This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH][committed] Expand trunc inline with SSE math on x86_64 and i?86
- From: Richard Guenther <rguenther at suse dot de>
- To: gcc-patches at gcc dot gnu dot org
- Date: Sun, 29 Oct 2006 17:42:23 +0100 (CET)
- Subject: [PATCH][committed] Expand trunc inline with SSE math on x86_64 and i?86
This expands C99 trunc inline using SSE instructions. On x86_64 and
for i686 SFmode we can do
double xa = fabs (x), x2;
if (!isless (xa, TWO52))
return x;
return (double)(long)x;
where i686 DFmode requires compensation code like
double xa = fabs (x), x2;
if (!isless (xa, TWO52))
return x;
xa2 = xa + TWO52 - TWO52;
Compensate:
if (xa2 > xa)
xa2 -= 1.0;
x2 = copysign (xa2, x);
return x2;
Bootstrapped and tested on {x86_64,i686}-linux-gnu. Applied to mainline
as preapproved on IRC by Roger.
This completes the series and as suggested I will go back and add
appropriate !optimize_size checks where not already done and also
audit the other intrinsics in the i386 backend.
Richard.
2006-10-29 Richard Guenther <rguenther@suse.de>
* config/i386/i386-protos.h (ix86_expand_trunc): Declare.
(ix86_expand_truncdf_32): Likewise.
* config/i386/i386.c (ix86_expand_trunc): New function expanding
trunc inline for SSE math and -fno-trapping-math and if not
optimizing for size.
(ix86_expand_truncdf_32): Same for DFmode on 32bit archs.
* config/i386/i386.md (btruncsf2, btruncdf2): Adjust expanders
for expanding btrunc inline for SSE math.
* gcc.target/i386/math-torture/trunc.c: New testcase.
Index: gcc/config/i386/i386-protos.h
===================================================================
--- gcc.orig/config/i386/i386-protos.h 2006-10-29 17:30:50.000000000 +0100
+++ gcc/config/i386/i386-protos.h 2006-10-29 17:35:41.000000000 +0100
@@ -164,6 +164,8 @@ extern void ix86_expand_floorceil (rtx,
extern void ix86_expand_floorceildf_32 (rtx, rtx, bool);
extern void ix86_expand_round (rtx, rtx);
extern void ix86_expand_rounddf_32 (rtx, rtx);
+extern void ix86_expand_trunc (rtx, rtx);
+extern void ix86_expand_truncdf_32 (rtx, rtx);
#ifdef TREE_CODE
extern void init_cumulative_args (CUMULATIVE_ARGS *, tree, rtx, tree);
Index: gcc/config/i386/i386.c
===================================================================
--- gcc.orig/config/i386/i386.c 2006-10-29 17:30:50.000000000 +0100
+++ gcc/config/i386/i386.c 2006-10-29 17:35:41.000000000 +0100
@@ -19639,6 +19639,100 @@ ix86_expand_rounddf_32 (rtx operand0, rt
emit_move_insn (operand0, res);
}
+/* Expand SSE sequence for computing trunc from OPERAND1 storing
+ into OPERAND0. */
+void
+ix86_expand_trunc (rtx operand0, rtx operand1)
+{
+ /* C code for SSE variant we expand below.
+ double xa = fabs (x), x2;
+ if (!isless (xa, TWO52))
+ return x;
+ return (double)(long)x;
+ */
+ enum machine_mode mode = GET_MODE (operand0);
+ rtx xa, xi, TWO52, label, res;
+
+ TWO52 = ix86_gen_TWO52 (mode);
+
+ /* Temporary for holding the result, initialized to the input
+ operand to ease control flow. */
+ res = gen_reg_rtx (mode);
+ emit_move_insn (res, operand1);
+
+ /* xa = abs (operand1) */
+ xa = ix86_expand_sse_fabs (res, NULL);
+
+ /* if (!isless (xa, TWO52)) goto label; */
+ label = ix86_expand_sse_compare_and_jump (UNLE, TWO52, xa, false);
+
+ /* x = (double)(long)x */
+ xi = gen_reg_rtx (mode == DFmode ? DImode : SImode);
+ expand_fix (xi, res, 0);
+ expand_float (res, xi, 0);
+
+ emit_label (label);
+ LABEL_NUSES (label) = 1;
+
+ emit_move_insn (operand0, res);
+}
+
+/* Expand SSE sequence for computing trunc from OPERAND1 storing
+ into OPERAND0. */
+void
+ix86_expand_truncdf_32 (rtx operand0, rtx operand1)
+{
+ enum machine_mode mode = GET_MODE (operand0);
+ rtx xa, mask, TWO52, label, one, res, smask;
+
+ /* C code for SSE variant we expand below.
