Merge pull request #3987 from xianyi/develop

Merge from develop branch for 0.3.23

CMakeLists.txt

@@ -8,7 +8,7 @@ project(OpenBLAS C ASM)
 
 set(OpenBLAS_MAJOR_VERSION 0)
 set(OpenBLAS_MINOR_VERSION 3)
-set(OpenBLAS_PATCH_VERSION 22)
+set(OpenBLAS_PATCH_VERSION 22.dev)
 
 set(OpenBLAS_VERSION "${OpenBLAS_MAJOR_VERSION}.${OpenBLAS_MINOR_VERSION}.${OpenBLAS_PATCH_VERSION}")
 

Changelog.txt

@@ -1,4 +1,21 @@
 OpenBLAS ChangeLog
+====================================================================
+Version 0.3.23
+ 01-Apr-2023
+ 
+ general:
+   - fixed a serious regression in GETRF/GETF2 and ZGETRF/ZGETF2 where
+     subnormal but nonzero data elements triggered the singularity flag
+   - fixed a long-standing bug in CSPR/ZSPR in single-threaded operation
+     for cases where elements of the X vector are real numbers (or 
+     complex with only the real part zero)
+   - fixed gmake builds with the option NO_LAPACK
+   - fixed a few instances in the gmake Makefiles where expressly
+     setting NO_LAPACK=0 or NO_LAPACKE=0 would have the opposite effect
+   
+x86_64:
+   - added further CPUID values for Intel Raptor Lake
+
 ====================================================================
 Version 0.3.22
  26-Mar-2023

Makefile.install

@@ -77,7 +77,7 @@ endif
 endif
 
 ifneq ($(OSNAME), AIX)
-ifndef NO_LAPACKE
+ifneq ($(NO_LAPACKE), 1)
 	@echo Copying LAPACKE header files to $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)
 	@-install -m644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapack.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapack.h"
 	@-install -m644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapacke.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke.h"
@@ -127,7 +127,7 @@ endif
 
 else
 #install on AIX has different options syntax
-ifndef NO_LAPACKE
+ifneq ($(NO_LAPACKE), 1)
 	@echo Copying LAPACKE header files to $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)
 	@-installbsd -c -m 644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapack.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapack.h"
 	@-installbsd -c -m 644 $(NETLIB_LAPACK_DIR)/LAPACKE/include/lapacke.h "$(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke.h"

Makefile.rule

@@ -3,7 +3,7 @@
 #
 
 # This library's version
-VERSION = 0.3.22
+VERSION = 0.3.22.dev
 
 # If you set the suffix, the library name will be libopenblas_$(LIBNAMESUFFIX).a
 # and libopenblas_$(LIBNAMESUFFIX).so. Meanwhile, the soname in shared library

cpuid_x86.c

@@ -1547,6 +1547,8 @@ int get_cpuname(void){
       case 11: //family 6 exmodel 11
         switch (model) {
           case 7: // Raptor Lake
+          case 10:
+          case 15:
             if(support_avx2())
               return CPUTYPE_HASWELL;
 	    if(support_avx())
@@ -2348,6 +2350,8 @@ int get_coretype(void){
       case 11:
 	switch (model) {
 	  case 7: // Raptor Lake
+          case 10:
+          case 15:
 #ifndef NO_AVX2
 	  if(support_avx2())
             return CORE_HASWELL;

driver/level2/Makefile

@@ -92,7 +92,7 @@ CBLASOBJS += \
 	ctrsv_RUU.$(SUFFIX) ctrsv_RUN.$(SUFFIX) ctrsv_RLU.$(SUFFIX) ctrsv_RLN.$(SUFFIX) \
 	ctrsv_CUU.$(SUFFIX) ctrsv_CUN.$(SUFFIX) ctrsv_CLU.$(SUFFIX) ctrsv_CLN.$(SUFFIX)
 
