Add nallocm().

Add nallocm(), which computes the real allocation size that would result
from the corresponding allocm() call.  nallocm() is a functional
superset of OS X's malloc_good_size(), in that it takes alignment
constraints into account.

5 files changed, 131 insertions, 25 deletions

diff --git a/doc/jemalloc.xml.in b/doc/jemalloc.xml.in
index e7cc628..6aa412a 100644
--- a/doc/jemalloc.xml.in
+++ b/doc/jemalloc.xml.in
@@ -41,6 +41,7 @@
     <refname>rallocm</refname>
     <refname>sallocm</refname>
     <refname>dallocm</refname>
+    <refname>nallocm</refname>
     -->
     <refpurpose>general purpose memory allocation functions</refpurpose>
   </refnamediv>
@@ -154,6 +155,12 @@
           <paramdef>void *<parameter>ptr</parameter></paramdef>
           <paramdef>int <parameter>flags</parameter></paramdef>
         </funcprototype>
+        <funcprototype>
+          <funcdef>int <function>nallocm</function></funcdef>
+          <paramdef>size_t *<parameter>rsize</parameter></paramdef>
+          <paramdef>size_t <parameter>size</parameter></paramdef>
+          <paramdef>int <parameter>flags</parameter></paramdef>
+        </funcprototype>
       </refsect2>
     </funcsynopsis>
   </refsynopsisdiv>
@@ -301,8 +308,9 @@ for (i = 0; i < nbins; i++) {
 
       <para>The <function>allocm<parameter/></function>,
       <function>rallocm<parameter/></function>,
-      <function>sallocm<parameter/></function>, and
-      <function>dallocm<parameter/></function> functions all have a
+      <function>sallocm<parameter/></function>,
+      <function>dallocm<parameter/></function>, and
+      <function>nallocm<parameter/></function> functions all have a
       <parameter>flags</parameter> argument that can be used to specify
       options.  The functions only check the options that are contextually
       relevant.  Use bitwise or (<code language="C">|</code>) operations to
@@ -351,7 +359,9 @@ for (i = 0; i < nbins; i++) {
       least <parameter>size</parameter> bytes of memory, sets
       <parameter>*ptr</parameter> to the base address of the allocation, and
       sets <parameter>*rsize</parameter> to the real size of the allocation if
-      <parameter>rsize</parameter> is not <constant>NULL</constant>.</para>
+      <parameter>rsize</parameter> is not <constant>NULL</constant>.  Behavior
+      is undefined if <parameter>size</parameter> is
+      <constant>0<constant>.</para>
 
       <para>The <function>rallocm<parameter/></function> function resizes the
       allocation at <parameter>*ptr</parameter> to be at least
@@ -364,7 +374,8 @@ for (i = 0; i < nbins; i++) {
       language="C"><parameter>size</parameter> +
       <parameter>extra</parameter>)</code> bytes, though inability to allocate
       the extra byte(s) will not by itself result in failure.  Behavior is
-      undefined if <code language="C">(<parameter>size</parameter> +
+      undefined if <parameter>size</parameter> is <constant>0<constant>, or if
+      <code language="C">(<parameter>size</parameter> +
       <parameter>extra</parameter> &gt;
       <constant>SIZE_T_MAX</constant>)</code>.</para>
 