+ double xa = fabs (x), x2;
+ if (!isless (xa, TWO52))
+ return x;
+ xa2 = xa + TWO52 - TWO52;
+ Compensate:
+ if (xa2 > xa)
+ xa2 -= 1.0;
+ x2 = copysign (xa2, x);
+ return x2;
+ */
+
+ TWO52 = ix86_gen_TWO52 (mode);
+
+ /* Temporary for holding the result, initialized to the input
+ operand to ease control flow. */
+ res = gen_reg_rtx (mode);
+ emit_move_insn (res, operand1);
+
+ /* xa = abs (operand1) */
+ xa = ix86_expand_sse_fabs (res, &smask);
+
+ /* if (!isless (xa, TWO52)) goto label; */
+ label = ix86_expand_sse_compare_and_jump (UNLE, TWO52, xa, false);
+
+ /* res = xa + TWO52 - TWO52; */
+ expand_simple_binop (mode, PLUS, xa, TWO52, res, 0, OPTAB_DIRECT);
+ expand_simple_binop (mode, MINUS, res, TWO52, res, 0, OPTAB_DIRECT);
+
+ /* generate 1.0 */
+ one = force_reg (mode, const_double_from_real_value (dconst1, mode));
+
+ /* Compensate: res = xa2 - (res > xa ? 1 : 0) */
+ mask = ix86_expand_sse_compare_mask (UNGT, res, xa, false);
+ emit_insn (gen_rtx_SET (VOIDmode, mask,
+ gen_rtx_AND (mode, mask, one)));
+ expand_simple_binop (mode, MINUS,
+ res, mask, res, 0, OPTAB_DIRECT);
+
+ /* res = copysign (res, operand1) */
+ ix86_sse_copysign_to_positive (res, res, force_reg (mode, operand1), smask);
+
+ emit_label (label);
+ LABEL_NUSES (label) = 1;
+
+ emit_move_insn (operand0, res);
+}
+
/* Expand SSE sequence for computing round from OPERAND1 storing
into OPERAND0. */
void
Index: gcc/config/i386/i386.md
===================================================================
--- gcc.orig/config/i386/i386.md 2006-10-29 17:31:30.000000000 +0100
+++ gcc/config/i386/i386.md 2006-10-29 17:37:09.000000000 +0100
@@ -18032,34 +18032,59 @@
(define_expand "btruncdf2"
[(use (match_operand:DF 0 "register_operand" ""))
(use (match_operand:DF 1 "register_operand" ""))]
- "TARGET_USE_FANCY_MATH_387
- && (!(TARGET_SSE2 && TARGET_SSE_MATH) || TARGET_MIX_SSE_I387)
- && flag_unsafe_math_optimizations"
-{
- rtx op0 = gen_reg_rtx (XFmode);
- rtx op1 = gen_reg_rtx (XFmode);
+ "(TARGET_USE_FANCY_MATH_387
+ && (!(TARGET_SSE2 && TARGET_SSE_MATH) || TARGET_MIX_SSE_I387)
+ && flag_unsafe_math_optimizations)
+ || (SSE_FLOAT_MODE_P (DFmode) && TARGET_SSE_MATH
+ && !flag_trapping_math
+ && !optimize_size)"
+{
+ if (SSE_FLOAT_MODE_P (DFmode) && TARGET_SSE_MATH
+ && !flag_trapping_math
+ && !optimize_size)
+ {
+ if (TARGET_64BIT)
+ ix86_expand_trunc (operand0, operand1);
+ else
+ ix86_expand_truncdf_32 (operand0, operand1);
+ }
+ else
+ {
+ rtx op0 = gen_reg_rtx (XFmode);
+ rtx op1 = gen_reg_rtx (XFmode);
- emit_insn (gen_extenddfxf2 (op1, operands[1]));
- emit_insn (gen_frndintxf2_trunc (op0, op1));
+ emit_insn (gen_extenddfxf2 (op1, operands[1]));
+ emit_insn (gen_frndintxf2_trunc (op0, op1));
- emit_insn (gen_truncxfdf2_i387_noop (operands[0], op0));
+ emit_insn (gen_truncxfdf2_i387_noop (operands[0], op0));
+ }
DONE;
})
(define_expand "btruncsf2"
[(use (match_operand:SF 0 "register_operand" ""))
(use (match_operand:SF 1 "register_operand" ""))]
- "TARGET_USE_FANCY_MATH_387
- && (!TARGET_SSE_MATH || TARGET_MIX_SSE_I387)
- && flag_unsafe_math_optimizations"
-{
- rtx op0 = gen_reg_rtx (XFmode);
- rtx op1 = gen_reg_rtx (XFmode);
+ "(TARGET_USE_FANCY_MATH_387
+ && (!TARGET_SSE_MATH || TARGET_MIX_SSE_I387)
+ && flag_unsafe_math_optimizations)
+ || (SSE_FLOAT_MODE_P (SFmode) && TARGET_SSE_MATH
+ && !flag_trapping_math
+ && !optimize_size)"
+{
+ if (SSE_FLOAT_MODE_P (SFmode) && TARGET_SSE_MATH
+ && !flag_trapping_math
+ && !optimize_size)
+ ix86_expand_trunc (operand0, operand1);
+ else
+ {
+ rtx op0 = gen_reg_rtx (XFmode);
+ rtx op1 = gen_reg_rtx (XFmode);
- emit_insn (gen_extendsfxf2 (op1, operands[1]));
- emit_insn (gen_frndintxf2_trunc (op0, op1));
+ emit_insn (gen_extendsfxf2 (op1, operands[1]));
+ emit_insn (gen_frndintxf2_trunc (op0, op1));
- emit_insn (gen_truncxfsf2_i387_noop (operands[0], op0));
+ emit_insn (gen_truncxfsf2_i387_noop (operands[0], op0));
+ }
DONE;
})
Index: gcc/testsuite/gcc.target/i386/math-torture/trunc.c
===================================================================
--- /dev/null 1970-01-01 00:00:00.000000000 +0000
+++ gcc/testsuite/gcc.target/i386/math-torture/trunc.c 2006-10-29 17:35:41.000000000 +0100
@@ -0,0 +1,14 @@
+/* { dg-do assemble } */
+
+float testlf (float x)
+{
+ return __builtin_truncf (x);
+}
+double testl (double x)
+{
+ return __builtin_trunc (x);
+}
+long double testll (long double x)
+{
+ return __builtin_truncl (x);
+}