibm.com>
* config/rs6000/rs6000-c.c: Add support for built-in functions
vector signed int vec_signed (vector float);
vector signed long long vec_signed (vector double);
vector signed int vec_signed2 (vector double, vector double);
vector signed int vec_signede (vector double);
vector signed int vec_signedo (vector double);
* config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
instruction generator.
* config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
UNSPEC_VSX_VSIGNED2): Add UNSPECS
(vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
Add define_insn.
(vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
vunsignede_v2df): Add define_expands.
* config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
VEC_UNSIGNEDO): Add definitions.
* config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
UNSPEC_VSX_VSIGNED2): Add UNSPECs.
(vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
(vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
* config/rs6000/altivec.h (vec_signed, vec_signed2,
vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
vec_unsignede, vec_unsignedo): Add builtin defines.
* config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
declaration.
* doc/extend.texi: Update the built-in documentation file for the
new built-in functions.
gcc/testsuite/ChangeLog:
2017-06-28 Carl Love <cel@us.ibm.com>
* gcc.target/powerpc/builtins-3-runnable.c(test_int_result,
test_unsigned_int_result, test_ll_int_result,
test_ll_unsigned_int_result): Add result checking functions, add
debug support.
(main): Add builtin function tests, .
Signed-off-by: Carl Love <carll@us.ibm.com>
---
gcc/config/rs6000/altivec.h | 8 +
gcc/config/rs6000/rs6000-builtin.def | 23 +++
gcc/config/rs6000/rs6000-c.c | 23 +++
gcc/config/rs6000/rs6000-protos.h | 1 +
gcc/config/rs6000/rs6000.c | 29 +++
gcc/config/rs6000/vsx.md | 194 +++++++++++++++++
gcc/doc/extend.texi | 14 ++
.../gcc.target/powerpc/builtins-3-runnable.c | 229 +++++++++++++++++++--
8 files changed, 502 insertions(+), 19 deletions(-)
diff --git a/gcc/config/rs6000/altivec.h b/gcc/config/rs6000/altivec.h
index d542315..417e143 100644
--- a/gcc/config/rs6000/altivec.h
+++ b/gcc/config/rs6000/altivec.h
@@ -181,6 +181,14 @@
#define vec_rlnm(a,b,c) (__builtin_vec_rlnm((a),((b)<<8)|(c)))
#define vec_rsqrt __builtin_vec_rsqrt
#define vec_rsqrte __builtin_vec_rsqrte
+#define vec_signed __builtin_vec_vsigned
+#define vec_signed2 __builtin_vec_vsigned2
+#define vec_signede __builtin_vec_vsignede
+#define vec_signedo __builtin_vec_vsignedo
+#define vec_unsigned __builtin_vec_vunsigned
+#define vec_unsigned2 __builtin_vec_vunsigned2
+#define vec_unsignede __builtin_vec_vunsignede
+#define vec_unsignedo __builtin_vec_vunsignedo
#define vec_vsubfp __builtin_vec_vsubfp
#define vec_subc __builtin_vec_subc
#define vec_vsubsws __builtin_vec_vsubsws
diff --git a/gcc/config/rs6000/rs6000-builtin.def b/gcc/config/rs6000/rs6000-builtin.def
index 963b9a8..9bdc2b4 100644
