[gcc(refs/users/meissner/heads/work119)] Fold sign or zero extension into vsx_extract from memory with constant element.

[gcc(refs/users/meissner/heads/work119)] Fold sign or zero extension into vsx_extract from memory with constant element.

[Michael Meissner]

https://gcc.gnu.org/g:5ee927b57725dcdf7d76c3295c6c2ed5c7aa7ee8

commit 5ee927b57725dcdf7d76c3295c6c2ed5c7aa7ee8
Author: Michael Meissner <meissner@linux.ibm.com>
Date:   Thu Apr 20 23:49:30 2023 -0400

    Fold sign or zero extension into vsx_extract from memory with constant element.
    
    This patch folds sign or zero convert operations into vsx_extract from memory
    where the element number is constant.
    
    2023-04-21   Michael Meissner  <meissner@linux.ibm.com>
    
    gcc/
    
            * config/rs6000/vsx.md (VSX_EXTRACT_ISIGN): New mode attribute.
            (vsx_extract_<mode>_load_to_udi): New insn.
            (vsx_extract_<mode>_load_to_sdi): New insn.
            (vsx_extract_v8hi_load_to_<su>si): New insn.
    
    gcc/testsuite/
    
            * gcc.target/powerpc/vec-extract-mem-char-1.c: New file.
            * gcc.target/powerpc/vec-extract-mem-int-1.c: New file.
            * gcc.target/powerpc/vec-extract-mem-short-1.c: New file.

Diff:
---
 gcc/config/rs6000/vsx.md                           | 71 ++++++++++++++++++++++
 .../gcc.target/powerpc/vec-extract-mem-char-1.c    | 22 +++++++
 .../gcc.target/powerpc/vec-extract-mem-int-1.c     | 37 +++++++++++
 .../gcc.target/powerpc/vec-extract-mem-short-1.c   | 37 +++++++++++
 4 files changed, 167 insertions(+)

diff --git a/gcc/config/rs6000/vsx.md b/gcc/config/rs6000/vsx.md
index 63f31980806..dadd44ffed8 100644
--- a/gcc/config/rs6000/vsx.md
+++ b/gcc/config/rs6000/vsx.md
@@ -207,6 +207,10 @@
 (define_mode_iterator VSX_EXTRACT_I2 [V16QI V8HI])
 (define_mode_iterator VSX_EXTRACT_I4 [V16QI V8HI V4SI V2DI])
 
+;; Iterator for vector extract/insert of small integer vectors that can be sign
+;; extended with the load.
+(define_mode_iterator VSX_EXTRACT_ISIGN  [V8HI V4SI])
+
 (define_mode_attr VSX_EXTRACT_WIDTH [(V16QI "b")
 		  		     (V8HI "h")
 				     (V4SI "w")])
@@ -3981,6 +3985,73 @@
    (set_attr "length" "4,4,8,4,8")
    (set_attr "isa" "*,*,*,<VSX_EX_ISA>,<VSX_EX_ISA>")])
 
