rs6000, add overloaded DFP quantize support
Checks
Commit Message
GCC maintainers:
The following patch adds four built-ins for the decimal floating point
(DFP) quantize instructions on rs6000. The built-ins are for 64-bit
and 128-bit DFP operands.
The patch also adds a test case for the new builtins.
The Patch has been tested on Power 10LE and Power 9 LE/BE.
Please let me know if the patch is acceptable for mainline. Thanks.
Carl Love
------------------------------------------
rs6000, add overloaded DFP quantize support
Add decimal floating point (DFP) quantize built-ins for both 64-bit DFP
and 128-DFP operands. In each case, there is an immediate version and a
variable version of the bult-in. The RM value is a 2-bit const int which
specifies the rounding mode to use. For the immediate versions of the
built-in, TE field is a 5-bit constant that specifies the value of the
ideal exponent for the result. The built-in specifications are:
__Decimal64 builtin_dfp_quantize (_Decimal64, _Decimal64,
const int RM)
__Decimal64 builtin_dfp_quantize (const int TE, _Decimal64,
const int)
__Decimal128 builtin_dfpq_quantize (_Decimal128, _Decimal128,
const int RM)
__Decimal128 builtin_dfpq_quantize (const int TE, _Decimal128,
const int)
A testcase is added for the new built-in definitions.
gcc/ChangeLog:
* config/rs6000/dfp.md: New UNSPECDQUAN.
(dfp_quan_<mode>, dfp_quan_i<mode>): New define_insn.
* config/rs6000/rs6000-builtins.def (__builtin_dfp_quantize_64,
__builtin_dfp_quantize_64i, __builtin_dfp_quantize_128,
__builtin_dfp_quantize_128i): New buit-in definitions.
* config/rs6000/rs6000-overload.def (__builtin_dfp_quantize,
__builtin_dfpq_quantize): New overloaded definitions.
gcc/testsuite/
* gcc.target/powerpc/builtin-dfp-quantize-runnable.c: New test
case.
---
gcc/config/rs6000/dfp.md | 25 ++-
gcc/config/rs6000/rs6000-builtins.def | 15 ++
gcc/config/rs6000/rs6000-overload.def | 12 ++
.../powerpc/builtin-dfp-quantize-runnable.c | 198 ++++++++++++++++++
4 files changed, 249 insertions(+), 1 deletion(-)
create mode 100644 gcc/testsuite/gcc.target/powerpc/builtin-dfp-quantize-runnable.c
Comments
Hi Carl,
on 2023/8/9 23:52, Carl Love wrote:
>
> GCC maintainers:
>
> The following patch adds four built-ins for the decimal floating point
> (DFP) quantize instructions on rs6000. The built-ins are for 64-bit
> and 128-bit DFP operands.
>
> The patch also adds a test case for the new builtins.
>
> The Patch has been tested on Power 10LE and Power 9 LE/BE.
>
> Please let me know if the patch is acceptable for mainline. Thanks.
>
> Carl Love
>
>
> ------------------------------------------
> rs6000, add overloaded DFP quantize support
>
> Add decimal floating point (DFP) quantize built-ins for both 64-bit DFP
> and 128-DFP operands. In each case, there is an immediate version and a
> variable version of the bult-in. The RM value is a 2-bit const int which
Nit: s/bult-in//built-in/
> specifies the rounding mode to use. For the immediate versions of the
> built-in, TE field is a 5-bit constant that specifies the value of the
> ideal exponent for the result. The built-in specifications are:
>
> __Decimal64 builtin_dfp_quantize (_Decimal64, _Decimal64,
> const int RM)
> __Decimal64 builtin_dfp_quantize (const int TE, _Decimal64,
> const int)
> __Decimal128 builtin_dfpq_quantize (_Decimal128, _Decimal128,
> const int RM)
> __Decimal128 builtin_dfpq_quantize (const int TE, _Decimal128,
> const int)
>
I noticed that the existing DFP bifs are directly using the insn
mnemonics, I perfer to keep consistent with them. So could we
have one function like unique external interface like dfp_quantize
for users' uses?
And we can have the underlying instances for it:
__Decimal64 builtin_dfp_dqua (_Decimal64, _Decimal64, const int RM)
__Decimal64 builtin_dfp_dquai (const int TE, _Decimal64, const int)
__Decimal128 builtin_dfp_dquaq (_Decimal128, _Decimal128, const int RM)
__Decimal128 builtin_dfp_dquaiq (const int TE, _Decimal128, const int)
Besides, this patch missed to update the documentation, please add them
in gcc//doc/extend.texi by searching "The following built-in functions
are available when hardware decimal floating point".