--- a/gcc/config/rs6000/rs6000-builtin.def
+++ b/gcc/config/rs6000/rs6000-builtin.def
@@ -1602,6 +1602,9 @@ BU_VSX_2 (CMPLE_U2DI, "cmple_u2di", CONST, vector_ngtuv2di)
BU_VSX_2 (FLOAT2_V2DI, "float2_v2di", CONST, float2_v2di)
BU_VSX_2 (UNS_FLOAT2_V2DI, "uns_float2_v2di", CONST, uns_float2_v2di)
+BU_VSX_2 (VEC_VSIGNED2_V2DF, "vsigned2_v2df", CONST, vsigned2_v2df)
+BU_VSX_2 (VEC_VUNSIGNED2_V2DF, "vunsigned2_v2df", CONST, vunsigned2_v2df)
+
/* VSX abs builtin functions. */
BU_VSX_A (XVABSDP, "xvabsdp", CONST, absv2df2)
BU_VSX_A (XVNABSDP, "xvnabsdp", CONST, vsx_nabsv2df2)
@@ -1693,6 +1696,16 @@ BU_VSX_1 (DOUBLEL_V4SI, "doublel_v4si", CONST, doublelv4si2)
BU_VSX_1 (DOUBLEL_V4SF, "doublel_v4sf", CONST, doublelv4sf2)
BU_VSX_1 (UNS_DOUBLEL_V4SI, "uns_doublel_v4si", CONST, unsdoublelv4si2)
+BU_VSX_1 (VEC_VSIGNED_V4SF, "vsigned_v4sf", CONST, vsx_xvcvspsxws)
+BU_VSX_1 (VEC_VSIGNED_V2DF, "vsigned_v2df", CONST, vsx_xvcvdpsxds)
+BU_VSX_1 (VEC_VSIGNEDE_V2DF, "vsignede_v2df", CONST, vsignede_v2df)
+BU_VSX_1 (VEC_VSIGNEDO_V2DF, "vsignedo_v2df", CONST, vsignedo_v2df)
+
+BU_VSX_1 (VEC_VUNSIGNED_V4SF, "vunsigned_v4sf", CONST, vsx_xvcvspsxws)
+BU_VSX_1 (VEC_VUNSIGNED_V2DF, "vunsigned_v2df", CONST, vsx_xvcvdpsxds)
+BU_VSX_1 (VEC_VUNSIGNEDE_V2DF, "vunsignede_v2df", CONST, vunsignede_v2df)
+BU_VSX_1 (VEC_VUNSIGNEDO_V2DF, "vunsignedo_v2df", CONST, vunsignedo_v2df)
+
/* VSX predicate functions. */
BU_VSX_P (XVCMPEQSP_P, "xvcmpeqsp_p", CONST, vector_eq_v4sf_p)
BU_VSX_P (XVCMPGESP_P, "xvcmpgesp_p", CONST, vector_ge_v4sf_p)
@@ -1782,6 +1795,8 @@ BU_VSX_OVERLOAD_2 (XXSPLTD, "xxspltd")
BU_VSX_OVERLOAD_2 (XXSPLTW, "xxspltw")
BU_VSX_OVERLOAD_2 (FLOAT2, "float2")
BU_VSX_OVERLOAD_2 (UNS_FLOAT2, "uns_float2")
+BU_VSX_OVERLOAD_2 (VSIGNED2, "vsigned2")
+BU_VSX_OVERLOAD_2 (VUNSIGNED2, "vu "VECTOR_UNIT_VSX_P (V4SFmode)"nsigned2")
/* 1 argument VSX overloaded builtin functions. */
BU_VSX_OVERLOAD_1 (DOUBLE, "double")
@@ -1797,6 +1812,14 @@ BU_VSX_OVERLOAD_1 (FLOAT, "float")
BU_VSX_OVERLOAD_1 (FLOATE, "floate")
BU_VSX_OVERLOAD_1 (FLOATO, "floato")
+BU_VSX_OVERLOAD_1 (VSIGNED, "vsigned")
+BU_VSX_OVERLOAD_1 (VSIGNEDE, "vsignede")
+BU_VSX_OVERLOAD_1 (VSIGNEDO, "vsignedo")
+
+BU_VSX_OVERLOAD_1 (VUNSIGNED, "vunsigned")
+BU_VSX_OVERLOAD_1 (VUNSIGNEDE, "vunsignede")
+BU_VSX_OVERLOAD_1 (VUNSIGNEDO, "vunsignedo")
+
/* VSX builtins that are handled as special cases. */
BU_VSX_OVERLOAD_X (LD, "ld")
BU_VSX_OVERLOAD_X (ST, "st")
diff --git a/gcc/config/rs6000/rs6000-c.c b/gcc/config/rs6000/rs6000-c.c
index cfd74ab..0752ef9 100644
--- a/gcc/config/rs6000/rs6000-c.c
+++ b/gcc/config/rs6000/rs6000-c.c
@@ -5521,6 +5521,29 @@ const struct altivec_builtin_types altivec_overloaded_builtins[] = {
{ P9V_BUILTIN_VEC_REVB, P9V_BUILTIN_XXBRH_V8HI,
RS6000_BTI_V8HI, RS6000_BTI_V8HI, 0, 0 },
+ { VSX_BUILTIN_VEC_VSIGNED, VSX_BUILTIN_VEC_VSIGNED_V4SF,
+ RS6000_BTI_V4SI, RS6000_BTI_V4SF, 0, 0 },
+ { VSX_BUILTIN_VEC_VSIGNED, VSX_BUILTIN_VEC_VSIGNED_V2DF,
+ RS6000_BTI_V2DI, RS6000_BTI_V2DF, 0, 0 },
