+2016-03-25 Richard Henderson <rth@redhat.com>
+
+ PR target/70120
+ * config/aarch64/aarch64.c (aarch64_asm_output_pool_epilogue): New.
+ * config/aarch64/aarch64-protos.h: Declare it.
+ * config/aarch64/aarch64.h (ASM_OUTPUT_POOL_EPILOGUE): New.
+
2016-03-25 Alan Modra <amodra@gmail.com>
PR target/70052
bool aarch64_operands_ok_for_ldpstp (rtx *, bool, enum machine_mode);
bool aarch64_operands_adjust_ok_for_ldpstp (rtx *, bool, enum machine_mode);
extern bool aarch64_nopcrelative_literal_loads;
+
+extern void aarch64_asm_output_pool_epilogue (FILE *, const char *,
+ tree, HOST_WIDE_INT);
+
#endif /* GCC_AARCH64_PROTOS_H */
return default_elf_select_rtx_section (mode, x, align);
}
+/* Implement ASM_OUTPUT_POOL_EPILOGUE. */
+void
+aarch64_asm_output_pool_epilogue (FILE *f, const char *, tree,
+ HOST_WIDE_INT offset)
+{
+ /* When using per-function literal pools, we must ensure that any code
+ section is aligned to the minimal instruction length, lest we get
+ errors from the assembler re "unaligned instructions". */
+ if ((offset & 3) && aarch64_can_use_per_function_literal_pools_p ())
+ ASM_OUTPUT_ALIGN (f, 2);
+}
+
/* Costs. */
/* Helper function for rtx cost calculation. Strip a shift expression
#define EXTRA_SPECS \
{ "asm_cpu_spec", ASM_CPU_SPEC }
+#define ASM_OUTPUT_POOL_EPILOGUE aarch64_asm_output_pool_epilogue
+
#endif /* GCC_AARCH64_H */
+2016-03-25 Richard Henderson <rth@redhat.com>
+
+ PR target/70120
+ * gcc.target/aarch64/pr70120-1.c: New.
+ * gcc.target/aarch64/pr70120-2.c: New.
+ * gcc.target/aarch64/pr70120-3.c: New.
+
2016-03-25 Patrick Palka <ppalka@gcc.gnu.org>
PR c++/62212
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-Og -fschedule-insns -mno-pc-relative-literal-loads -g" } */
+
+typedef short v32u16 __attribute__ ((vector_size (32)));
+typedef int v32u32 __attribute__ ((vector_size (32)));
+typedef long v32u64 __attribute__ ((vector_size (32)));
+typedef __int128 u128;
+typedef __int128 v32u128 __attribute__ ((vector_size (32)));
+
+int
+foo(int u16_0, int u32_0, int u64_0, u128 u128_0, int u16_1, int u32_1, int u64_1, u128 u128_1, v32u16 v32u16_0, v32u32 v32u32_0, v32u64 v32u64_0, v32u128 v32u128_0, v32u16 v32u16_1, v32u32 v32u32_1, v32u64 v32u64_1, v32u128 v32u128_1)
+{
+ v32u32_1 ^= (v32u32) ~ v32u64_0;
+ v32u32_1 %= (v32u32) - v32u16_1 | 1;
+ v32u16_1 -= (v32u16) v32u16_1;
+ v32u64_0 *= (v32u64){~ u128_0, v32u16_1[5], v32u16_0[15], v32u32_1[4]};
+ v32u16_0 /= (v32u16){0x574c, ~u128_1, v32u128_1[0], u64_1, v32u64_0[1], v32u64_1[2], 0, 0x8ce6, u128_1, 0x5e69} |1;
+ return v32u16_0[0] + v32u16_0[6] + v32u16_0[8] + v32u16_0[9] + v32u32_0[0] + v32u32_0[1] + v32u32_0[2] + v32u32_0[3] + v32u32_0[4] + v32u32_0[6] + v32u64_0[0] + v32u64_0[2] + v32u64_0[3] + v32u128_0[0] + v32u128_0[1] + v32u32_1[0] + v32u32_1[2] + v32u64_1[2] + v32u64_1[3] + v32u128_1[1];
+}
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-Og -freorder-functions -g3 -mcmodel=large" } */
+
+typedef short v32u16 __attribute__ ((vector_size (32)));
