Bump architecture baseline from sm_35 to sm_50 [#30]

..., so that things keep working with CUDA 12.0+.
SourceryTools · Sep 4, 2023 · 1321670 · 1321670
1 parent 1b5946d
commit 1321670
Show file tree

Hide file tree

Showing 21 changed files with 107 additions and 84 deletions.
diff --git a/README b/README
@@ -53,22 +53,17 @@ Individual nvptx-tools depend on CUDA/PTX as follows:
 
     Unless overridden with the '-m' option, the target architecture used for
     'ptxas' verification ('--gpu-name' option to 'ptxas') is deduced from the
-    '.target' directive in the input's preamble.  This means that it's the
-    responsibility of the producer (nvptx-none GCC toolchain) that the PTX code
-    works with the CUDA version used.
+    '.target' directive in the input's preamble.  As support for the Kepler
+    microarchitecture in 'ptxas' has been removed in CUDA 11.0 through 12.0,
+    sm_3x code is verified with '--gpu-name sm_50'.
 
   * nvptx-none-run
 
     Uses the CUDA Driver library ('libcuda') to run PTX binaries.
 
+Support for sm_50 was introduced with PTX ISA 4.0, CUDA 6.0, driver r331, and
+remains supported at this time (CUDA 12.2.2).
 
-For the testsuite (only), the baseline dependency is sm_35 functionality (or
-higher).
-
-Support for sm_35 has been introduced with PTX ISA 3.1, CUDA 5.0, driver r302,
-has been deprecated in CUDA 10.2 (see
-<https://docs.nvidia.com/cuda/archive/10.2/cuda-toolkit-release-notes/#deprecated-features>),
-but has not yet been removed.
 
 Certain testcases for the nvptx-tools have dependencies as follows:
 

diff --git a/nvptx-as.cc b/nvptx-as.cc
@@ -1249,26 +1249,34 @@ This program has absolutely no warranty.\n",
 	  target_arg = preamble_target_arg;
 
 	  if ((strcmp ("sm_30", target_arg) == 0)
-	      || (strcmp ("sm_32", target_arg) == 0))
+	      || (strcmp ("sm_32", target_arg) == 0)
+	      || (strcmp ("sm_35", target_arg) == 0)
+	      || (strcmp ("sm_37", target_arg) == 0))
 	    {
-	      /* Starting with CUDA 11.0, "Support for Kepler 'sm_30' and
-		 'sm_32' architecture based products is dropped", and these may
-		 no longer be specified in '--gpu-name' of 'ptxas':
+	      /* In CUDA 11.0, "Support for Kepler 'sm_30' and 'sm_32'
+		 architecture based products is dropped", and in CUDA 12.0,
+		 "Kepler architecture support is removed" (that is, sm_35,
+		 sm_37), and these may no longer be specified in '--gpu-name'
+		 of 'ptxas':
 
 		     ptxas fatal   : Value 'sm_30' is not defined for option 'gpu-name'
 
 		     ptxas fatal   : Value 'sm_32' is not defined for option 'gpu-name'
 
+		     ptxas fatal   : Value 'sm_35' is not defined for option 'gpu-name'
+
+		     ptxas fatal   : Value 'sm_37' is not defined for option 'gpu-name'
+
 		 ..., but we need to continue supporting GCC emitting
 		 '.target sm_30' code, for example.
 
 		 Detecting the CUDA/'ptxas' version and the supported
 		 '--gpu-name' options is clumsy, so in this case, just use
-		 'sm_35', which is the baseline supported by all current CUDA
+		 'sm_50', which is the baseline supported by all current CUDA
 		 versions down to CUDA 6.5, at least.  */
 	      if (verbose)
 		fprintf (stderr, "Verifying %s code", target_arg);
-	      target_arg = "sm_35";
+	      target_arg = "sm_50";
 	      if (verbose)
 		fprintf (stderr, " with %s code generation.\n", target_arg);
 	    }

diff --git a/test/GLOBAL_FUNCTION_DECL_f.s b/test/GLOBAL_FUNCTION_DECL_f.s
@@ -1,6 +1,6 @@
 // BEGIN PREAMBLE
-.version	3.1
-.target	sm_35
+.version 4.0
+.target sm_50
 .address_size 64
 // END PREAMBLE
 

diff --git a/test/GLOBAL_FUNCTION_DEF_f.s b/test/GLOBAL_FUNCTION_DEF_f.s
@@ -1,6 +1,6 @@
 // BEGIN PREAMBLE
-.version	3.1
-.target	sm_35
+.version 4.0
+.target sm_50
 .address_size 64
 // END PREAMBLE
 

