git: 73b83478c14b - main - biology/hyphy: Fix build with clang 13.0.0
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Mon, 21 Feb 2022 20:13:50 UTC
The branch main has been updated by jrm:
URL: https://cgit.FreeBSD.org/ports/commit/?id=73b83478c14b99126db25691ed42c50a36f9c005
commit 73b83478c14b99126db25691ed42c50a36f9c005
Author: Joseph Mingrone <jrm@FreeBSD.org>
AuthorDate: 2022-02-21 19:57:10 +0000
Commit: Joseph Mingrone <jrm@FreeBSD.org>
CommitDate: 2022-02-21 20:13:16 +0000
biology/hyphy: Fix build with clang 13.0.0
clang 13 complains about duplicate loop unrolling pragmas for clang and
CGG. According to [1], clang supports #pragma GCC unroll with identical
semantics to #pragma clang loop unroll, so remove instances of #pragma
clang loop unroll and #pragma unroll when they appear with #pragma GCC
unroll for the same loop.
[1]
https://clang.llvm.org/docs/AttributeReference.html#pragma-unroll-pragma-nounroll
Sponsored by: The FreeBSD Foundation
---
.../patch-src_core_include_function__templates.h | 10 ++
biology/hyphy/files/patch-src_core_matrix.cpp | 106 ++++++++++++++
.../hyphy/files/patch-src_core_tree__evaluator.cpp | 154 +++++++++++++++++++++
3 files changed, 270 insertions(+)
diff --git a/biology/hyphy/files/patch-src_core_include_function__templates.h b/biology/hyphy/files/patch-src_core_include_function__templates.h
new file mode 100644
index 000000000000..7ca06b1cc56a
--- /dev/null
+++ b/biology/hyphy/files/patch-src_core_include_function__templates.h
@@ -0,0 +1,10 @@
+--- src/core/include/function_templates.h.orig 2022-02-21 19:44:45 UTC
++++ src/core/include/function_templates.h
+@@ -156,7 +156,6 @@ void ArrayForEach(ARG_TYPE *array, unsigned long dimen
+ template <typename ARG_TYPE>
+ void InitializeArray(ARG_TYPE *array, unsigned long dimension,
+ ARG_TYPE &&value) {
+- #pragma clang loop unroll_count(8)
+ #pragma GCC unroll 4
+ for (unsigned long i = 0UL; i < dimension; i++) {
+ array[i] = value;
diff --git a/biology/hyphy/files/patch-src_core_matrix.cpp b/biology/hyphy/files/patch-src_core_matrix.cpp
new file mode 100644
index 000000000000..0d5720b2751b
--- /dev/null
+++ b/biology/hyphy/files/patch-src_core_matrix.cpp
@@ -0,0 +1,106 @@
+--- src/core/matrix.cpp.orig 2022-02-21 19:46:34 UTC
++++ src/core/matrix.cpp
+@@ -3391,7 +3391,6 @@ void _Matrix::AddMatrix (_Matrix& storage, _Matrix
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ #pragma GCC ivdep
+ #pragma ivdep
+ for (long idx = 0; idx < upto; idx+=16) {
+@@ -3410,7 +3409,6 @@ void _Matrix::AddMatrix (_Matrix& storage, _Matrix
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ #pragma GCC ivdep
+ #pragma ivdep
+ for (long idx = 0; idx < upto; idx+=8) {
+@@ -3440,7 +3438,6 @@ void _Matrix::AddMatrix (_Matrix& storage, _Matrix
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long idx = 0; idx < upto; idx+=16) {
+ CELL_OP (idx);
+ CELL_OP (idx+4);
+@@ -3454,7 +3451,6 @@ void _Matrix::AddMatrix (_Matrix& storage, _Matrix
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ #pragma GCC ivdep
+ #pragma ivdep
+ for (long idx = 0; idx < upto; idx+=8) {
+@@ -3895,7 +3891,6 @@ void _Matrix::Multiply (_Matrix& storage, _Matrix
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long k = 0; k < vDim; k+=4) {
+ __m256d D4, B4;
+ DO_GROUP_OP (D4, B4, k);
+@@ -3946,7 +3941,6 @@ void _Matrix::Multiply (_Matrix& storage, _Matrix
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long k = 0; k < vDim; k+=2) {
+ __m128d D4, B4;
+ DO_GROUP_OP1 (D4, B4, k);
+@@ -3998,7 +3992,6 @@ void _Matrix::Multiply (_Matrix& storage, _Matrix
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long k = 0; k < vDim; k+=2) {
+ float64x2_t D4, B4;
+ DO_GROUP_OP1 (D4, B4, k);
+@@ -4163,7 +4156,6 @@ void _Matrix::Multiply (_Matrix& storage, _Matrix
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long k = 0; k < dimm4; k+=4) {