-ifndef NO_LAPACK
+ifneq ($(NO_LAPACK), 1)
 CBLASOBJS += \
 	cspmv_U.$(SUFFIX) cspmv_L.$(SUFFIX) \
 	cspr_U.$(SUFFIX)  cspr_L.$(SUFFIX)  \

driver/level2/zspr_k.c

@@ -53,7 +53,7 @@ int CNAME(BLASLONG m, FLOAT alpha_r, FLOAT alpha_i,
 
   for (i = 0; i < m; i++){
 #ifndef LOWER
-    if ((X[i * 2 + 0] != ZERO) && (X[i * 2 + 1] != ZERO)) {
+    if ((X[i * 2 + 0] != ZERO) || (X[i * 2 + 1] != ZERO)) {
       AXPYU_K(i + 1, 0, 0,
 	      alpha_r * X[i * 2 + 0] - alpha_i * X[i * 2 + 1],
 	      alpha_i * X[i * 2 + 0] + alpha_r * X[i * 2 + 1],
@@ -61,7 +61,7 @@ int CNAME(BLASLONG m, FLOAT alpha_r, FLOAT alpha_i,
     }
     a += (i + 1) * 2;
 #else
-    if ((X[i * 2 + 0] != ZERO) && (X[i * 2 + 1] != ZERO)) {
+    if ((X[i * 2 + 0] != ZERO) || (X[i * 2 + 1] != ZERO)) {
       AXPYU_K(m - i, 0, 0,
 	      alpha_r * X[i * 2 + 0] - alpha_i * X[i * 2 + 1],
 	      alpha_i * X[i * 2 + 0] + alpha_r * X[i * 2 + 1],

interface/Makefile

@@ -92,9 +92,8 @@ CBLAS2OBJS    = \
 		cgemv.$(SUFFIX) cgeru.$(SUFFIX) cgerc.$(SUFFIX) \
 		ctrsv.$(SUFFIX) ctrmv.$(SUFFIX) \
 		csyr2.$(SUFFIX) cgbmv.$(SUFFIX) \
-		csbmv.$(SUFFIX) cspmv.$(SUFFIX) \
+		csbmv.$(SUFFIX) \
-		cspr.$(SUFFIX)  cspr2.$(SUFFIX) \
+		cspr2.$(SUFFIX) \
-		csymv.$(SUFFIX) csyr.$(SUFFIX)  \
 		ctbsv.$(SUFFIX) ctbmv.$(SUFFIX) \
 		ctpsv.$(SUFFIX) ctpmv.$(SUFFIX) \
 		chemv.$(SUFFIX) chbmv.$(SUFFIX) \
@@ -122,9 +121,8 @@ ZBLAS2OBJS    = \
 		zgemv.$(SUFFIX) zgeru.$(SUFFIX) zgerc.$(SUFFIX) \
 		ztrsv.$(SUFFIX) ztrmv.$(SUFFIX) \
 		zsyr2.$(SUFFIX) zgbmv.$(SUFFIX) \
-		zsbmv.$(SUFFIX) zspmv.$(SUFFIX) \
+		zsbmv.$(SUFFIX) \
-		zspr.$(SUFFIX)  zspr2.$(SUFFIX) \
+		zspr2.$(SUFFIX) \
-		zsymv.$(SUFFIX) zsyr.$(SUFFIX)  \
 		ztbsv.$(SUFFIX) ztbmv.$(SUFFIX) \
 		ztpsv.$(SUFFIX) ztpmv.$(SUFFIX) \
 		zhemv.$(SUFFIX) zhbmv.$(SUFFIX) \
@@ -447,7 +445,8 @@ QLAPACKOBJS	= \
 CLAPACKOBJS	= \
 	cgetrf.$(SUFFIX) cgetrs.$(SUFFIX) cpotrf.$(SUFFIX) cgetf2.$(SUFFIX) \
 	cpotf2.$(SUFFIX) claswp.$(SUFFIX) cgesv.$(SUFFIX) clauu2.$(SUFFIX) \
-	clauum.$(SUFFIX) ctrti2.$(SUFFIX) ctrtri.$(SUFFIX) ctrtrs.$(SUFFIX)
+	clauum.$(SUFFIX) ctrti2.$(SUFFIX) ctrtri.$(SUFFIX) ctrtrs.$(SUFFIX) \
+	cspr.$(SUFFIX) cspmv.$(SUFFIX) csymv.$(SUFFIX) csyr.$(SUFFIX)
 