diff --git a/test/as/bare-1.s b/test/as/bare-1.s
@@ -1,3 +1,3 @@
-.version 3.1
-.target sm_35
+.version 4.0
+.target sm_50
 .address_size 64
diff --git a/test/as/debug/directive-file-1.o.golden b/test/as/debug/directive-file-1.o.golden
@@ -1,8 +1,8 @@
 //     $ [nvptx-none-gcc] directive-file-1.c -o directive-file-1.s -S -g
 //     $ [nvptx-none-as] directive-file-1.s -o directive-file-1.o.golden
 // BEGIN PREAMBLE
-.version 3.1
-.target sm_35
+.version 4.0
+.target sm_50
 .address_size 64
 // END PREAMBLE
 // BEGIN GLOBAL FUNCTION DECL: f

diff --git a/test/as/debug/directive-file-1.s b/test/as/debug/directive-file-1.s
@@ -3,8 +3,8 @@
 
 
 // BEGIN PREAMBLE
-	.version	3.1
-	.target	sm_35
+	.version 4.0
+	.target	sm_50
 	.address_size 64
 // END PREAMBLE
 

diff --git a/test/as/ptxas/dummy_ptxas_log.golden b/test/as/ptxas/dummy_ptxas_log.golden
@@ -4,5 +4,5 @@ dummy ptxas
   arg: /dev/null
   arg: /dev/null
   arg: --gpu-name
-  arg: sm_35
+  arg: sm_50
   arg: -O0
diff --git a/test/as/ptxas/invoke-1.test b/test/as/ptxas/invoke-1.test
@@ -55,105 +55,125 @@ Implicit '--verify', '-m sm_2020'
 
 RUN: rm -f %t*
 RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %S/../bare-1.s -m sm_2020 > %t.stdout 2> %t.stderr
-RUN: sed -e 's|sm_35|sm_2020|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
+RUN: sed -e 's|sm_50|sm_2020|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
 RUN: cmp %t.dummy_ptxas_log.golden %t.dummy_ptxas_log
 RUN: ! test -s %t.stdout
-RUN: sed -e 's|sm_35|sm_2020|g' < %S/verbose_stderr.golden > %t.stderr.golden
+RUN: sed -e 's|sm_50|sm_2020|g' < %S/verbose_stderr.golden > %t.stderr.golden
 RUN: cmp %t.stderr.golden %t.stderr
 
 
 Implicit '--verify', preamble '.target sm_2022'
 
 RUN: rm -f %t*
-RUN: sed -e 's|sm_35|sm_2022|g' < %S/../bare-1.s > %t.bare-1.s
+RUN: sed -e 's|sm_50|sm_2022|g' < %S/../bare-1.s > %t.bare-1.s
 RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %t.bare-1.s > %t.stdout 2> %t.stderr
-RUN: sed -e 's|sm_35|sm_2022|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
+RUN: sed -e 's|sm_50|sm_2022|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
 RUN: cmp %t.dummy_ptxas_log.golden %t.dummy_ptxas_log
 RUN: ! test -s %t.stdout
-RUN: sed -e 's|sm_35|sm_2022|g' < %S/verbose_stderr.golden > %t.stderr.golden
+RUN: sed -e 's|sm_50|sm_2022|g' < %S/verbose_stderr.golden > %t.stderr.golden
 RUN: cmp %t.stderr.golden %t.stderr
 
 
-Special handling re CUDA 11.0, "Support for Kepler 'sm_30' and 'sm_32' architecture based products is dropped"
+Special handling re CUDA 11.0, "Support for Kepler 'sm_30' and 'sm_32' architecture based products is dropped", and CUDA 12.0, "Kepler architecture support is removed" (that is, sm_35, sm_37).
 
 Preamble '.target sm_3': doesn't exist; not special-cased.
 
 RUN: rm -f %t*
-RUN: sed -e 's|sm_35|sm_3|g' < %S/../bare-1.s > %t.bare-1.s
+RUN: sed -e 's|sm_50|sm_3|g' < %S/../bare-1.s > %t.bare-1.s
 RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %t.bare-1.s > %t.stdout 2> %t.stderr
-RUN: sed -e 's|sm_35|sm_3|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
+RUN: sed -e 's|sm_50|sm_3|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
 RUN: cmp %t.dummy_ptxas_log.golden %t.dummy_ptxas_log
 RUN: ! test -s %t.stdout
-RUN: sed -e 's|sm_35|sm_3|g' < %S/verbose_stderr.golden > %t.stderr.golden
+RUN: sed -e 's|sm_50|sm_3|g' < %S/verbose_stderr.golden > %t.stderr.golden
 RUN: cmp %t.stderr.golden %t.stderr
 
-Preamble '.target sm_30': special-cased to '--gpu-name sm_35'.
+Preamble '.target sm_30': special-cased to '--gpu-name sm_50'.
 