> A testcase is added for the new built-in definitions.
>
> gcc/ChangeLog:
> * config/rs6000/dfp.md: New UNSPECDQUAN.
> (dfp_quan_<mode>, dfp_quan_i<mode>): New define_insn.
> * config/rs6000/rs6000-builtins.def (__builtin_dfp_quantize_64,
> __builtin_dfp_quantize_64i, __builtin_dfp_quantize_128,
> __builtin_dfp_quantize_128i): New buit-in definitions.
> * config/rs6000/rs6000-overload.def (__builtin_dfp_quantize,
> __builtin_dfpq_quantize): New overloaded definitions.
>
> gcc/testsuite/
> * gcc.target/powerpc/builtin-dfp-quantize-runnable.c: New test
> case.
> ---
> gcc/config/rs6000/dfp.md | 25 ++-
> gcc/config/rs6000/rs6000-builtins.def | 15 ++
> gcc/config/rs6000/rs6000-overload.def | 12 ++
> .../powerpc/builtin-dfp-quantize-runnable.c | 198 ++++++++++++++++++
> 4 files changed, 249 insertions(+), 1 deletion(-)
> create mode 100644 gcc/testsuite/gcc.target/powerpc/builtin-dfp-quantize-runnable.c
>
> diff --git a/gcc/config/rs6000/dfp.md b/gcc/config/rs6000/dfp.md
> index 5ed8a73ac51..254c22a5c20 100644
> --- a/gcc/config/rs6000/dfp.md
> +++ b/gcc/config/rs6000/dfp.md
> @@ -271,7 +271,8 @@
> UNSPEC_DIEX
> UNSPEC_DSCLI
> UNSPEC_DTSTSFI
> - UNSPEC_DSCRI])
> + UNSPEC_DSCRI
> + UNSPEC_DQUAN])
>
> (define_code_iterator DFP_TEST [eq lt gt unordered])
>
> @@ -395,3 +396,25 @@
> "dscri<q> %0,%1,%2"
> [(set_attr "type" "dfp")
> (set_attr "size" "<bits>")])
> +
> +(define_insn "dfp_quan_<mode>"
s/quan/dqua/ to aglin with the others.
> + [(set (match_operand:DDTD 0 "gpc_reg_operand" "=d")
> + (unspec:DDTD [(match_operand:DDTD 1 "gpc_reg_operand" "d")
> + (match_operand:DDTD 2 "gpc_reg_operand" "d")
> + (match_operand:QI 3 "immediate_operand" "i")]
> + UNSPEC_DQUAN))]
> + "TARGET_DFP"
> + "dqua<q> %0,%1,%2,%3"
> + [(set_attr "type" "dfp")
> + (set_attr "size" "<bits>")])
> +
> +(define_insn "dfp_quan_i<mode>"
also s/quan_i/dquai_/
> + [(set (match_operand:DDTD 0 "gpc_reg_operand" "=d")
> + (unspec:DDTD [(match_operand:SI 1 "const_int_operand" "n")
> + (match_operand:DDTD 2 "gpc_reg_operand" "d")
> + (match_operand:SI 3 "immediate_operand" "i")]
> + UNSPEC_DQUAN))]
> + "TARGET_DFP"
> + "dquai<q> %1,%0,%2,%3"
> + [(set_attr "type" "dfp")
> + (set_attr "size" "<bits>")])
> diff --git a/gcc/config/rs6000/rs6000-builtins.def b/gcc/config/rs6000/rs6000-builtins.def
> index 35c4cdf74c5..36a56311643 100644
> --- a/gcc/config/rs6000/rs6000-builtins.def
> +++ b/gcc/config/rs6000/rs6000-builtins.def
> @@ -2983,6 +2983,21 @@
> const unsigned long long __builtin_unpack_dec128 (_Decimal128, const int<1>);
> UNPACK_TD unpacktd {}
>
> + const _Decimal64 __builtin_dfp_quantize_64 (_Decimal64, _Decimal64, \
> + const int<2>);
> + DFPQUAN_64 dfp_quan_dd {}
Update with builtin_dfp_dqua and dfp_dqua_dd accordingly, similar for the below ones.