 #ZLAPACKOBJS	= \
 #	zgetrf.$(SUFFIX) zgetrs.$(SUFFIX) zpotrf.$(SUFFIX) zgetf2.$(SUFFIX) \
@@ -458,8 +457,8 @@ CLAPACKOBJS	= \
 ZLAPACKOBJS	= \
 	zgetrf.$(SUFFIX) zgetrs.$(SUFFIX) zpotrf.$(SUFFIX) zgetf2.$(SUFFIX) \
 	zpotf2.$(SUFFIX) zlaswp.$(SUFFIX) zgesv.$(SUFFIX)  zlauu2.$(SUFFIX) \
-	zlauum.$(SUFFIX) ztrti2.$(SUFFIX) ztrtri.$(SUFFIX) ztrtrs.$(SUFFIX)
+	zlauum.$(SUFFIX) ztrti2.$(SUFFIX) ztrtri.$(SUFFIX) ztrtrs.$(SUFFIX) \
-
+	zspr.$(SUFFIX) zspmv.$(SUFFIX) zsymv.$(SUFFIX) zsyr.$(SUFFIX)
 
 XLAPACKOBJS	= \
 	xgetf2.$(SUFFIX) xgetrf.$(SUFFIX) xlauu2.$(SUFFIX) xlauum.$(SUFFIX) \
@@ -1021,7 +1020,7 @@ dsymv.$(SUFFIX) dsymv.$(PSUFFIX) : symv.c
 qsymv.$(SUFFIX) qsymv.$(PSUFFIX) : symv.c
 	$(CC) -c $(CFLAGS) $< -o $(@F)
 
-ifndef NO_LAPACK
+ifneq ($(NO_LAPACK), 1)
 csymv.$(SUFFIX) csymv.$(PSUFFIX) : zsymv.c
 	$(CC) -c $(CFLAGS) $< -o $(@F)
 
@@ -1041,7 +1040,7 @@ dsyr.$(SUFFIX) dsyr.$(PSUFFIX) : syr.c
 qsyr.$(SUFFIX) qsyr.$(PSUFFIX) : syr.c
 	$(CC) -c $(CFLAGS) $< -o $(@F)
 
-ifndef NO_LAPACK
+ifneq ($(NO_LAPACK), 1)
 csyr.$(SUFFIX) csyr.$(PSUFFIX) : zsyr.c
 	$(CC) -c $(CFLAGS) $< -o $(@F)
 
@@ -1115,7 +1114,7 @@ dspmv.$(SUFFIX) dspmv.$(PSUFFIX) : spmv.c
 qspmv.$(SUFFIX) qspmv.$(PSUFFIX) : spmv.c
 	$(CC) -c $(CFLAGS) $< -o $(@F)
 
-ifndef NO_LAPACK
+ifneq ($(NO_LAPACK), 1)
 cspmv.$(SUFFIX) cspmv.$(PSUFFIX) : zspmv.c
 	$(CC) -c $(CFLAGS) $< -o $(@F)
 
@@ -1135,7 +1134,7 @@ dspr.$(SUFFIX) dspr.$(PSUFFIX) : spr.c
 qspr.$(SUFFIX) qspr.$(PSUFFIX) : spr.c
 	$(CC) -c $(CFLAGS) $< -o $(@F)
 