+ { VSX_BUILTIN_VEC_VSIGNEDE, VSX_BUILTIN_VEC_VSIGNEDE_V2DF,
+ RS6000_BTI_V4SI, RS6000_BTI_V2DF, 0, 0 },
+ { VSX_BUILTIN_VEC_VSIGNEDO, VSX_BUILTIN_VEC_VSIGNEDO_V2DF,
+ RS6000_BTI_V4SI, RS6000_BTI_V2DF, 0, 0 },
+ { VSX_BUILTIN_VEC_VSIGNED2, VSX_BUILTIN_VEC_VSIGNED2_V2DF,
+ RS6000_BTI_V4SI, RS6000_BTI_V2DF, RS6000_BTI_V2DF, 0 },
+
+ { VSX_BUILTIN_VEC_VUNSIGNED, VSX_BUILTIN_VEC_VUNSIGNED_V4SF,
+ RS6000_BTI_unsigned_V4SI, RS6000_BTI_V4SF, 0, 0 },
+ { VSX_BUILTIN_VEC_VUNSIGNED, VSX_BUILTIN_VEC_VUNSIGNED_V2DF,
+ RS6000_BTI_unsigned_V2DI, RS6000_BTI_V2DF, 0, 0 },
+ { VSX_BUILTIN_VEC_VUNSIGNEDE, VSX_BUILTIN_VEC_VUNSIGNEDE_V2DF,
+ RS6000_BTI_unsigned_V4SI, RS6000_BTI_V2DF, 0, 0 },
+ { VSX_BUILTIN_VEC_VUNSIGNEDO, VSX_BUILTIN_VEC_VUNSIGNEDO_V2DF,
+ RS6000_BTI_unsigned_V4SI, RS6000_BTI_V2DF, 0, 0 },
+ { VSX_BUILTIN_VEC_VUNSIGNED2, VSX_BUILTIN_VEC_VUNSIGNED2_V2DF,
+ RS6000_BTI_unsigned_V4SI, RS6000_BTI_V2DF,
+ RS6000_BTI_V2DF, 0 },
+
/* Crypto builtins. */
{ CRYPTO_BUILTIN_VPERMXOR, CRYPTO_BUILTIN_VPERMXOR_V16QI,
RS6000_BTI_unsigned_V16QI, RS6000_BTI_unsigned_V16QI,
diff --git a/gcc/config/rs6000/rs6000-protos.h b/gcc/config/rs6000/rs6000-protos.h
index e313204..44fdf8a 100644
--- a/gcc/config/rs6000/rs6000-protos.h
+++ b/gcc/config/rs6000/rs6000-protos.h
@@ -73,6 +73,7 @@ extern void rs6000_ "VECTOR_UNIT_VSX_P (V4SFmode)"expand_extract_even (rtx, rtx, rtx);
extern void rs6000_expand_interleave (rtx, rtx, rtx, bool);
extern void rs6000_scale_v2df (rtx, rtx, int);
extern void rs6000_generate_float2_code (bool, rtx, rtx, rtx);
+extern void rs6000_generate_vsigned2_code (bool, rtx, rtx, rtx);
extern int expand_block_clear (rtx[]);
extern int expand_block_move (rtx[]);
extern bool expand_block_compare (rtx[]);
diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index 79dccba..181794d 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -41074,6 +41074,35 @@ rs6000_generate_float2_code (bool signed_convert, rtx dst, rtx src1, rtx src2)
emit_insn (gen_p8_vmrgew_v4sf (dst, rtx_tmp3, rtx_tmp2));
}
+void
+rs6000_generate_vsigned2_code (bool signed_convert, rtx dst, rtx src1,
+ rtx src2)
+{
+ rtx rtx_tmp0, rtx_tmp1, rtx_tmp2, rtx_tmp3;
+
+ rtx_tmp0 = gen_reg_rtx (V2DFmode);
+ rtx_tmp1 = gen_reg_rtx (V2DFmode);
+
+ emit_insn (gen_vsx_xxpermdi_v2df (rtx_tmp0, src1, src2, GEN_INT (0)));
+ emit_insn (gen_vsx_xxpermdi_v2df (rtx_tmp1, src1, src2, GEN_INT (3)));
+
+ rtx_tmp2 = gen_reg_rtx (V4SImode);
+ rtx_tmp3 = gen_reg_rtx (V4SImode);
+
+ if (signed_convert)
+ {
+ emit_insn (gen_vsx_xvcvdpsxws (rtx_tmp2, rtx_tmp0));
+ emit_insn (gen_vsx_xvcvdpsxws (rtx_tmp3, rtx_tmp1));
+ }
+ else
+ {
+ emit_insn (gen_vsx_xvcvdpuxws (rtx_tmp2, rtx_tmp0));
+ emit_insn (gen_vsx_xvcvdpuxws (rtx_tmp3, rtx_tmp1));
+ }
+
+ emit_insn (gen_p8_vmrgew_v4si (dst, rtx_tmp2, rtx_tmp3));
+}
+
/* Implement the TARGET_OPTAB_SUPPORTED_P hook. */
static bool
diff --git a/gcc/config/rs6000/vsx.md b/gcc/config/rs6000/vsx.md
index 4d73f0a..8f98a34 100644