> +
> + const _Decimal64 __builtin_dfp_quantize_64i (const int<5>, _Decimal64, \
> + const int<2>);
> + DFPQUAN_64i dfp_quan_idd {}
> +
> + const _Decimal128 __builtin_dfp_quantize_128 (_Decimal128, _Decimal128, \
> + const int<2>);
> + DFPQUAN_128 dfp_quan_td {}
> +
> + const _Decimal128 __builtin_dfp_quantize_128i (const int<5>, _Decimal128, \
> + const int<2>);
> + DFPQUAN_128i dfp_quan_itd {}
>
> [crypto]
> const vull __builtin_crypto_vcipher (vull, vull);
> diff --git a/gcc/config/rs6000/rs6000-overload.def b/gcc/config/rs6000/rs6000-overload.def
> index b83946f5ad8..3bb1bedd69d 100644
> --- a/gcc/config/rs6000/rs6000-overload.def
> +++ b/gcc/config/rs6000/rs6000-overload.def
> @@ -195,6 +195,18 @@
> unsigned long long __builtin_cmpb (unsigned long long, unsigned long long);
> CMPB
>
> +[DFPQUAN_64, dfp_quan_64, __builtin_dfp_quantize]
> + _Decimal64 __builtin_dfp_quantize (_Decimal64, _Decimal64, const int);
> + DFPQUAN_64
> + _Decimal64 __builtin_dfp_quantize (const int, _Decimal64, const int);
> + DFPQUAN_64i
> +
> +[DFPQUAN_128, dfp_quan_128, __builtin_dfpq_quantize]
> + _Decimal128 __builtin_dfpq_quantize (_Decimal128, _Decimal128, const int);
> + DFPQUAN_128
> + _Decimal128 __builtin_dfpq_quantize (const int, _Decimal128, const int);
> + DFPQUAN_128i
> +
> [VEC_ABS, vec_abs, __builtin_vec_abs]
> vsc __builtin_vec_abs (vsc);
> ABS_V16QI
> diff --git a/gcc/testsuite/gcc.target/powerpc/builtin-dfp-quantize-runnable.c b/gcc/testsuite/gcc.target/powerpc/builtin-dfp-quantize-runnable.c
> new file mode 100644
> index 00000000000..f4ba7236fe3
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/powerpc/builtin-dfp-quantize-runnable.c
Nit: Since there is PR93448, use its name as test case name?
Also add one line as PR marker in commit log.
> @@ -0,0 +1,198 @@
> +/* { dg-do run } */
> +/* { dg-require-effective-target dfp } */> +/* { dg-options "-O2 -save-temps" } */
Since this is for run, I think what you want are:
/* { dg-require-effective-target dfp_hw } */
/* { dg-require-effective-target has_arch_pwr6 } */
/* { dg-options "-mhard-float -O2 -save-temps" } */
> +
> +
> +#define DEBUG 0
> +
> +#ifdef DEBUG
> +#include <stdio.h>
> +#endif
> +#include <float.h>
> +
> +void abort (void);
> +
> +int main()
> +{
> +#define IMM2 2
> +#define IMM3 3
> +#define IMM4 4
> +
> + _Decimal64 srcA_dfp64, srcB_dfp64;
> + _Decimal64 result_dfp64;
> + _Decimal64 expected_result_dfp64;
> + _Decimal128 srcA_dfp128, srcB_dfp128;
> + _Decimal128 result_dfp128;
> + _Decimal128 expected_result_dfp128;
> +
> + /* Third argument of quantize built-ins is the rounding mode value (RMC).
> +
> + RMC Rounding Mode
> + 00 Round toward 0
> + 01 Round to -infinity
> + 10 Round to nearest, ties toward 0
> + 11 Round according to DRN */
As ISA doc, the description should be:
Primary
RMC Rounding Mode
00 Round to nearest, ties to even
01 Round toward 0
10 Round to nearest, ties away from 0
