https://github.com/frasercrmck closed
https://github.com/llvm/llvm-project/pull/129045
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
https://github.com/arsenm approved this pull request.
https://github.com/llvm/llvm-project/pull/129045
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
@@ -0,0 +1,4 @@
+__attribute__((weak)) _CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE
+__clc_rsqrt(__CLC_GENTYPE val) {
+ return __CLC_FP_LIT(1.0) / __clc_sqrt(val);
frasercrmck wrote:
Good point. Perhaps we just use the elementwise sqrt builtin. I'm not sure I
want to t
https://github.com/frasercrmck updated
https://github.com/llvm/llvm-project/pull/129045
>From 4efd607ca71b91847c9271740df7aab4534fe01b Mon Sep 17 00:00:00 2001
From: Fraser Cormack
Date: Thu, 27 Feb 2025 12:48:25 +
Subject: [PATCH 1/3] [libclc] Move rsqrt to the CLC library
This also adds
@@ -0,0 +1,4 @@
+__attribute__((weak)) _CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE
+__clc_rsqrt(__CLC_GENTYPE val) {
+ return __CLC_FP_LIT(1.0) / __clc_sqrt(val);
arsenm wrote:
```suggestion
__clc_rsqrt(__CLC_GENTYPE val) {
#pragma clang fp contract(fast)
return __
@@ -0,0 +1,4 @@
+__attribute__((weak)) _CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE
+__clc_rsqrt(__CLC_GENTYPE val) {
+ return __CLC_FP_LIT(1.0) / __clc_sqrt(val);
arsenm wrote:
For this to work as you want, probably needs to directly use the sqrt builtin
(or really, w
https://github.com/frasercrmck created
https://github.com/llvm/llvm-project/pull/129045
This also adds missing half variants to certain targets.
It also optimizes some targets' implementations to perform the operation
directly in vector types, as opposed to scalarizing.
>From 4efd607ca71b9184