+;; Extract a V16QI/V8HI/V4SI element from memory with a constant element number
+;; and zero extend it to DImode.
+(define_insn_and_split "*vsx_extract_<mode>_load_to_udi"
+  [(set (match_operand:DI 0 "register_operand" "=r,r,r,<VSX_EX>,<VSX_EX>")
+	(zero_extend:DI
+	 (vec_select:<VEC_base>
+	  (match_operand:VSX_EXTRACT_I 1 "memory_operand" "m,o,m,Z,Q")
+	  (parallel [(match_operand:QI 2 "<VSX_EXTRACT_PREDICATE>" "0,n,n,0,n")]))))
+   (clobber (match_scratch:DI 3 "=X,X,&b,X,&b"))]
+  "VECTOR_MEM_VSX_P (<MODE>mode) && TARGET_POWERPC64"
+  "#"
+  "&& 1"
+  [(set (match_dup 0)
+	(zero_extend:DI (match_dup 4)))]
+{
+  operands[4] = rs6000_adjust_vec_address (operands[0], operands[1], operands[2],
+					   operands[3], <VEC_base>mode);
+}
+  [(set_attr "type" "load,load,load,fpload,fpload")
+   (set_attr "length" "4,4,8,4,8")
+   (set_attr "isa" "*,*,*,<VSX_EX_ISA>,<VSX_EX_ISA>")])
+
+;; Extract a V16QI/V8HI/V4SI element from memory with a constant element number
+;; and sign extend it to DImode.
+(define_insn_and_split "*vsx_extract_<mode>_load_to_sdi"
+  [(set (match_operand:DI 0 "register_operand" "=r,r,r,<VSX_EX>,<VSX_EX>")
+	(sign_extend:DI
+	 (vec_select:<VEC_base>
+	  (match_operand:VSX_EXTRACT_ISIGN 1 "memory_operand" "m,o,m,Z,Q")
+	  (parallel [(match_operand:QI 2 "<VSX_EXTRACT_PREDICATE>" "0,n,n,0,n")]))))
+   (clobber (match_scratch:DI 3 "=X,X,&b,X,&b"))]
+  "VECTOR_MEM_VSX_P (<MODE>mode) && TARGET_POWERPC64"
+  "#"
+  "&& 1"
+  [(set (match_dup 0)
+	(sign_extend:DI (match_dup 4)))]
+{
+  operands[4] = rs6000_adjust_vec_address (operands[0], operands[1], operands[2],
+					   operands[3], <VEC_base>mode);
+}
+  [(set_attr "type" "load,load,load,fpload,fpload")
+   (set_attr "length" "4,4,12,8,12")
+   (set_attr "isa" "*,*,*,<VSX_EX_ISA>,<VSX_EX_ISA>")])
+
+;; Extract a V8HI element from memory with a constant element number
+;; and zero or sign extend it to SImode.
+(define_insn_and_split "*vsx_extract_v8hi_load_to_<su>si"
+  [(set (match_operand:SI 0 "register_operand" "=r,r,r,v,v")
+	(any_extend:SI
+	 (vec_select:HI
+	  (match_operand:V8HI 1 "memory_operand" "m,o,m,Z,Q")
+	  (parallel [(match_operand:QI 2 "const_0_to_7_operand" "0,n,n,0,n")]))))
+   (clobber (match_scratch:DI 3 "=X,X,&b,X,&b"))]
+  "VECTOR_MEM_VSX_P (V8HImode) && TARGET_POWERPC64"
+  "#"
+  "&& 1"
+  [(set (match_dup 0)
+	(any_extend:SI (match_dup 4)))]
+{
+  operands[4] = rs6000_adjust_vec_address (operands[0], operands[1], operands[2],
+					   operands[3], HImode);
+}
+  [(set_attr "type" "load,load,load,fpload,fpload")
+   (set_attr "length" "4,4,12,8,12")
+   (set_attr "isa" "*,*,*,p9v,p9v")])
+
+
 ;; Variable V16QI/V8HI/V4SI extract from a register
 (define_insn_and_split "vsx_extract_<mode>_var"
   [(set (match_operand:<VEC_base> 0 "gpc_reg_operand" "=r,r")
diff --git a/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-char-1.c b/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-char-1.c
new file mode 100644
index 00000000000..fa75c4c2a83
--- /dev/null
+++ b/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-char-1.c
@@ -0,0 +1,22 @@
+/* { dg-do compile { target lp64 } } */
+/* { dg-options "-O2 -mvsx" } */
