Skip to contentSkip to conversation

CGEN unordered fpu compares + fixes

PATCH New Stafford Horne 3 participants 7 messages Jun 1, 2019 – Jun 12, 2019

lore
Share & tools
lore
#0 lore raw

cover CGEN unordered fpu compares + fixes

Hello,

As I am working on openrisc which uses cgen in binutils I have needed to make these changes.

The Unordered compares is needed for FPU unordered (NaN detecting) comparisons. These have been implemented and tested in OpenRISC (both simulation and assembler (running on FPGA hardware)).

The gen-doc changes come from me noticing the docs [0] on the cgen website are very much out of date.

Does anyone know the process to get the regenerated docs posted? I think they are beneficial and with some more updates they could be really good.

[0] https://sourceware.org/cgen/gen-doc/

Stafford Horne (2):
cgen: Add unordered compare operation
gen-doc: Updates for latest cpu definitions

 gen-all-doc   | 46 ++++++++++++++++++++++++++++++++--------------
 html.scm      |  2 +-
 rtl-c.scm     |  4 ++++
 rtx-funcs.scm |  7 +++++++
 4 files changed, 44 insertions(+), 15 deletions(-)
 mode change 100644 => 100755 gen-all-doc
Signature · 1 line
2.21.0
#1 lore raw

PATCH 1/2 cgen: Add unordered compare operation

On OpenRISC we have added FPU unordered comparison operations (NaN detection). This patch adds the unordered operation which will generate c hooks to be implemented in simulators.

An unordered comparison can now be defined as:

; Compare unordered (set flag if either r1 or r2 is NaN) (set BI flag (unordered WI r1 r2))

; Compare unordered, greater than or equal
(set BI flag (or (unordered WI r1 r2)

		   (ge WI r1 r2)))

ChangeLog:

yyyy-mm-dd Stafford Horne <[email protected]>

* rtl-c.scm (unordered): New comparison operation.
* rtx-funcs.scm (unordered): New rtx node.

---
 rtl-c.scm     | 4 ++++
 rtx-funcs.scm | 7 +++++++
 2 files changed, 11 insertions(+)
rtl-c.scm +4 -0
index 1a4b8b0..4d4d671 100644
@@ -1897,6 +1897,10 @@
 (define-fn geu (*estate* options mode s1 s2)
   (s-cmpop *estate* 'geu ">=" mode s1 s2)
 )
+(define-fn unordered (*estate* options mode s1 s2)
+  (s-cmpop *estate* 'unordered #f mode s1 s2)
+)
+
 
 (define-fn member (*estate* options mode value set)
   ;; NOTE: There are multiple evalutions of VALUE in the generated code.
rtx-funcs.scm +7 -0
index 6c28735..f16864a 100644
@@ -1086,6 +1086,13 @@
      COMPARE
      #f
 )
+; Detect NaNs
+(drn (unordered &options &mode s1 s2)
+     BI
+     (OPTIONS ANYNUMMODE RTX RTX) (NA NA MATCHEXPR MATCH2)
+     COMPARE
+     #f
+)
 
 ; Set membership.
 ; Useful in ifield assertions.
Signature · 1 line
2.21.0
#2 lore raw

PATCH 2/2 gen-doc: Updates for latest cpu definitions

In reply to Stafford Horne

Doc generation was no longer working due to several issues.

- openrisc.cpu no longer exists (its not cpu/or1k.cpu in binutils-gdb)
- Relavive paths for archfile no longer work due to wrong cwd.
- Many architectures have been added and removed.

I have be able to get this working and added all the architectures I could get working with the below changes. I have posted the results here:

http://shorne.noip.me/downloads/gen-doc/

ChangeLog:

yyyy-mm-dd Stafford Horne <[email protected]>

* gen-all-doc (archs): Add new archs and update locations. (archfile): Prepend cgendir.
(index.html gen): Change name openrisc to or1k.
* html.scm (gen-html-trailer): Change redhat.com to sourceware.org.

---
 gen-all-doc | 46 ++++++++++++++++++++++++++++++++--------------
 html.scm    |  2 +-
 2 files changed, 33 insertions(+), 15 deletions(-)
 mode change 100644 => 100755 gen-all-doc
gen-all-doc +32 -14
old mode 100644
new mode 100755
index e2a49b6..5db3200
@@ -18,7 +18,7 @@
 # <http://www.gnu.org/licenses/>.
 #
 
-# Run this script in the src/cgen directory.
+# Run this script in the binutils-gdb/cgen directory.
 