 RUN: rm -f %t*
-RUN: sed -e 's|sm_35|sm_30|g' < %S/../bare-1.s > %t.bare-1.s
+RUN: sed -e 's|sm_50|sm_30|g' < %S/../bare-1.s > %t.bare-1.s
 RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %t.bare-1.s > %t.stdout 2> %t.stderr
 RUN: cmp %S/dummy_ptxas_log.golden %t.dummy_ptxas_log
 RUN: ! test -s %t.stdout
 RUN: cmp %S/verbose_stderr-sm_30.golden %t.stderr
 
+Preamble '.target sm_30x': doesn't exist; not special-cased.
+
+RUN: rm -f %t*
+RUN: sed -e 's|sm_50|sm_30x|g' < %S/../bare-1.s > %t.bare-1.s
+RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %t.bare-1.s > %t.stdout 2> %t.stderr
+RUN: sed -e 's|sm_50|sm_30x|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
+RUN: cmp %t.dummy_ptxas_log.golden %t.dummy_ptxas_log
+RUN: ! test -s %t.stdout
+RUN: sed -e 's|sm_50|sm_30x|g' < %S/verbose_stderr.golden > %t.stderr.golden
+RUN: cmp %t.stderr.golden %t.stderr
+
 Preamble '.target sm_31': doesn't exist; not special-cased.
 
 RUN: rm -f %t*
-RUN: sed -e 's|sm_35|sm_31|g' < %S/../bare-1.s > %t.bare-1.s
+RUN: sed -e 's|sm_50|sm_31|g' < %S/../bare-1.s > %t.bare-1.s
 RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %t.bare-1.s > %t.stdout 2> %t.stderr
-RUN: sed -e 's|sm_35|sm_31|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
+RUN: sed -e 's|sm_50|sm_31|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
 RUN: cmp %t.dummy_ptxas_log.golden %t.dummy_ptxas_log
 RUN: ! test -s %t.stdout
-RUN: sed -e 's|sm_35|sm_31|g' < %S/verbose_stderr.golden > %t.stderr.golden
+RUN: sed -e 's|sm_50|sm_31|g' < %S/verbose_stderr.golden > %t.stderr.golden
 RUN: cmp %t.stderr.golden %t.stderr
 
-Preamble '.target sm_32': special-cased to '--gpu-name sm_35'.
+Preamble '.target sm_32': special-cased to '--gpu-name sm_50'.
 
 RUN: rm -f %t*
-RUN: sed -e 's|sm_35|sm_32|g' < %S/../bare-1.s > %t.bare-1.s
+RUN: sed -e 's|sm_50|sm_32|g' < %S/../bare-1.s > %t.bare-1.s
 RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %t.bare-1.s > %t.stdout 2> %t.stderr
 RUN: cmp %S/dummy_ptxas_log.golden %t.dummy_ptxas_log
 RUN: ! test -s %t.stdout
 RUN: sed -e 's|sm_30|sm_32|g' < %S/verbose_stderr-sm_30.golden > %t.stderr.golden
 RUN: cmp %t.stderr.golden %t.stderr
 
-Preamble '.target sm_35': not special-cased.
+Preamble '.target sm_32x': doesn't exist; not special-cased.
 
 RUN: rm -f %t*
-RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %S/../bare-1.s > %t.stdout 2> %t.stderr
-RUN: cmp %S/dummy_ptxas_log.golden %t.dummy_ptxas_log
+RUN: sed -e 's|sm_50|sm_32x|g' < %S/../bare-1.s > %t.bare-1.s
+RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %t.bare-1.s > %t.stdout 2> %t.stderr
+RUN: sed -e 's|sm_50|sm_32x|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
+RUN: cmp %t.dummy_ptxas_log.golden %t.dummy_ptxas_log
 RUN: ! test -s %t.stdout
-RUN: cmp %S/verbose_stderr.golden %t.stderr
+RUN: sed -e 's|sm_50|sm_32x|g' < %S/verbose_stderr.golden > %t.stderr.golden
+RUN: cmp %t.stderr.golden %t.stderr
 
-Preamble '.target sm_37': not special-cased.
+Preamble '.target sm_35': special-cased to '--gpu-name sm_50'.
 