+typedef int v32u32 __attribute__ ((vector_size (32)));
+typedef long v32u64 __attribute__ ((vector_size (32)));
+typedef __int128 u128;
+typedef __int128 v32u128 __attribute__ ((vector_size (32)));
+
+int
+foo (int u16_0, int u32_0, int u64_0, u128 u128_0, int u16_1, int u32_1, v32u16 v32u16_0, v32u32 v32u32_0, v32u64 v32u64_0, v32u128 v32u128_0, v32u16 v32u16_1, v32u32 v32u32_1, v32u64 v32u64_1, v32u128 v32u128_1)
+{
+ u128_0 <<= 0x6c;
+ v32u16_1 %= (v32u16) { 1, 64, 0xf294, 0, u32_1, v32u32_1[6], ~u128_0, 0x2912, v32u32_0[2]} | 1;
+ v32u16_0 ^= (v32u16){-v32u16_1[11], -u32_1, 64, ~u128_0, 0, 1, 64, ~u64_0, 0};
+ return u16_0 + u32_0 + u16_1 + v32u16_0[0] + v32u32_0[1] + v32u32_0[2] + v32u32_0[4] + v32u32_0[6] + v32u64_0[0] + v32u64_0[1] + v32u64_0[2] + v32u64_0[3] + v32u128_0[0] + v32u128_0[1] + v32u16_1[0] + v32u32_1[7] + v32u64_1[0] + v32u64_1[1] + v32u64_1[2] + v32u64_1[3] + v32u128_1[0] + v32u128_1[1];
+}
--- /dev/null
+/* { dg-do compile } */
+/* { dg-options "-O1 -ffunction-sections -mno-pc-relative-literal-loads -g3" } */
+
+typedef unsigned short u16;
+typedef unsigned short v32u16 __attribute__((vector_size(32)));
+typedef unsigned int u32;
+typedef unsigned int v32u32 __attribute__((vector_size(32)));
+typedef unsigned long long u64;
+typedef unsigned long long v32u64 __attribute__((vector_size(32)));
+typedef unsigned __int128 u128;
+typedef unsigned __int128 v32u128 __attribute__((vector_size(32)));
+u128 __attribute__((noinline, noclone))
+foo(u16 u16_0, u32 u32_0, u64 u64_0, u128 u128_0, u16 u16_1, u32 u32_1, u64 u64_1, u128 u128_1, v32u16 v32u16_0, v32u32 v32u32_0, v32u64 v32u64_0, v32u128 v32u128_0, v32u16 v32u16_1, v32u32 v32u32_1, v32u64 v32u64_1, v32u128 v32u128_1)
+{
+ v32u128_1 %= (v32u128)v32u32_1 | 1;
+ u16_1 /= ((u32)~(u128)(((u128)0xa1 << 0))) | 1;
+ v32u32_0 += (v32u32){(u16)v32u16_1[9], (u16)v32u16_1[14], (u32)-v32u32_0[7], ((u64)(u32)(((u128)0x43bc59e9 << 0))), ((u32)(u32)(((u128)0x14a47ba8f240a6 << 0))), (u128)v32u128_1[1], (u16)u16_1, (u64)-u64_1};
+ return u16_0 + u32_0 + u64_0 + u128_0 + u16_1 + u32_1 + u64_1 + u128_1 + v32u16_0[0] + v32u16_0[1] + v32u16_0[2] + v32u16_0[3] + v32u16_0[4] + v32u16_0[5] + v32u16_0[6] + v32u16_0[7] + v32u16_0[8] + v32u16_0[9] + v32u16_0[10] + v32u16_0[11] + v32u16_0[12] + v32u16_0[13] + v32u16_0[14] + v32u16_0[15] + v32u32_0[0] + v32u32_0[1] + v32u32_0[2] + v32u32_0[3] + v32u32_0[4] + v32u32_0[5] + v32u32_0[6] + v32u32_0[7] + v32u64_0[0] + v32u64_0[1] + v32u64_0[2] + v32u64_0[3] + v32u128_0[0] + v32u128_0[1] + v32u16_1[0] + v32u16_1[1] + v32u16_1[2] + v32u16_1[3] + v32u16_1[4] + v32u16_1[5] + v32u16_1[6] + v32u16_1[7] + v32u16_1[8] + v32u16_1[9] + v32u16_1[10] + v32u16_1[11] + v32u16_1[12] + v32u16_1[13] + v32u16_1[14] + v32u16_1[15] + v32u32_1[0] + v32u32_1[1] + v32u32_1[2] + v32u32_1[3] + v32u32_1[4] + v32u32_1[5] + v32u32_1[6] + v32u32_1[7] + v32u64_1[0] + v32u64_1[1] + v32u64_1[2] + v32u64_1[3] + v32u128_1[0] + v32u128_1[1];
+}
+int main()
+{
+}