 # Exit on any error.
 set -e
@@ -26,28 +26,46 @@ set -e
 # For debugging.
 set -x
 
-# ??? Some are missing, need to add them after some research.
-#../cpu/m32c.cpu
-#../cpu/xc16x.cpu
+# ??? Broken archs not in list, grouped by error
+# h-prefixreg error
+#../cpu/cris.cpu
+# Error: unexpected operand: #f
+# ../cpu/m32c.cpu
+# ../cpu/xc16x.cpu
+# cpu/ia32.cpu
+# sh64-compact.cpu:316:3: Error: Instruction has opcode bits outside of its mask
+# ../cpu/sh.cpu
+# ia64.cpu:924:invalid argument to %sym: (if (eq? (string-length "") 0) "" ".")
+# ../cpu/ia64.cpu
+# sparc.cpu:389:1: define-hardware: hardware already defined: h-gr
+# cpu/sparc.cpu
+
+# Archs in parent directly should be from the binutils-gdb project. Others are
+# hosted here in cgen.
+
 archs="
-cpu/fr30.cpu
-cpu/arm.cpu
+../cpu/epiphany.cpu
+../cpu/fr30.cpu
 ../cpu/frv.cpu
-cpu/i960.cpu
-cpu/ip2k.cpu
+../cpu/ip2k.cpu
 ../cpu/iq2000.cpu
 ../cpu/lm32.cpu
 ../cpu/m32r.cpu
-cpu/mep.cpu
+../cpu/mep.cpu
 ../cpu/mt.cpu
-cpu/openrisc.cpu
+../cpu/or1k.cpu
+../cpu/xstormy16.cpu
+cpu/arm.cpu
+cpu/i960.cpu
+cpu/m68k.cpu
+cpu/play.cpu
+cpu/powerpc.cpu
 cpu/sh.cpu
-cpu/xstormy16.cpu
 "
 
 if [ ! -f sim.scm ]
 then
-    echo "Not in the src/cgen directory." >& 2
+    echo "Not in the cgen source directory." >& 2
     exit 1
 fi
 
@@ -66,7 +84,7 @@ export cgendir=`pwd`
 
   for a in $archs
   do
-    archfile=../$a
+    archfile=$cgendir/$a
     arch=$(basename $archfile .cpu)
 
     case $arch in
@@ -129,7 +147,7 @@ EOF
     arm-thumb.*) name="ARM Thumb" ;;
     frv-1.*) name="FRV FR400" ;;
     frv-2.*) name="FRV FR500" ;;
-    openrisc.*) name="Openrisc" ;;
+    or1k.*) name="OpenRISC" ;;
     xstormy16.*) name="Xstormy16" ;;
     *) name=`echo ${html/.html} | tr $lower $upper` ;;
     esac
html.scm +1 -1
index 3d6bb4a..0a57a73 100644
@@ -111,7 +111,7 @@ See the input .cpu file(s) for copyright information.
    "This documentation was machine generated from the cgen cpu description\n"
    "files for this architecture.\n"
    "<br>\n"
-   "<a href=\"http://sources.redhat.com/cgen/\">http://sources.redhat.com/cgen/</a>\n"
+   "<a href=\"https://sourceware.org/cgen/\">https://sourceware.org/cgen/</a>\n"
    "</body>\n"
    "</html>\n"
    )
Signature · 1 line
2.21.0
#3 lore raw

cover CGEN unordered fpu compares + fixes

In reply to Stafford Horne

Hello,

Did anyone have a chance to look at this? The binutils patches that require this are approved, but I rather not commit them unless this I get OK on these patches and push them.

-Stafford

Quoted 31 lines · Stafford Horne

Hello,

As I am working on openrisc which uses cgen in binutils I have needed to make these changes.

The Unordered compares is needed for FPU unordered (NaN detecting) comparisons. These have been implemented and tested in OpenRISC (both simulation and assembler (running on FPGA hardware)).

The gen-doc changes come from me noticing the docs [0] on the cgen website are very much out of date.

Does anyone know the process to get the regenerated docs posted? I think they are beneficial and with some more updates they could be really good.

[0] https://sourceware.org/cgen/gen-doc/

Stafford Horne (2):
cgen: Add unordered compare operation
gen-doc: Updates for latest cpu definitions

 gen-all-doc   | 46 ++++++++++++++++++++++++++++++++--------------
 html.scm      |  2 +-
 rtl-c.scm     |  4 ++++
 rtx-funcs.scm |  7 +++++++
 4 files changed, 44 insertions(+), 15 deletions(-)
 mode change 100644 => 100755 gen-all-doc
Signature · 1 line
2.21.0
#4 lore raw

cover CGEN unordered fpu compares + fixes

Did anyone have a chance to look at this? The binutils patches that require this are approved, but I rather not commit them unless this I get OK on these patches and push them.

I'm not a CGEN maintainer, but FWIW both patches look good to me.

#5 lore raw

cover CGEN unordered fpu compares + fixes

In reply to Stafford Horne

Hi -

Quoted 3 lines

Did anyone have a chance to look at this? The binutils patches that require this are approved, but I rather not commit them unless this I get OK on these patches and push them.

These are merged, thanks.

- FChE

#6 lore raw

cover CGEN unordered fpu compares + fixes

Quoted 7 lines · Frank Ch. Eigler

Hi -

Quoted 3 lines

Did anyone have a chance to look at this? The binutils patches that require this are approved, but I rather not commit them unless this I get OK on these patches and push them.

These are merged, thanks.

Thanks,

Note, I noticed the changelog was not updated. I usually do that as part of the final phase up pushing the changes to git.

Or did we stop maintaining the change log when moving to git (I would have no problem with that myself ;) )?

-Stafford