+ __m256d D4, B4;
+ DO_GROUP_OP (D4, B4, k);
+@@ -4271,7 +4263,6 @@ void _Matrix::Multiply (_Matrix& storage, _Matrix
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long k = 0; k < dimm4; k+=2) {
+ __m128d D4, B4;
+ DO_GROUP_OP1 (D4, B4, k);
+@@ -4380,7 +4371,6 @@ void _Matrix::Multiply (_Matrix& storage, _Matrix
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long k = 0; k < dimm4; k+=2) {
+ float64x2_t D4, B4;
+ DO_GROUP_OP1 (D4, B4, k);
+@@ -4441,7 +4431,6 @@ void _Matrix::Multiply (_Matrix& storage, _Matrix
+ #pragma GCC unroll 8
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long k = 0, column = j*hDim; k < vDim; k++, column ++) {
+ resCell += row[k] * secondArg.theData [column];
+ }
+@@ -5822,7 +5811,6 @@ _Matrix* _Matrix::Exponentiate (hyFloat scale_to, b
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long c = from; c < compressedIndex[r]; c++, i++) {
+ theIndex[i] = compressedIndex[c+hDim] * vDim + r;
+ }
+@@ -6988,7 +6976,6 @@ hyFloat _Matrix::Sqr (hyFloat* _hprestrict_ sta
+ #pragma GCC unroll 4
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ #pragma GCC ivdep
+ #pragma ivdep
+ for (long k = 0; k < loopBound; k+=4) {
diff --git a/biology/hyphy/files/patch-src_core_tree__evaluator.cpp b/biology/hyphy/files/patch-src_core_tree__evaluator.cpp
new file mode 100644
index 000000000000..3adcafaf5866
--- /dev/null
+++ b/biology/hyphy/files/patch-src_core_tree__evaluator.cpp
@@ -0,0 +1,154 @@
+--- src/core/tree_evaluator.cpp.orig 2022-02-21 19:48:41 UTC
++++ src/core/tree_evaluator.cpp
+@@ -127,7 +127,6 @@ inline double _sse_sum_2 (__m128d const & x) {
+ template<long D> inline void __ll_handle_matrix_transpose (hyFloat const * __restrict transitionMatrix, hyFloat * __restrict tMatrixT) {
+ long i = 0L;
+ for (long r = 0L; r < D; r++) {
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long c = 0L; c < D; c++, i++) {
+ tMatrixT[c*D+r] = transitionMatrix[i];
+@@ -154,7 +153,6 @@ template<long D> inline bool __ll_handle_conditional_a
+ }
+ }*/
+
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long k = 0L; k < D; k++) {
+ parentConditionals[k] *= tMatrix[siteState+D*k];
+@@ -172,7 +170,6 @@ template<long D> inline bool __ll_handle_conditional_a
+ } else {
+ if (tcc) {
+ if (__builtin_expect((tcc->list_data[currentTCCIndex] & bitMaskArray.masks[currentTCCBit]) > 0 && siteID > siteFrom,0)) {
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long k = 0L; k < D; k++) {
+ childVector[k] = lastUpdatedSite[k];
+@@ -200,7 +197,6 @@ inline bool __ll_handle_conditional_array_initializati
+ }
+ if (__builtin_expect(siteState >= 0L,1)) {
+ // a single character state; sweep down the appropriate column
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long k = 0L; k < D; k++) {
+ parentConditionals[k] *= tMatrix[siteState+D*k];
+@@ -212,7 +208,6 @@ inline bool __ll_handle_conditional_array_initializati
+ } else {
+ if (tcc) {
+ if (__builtin_expect((tcc->list_data[currentTCCIndex] & bitMaskArray.masks[currentTCCBit]) > 0 && siteID > siteFrom,0)) {
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long k = 0L; k < D; k++) {
+ childVector[k] = lastUpdatedSite[k];
+@@ -542,7 +537,6 @@ template<long D> inline void __ll_product_sum_loop (hy
+ #pragma GCC unroll 8
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long c = 0; c < D; c++)
+ accumulator += tMatrix[c] * childVector[c];
+
+@@ -558,7 +552,6 @@ inline void __ll_product_sum_loop_generic (hyFloat con
+ #pragma GCC unroll 8
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long c = 0; c < D; c++)
+ accumulator += tMatrix[c] * childVector[c];
+
+@@ -595,7 +588,6 @@ template<long D, bool ADJUST> inline void __ll_loop_ha
+ fprintf (stderr, "UP %ld (%ld) %lg\n", didScale, parentCode, scaler);
+ }*/
+ if (didScale) {