@@ -374,6 +385,15 @@ for (i = 0; i < nbins; i++) {
       <para>The <function>dallocm<parameter/></function> function causes the
       memory referenced by <parameter>ptr</parameter> to be made available for
       future allocations.</para>
+
+      <para>The <function>nallocm<parameter/></function> function allocates no
+      memory, but it performs the same size computation as the
+      <function>allocm<parameter/></function> function, and if
+      <parameter>rsize</parameter> is not <constant>NULL</constant> it sets
+      <parameter>*rsize</parameter> to the real size of the allocation that
+      would result from the equivalent <function>allocm<parameter/></function>
+      function call.  Behavior is undefined if
+      <parameter>size</parameter> is <constant>0<constant>.</para>
     </refsect2>
   </refsect1>
   <refsect1 id="tuning">
@@ -1857,11 +1877,13 @@ malloc_conf = "xmalloc:true";]]></programlisting>
       <title>Experimental API</title>
       <para>The <function>allocm<parameter/></function>,
       <function>rallocm<parameter/></function>,
-      <function>sallocm<parameter/></function>, and
-      <function>dallocm<parameter/></function> functions return
+      <function>sallocm<parameter/></function>,
+      <function>dallocm<parameter/></function>, and
+      <function>nallocm<parameter/></function> functions return
       <constant>ALLOCM_SUCCESS</constant> on success; otherwise they return an
-      error value.  The <function>allocm<parameter/></function> and
-      <function>rallocm<parameter/></function> functions will fail if:
+      error value.  The <function>allocm<parameter/></function>,
+      <function>rallocm<parameter/></function>, and
+      <function>nallocm<parameter/></function> functions will fail if:
         <variablelist>
           <varlistentry>
             <term><errorname>ALLOCM_ERR_OOM</errorname></term>
diff --git a/include/jemalloc/internal/jemalloc_internal.h.in b/include/jemalloc/internal/jemalloc_internal.h.in
index c21c218..aa07346 100644
--- a/include/jemalloc/internal/jemalloc_internal.h.in
+++ b/include/jemalloc/internal/jemalloc_internal.h.in
@@ -548,10 +548,6 @@ sa2u(size_t size, size_t alignment, size_t *run_size_p)
 	 *     96 |  1100000 |  32
 	 *    144 | 10100000 |  32
 	 *    192 | 11000000 |  64
-	 *
-	 * Depending on runtime settings, it is possible that arena_malloc()
-	 * will further round up to a power of two, but that never causes
-	 * correctness issues.
 	 */
 	usize = (size + (alignment - 1)) & (-alignment);
 	/*
diff --git a/include/jemalloc/jemalloc.h.in b/include/jemalloc/jemalloc.h.in
index 580a5ec..428c0d3 100644
--- a/include/jemalloc/jemalloc.h.in
+++ b/include/jemalloc/jemalloc.h.in
@@ -59,6 +59,7 @@ int	JEMALLOC_P(rallocm)(void **ptr, size_t *rsize, size_t size,
 int	JEMALLOC_P(sallocm)(const void *ptr, size_t *rsize, int flags)
     JEMALLOC_ATTR(nonnull(1));
 int	JEMALLOC_P(dallocm)(void *ptr, int flags) JEMALLOC_ATTR(nonnull(1));
+int	JEMALLOC_P(nallocm)(size_t *rsize, size_t size, int flags);
 
 #ifdef __cplusplus
 };
diff --git a/src/jemalloc.c b/src/jemalloc.c
index ccc3a20..34fd1aa 100644
--- a/src/jemalloc.c
+++ b/src/jemalloc.c
@@ -1586,6 +1586,28 @@ JEMALLOC_P(dallocm)(void *ptr, int flags)
 	return (ALLOCM_SUCCESS);
 }
 
+JEMALLOC_ATTR(visibility("default"))
+int
+JEMALLOC_P(nallocm)(size_t *rsize, size_t size, int flags)
+{
+	size_t usize;
+	size_t alignment = (ZU(1) << (flags & ALLOCM_LG_ALIGN_MASK)
+	    & (SIZE_T_MAX-1));
+
+	assert(size != 0);
+
+	if (malloc_init())
+		return (ALLOCM_ERR_OOM);
+
+	usize = (alignment == 0) ? s2u(size) : sa2u(size, alignment, NULL);
+	if (usize == 0)
+		return (ALLOCM_ERR_OOM);
+
+	if (rsize != NULL)
+		*rsize = usize;
+	return (ALLOCM_SUCCESS);
+}
+
 /*
  * End non-standard functions.
  */
diff --git a/test/allocm.c b/test/allocm.c
index 59d0002..762e350 100644
--- a/test/allocm.c
+++ b/test/allocm.c
@@ -15,24 +15,33 @@ main(void)
 {
 	int r;
 	void *p;
-	size_t sz, alignment, total, tsz;
+	size_t nsz, rsz, sz, alignment, total;
 	unsigned i;
 	void *ps[NITER];
 
 	fprintf(stderr, "Test begin\n");
 