--- a/gcc/config/rs6000/vsx.md
+++ b/gcc/config/rs6000/vsx.md
@@ -358,6 +358,8 @@
UNSPEC_VSX_XVCVDPSXDS
UNSPEC_VSX_XVCVDPUXDS
UNSPEC_VSX_SIGN_EXTEND
+ UNSPEC_VSX_XVCVSPSXWS
+ UNSPEC_VSX_XVCVSPSXDS
UNSPEC_VSX_VSLO
UNSPEC_VSX_EXTRACT
UNSPEC_VSX_SXEXPDP
@@ -370,6 +372,7 @@
UNSPEC_VSX_VIEXP
UNSPEC_VSX_VTSTDC
UNSPEC_VSX_VEC_INIT
+ UNSPEC_VSX_VSIGNED2
UNSPEC_LXVL
UNSPEC_STXVL
UNSPEC_VCLZLSBB
@@ -1864,6 +1867,8 @@
DONE;
})
+;; convert vector of 64-bit floating point numbers to vector of
+;; 64-bit signed integer
(define_insn "vsx_xvcvdpsxds"
[(set (match_operand:V2DI 0 "vsx_register_operand" "=wa")
(unspec:V2DI [(match_operand:V2DF 1 "vsx_register_operand" "wa")]
@@ -1872,6 +1877,18 @@
"xvcvdpsxds %x0,%x1"
[(set_attr "type" "vecdouble")])
+;; convert vector of 32-bit floating point numbers to vector of
+;; 32-bit signed integer
+(define_insn "vsx_xvcvspsxws"
+ [(set (match_operand:V4SI 0 "vsx_register_operand" "=wa")
+ (unspec:V4SI [(match_operand:V4SF 1 "vsx_register_operand" "wa")]
+ UNSPEC_VSX_XVCVSPSXWS))]
+ "VECTOR_UNIT_VSX_P (V4SFmode)"
+ "xvcvspsxws %x0,%x1"
+ [(set_attr "type" "vecfloat")])
+
+;; convert vector of 64-bit floating point numbers to vector of
+;; 64-bit unsigned integer
(define_expand "vsx_xvcvdpuxds_scale"
[(match_operand:V2DI 0 "vsx_register_operand" "")
(match_operand:V2DF 1 "vsx_register_operand" "")
@@ -1893,6 +1910,16 @@
DONE;
})
+;; convert vector of 32-bit floating point numbers to vector of
+;; 32-bit unsigned integer
+(define_insn "vsx_xvcvspuxws"
+ [(set (match_operand:V4SI 0 "vsx_register_operand" "=wa")
+ (unspec:V4SI [(match_operand:V4SF 1 "vsx_register_operand" "wa")]
+ UNSPEC_VSX_XVCVSPSXWS))]
+ "VECTOR_UNIT_VSX_P (V4SFmode)"
+ "xvcvspuxws %x0,%x1"
+ [(set_attr "type" "vecfloat")])
+
(define_insn "vsx_xvcvdpuxds"
[(set (match_operand:V2DI 0 "vsx_register_operand" "=wa")
(unspec:V2DI [(match_operand:V2DF 1 "vsx_register_operand" "wa")]
@@ -2137,6 +2164,173 @@
DONE;
})
+;; Generate vsigned2
+;; convert two double float vectors to a vector of single precision ints
+(define_expand "vsigned2_v2df"
+ [(match_operand:V4SI 0 "register_operand" "=wa")
+ (unspec:V4SI [(match_operand:V2DF 1 "register_operand" "wa")
+ (match_operand:V2DF 2 "register_operand" "wa")]
+ UNSPEC_VSX_VSIGNED2)]
+ "TARGET_VSX"
+{
+ rtx rtx_src1, rtx_src2, rtx_dst;
+ bool signed_convert=true;
+
+ rtx_dst = operands[0];
+ rtx_src1 = operands[1];
+ rtx_src2 = operands[2];
+
+ rs6000_generate_vsigned2_code (signed_convert, rtx_dst, rtx_src1, rtx_src2);
+ DONE;
+})
+
+;; Generate vsignedo_v2df
+;; signed double float to int convert odd word
+(define_expand "vsignedo_v2df"
+ [(set (match_operand:V4SI 0 "register_operand" "=wa")
+ (match_operand:V2DF 1 "register_operand" "wa"))]
+ "TARGET_VSX"
+{
+ if (VECTOR_ELT_ORDER_BIG)
+ {
+ rtx rtx_tmp;
+ rtx rtx_val = GEN_INT (12);
+ rtx_tmp = gen_reg_rtx (V4SImode);
+
+ emit_insn (gen_vsx_xvcvdpsxws (rtx_tmp, operands[1]));
+
+ /* Big endian word numbering for words in operand is 0 1 2 3.