 RUN: rm -f %t*
-RUN: sed -e 's|sm_35|sm_37|g' < %S/../bare-1.s > %t.bare-1.s
+RUN: sed -e 's|sm_50|sm_35|g' < %S/../bare-1.s > %t.bare-1.s
 RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %t.bare-1.s > %t.stdout 2> %t.stderr
-RUN: sed -e 's|sm_35|sm_37|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
-RUN: cmp %t.dummy_ptxas_log.golden %t.dummy_ptxas_log
+RUN: cmp %S/dummy_ptxas_log.golden %t.dummy_ptxas_log
 RUN: ! test -s %t.stdout
-RUN: sed -e 's|sm_35|sm_37|g' < %S/verbose_stderr.golden > %t.stderr.golden
+RUN: sed -e 's|sm_30|sm_35|g' < %S/verbose_stderr-sm_30.golden > %t.stderr.golden
 RUN: cmp %t.stderr.golden %t.stderr
 
-Preamble '.target sm_30x': doesn't exist; not special-cased.
+Preamble '.target sm_37': special-cased to '--gpu-name sm_50'.
 
 RUN: rm -f %t*
-RUN: sed -e 's|sm_35|sm_30x|g' < %S/../bare-1.s > %t.bare-1.s
+RUN: sed -e 's|sm_50|sm_37|g' < %S/../bare-1.s > %t.bare-1.s
 RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %t.bare-1.s > %t.stdout 2> %t.stderr
-RUN: sed -e 's|sm_35|sm_30x|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
-RUN: cmp %t.dummy_ptxas_log.golden %t.dummy_ptxas_log
+RUN: cmp %S/dummy_ptxas_log.golden %t.dummy_ptxas_log
 RUN: ! test -s %t.stdout
-RUN: sed -e 's|sm_35|sm_30x|g' < %S/verbose_stderr.golden > %t.stderr.golden
+RUN: sed -e 's|sm_30|sm_37|g' < %S/verbose_stderr-sm_30.golden > %t.stderr.golden
 RUN: cmp %t.stderr.golden %t.stderr
 
-Preamble '.target sm_32x': doesn't exist; not special-cased.
+Preamble '.target sm_50': not special-cased.
+
+RUN: rm -f %t*
+RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %S/../bare-1.s > %t.stdout 2> %t.stderr
+RUN: cmp %S/dummy_ptxas_log.golden %t.dummy_ptxas_log
+RUN: ! test -s %t.stdout
+RUN: cmp %S/verbose_stderr.golden %t.stderr
+
+Preamble '.target sm_52': not special-cased.
 
 RUN: rm -f %t*
-RUN: sed -e 's|sm_35|sm_32x|g' < %S/../bare-1.s > %t.bare-1.s
+RUN: sed -e 's|sm_50|sm_52|g' < %S/../bare-1.s > %t.bare-1.s
 RUN: %dummy_ptxas_path DUMMY_PTXAS_LOG=%t.dummy_ptxas_log %target_as_cmd -v -o /dev/null %t.bare-1.s > %t.stdout 2> %t.stderr
-RUN: sed -e 's|sm_35|sm_32x|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
+RUN: sed -e 's|sm_50|sm_52|g' < %S/dummy_ptxas_log.golden > %t.dummy_ptxas_log.golden
 RUN: cmp %t.dummy_ptxas_log.golden %t.dummy_ptxas_log
 RUN: ! test -s %t.stdout
-RUN: sed -e 's|sm_35|sm_32x|g' < %S/verbose_stderr.golden > %t.stderr.golden
+RUN: sed -e 's|sm_50|sm_52|g' < %S/verbose_stderr.golden > %t.stderr.golden
 RUN: cmp %t.stderr.golden %t.stderr
diff --git a/test/as/ptxas/verbose_stderr-sm_30.golden b/test/as/ptxas/verbose_stderr-sm_30.golden
@@ -1,2 +1,2 @@
-Verifying sm_30 code with sm_35 code generation.
- ptxas -c -o /dev/null /dev/null --gpu-name sm_35 -O0
+Verifying sm_30 code with sm_50 code generation.
+ ptxas -c -o /dev/null /dev/null --gpu-name sm_50 -O0
diff --git a/test/as/ptxas/verbose_stderr.golden b/test/as/ptxas/verbose_stderr.golden
@@ -1 +1 @@
- ptxas -c -o /dev/null /dev/null --gpu-name sm_35 -O0
+ ptxas -c -o /dev/null /dev/null --gpu-name sm_50 -O0
diff --git a/test/nm/1-1-C++.s b/test/nm/1-1-C++.s
@@ -2,8 +2,8 @@
 // ..., and then manually simplify.
 