-ifndef NO_LAPACK
+ifneq ($(NO_LAPACK), 1)
 cspr.$(SUFFIX) cspr.$(PSUFFIX) : zspr.c
 	$(CC) -c $(CFLAGS) $< -o $(@F)
 

lapack/getf2/getf2_k.c

@@ -100,16 +100,21 @@ blasint CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa,
       jp--;
       temp1 = *(b + jp);
 
-      //if (temp1 != ZERO) {
+      if (temp1 != ZERO) {
+#if defined(DOUBLE)
 	if (fabs(temp1) >= DBL_MIN ) {
-	temp1 = dp1 / temp1;
+#else
+	if (fabs(temp1) >= FLT_MIN ) {
+#endif
+  	  temp1 = dp1 / temp1;
 
-	if (jp != j) {
+	  if (jp != j) {
-	  SWAP_K(j + 1, 0, 0, ZERO, a + j, lda, a + jp, lda, NULL, 0);
+	    SWAP_K(j + 1, 0, 0, ZERO, a + j, lda, a + jp, lda, NULL, 0);
-	}
+	  }
-	if (j + 1 < m) {
+	  if (j + 1 < m) {
-	  SCAL_K(m - j - 1, 0, 0, temp1, b + j + 1, 1, NULL, 0, NULL, 0);
+	    SCAL_K(m - j - 1, 0, 0, temp1, b + j + 1, 1, NULL, 0, NULL, 0);
-	}
+	  }
+        }
       } else {
 	if (!info) info = j + 1;
       }

lapack/getf2/zgetf2_k.c

@@ -106,30 +106,34 @@ blasint CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa,
       temp1 = *(b + jp * 2 + 0);
       temp2 = *(b + jp * 2 + 1);
 
-  //    if ((temp1 != ZERO) || (temp2 != ZERO)) {
+      if ((temp1 != ZERO) || (temp2 != ZERO)) {
-	if ((fabs(temp1) >= DBL_MIN) && (fabs(temp2) >= DBL_MIN)) {
+#if defined(DOUBLE)
-		
+	if ((fabs(temp1) >= DBL_MIN) || (fabs(temp2) >= DBL_MIN)) {
-	if (jp != j) {
+#else
-	  SWAP_K(j + 1, 0, 0, ZERO, ZERO, a + j * 2, lda,
+	if ((fabs(temp1) >= FLT_MIN) || (fabs(temp2) >= FLT_MIN)) {
+#endif		
+	  if (jp != j) {
+	    SWAP_K(j + 1, 0, 0, ZERO, ZERO, a + j * 2, lda,
 		 a + jp * 2, lda, NULL, 0);
-	}
+	  }
 
-	if (fabs(temp1) >= fabs(temp2)){
+	  if (fabs(temp1) >= fabs(temp2)){
-	  ratio = temp2 / temp1;
+	    ratio = temp2 / temp1;
-	  den = dp1 /(temp1 * ( 1 + ratio * ratio));
+	    den = dp1 /(temp1 * ( 1 + ratio * ratio));
-	  temp3 =  den;
+	    temp3 =  den;
-	  temp4 = -ratio * den;
+	    temp4 = -ratio * den;
-	} else {
+	  } else {
-	  ratio = temp1 / temp2;
+	    ratio = temp1 / temp2;
-	  den = dp1 /(temp2 * ( 1 + ratio * ratio));
+	    den = dp1 /(temp2 * ( 1 + ratio * ratio));
-	  temp3 =  ratio * den;
+	    temp3 =  ratio * den;
-	  temp4 = -den;
+	    temp4 = -den;
-	}
+	  }
 
-	if (j + 1 < m) {
+	  if (j + 1 < m) {
-	  SCAL_K(m - j - 1, 0, 0, temp3, temp4,
+	    SCAL_K(m - j - 1, 0, 0, temp3, temp4,
-		 b + (j + 1) * 2, 1, NULL, 0, NULL, 0);
+	  	 b + (j + 1) * 2, 1, NULL, 0, NULL, 0);
-	}
+	  }
+        }
       } else {
 	if (!info) info = j + 1;
       }