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long c = 0; c < D; c++) {
+ parentConditionals [c] *= scaler;
+@@ -626,7 +618,6 @@ template<long D, bool ADJUST> inline void __ll_loop_ha
+ }*/
+
+ if (didScale) {
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long c = 0; c < D; c++) {
+ parentConditionals [c] *= scaler;
+@@ -657,7 +648,6 @@ template<bool ADJUST> inline void __ll_loop_handle_sca
+ hyFloat scaler = _computeBoostScaler(scalingAdjustments [parentCode*siteCount + siteID] * _lfScalerUpwards, sum, didScale);
+
+ if (didScale) {
+- #pragma unroll(8)
+ #pragma GCC unroll 8
+ for (long c = 0; c < D; c++) {
+ parentConditionals [c] *= scaler;
+@@ -679,7 +669,6 @@ template<bool ADJUST> inline void __ll_loop_handle_sca
+ hyFloat scaler = _computeReductionScaler (scalingAdjustments [parentCode*siteCount + siteID] * _lfScalingFactorThreshold, sum, didScale);
+
+ if (didScale) {
+- #pragma unroll(8)
+ #pragma GCC unroll 8
+ for (long c = 0; c < D; c++) {
+ parentConditionals [c] *= scaler;
+@@ -707,7 +696,6 @@ template<long D> inline void __ll_loop_handle_leaf_cas
+ } else {
+ for (long k = siteFrom; k < siteTo; k++, pp += D) {
+ hyFloat lsf = localScalingFactor[k];
+-#pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long s = 0; s < D; s++) {
+ pp[s] = lsf;
+@@ -1585,7 +1573,6 @@ hyFloat _TheTree::ComputeTreeBlockByBranch (
+ accumulator = rootConditionals[rootIndex + rootState] * theProbs[rootState];
+ rootIndex += alphabetDimension;
+ } else {
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long p = 0; p < alphabetDimension; p++,rootIndex++) {
+ accumulator += rootConditionals[rootIndex] * theProbs[p];
+@@ -1644,7 +1631,6 @@ template<long D> inline bool __lcache_loop_preface (bo
+ long siteState = lNodeFlags[nodeCode*siteCount + siteOrdering.list_data[siteID]] ;
+ if (siteState >= 0L) {
+ unsigned long target_index = siteState;
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long k = 0L; k < D; k++, target_index+=D) {
+ parentConditionals[k] *= tMatrix[target_index];
+@@ -1659,7 +1645,6 @@ template<long D> inline bool __lcache_loop_preface (bo
+ if ((tcc->list_data[currentTCCIndex] & bitMaskArray.masks[currentTCCBit]) > 0 && siteID > siteFrom)
+ // the value of this conditional vector needs to be copied from a previously stored site
+ // subtree duplication
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long k = 0UL; k < D; k++) {
+ childVector[k] = lastUpdatedSite[k];
+@@ -1688,7 +1673,6 @@ inline bool __lcache_loop_preface_generic (bool isLeaf
+ long siteState = lNodeFlags[nodeCode*siteCount + siteOrdering.list_data[siteID]] ;
+ if (siteState >= 0L) {
+ unsigned long target_index = siteState;
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long k = 0L; k < D; k++, target_index+=D) {
+ parentConditionals[k] *= tMatrix[target_index];
+@@ -1704,7 +1688,6 @@ inline bool __lcache_loop_preface_generic (bool isLeaf
+ if ((tcc->list_data[currentTCCIndex] & bitMaskArray.masks[currentTCCBit]) > 0 && siteID > siteFrom)
+ // the value of this conditional vector needs to be copied from a previously stored site
+ // subtree duplication
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (long k = 0UL; k < D; k++) {
+ childVector[k] = lastUpdatedSite[k];
+@@ -1966,7 +1949,6 @@ void _TheTree::ComputeBranchCache (
+ unsigned long k3 = 0UL;
+ for (unsigned long k = siteFrom; k < siteTo; k++) {
+ hyFloat scaler = localScalingFactor[k];
+- #pragma unroll(4)
+ #pragma GCC unroll 4
+ for (unsigned long k2 = 0UL; k2 < alphabetDimension; k2++, k3++) {
+ parentConditionals [k3] = scaler;
+@@ -2474,7 +2456,6 @@ void _TheTree::ComputeBranchCache (
+ #pragma GCC unroll 8
+ #pragma clang loop vectorize(enable)
+ #pragma clang loop interleave(enable)
+- #pragma clang loop unroll(enable)
+ for (long k = 0; k < alphabetDimension; k++) {
+ sum += parentConditionals[k];
+ }