+ take (operand[1] operand[1]) and shift left one word
+ 0 1 2 3 0 1 2 3 => 1 2 3 0
+ Words 1 and 3 are now are now where they need to be for result. */
+
+ emit_insn (gen_altivec_vsldoi_v4si (operands[0], rtx_tmp,
+ rtx_tmp, rtx_val));
+ }
+ else
+ /* Little endian word numbering for operand is 3 2 1 0.
+ Result words 3 and 1 are where they need to be. */
+ emit_insn (gen_vsx_xvcvdpsxws (operands[0], operands[1]));
+
+ DONE;
+}
+ [(set_attr "type" "veccomplex")])
+
+;; Generate vsignede_v2df
+;; signed double float to int even word
+(define_expand "vsignede_v2df"
+ [(set (match_operand:V4SI 0 "register_operand" "=v")
+ (match_operand:V2DF 1 "register_operand" "v"))]
+ "TARGET_VSX"
+{
+ if (VECTOR_ELT_ORDER_BIG)
+ /* Big endian word numbering for words in operand is 0 1
+ Result words 0 is where they need to be. */
+ emit_insn (gen_vsx_xvcvdpsxws (operands[0], operands[1]));
+
+ else
+ {
+ rtx rtx_tmp;
+ rtx rtx_val = GEN_INT (12);
+ rtx_tmp = gen_reg_rtx (V4SImode);
+
+ emit_insn (gen_vsx_xvcvdpsxws (rtx_tmp, operands[1]));
+
+ /* Little endian word numbering for operand is 3 2 1 0.
+ take (operand[1] operand[1]) and shift left three words
+ 0 1 2 3 0 1 2 3 => 3 0 1 2
+ Words 0 and 2 are now where they need to be for the result. */
+ emit_insn (gen_altivec_vsldoi_v4si (operands[0], rtx_tmp,
+ rtx_tmp, rtx_val));
+ }
+ DONE;
+}
+ [(set_attr "type" "veccomplex")])
+
+;; Generate unsigned2
+;; convert two double float vectors to a vector of single precision
+;; unsigned ints
+(define_expand "vunsigned2_v2df"
+[(match_operand:V4SI 0 "register_operand" "=v")
+ (unspec:V4SI [(match_operand:V2DF 1 "register_operand" "v")
+ (match_operand:V2DF 2 "register_operand" "v")]
+ UNSPEC_VSX_VSIGNED2)]
+ "TARGET_VSX"
+{
+ rtx rtx_src1, rtx_src2, rtx_dst;
+ bool signed_convert=false;
+
+ rtx_dst = operands[0];
+ rtx_src1 = operands[1];
+ rtx_src2 = operands[2];
+
+ rs6000_generate_vsigned2_code (signed_convert, rtx_dst, rtx_src1, rtx_src2);
+ DONE;
+})
+
+;; Generate vunsignedo_v2df
+;; unsigned double float to int convert odd word
+(define_expand "vunsignedo_v2df"
+ [(set (match_operand:V4SI 0 "register_operand" "=v")
+ (match_operand:V2DF 1 "register_operand" "v"))]
+ "TARGET_VSX"
+{
+ if (VECTOR_ELT_ORDER_BIG)
+ {
+ rtx rtx_tmp;
+ rtx rtx_val = GEN_INT (12);
+ rtx_tmp = gen_reg_rtx (V4SImode);
+
+ emit_insn (gen_vsx_xvcvdpuxws (rtx_tmp, operands[1]));
+
+ /* Big endian word numbering for words in operand is 0 1 2 3.
+ take (operand[1] operand[1]) and shift left one word
+ 0 1 2 3 0 1 2 3 => 1 2 3 0
+ Words 1 and 3 are now are now where they need to be for result. */
+
+ emit_insn (gen_altivec_vsldoi_v4si (operands[0], rtx_tmp,
+ rtx_tmp, rtx_val));
+ }
+ else
+ /* Little endian word numbering for operand is 3 2 1 0.
+ Result words 3 and 1 are where they need to be. */
+ emit_insn (gen_vsx_xvcvdpuxws (operands[0], operands[1]));
+
+ DONE;
+}
+ [(set_attr "type" "veccomplex")])
+
+;; Generate vunsignede_v2df
+;; unsigned double float to int even word
+(define_expand "vunsignede_v2df"
+ [(set (match_operand:V4SI 0 "register_operand" "=v")
+ (match_operand:V2DF 1 "register_operand" "v"))]
+ "TARGET_VSX"
+{
+ if (VECTOR_ELT_ORDER_BIG)
+ /* Big endian word numbering for words in operand is 0 1
+ Result words 0 is where they need to be. */
+ emit_insn (gen_vsx_xvcvdpuxws (operands[0], operands[1]));
+
+ else
+ {
+ rtx rtx_tmp;
+ rtx rtx_val = GEN_INT (12);
+ rtx_tmp = gen_reg_rtx (V4SImode);
+
+ emit_insn (gen_vsx_xvcvdpuxws (rtx_tmp, operands[1]));
+
+ /* Little endian word numbering for operand is 3 2 1 0.