11 Round according to DRN
BR,
Kewen
@@ -271,7 +271,8 @@
UNSPEC_DIEX
UNSPEC_DSCLI
UNSPEC_DTSTSFI
- UNSPEC_DSCRI])
+ UNSPEC_DSCRI
+ UNSPEC_DQUAN])
(define_code_iterator DFP_TEST [eq lt gt unordered])
@@ -395,3 +396,25 @@
"dscri<q> %0,%1,%2"
[(set_attr "type" "dfp")
(set_attr "size" "<bits>")])
+
+(define_insn "dfp_quan_<mode>"
+ [(set (match_operand:DDTD 0 "gpc_reg_operand" "=d")
+ (unspec:DDTD [(match_operand:DDTD 1 "gpc_reg_operand" "d")
+ (match_operand:DDTD 2 "gpc_reg_operand" "d")
+ (match_operand:QI 3 "immediate_operand" "i")]
+ UNSPEC_DQUAN))]
+ "TARGET_DFP"
+ "dqua<q> %0,%1,%2,%3"
+ [(set_attr "type" "dfp")
+ (set_attr "size" "<bits>")])
+
+(define_insn "dfp_quan_i<mode>"
+ [(set (match_operand:DDTD 0 "gpc_reg_operand" "=d")
+ (unspec:DDTD [(match_operand:SI 1 "const_int_operand" "n")
+ (match_operand:DDTD 2 "gpc_reg_operand" "d")
+ (match_operand:SI 3 "immediate_operand" "i")]
+ UNSPEC_DQUAN))]
+ "TARGET_DFP"
+ "dquai<q> %1,%0,%2,%3"
+ [(set_attr "type" "dfp")
+ (set_attr "size" "<bits>")])
@@ -2983,6 +2983,21 @@
const unsigned long long __builtin_unpack_dec128 (_Decimal128, const int<1>);
UNPACK_TD unpacktd {}
+ const _Decimal64 __builtin_dfp_quantize_64 (_Decimal64, _Decimal64, \
+ const int<2>);
+ DFPQUAN_64 dfp_quan_dd {}
+
+ const _Decimal64 __builtin_dfp_quantize_64i (const int<5>, _Decimal64, \
+ const int<2>);
+ DFPQUAN_64i dfp_quan_idd {}
+
+ const _Decimal128 __builtin_dfp_quantize_128 (_Decimal128, _Decimal128, \
+ const int<2>);
+ DFPQUAN_128 dfp_quan_td {}
+
+ const _Decimal128 __builtin_dfp_quantize_128i (const int<5>, _Decimal128, \
+ const int<2>);
+ DFPQUAN_128i dfp_quan_itd {}
[crypto]
const vull __builtin_crypto_vcipher (vull, vull);
@@ -195,6 +195,18 @@
unsigned long long __builtin_cmpb (unsigned long long, unsigned long long);
CMPB
+[DFPQUAN_64, dfp_quan_64, __builtin_dfp_quantize]
+ _Decimal64 __builtin_dfp_quantize (_Decimal64, _Decimal64, const int);
+ DFPQUAN_64
+ _Decimal64 __builtin_dfp_quantize (const int, _Decimal64, const int);
+ DFPQUAN_64i
+
+[DFPQUAN_128, dfp_quan_128, __builtin_dfpq_quantize]
+ _Decimal128 __builtin_dfpq_quantize (_Decimal128, _Decimal128, const int);
+ DFPQUAN_128
+ _Decimal128 __builtin_dfpq_quantize (const int, _Decimal128, const int);
+ DFPQUAN_128i
+
[VEC_ABS, vec_abs, __builtin_vec_abs]
vsc __builtin_vec_abs (vsc);
ABS_V16QI
new file mode 100644
@@ -0,0 +1,198 @@
+/* { dg-do run } */
+/* { dg-require-effective-target dfp } */
+/* { dg-options "-O2 -save-temps" } */
+
+
+#define DEBUG 0
+
+#ifdef DEBUG
+#include <stdio.h>
+#endif
+#include <float.h>
+
+void abort (void);
+
+int main()
+{
+#define IMM2 2
+#define IMM3 3
+#define IMM4 4
+
+ _Decimal64 srcA_dfp64, srcB_dfp64;
+ _Decimal64 result_dfp64;
+ _Decimal64 expected_result_dfp64;
+ _Decimal128 srcA_dfp128, srcB_dfp128;
+ _Decimal128 result_dfp128;
+ _Decimal128 expected_result_dfp128;
+
+ /* Third argument of quantize built-ins is the rounding mode value (RMC).