+/* { dg-require-effective-target powerpc_vsx_ok } */
+
+/* Test to verify that the vec_extract with constant element numbers can load
+   QImode and fold the sign/extension into the load.  */
+
+#include <altivec.h>
+
+unsigned long long
+extract_uns_v16qi_0 (vector unsigned char *p)
+{
+  return vec_extract (*p, 0);		/* lbz, no rlwinm.  */
+}
+
+unsigned long long
+extract_uns_v16qi_1 (vector unsigned char *p)
+{
+  return vec_extract (*p, 1);		/* lbz, no rlwinm.  */
+}
+
+/* { dg-final { scan-assembler-not {\mrlwinm\M} } } */
diff --git a/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-int-1.c b/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-int-1.c
new file mode 100644
index 00000000000..e81ab4954ae
--- /dev/null
+++ b/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-int-1.c
@@ -0,0 +1,37 @@
+/* { dg-do compile { target lp64 } } */
+/* { dg-options "-O2 -mvsx" } */
+/* { dg-require-effective-target powerpc_vsx_ok } */
+
+/* Test to verify that the vec_extract with constant element numbers can load
+   SImode and fold the sign/extension into the load.  */
+
+#include <altivec.h>
+
+long long
+extract_sign_v4si_0 (vector int *p)
+{
+  return vec_extract (*p, 0);          /* lwa, no extsw.  */
+}
+
+long long
+extract_sign_v4si_1 (vector int *p)
+{
+  return vec_extract (*p, 1);          /* lwa, no extsw.  */
+}
+
+unsigned long long
+extract_uns_v4si_0 (vector unsigned int *p)
+{
+  return vec_extract (*p, 0);          /* lwz, no rldicl.  */
+}
+
+unsigned long long
+extract_uns_v4si_1 (vector unsigned int *p)
+{
+  return vec_extract (*p, 1);          /* lwz, no rldicl.  */
+}
+
+/* { dg-final { scan-assembler-times {\mlwa\M}   2 } } */
+/* { dg-final { scan-assembler-times {\mlwz\M}   2 } } */
+/* { dg-final { scan-assembler-not   {\mextsw\M}   } } */
+/* { dg-final { scan-assembler-not   {\mrldicl\M}  } } */
diff --git a/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-short-1.c b/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-short-1.c
new file mode 100644
index 00000000000..3c834a77948
--- /dev/null
+++ b/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-short-1.c
@@ -0,0 +1,37 @@
+/* { dg-do compile { target lp64 } } */
+/* { dg-options "-O2 -mvsx" } */
+/* { dg-require-effective-target powerpc_vsx_ok } */
+
+/* Test to verify that the vec_extract with constant element numbers can load
+   SImode and fold the sign/extension into the load.  */
+
+#include <altivec.h>
+
+long long
+extract_sign_v8hi_0 (vector short *p)
+{
+  return vec_extract (*p, 0);          /* lwa, no extsw.  */
+}
+
+long long
+extract_sign_v8hi_1 (vector short *p)
+{
+  return vec_extract (*p, 1);          /* lwa, no extsw.  */
+}
+
+unsigned long long
+extract_uns_v8hi_0 (vector unsigned short *p)
+{
+  return vec_extract (*p, 0);          /* lwz, no rlwinm.  */
+}
+
+unsigned long long
+extract_uns_v8hi_1 (vector unsigned short *p)
+{
+  return vec_extract (*p, 1);          /* lwz, no rlwinm.  */
+}
+
+/* { dg-final { scan-assembler-times {\mlha\M}    2 } } */
+/* { dg-final { scan-assembler-times {\mlhz\M}    2 } } */
+/* { dg-final { scan-assembler-not   {\mextsh\M}    } } */
+/* { dg-final { scan-assembler-not   {\mrlwinm\M}   } } */

