[PATCH] AArch32 v8 NEON intrinsics for numeric max/min and directed rounding to integral

[Graham Hunter]

(Without the disclaimer this time -- oops)

Hi James,

Thanks for reviewing. (Tim as well)

Updated patch attached, have retested and confirmed it all still passes.

>  Do you need help committing?

Yes please.

-Graham

From:  James Molloy <james at jamesmolloy.co.uk>
Date:  Friday, 5 September 2014 13:35
To:  Admin <Graham.Hunter at arm.com>
Cc:  llvm cfe <cfe-commits at cs.uiuc.edu>
Subject:  Re: [PATCH] AArch32 v8 NEON intrinsics for numeric max/min and
directed rounding to integral


Hi Graham,
Thanks for working on this. The implementation looks fine to me, but in
your testcase:

+float32x2_t test_vrndx_f32(float32x2_t a) {
+  // CHECK-LABEL: test_vrndx_f32
+  // CHECK-LABEL: call <2 x float> @llvm.arm.neon.vrintx.v2f32(<2 x
float> %a)
+  return vrndx_f32(a);
+}


Only the first should be a CHECK-LABEL. The other should be a CHECK.

CHECK-LABELs are only hints to FileCheck, used to help isolate individual
testcases and prevent accidental crosstalk (matching one CHECK: line in
another function accidentally).

With those changed, LGTM.

Do you need help committing?

Cheers,

James



On 5 September 2014 12:12, Graham Hunter <Graham.Hunter at arm.com> wrote:

Hi,

This patch adds support for the 32bit numeric max/min and directed
round-to-integral NEON intrinsics that were added as part of v8, along
with unit tests.

-Graham

_______________________________________________
cfe-commits mailing list
cfe-commits at cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits





-------------- next part --------------
A non-text attachment was scrubbed...
Name: vrint_and_maxmin_intrinsics_v2.patch
Type: application/octet-stream
Size: 8814 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140905/7a72a5d6/attachment.obj>