-	sz = 0;
-	r = JEMALLOC_P(allocm)(&p, &sz, 42, 0);
+	sz = 42;
+	nsz = 0;
+	r = JEMALLOC_P(nallocm)(&nsz, sz, 0);
+	if (r != ALLOCM_SUCCESS) {
+		fprintf(stderr, "Unexpected nallocm() error\n");
+		abort();
+	}
+	rsz = 0;
+	r = JEMALLOC_P(allocm)(&p, &rsz, sz, 0);
 	if (r != ALLOCM_SUCCESS) {
 		fprintf(stderr, "Unexpected allocm() error\n");
 		abort();
 	}
-	if (sz < 42)
+	if (rsz < sz)
 		fprintf(stderr, "Real size smaller than expected\n");
+	if (nsz != rsz)
+		fprintf(stderr, "nallocm()/allocm() rsize mismatch\n");
 	if (JEMALLOC_P(dallocm)(p, 0) != ALLOCM_SUCCESS)
 		fprintf(stderr, "Unexpected dallocm() error\n");
 
-	r = JEMALLOC_P(allocm)(&p, NULL, 42, 0);
+	r = JEMALLOC_P(allocm)(&p, NULL, sz, 0);
 	if (r != ALLOCM_SUCCESS) {
 		fprintf(stderr, "Unexpected allocm() error\n");
 		abort();
@@ -40,11 +49,20 @@ main(void)
 	if (JEMALLOC_P(dallocm)(p, 0) != ALLOCM_SUCCESS)
 		fprintf(stderr, "Unexpected dallocm() error\n");
 
-	r = JEMALLOC_P(allocm)(&p, NULL, 42, ALLOCM_ZERO);
+	nsz = 0;
+	r = JEMALLOC_P(nallocm)(&nsz, sz, ALLOCM_ZERO);
+	if (r != ALLOCM_SUCCESS) {
+		fprintf(stderr, "Unexpected nallocm() error\n");
+		abort();
+	}
+	rsz = 0;
+	r = JEMALLOC_P(allocm)(&p, &rsz, sz, ALLOCM_ZERO);
 	if (r != ALLOCM_SUCCESS) {
 		fprintf(stderr, "Unexpected allocm() error\n");
 		abort();
 	}
+	if (nsz != rsz)
+		fprintf(stderr, "nallocm()/allocm() rsize mismatch\n");
 	if (JEMALLOC_P(dallocm)(p, 0) != ALLOCM_SUCCESS)
 		fprintf(stderr, "Unexpected dallocm() error\n");
 
@@ -55,12 +73,22 @@ main(void)
 	alignment = 0x80000000LU;
 	sz        = 0x80000000LU;
 #endif
-	r = JEMALLOC_P(allocm)(&p, NULL, sz, ALLOCM_ALIGN(alignment));
+	nsz = 0;
+	r = JEMALLOC_P(nallocm)(&nsz, sz, ALLOCM_ALIGN(alignment));
+	if (r == ALLOCM_SUCCESS) {
+		fprintf(stderr,
+		    "Expected error for nallocm(&nsz, %zu, 0x%x)\n",
+		    sz, ALLOCM_ALIGN(alignment));
+	}
+	rsz = 0;
+	r = JEMALLOC_P(allocm)(&p, &rsz, sz, ALLOCM_ALIGN(alignment));
 	if (r == ALLOCM_SUCCESS) {
 		fprintf(stderr,
 		    "Expected error for allocm(&p, %zu, 0x%x)\n",
 		    sz, ALLOCM_ALIGN(alignment));
 	}
+	if (nsz != rsz)
+		fprintf(stderr, "nallocm()/allocm() rsize mismatch\n");
 