[gcc(refs/users/meissner/heads/work119)] Fold sign or zero extension into vsx_extract from memory with constant element.

[Michael Meissner]

https://gcc.gnu.org/g:a94a85f32a3276600c15ad490c208df6f5e41c75

commit a94a85f32a3276600c15ad490c208df6f5e41c75
Author: Michael Meissner <meissner@linux.ibm.com>
Date:   Thu Apr 20 23:52:54 2023 -0400

    Fold sign or zero extension into vsx_extract from memory with constant element.
    
    This patch folds conversion to floating point of vsx_extract from memory of V4SI
    elements where the element number is constant.  This code optimizes things so it
    will load the integer with LFIWAX or LFIWZX directly into a vector register
    rather than loading it into a GPR and doing a direct move operation.
    
    2023-04-21   Michael Meissner  <meissner@linux.ibm.com>
    
    gcc/
    
            * config/rs6000/vsx.md (SIGN_ZERO_EXTEND): New mode attribute.
            (vsx_extract_v4si_load_to_<uns><mode>): New insn.
    
    gcc/testsuite/
    
            * gcc.target/powerpc/vec-extract-mem-int-2.c: New file.

Diff:
---
 gcc/config/rs6000/vsx.md                           | 32 ++++++++++++++++++
 .../gcc.target/powerpc/vec-extract-mem-int-2.c     | 38 ++++++++++++++++++++++
 2 files changed, 70 insertions(+)

diff --git a/gcc/config/rs6000/vsx.md b/gcc/config/rs6000/vsx.md
index dadd44ffed8..bb06abceb00 100644
--- a/gcc/config/rs6000/vsx.md
+++ b/gcc/config/rs6000/vsx.md
@@ -251,6 +251,10 @@
 			       (TF "TARGET_FLOAT128_HW
 				    && FLOAT128_IEEE_P (TFmode)")])
 
+;; Whether to use SIGN or ZERO when depending on the floating point conversion.
+(define_code_attr SIGN_ZERO_EXTEND [(float          "SIGN_EXTEND")
+				    (unsigned_float "ZERO_EXTEND")])
+
 ;; Iterator for the 2 short vector types to do a splat from an integer
 (define_mode_iterator VSX_SPLAT_I [V16QI V8HI])
 
@@ -4051,6 +4055,34 @@
    (set_attr "length" "4,4,12,8,12")
    (set_attr "isa" "*,*,*,p9v,p9v")])
 
+;; Extract a V4SI element from memory with constant element number and convert
+;; it to SFmode or DFmode using either signed or unsigned conversion.
+(define_insn_and_split "*vsx_extract_v4si_load_to_<uns><mode>"
+  [(set (match_operand:SFDF 0 "register_operand" "=wa,wa")
+	(any_float:SFDF
+	 (vec_select:SI
+	  (match_operand:V4SI 1 "memory_operand" "m,m")
+	  (parallel [(match_operand:QI 2 "const_0_to_3_operand" "n,n")]))))
+   (clobber (match_scratch:DI 3 "=&b,&b"))
+   (clobber (match_scratch:DI 4 "=f,v"))]
+  "VECTOR_MEM_VSX_P (V4SImode) && TARGET_POWERPC64"
+  "#"
+  "&& 1"
+  [(set (match_dup 4)
+	(match_dup 5))
+   (set (match_dup 0)
+	(float:SFDF (match_dup 4)))]
+{
+  if (GET_CODE (operands[4]) == SCRATCH)
+    operands[4] = gen_reg_rtx (DImode);
+
+  rtx new_mem = rs6000_adjust_vec_address (operands[4], operands[1], operands[2],
+					   operands[3], SImode);
+  operands[5] = gen_rtx_<SIGN_ZERO_EXTEND> (DImode, new_mem);
+}
+  [(set_attr "type" "fpload")
+   (set_attr "length" "12")
+   (set_attr "isa" "*,p8v")])
 
 ;; Variable V16QI/V8HI/V4SI extract from a register
 (define_insn_and_split "vsx_extract_<mode>_var"
diff --git a/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-int-2.c b/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-int-2.c
new file mode 100644
index 00000000000..91e85bf5a5b
--- /dev/null
+++ b/gcc/testsuite/gcc.target/powerpc/vec-extract-mem-int-2.c
@@ -0,0 +1,38 @@
+/* { dg-do compile { target lp64 } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-O2 -mdejagnu-cpu=power8" } */
+
+/* Test to verify that the vec_extract with constant element numbers can load
+   SImode and convert the value to float, and double by loading the value
+   directly into a vector register, and not loading up the GPRs first.  */
+
+#include <altivec.h>
+
+float
+extract_float_sign_v4si_0 (vector int *p)
+{
+  return vec_extract (*p, 0);		/* lfiwax or lxsiwax.  */
+}
+
+float
+extract_float_sign_v4si_1 (vector int *p)
+{
+  return vec_extract (*p, 1);		/* lfiwax or lxsiwax.  */
+}
+
+double
+extract_double_uns_v4si_0 (vector unsigned int *p)
+{
+  return vec_extract (*p, 0);		/* lfiwzx or lxsiwzx.  */
+}
+
+double
+extract_double_uns_v4si_3 (vector unsigned int *p)
+{
+  return vec_extract (*p, 3);		/* lfiwzx or lxsiwzx.  */
+}
+
+/* { dg-final { scan-assembler-times {\mlfiwax\M|\mlxsiwax\M}  2 } } */
+/* { dg-final { scan-assembler-times {\mlfiwzx\M|\mlxsiwzx\M}  2 } } */
+/* { dg-final { scan-assembler-not   {\mlw[az]\M}                } } */
+/* { dg-final { scan-assembler-not   {\mmtvsrw[sz]\M}            } } */