[PATCH][GCC][ARM] Adjust costs so udiv is preferred over sdiv when both are valid. [Patch (2/2)]

[Tamar Christina <Tamar.Christina@arm.com>]

[-- Attachment #1: Type: text/plain, Size: 1323 bytes --]

Hi All, 

This patch adjusts the cost model so that when both sdiv and udiv are possible
it prefers udiv over sdiv. This was done by making sdiv slightly more expensive
instead of making udiv cheaper to keep the baseline costs of a division the same
as before.

Similar to aarch64 this patch along with my other two related mid-end changes
makes a big difference in division by constants.

Given:

int f2(int x)
{
  return ((x * x) % 300) + ((x * x) / 300);
}

we now generate

f2:
	mul	r3, r0, r0
	mov	r0, r3
	ldr	r1, .L3
	umull	r2, r3, r0, r1
	lsr	r2, r3, #5
	add	r3, r2, r2, lsl #2
	rsb	r3, r3, r3, lsl #4
	sub	r0, r0, r3, lsl #2
	add	r0, r0, r2
	bx	lr

as opposed to

f2:
	mul	r3, r0, r0
	mov	r0, r3
	ldr	r3, .L4
	push	{r4, r5}
	smull	r4, r5, r0, r3
	asr	r3, r0, #31
	rsb	r3, r3, r5, asr #5
	add	r2, r3, r3, lsl #2
	rsb	r2, r2, r2, lsl #4
	sub	r0, r0, r2, lsl #2
	add	r0, r0, r3
	pop	{r4, r5}
	bx	lr

Bootstrapped and reg tested on arm-none-eabi
with no regressions.

OK for trunk?

Thanks,
Tamar


gcc/
2017-05-02  Tamar Christina  <tamar.christina@arm.com>

	* config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive than udiv.


gcc/testsuite/
2017-05-02  Tamar Christina  <tamar.christina@arm.com>

	* gcc.target/arm/sdiv_costs_1.c: New.

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: sdiv-arm.patch --]
[-- Type: text/x-patch; name="sdiv-arm.patch", Size: 1911 bytes --]

diff --git a/gcc/config/arm/arm.c b/gcc/config/arm/arm.c
index b24143e32e2f100000f3b150f7ed0df4fabb3cc8..ecc7688b1db6309a4dd694a8e254e64abe14d7e3 100644
--- a/gcc/config/arm/arm.c
+++ b/gcc/config/arm/arm.c
@@ -9258,6 +9258,8 @@ arm_rtx_costs_internal (rtx x, enum rtx_code code, enum rtx_code outer_code,
 	*cost += COSTS_N_INSNS (speed_p ? extra_cost->mult[0].idiv : 0);
       else
 	*cost = LIBCALL_COST (2);
+
+      *cost += (code == DIV ? 1 : 0);
       return false;	/* All arguments must be in registers.  */
 
     case MOD:
@@ -9280,7 +9282,7 @@ arm_rtx_costs_internal (rtx x, enum rtx_code code, enum rtx_code outer_code,
 
     /* Fall-through.  */
     case UMOD:
-      *cost = LIBCALL_COST (2);
+      *cost = LIBCALL_COST (2) + (code == MOD ? 1 : 0);
       return false;	/* All arguments must be in registers.  */
 
     case ROTATE:
diff --git a/gcc/testsuite/gcc.target/arm/sdiv_costs_1.c b/gcc/testsuite/gcc.target/arm/sdiv_costs_1.c
new file mode 100644
index 0000000000000000000000000000000000000000..76086ab9ce28fceb37a4e8a615a38923fa7b985a
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/sdiv_costs_1.c
@@ -0,0 +1,38 @@
+/* { dg-do compile } */
+/* { dg-options "-O3 -march=armv8-a" } */
+
+/* Both sdiv and udiv can be used here, so prefer udiv.  */
+int f1 (unsigned char *p)
+{
+  return 100 / p[1];
+}
+
+int f2 (unsigned char *p, unsigned short x)
+{
+  return x / p[0];
+}
+
+int f3 (unsigned char *p, int x)
+{
+  x &= 0x7fffffff;
+  return x / p[0];
+}
+
+int f5 (unsigned char *p, unsigned short x)
+{
+  return x % p[0];
+}
+
+/* This should only generate signed divisions.  */
+int f4 (unsigned char *p)
+{
+  return -100 / p[1];
+}
+
+int f6 (unsigned char *p, short x)
+{
+  return x % p[0];
+}
+
+/* { dg-final { scan-assembler-times "udiv\tr\[0-9\]+, r\[0-9\]+, r\[0-9\]+" 4 } } */
+/* { dg-final { scan-assembler-times "sdiv\tr\[0-9\]+, r\[0-9\]+, r\[0-9\]+" 2 } } */