+ take (operand[1] operand[1]) and shift left three words
+ 0 1 2 3 0 1 2 3 => 3 0 1 2
+ Words 0 and 2 are now where they need to be for the result. */
+ emit_insn (gen_altivec_vsldoi_v4si (operands[0], rtx_tmp,
+ rtx_tmp, rtx_val));
+ }
+ DONE;
+}
+ [(set_attr "type" "veccomplex")])
+
;; Only optimize (float (fix x)) -> frz if we are in fast-math mode, since
;; since the xvrdpiz instruction does not truncate the value if the floating
;; point value is < LONG_MIN or > LONG_MAX.
diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi
index 43f9ecf..f87c1f3 100644
--- a/gcc/doc/extend.texi
+++ b/gcc/doc/extend.texi
@@ -16645,6 +16645,13 @@ vector bool char vec_sel (vector bool char,
vector bool char,
vector unsigned char);
+vector signed long long vec_signed (vector double);
+vector signed int vec_signed (vector float);
+
+vector signed int vec_signede (vector double);
+vector signed int vec_signedo (vector double);
+vector signed int vec_signed2 (vector double, vector double);
+
vector signed char vec_sl (vector signed char,
vector unsigned char);
vector unsigned char vec_sl (vector unsigned char,
@@ -17173,6 +17180,13 @@ vector signed int vec_sums (vector signed int, vector signed int);
vector float vec_trunc (vector float);
+vector signed long long vec_unsigned (vector double);
+vector signed int vec_unsigned (vector float);
+
+vector signed int vec_unsignede (vector double);
+vector signed int vec_unsignedo (vector double);
+vector signed int vec_unsigned2 (vector double, vector double);
+
vector signed short vec_unpackh (vector signed char);
vector bool short vec_unpackh (vector bool char);
vector signed int vec_unpackh (vector signed short);
diff --git a/gcc/testsuite/gcc.target/powerpc/builtins-3-runnable.c b/gcc/testsuite/gcc.target/powerpc/builtins-3-runnable.c
index 08b7a53..17bb9b3 100644
--- a/gcc/testsuite/gcc.target/powerpc/builtins-3-runnable.c
+++ b/gcc/testsuite/gcc.target/powerpc/builtins-3-runnable.c
@@ -5,53 +5,181 @@
#include <altivec.h> // vector
+#ifdef DEBUG
+#include <stdio.h>
+#endif
+
#define ALL 1
#define EVEN 2
#define ODD 3
void abort (void);
-void test_result_sp(int check, vector float vec_result, vector float vec_expected)
+void test_int_result(int check, vector int vec_result, vector int vec_expected)
{
int i;
- for(i = 0; i<4; i++) {
- switch (check) {
- case ALL:
- break;
- case EVEN:
- if (i%2 == 0)
+ for (i = 0; i < 4; i++) {
+ switch (check) {
+ case ALL:
break;
- else
- continue;
- case ODD:
- if (i%2 != 0)
+ case EVEN:
+ if (i%2 == 0)
+ break;
+ else
+ continue;
+ case ODD:
+ if (i%2 != 0)
+ break;
+ else
+ continue;
+ }
+
+ if (vec_result[i] != vec_expected[i]) {
+#ifdef DEBUG
+ printf("Test_int_result: ");
+ printf("vec_result[%d] (%d) != vec_expected[%d] (%d)\n",
+ i, vec_result[i], i, vec_expected[i]);
+#else
+ abort();
+#endif
+ }
+ }
+}
+
+void test_unsigned_int_result(int check, vector unsigned int vec_result,
+ vector unsigned int vec_expected)
+{
+ int i;
+
+ for (i = 0; i < 4; i++) {
+ switch (check) {
+ case ALL:
break;
- else
- continue;
+ case EVEN:
+ if (i%2 == 0)
+ break;
+ else
+ continue;
+ case ODD:
+ if (i%2 != 0)
+ break;
+ else
+ continue;
+ }
+
+ if (vec_result[i] != vec_expected[i]) {
+#ifdef DEBUG
+ printf("Test_unsigned int_result: ");
+ printf("vec_result[%d] (%d) != vec_expected[%d] (%d)\n",
+ i, vec_result[i], i, vec_expected[i]);
+#else
+ abort();
+#endif
+ }
+
}
+}
- if (vec_result[i] != vec_expected[i])
- abort();
+void test_ll_int_result(vector long long int vec_result,
+ vector long long int vec_expected)
+{
+ int i;
+
+ for (i = 0; i < 2; i++)
+ if (vec_result[i] != vec_expected[i]) {
+#ifdef DEBUG
+ printf("Test_ll_int_result: ");
+ printf("vec_result[%d] (%lld) != vec_expected[%d] (%lld)\n",
+ i, vec_result[i], i, vec_expected[i]);
+#else
+ abort();