 // BEGIN PREAMBLE
-	.version	3.1
-	.target	sm_35
+	.version 4.0
+	.target	sm_50
 	.address_size 64
 // END PREAMBLE
 

diff --git a/test/nm/1-1-C.s b/test/nm/1-1-C.s
@@ -2,8 +2,8 @@
 // ..., and then manually simplify.
 
 // BEGIN PREAMBLE
-	.version	3.1
-	.target	sm_35
+	.version 4.0
+	.target	sm_50
 	.address_size 64
 // END PREAMBLE
 

diff --git a/test/nm/1-2-C++.s b/test/nm/1-2-C++.s
@@ -2,8 +2,8 @@
 // ..., and then manually simplify.
 
 // BEGIN PREAMBLE
-	.version	3.1
-	.target	sm_35
+	.version 4.0
+	.target	sm_50
 	.address_size 64
 // END PREAMBLE
 

diff --git a/test/nm/1-2-C.s b/test/nm/1-2-C.s
@@ -2,8 +2,8 @@
 // ..., and then manually simplify.
 
 // BEGIN PREAMBLE
-	.version	3.1
-	.target	sm_35
+	.version 4.0
+	.target	sm_50
 	.address_size 64
 // END PREAMBLE
 

diff --git a/test/nm/libc/libc_a-dtoa.o b/test/nm/libc/libc_a-dtoa.o
@@ -1,6 +1,6 @@
 // BEGIN PREAMBLE
-.version 3.1
-.target sm_35
+.version 4.0
+.target sm_50
 .address_size 64
 // END PREAMBLE
 // BEGIN FUNCTION DECL: quorem

diff --git a/test/nm/libc/libc_a-environ.o b/test/nm/libc/libc_a-environ.o
@@ -1,6 +1,6 @@
 // BEGIN PREAMBLE
-.version 3.1
-.target sm_35
+.version 4.0
+.target sm_50
 .address_size 64
 // END PREAMBLE
 // BEGIN VAR DEF: initial_env

diff --git a/test/nm/libc/libc_a-memset.o b/test/nm/libc/libc_a-memset.o
@@ -1,6 +1,6 @@
 // BEGIN PREAMBLE
-.version 3.1
-.target sm_35
+.version 4.0
+.target sm_50
 .address_size 64
 // END PREAMBLE
 // BEGIN GLOBAL FUNCTION DECL: memset

diff --git a/test/run/exit-argc_minus_five.nvptx b/test/run/exit-argc_minus_five.nvptx
@@ -1,8 +1,8 @@
 //     $ [nvptx-none-gcc] exit-argc_minus_five.c -o exit-argc_minus_five.nvptx -O3
 // ..., and then manually simplify.
 
-.version 3.1
-.target sm_35
+.version 4.0
+.target sm_50
 .address_size 64
 .visible .entry __main (.param .u64 %in_ar0, .param .u32 %in_ar1, .param .u64 %in_ar2)
 {

diff --git a/test/run/print-argv-1.nvptx b/test/run/print-argv-1.nvptx
@@ -1,8 +1,8 @@
 //     $ [nvptx-none-gcc] print-argv-1.c -o print-argv-1.nvptx -O3
 // ..., and then manually simplify.
 
-.version 3.1
-.target sm_35
+.version 4.0
+.target sm_50
 .address_size 64
 .extern .func (.param .u32 %value_out) vprintf (.param .u64 %in_ar0, .param .u64 %in_ar1);
 .const .align 1 .u8 $LC0[8] =

diff --git a/test/run/sigsegv-1.nvptx b/test/run/sigsegv-1.nvptx
@@ -1,8 +1,8 @@
 //     $ [nvptx-none-gcc] sigsegv-1.c -o sigsegv-1.nvptx -O3
 // ..., and then manually simplify.
 
-.version 3.1
-.target sm_35
+.version 4.0
+.target sm_50
 .address_size 64
 .visible .entry __main (.param .u64 %in_ar0, .param .u32 %in_ar1, .param .u64 %in_ar2)
 {