 #if LG_SIZEOF_PTR == 3
 	alignment = 0x4000000000000000LLU;
@@ -69,7 +97,12 @@ main(void)
 	alignment = 0x40000000LU;
 	sz        = 0x84000001LU;
 #endif
-	r = JEMALLOC_P(allocm)(&p, NULL, sz, ALLOCM_ALIGN(alignment));
+	nsz = 0;
+	r = JEMALLOC_P(nallocm)(&nsz, sz, ALLOCM_ALIGN(alignment));
+	if (r != ALLOCM_SUCCESS)
+		fprintf(stderr, "Unexpected nallocm() error\n");
+	rsz = 0;
+	r = JEMALLOC_P(allocm)(&p, &rsz, sz, ALLOCM_ALIGN(alignment));
 	if (r == ALLOCM_SUCCESS) {
 		fprintf(stderr,
 		    "Expected error for allocm(&p, %zu, 0x%x)\n",
@@ -82,12 +115,22 @@ main(void)
 #else
 	sz   = 0xfffffff0LU;
 #endif
-	r = JEMALLOC_P(allocm)(&p, NULL, sz, ALLOCM_ALIGN(alignment));
+	nsz = 0;
+	r = JEMALLOC_P(nallocm)(&nsz, sz, ALLOCM_ALIGN(alignment));
+	if (r == ALLOCM_SUCCESS) {
+		fprintf(stderr,
+		    "Expected error for nallocm(&nsz, %zu, 0x%x)\n",
+		    sz, ALLOCM_ALIGN(alignment));
+	}
+	rsz = 0;
+	r = JEMALLOC_P(allocm)(&p, &rsz, sz, ALLOCM_ALIGN(alignment));
 	if (r == ALLOCM_SUCCESS) {
 		fprintf(stderr,
 		    "Expected error for allocm(&p, %zu, 0x%x)\n",
 		    sz, ALLOCM_ALIGN(alignment));
 	}
+	if (nsz != rsz)
+		fprintf(stderr, "nallocm()/allocm() rsize mismatch\n");
 
 	for (i = 0; i < NITER; i++)
 		ps[i] = NULL;
@@ -101,21 +144,43 @@ main(void)
 		    sz < 3 * alignment && sz < (1U << 31);
 		    sz += (alignment >> (LG_SIZEOF_PTR-1)) - 1) {
 			for (i = 0; i < NITER; i++) {
-				r = JEMALLOC_P(allocm)(&ps[i], NULL, sz,
+				nsz = 0;
+				r = JEMALLOC_P(nallocm)(&nsz, sz,
+				    ALLOCM_ALIGN(alignment) | ALLOCM_ZERO);
+				if (r != ALLOCM_SUCCESS) {
+					fprintf(stderr,
+					    "nallocm() error for size %zu"
+					    " (0x%zx): %d\n",
+					    sz, sz, r);
+					exit(1);
+				}
+				rsz = 0;
+				r = JEMALLOC_P(allocm)(&ps[i], &rsz, sz,
 				    ALLOCM_ALIGN(alignment) | ALLOCM_ZERO);
 				if (r != ALLOCM_SUCCESS) {
 					fprintf(stderr,
-					    "Error for size %zu (0x%zx): %d\n",
+					    "allocm() error for size %zu"
+					    " (0x%zx): %d\n",
 					    sz, sz, r);
 					exit(1);
 				}
+				if (rsz < sz) {
+					fprintf(stderr,
+					    "Real size smaller than"
+					    " expected\n");
+				}
+				if (nsz != rsz) {
+					fprintf(stderr,
+					    "nallocm()/allocm() rsize"
+					    " mismatch\n");
+				}
 				if ((uintptr_t)p & (alignment-1)) {
 					fprintf(stderr,
 					    "%p inadequately aligned for"
 					    " alignment: %zu\n", p, alignment);
 				}
-				JEMALLOC_P(sallocm)(ps[i], &tsz, 0);
-				total += tsz;
+				JEMALLOC_P(sallocm)(ps[i], &rsz, 0);
+				total += rsz;
 				if (total >= (MAXALIGN << 1))
 					break;
 			}