+#endif
+ }
+}
+
+void test_ll_unsigned_int_result(vector long long unsigned int vec_result,
+ vector long long unsigned int vec_expected)
+{
+ int i;
+
+ for (i = 0; i < 2; i++)
+ if (vec_result[i] != vec_expected[i]) {
+#ifdef DEBUG
+ printf("Test_ll_unsigned_int_result: ");
+ printf("vec_result[%d] (%lld) != vec_expected[%d] (%lld)\n",
+ i, vec_result[i], i, vec_expected[i]);
+#else
+ abort();
+#endif
+ }
+}
+
+void test_result_sp(int check, vector float vec_result,
+ vector float vec_expected)
+{
+ int i;
+ for(i = 0; i<4; i++) {
+
+ switch (check) {
+ case ALL:
+ break;
+ case EVEN:
+ if (i%2 == 0)
+ break;
+ else
+ continue;
+ case ODD:
+ if (i%2 != 0)
+ break;
+ else
+ continue;
+ }
+
+ if (vec_result[i] != vec_expected[i]) {
+#ifdef DEBUG
+ printf("Test_result_sp: ");
+ printf("vec_result[%d] (%lld) != vec_expected[%d] (%lld)\n",
+ i, vec_result[i], i, vec_expected[i]);
+#else
+ abort();
+#endif
+ }
}
}
void test_result_dp(vector double vec_result, vector double vec_expected)
{
- if (vec_result[0] != vec_expected[0])
+ if (vec_result[0] != vec_expected[0]) {
+#ifdef DEBUG
+ printf("Test_result_dp: ");
+ printf("vec_result[0] (%lld) != vec_expected[0] (%lld)\n",
+ vec_result[0], vec_expected[0]);
+#else
abort();
+#endif
+ }
- if (vec_result[1] != vec_expected[1])
+ if (vec_result[1] != vec_expected[1]) {
+#ifdef DEBUG
+ printf("Test_result_dp: ");
+ printf("vec_result[1] (%lld) != vec_expected[1] (%lld)\n",
+ vec_result[1], vec_expected[1]);
+#else
abort();
+#endif
+ }
}
int main()
{
int i;
- vector unsigned int vec_unint;
- vector signed int vec_int;
+ vector unsigned int vec_unint, vec_uns_int_expected, vec_uns_int_result;
+ vector signed int vec_int, vec_int_expected, vec_int_result;
vector long long int vec_ll_int0, vec_ll_int1;
+ vector long long int vec_ll_int_expected, vec_ll_int_result;
vector long long unsigned int vec_ll_uns_int0, vec_ll_uns_int1;
+ vector long long unsigned int vec_ll_uns_int_expected, vec_ll_uns_int_result;
vector float vec_flt, vec_flt_result, vec_flt_expected;
vector double vec_dble0, vec_dble1, vec_dble_result, vec_dble_expected;
@@ -163,4 +291,67 @@ int main()
vec_flt_expected = (vector float){0.00, 34.00, 0.00, 97.00};
vec_flt_result = vec_floato (vec_dble0);
test_result_sp(ODD, vec_flt_result, vec_flt_expected);
+
+ /* Convert single precision float to int */
+ vec_flt = (vector float){-14.30, 34.00, 22.00, 97.00};
+ vec_int_expected = (vector signed int){-14, 34, 22, 97};
+ vec_int_result = vec_signed (vec_flt);
+ test_int_result (ALL, vec_int_result, vec_int_expected);
+
+ /* Convert double precision float to long long int */
+ vec_dble0 = (vector double){-124.930, 81234.49};
+ vec_ll_int_expected = (vector long long signed int){-124, 81234};
+ vec_ll_int_result = vec_signed (vec_dble0);
+ test_ll_int_result (vec_ll_int_result, vec_ll_int_expected);
+
+ /* Convert two double precision vector float to vector int */
+ vec_dble0 = (vector double){-124.930, 81234.49};
+ vec_dble1 = (vector double){-24.370, 8354.99};
+ vec_int_expected = (vector signed int){-124, 81234, -24, 8354};
+ vec_int_result = vec_signed2 (vec_dble0, vec_dble1);
+ test_int_result (ALL, vec_int_result, vec_int_expected);
+
+ /* Convert double precision vector float to vector int, even words */
+ vec_dble0 = (vector double){-124.930, 81234.49};
+ vec_int_expected = (vector signed int){-124, 0, 81234, 0};
+ vec_int_result = vec_signede (vec_dble0);
+ test_int_result (EVEN, vec_int_result, vec_int_expected);
+
+ /* Convert double precision vector float to vector int, odd words */
+ vec_dble0 = (vector double){-124.930, 81234.49};
+ vec_int_expected = (vector signed int){0, -124, 0, 81234};
+ vec_int_result = vec_signedo (vec_dble0);
+ test_int_result (ODD, vec_int_result, vec_int_expected);
+