+
+ RMC Rounding Mode
+ 00 Round toward 0
+ 01 Round to -infinity
+ 10 Round to nearest, ties toward 0
+ 11 Round according to DRN */
+
+
+ /* Tests for quantize with 64-bit DFP variable. */
+ srcA_dfp64 = 100.0df;
+ srcB_dfp64 = 300.456789df;
+ expected_result_dfp64 = 300.5df;
+
+ result_dfp64 = __builtin_dfp_quantize (srcA_dfp64, srcB_dfp64, 0x0);
+
+ if (result_dfp64 != expected_result_dfp64)
+#if DEBUG
+ printf("DFP 64-bit quantize of variable, RMC = 0 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ srcA_dfp64 = 100.00df;
+ srcB_dfp64 = 300.456789df;
+ expected_result_dfp64 = 300.45df;
+
+ result_dfp64 = __builtin_dfp_quantize (srcA_dfp64, srcB_dfp64, 0x1);
+
+ if (result_dfp64 != expected_result_dfp64)
+#if DEBUG
+ printf("DFP 64-bit quantize of variable, RMC = 1 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ srcA_dfp64 = 100.001df;
+ srcB_dfp64 = 3001.456789df;
+ expected_result_dfp64 = 3001.457df;
+
+ result_dfp64 = __builtin_dfp_quantize (srcA_dfp64, srcB_dfp64, 0x2);
+
+ if (result_dfp64 != expected_result_dfp64)
+#if DEBUG
+ printf("DFP 64-bit quantize of variable, RMC = 2 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ /* Tests for 64-bit quantize with immediate value. */
+
+ srcB_dfp64 = 10.4567df;
+ expected_result_dfp64 = 000.0df;
+
+ result_dfp64 = __builtin_dfp_quantize (IMM2, srcB_dfp64, 0x0);
+
+ if (result_dfp64 != expected_result_dfp64)
+#if DEBUG
+ printf("DFP 64-bit quantize immediate, RMC = 0 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ srcB_dfp64 = 104567.891df;
+ expected_result_dfp64 = 100000.0df;
+
+ result_dfp64 = __builtin_dfp_quantize (IMM4, srcB_dfp64, 0x1);
+
+ if (result_dfp64 != expected_result_dfp64)
+#if DEBUG
+ printf("DFP 64-bit quantize immediate, RMC = 1 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ srcB_dfp64 = 109876.54321df;
+ expected_result_dfp64 = 109900.0df;
+
+ result_dfp64 = __builtin_dfp_quantize (IMM2, srcB_dfp64, 0x2);
+
+ if (result_dfp64 != expected_result_dfp64)
+#if DEBUG
+ printf("DFP 64-bit quantize immediate, RMC = 2 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ /* Tests for quantize 128-bit DFP variable. */
+ srcA_dfp128 = 0.018df;
+ srcB_dfp128 = 50000.18345df;
+ expected_result_dfp128 = 50000.180df;
+
+ result_dfp128 = __builtin_dfpq_quantize (srcA_dfp128, srcB_dfp128, 0x0);
+
+ if (result_dfp128 != expected_result_dfp128)
+#if DEBUG
+ printf("DFP 128-bit quantize variable, RMC = 0 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ srcA_dfp128 = 8.01df;
+ srcB_dfp128 = 50000.18345df;
+ expected_result_dfp128 = 50000.18df;
+
+ result_dfp128 = __builtin_dfpq_quantize (srcA_dfp128, srcB_dfp128, 0x1);
+
+ if (result_dfp128 != expected_result_dfp128)
+#if DEBUG
+ printf("DFP 128-bit quantize variable, RMC = 1 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ srcA_dfp128 = 0.1234df;
+ srcB_dfp128 = 50000.18346789df;
+ expected_result_dfp128 = 50000.1800df;
+
+ result_dfp128 = __builtin_dfpq_quantize (srcA_dfp128, srcB_dfp128, 0x2);
+
+ if (result_dfp128 != expected_result_dfp128)
+#if DEBUG
+ printf("DFP 128-bit quantize variable, RMC = 2 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ /* Tests for 128-bit quantize with immediate value. */
+ srcB_dfp128 = 1234.18345df;
+ expected_result_dfp128 = 1200.0df;
+
+ result_dfp128 = __builtin_dfpq_quantize (IMM2, srcB_dfp128, 0x0);
+
+ if (result_dfp128 != expected_result_dfp128)
+#if DEBUG
+ printf("DFP 128-bit quantize immediate, RMC = 0 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ srcB_dfp128 = 123456.18345df;
+ expected_result_dfp128 = 120000.0df;
+
+ result_dfp128 = __builtin_dfpq_quantize (IMM4, srcB_dfp128, 0x1);
+
+ if (result_dfp128 != expected_result_dfp128)
+#if DEBUG
+ printf("DFP 128-bit quantize immediate, RMC = 1 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ srcB_dfp128 = 12361834.5df;
+ expected_result_dfp128 = 12362000.0df;
+
+ result_dfp128 = __builtin_dfpq_quantize (IMM3, srcB_dfp128, 0x2);
+
+ if (result_dfp128 != expected_result_dfp128)
+#if DEBUG
+ printf("DFP 128-bit quantize immediate, RMC = 2 result does not match expected result\n");
+#else
+ abort();
+#endif
+
+ return 0;
+}
+
+/* { dg-final { scan-assembler-times {\mdqua\M} 3 } } */
+/* { dg-final { scan-assembler-times {\mdquai\M} 3 } } */
+/* { dg-final { scan-assembler-times {\mdquaq\M} 3 } } */
+/* { dg-final { scan-assembler-times {\mdquaiq\M} 3 } } */