+ /* Convert double precision float to long long unsigned int */
+ vec_dble0 = (vector double){124.930, 8134.49};
+ vec_ll_uns_int_expected = (vector long long unsigned int){124, 8134};
+ vec_ll_uns_int_result = vec_unsigned (vec_dble0);
+ test_ll_unsigned_int_result (vec_ll_uns_int_result,
+ vec_ll_uns_int_expected);
+
+ /* Convert two double precision vector float to vector unsigned int */
+ vec_dble0 = (vector double){124.930, 8134.49};
+ vec_dble1 = (vector double){24.370, 834.99};
+ vec_uns_int_expected = (vector unsigned int){124, 8134, 24, 834};
+ vec_uns_int_result = vec_unsigned2 (vec_dble0, vec_dble1);
+ test_unsigned_int_result (ALL, vec_uns_int_result,
+ vec_uns_int_expected);
+
+ /* Convert double precision vector float to vector unsigned int,
+ even words */
+ vec_dble0 = (vector double){3124.930, 8234.49};
+ vec_uns_int_expected = (vector unsigned int){3124, 0, 8234, 0};
+ vec_uns_int_result = vec_unsignede (vec_dble0);
+ test_unsigned_int_result (EVEN, vec_uns_int_result,
+ vec_uns_int_expected);
+
+ /* Convert double precision vector float to vector unsigned int,
+ odd words */
+ vec_dble0 = (vector double){1924.930, 81234.49};
+ vec_uns_int_expected = (vector unsigned int){0, 1924, 0, 81234};
+ vec_uns_int_result = vec_unsignedo (vec_dble0);
+ test_unsigned_int_result (ODD, vec_uns_int_result,
+ vec_uns_int_expected);
}
+
--
1.9.1
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH, rs6000] Signed builtin support
2017-06-28 15:37 [PATCH, rs6000] Signed builtin support Carl Love
@ 2017-06-28 22:05 ` Segher Boessenkool
2017-06-29 12:59 ` Bill Schmidt
0 siblings, 1 reply; 3+ messages in thread
From: Segher Boessenkool @ 2017-06-28 22:05 UTC (permalink / raw)
To: Carl Love; +Cc: gcc-patches, David Edelsohn, Bill Schmidt
Hi Carl,
On Wed, Jun 28, 2017 at 08:37:35AM -0700, Carl Love wrote:
> The following patch adds support for the vec_signed, vec_signede,
> vec_signedo and vec_signed2 builtins.
Are those names ABI-dictated?
> * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
> UNSPEC_VSX_VSIGNED2): Add UNSPECS
Missing "."
> * gcc.target/powerpc/builtins-3-runnable.c(test_int_result,
> test_unsigned_int_result, test_ll_int_result,
> test_ll_unsigned_int_result): Add result checking functions, add
> debug support.
Space before "(".
> (main): Add builtin function tests, .
Something is missing here?
> Signed-off-by: Carl Love <carll@us.ibm.com>
We don't use those.
> +(define_expand "vunsignede_v2df"
> + /* Little endian word numbering for operand is 3 2 1 0.
> + take (operand[1] operand[1]) and shift left three words
> + 0 1 2 3 0 1 2 3 => 3 0 1 2
> + Words 0 and 2 are now where they need to be for the result. */
Should use tabs for eight leading spaces.
Okay for trunk with those nits fixed. Thanks!
Segher
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH, rs6000] Signed builtin support
2017-06-28 22:05 ` Segher Boessenkool
@ 2017-06-29 12:59 ` Bill Schmidt
0 siblings, 0 replies; 3+ messages in thread
From: Bill Schmidt @ 2017-06-29 12:59 UTC (permalink / raw)
To: Segher Boessenkool; +Cc: Carl Love, gcc-patches, David Edelsohn
On Jun 28, 2017, at 4:48 PM, Segher Boessenkool <segher@kernel.crashing.org> wrote:
>
> Hi Carl,
>
> On Wed, Jun 28, 2017 at 08:37:35AM -0700, Carl Love wrote:
>> The following patch adds support for the vec_signed, vec_signede,
>> vec_signedo and vec_signed2 builtins.
>
> Are those names ABI-dictated?
Unfortunately, yes. I didn't like those names much, either, but it's where we ended up.
Bill
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2017-06-29 12:59 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-06-28 15:37 [PATCH, rs6000] Signed builtin support Carl Love
2017-06-28 22:05 ` Segher Boessenkool
2017-06-29 12:59 ` Bill Schmidt
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).